arXiv Daily Digest - 2026-02-26

Low-resource automatic speech recognition (ASR) continues to pose significant challenges, primarily due to the limited availability of transcribed data for numerous languages. While a wealth of spoken content is accessible in television dramas and online videos, Taiwanese Hokkien exemplifies this issue, with transcriptions often being scarce and the majority of available subtitles provided only in Mandarin. To address this deficiency, we introduce TG-ASR for Taiwanese Hokkien drama speech recognition, a translation-guided ASR framework that utilizes multilingual translation embeddings to enhance recognition performance in low-resource environments. The framework is centered around the parallel gated cross-attention (PGCA) mechanism, which adaptively integrates embeddings from various auxiliary languages into the ASR decoder. This mechanism facilitates robust cross-linguistic semantic guidance while ensuring stable optimization and minimizing interference between languages. To support ongoing research initiatives, we present YT-THDC, a 30-hour corpus of Taiwanese Hokkien drama speech with aligned Mandarin subtitles and manually verified Taiwanese Hokkien transcriptions. Comprehensive experiments and analyses identify the auxiliary languages that most effectively enhance ASR performance, achieving a 14.77% relative reduction in character error rate and demonstrating the efficacy of translation-guided learning for underrepresented languages in practical applications.

Metro trains often operate in highly complex environments, characterized by illumination variations, high-speed motion, and adverse weather conditions. These factors pose significant challenges for visual perception systems, especially those relying solely on conventional RGB cameras. To tackle these difficulties, we explore the integration of event cameras into the perception system, leveraging their advantages in low-light conditions, high-speed scenarios, and low power consumption. Specifically, we focus on Kilometer Marker Recognition (KMR), a critical task for autonomous metro localization under GNSS-denied conditions. In this context, we propose a robust baseline method based on a pre-trained RGB OCR foundation model, enhanced through multi-modal adaptation. Furthermore, we construct the first large-scale RGB-Event dataset, EvMetro5K, containing 5,599 pairs of synchronized RGB-Event samples, split into 4,479 training and 1,120 testing samples. Extensive experiments on EvMetro5K and other widely used benchmarks demonstrate the effectiveness of our approach for KMR. Both the dataset and source code will be released on https://github.com/Event-AHU/EvMetro5K_benchmark

Integrated Development Environments shape developers' daily experience, yet the empirical study of their usability and user experience (UX) remains limited. This work presents an LLM-assisted approach to detecting UX smells in Visual Studio Code by mining and classifying user-reported issues from the GitHub repository. Using a validated taxonomy and expert review, we identified recurring UX problems that affect the developer experience. Our results show that the majority of UX smells are concentrated in informativeness, clarity, intuitiveness, and efficiency, qualities that developers value most.

Disease progression modeling provides a robust framework to identify long-term disease trajectories from short-term biomarker data. It is a valuable tool to gain a deeper understanding of diseases with a long disease trajectory, such as Alzheimer's disease. A key limitation of most disease progression models is that they are specific to a single data type (e.g., continuous data), thereby limiting their applicability to heterogeneous, real-world datasets. To address this limitation, we propose the Mixed Events model, a novel disease progression model that handles both discrete and continuous data types. This model is implemented within the Subtype and Stage Inference (SuStaIn) framework, resulting in Mixed-SuStaIn, enabling subtype and progression modeling. We demonstrate the effectiveness of Mixed-SuStaIn through simulation experiments and real-world data from the Alzheimer's Disease Neuroimaging Initiative, showing that it performs well on mixed datasets. The code is available at: https://github.com/ucl-pond/pySuStaIn.

As the complexity of the HPC storage stack rapidly grows, domain scientists face increasing challenges in effectively utilizing HPC storage systems to achieve their desired I/O performance. To identify and address I/O issues, scientists largely rely on I/O experts to analyze their I/O traces and provide insights into potential problems. However, with a limited number of I/O experts and the growing demand for data-intensive applications, inaccessibility has become a major bottleneck, hindering scientists from maximizing their productivity. Rapid advances in LLMs make it possible to build an automated tool that brings trustworthy I/O performance diagnosis to domain scientists. However, key challenges remain, such as the inability to handle long context windows, a lack of accurate domain knowledge about HPC I/O, and the generation of hallucinations during complex interactions.In this work, we propose IOAgent as a systematic effort to address these challenges. IOAgent integrates a module-based pre-processor, a RAG-based domain knowledge integrator, and a tree-based merger to accurately diagnose I/O issues from a given Darshan trace file. Similar to an I/O expert, IOAgent provides detailed justifications and references for its diagnoses and offers an interactive interface for scientists to ask targeted follow-up questions. To evaluate IOAgent, we collected a diverse set of labeled job traces and released the first open diagnosis test suite, TraceBench. Using this test suite, we conducted extensive evaluations, demonstrating that IOAgent matches or outperforms state-of-the-art I/O diagnosis tools with accurate and useful diagnosis results. We also show that IOAgent is not tied to specific LLMs, performing similarly well with both proprietary and open-source LLMs. We believe IOAgent has the potential to become a powerful tool for scientists navigating complex HPC I/O subsystems in the future.

Bayesian deep learning (BDL) has emerged as a principled approach to produce reliable uncertainty estimates by integrating deep neural networks with Bayesian inference, and the selection of informative prior distributions remains a significant challenge. Various function-space variational inference (FSVI) regularisation methods have been presented, assigning meaningful priors over model predictions. However, these methods typically rely on a Gaussian prior, which fails to capture the heavy-tailed statistical characteristics inherent in neural network outputs. By contrast, this work proposes a novel function-space empirical Bayes regularisation framework -- termed ST-FS-EB -- which employs heavy-tailed Student's $t$ priors in both parameter and function spaces. Also, we approximate the posterior distribution through variational inference (VI), inducing an evidence lower bound (ELBO) objective based on Monte Carlo (MC) dropout. Furthermore, the proposed method is evaluated against various VI-based BDL baselines, and the results demonstrate its robust performance in in-distribution prediction, out-of-distribution (OOD) detection and handling distribution shifts.

Diversity has been gaining interest in the NLP community in recent years. At the same time, state-of-the-art transformer models such as ModernBERT use very large pre-training datasets, which are driven by size rather than by diversity. This summons for an investigation of the impact of diversity on the ModernBERT pre-training. We do so in this study, with the express intent of reducing pre-training dataset size, while retaining at least comparable performance. We compare diversity-driven sampling algorithms, so as to pick the best one. We find that diversity-driven sampling allows in some tasks to gain 10 points relative to randomly-sampled pre-training data of commensurate size. We also see that a model pre-trained for 483h on a diversity-driven dataset of 150M tokens can yield a commensurate performance to a model pre-trained for 1,775h on a randomly-driven dataset of 2.4B tokens.

In recent years, the machine learning community has increasingly embraced the optimal transport (OT) framework for modeling distributional relationships. In this work, we introduce a sample-based neural solver for computing the Wasserstein geodesic between a source and target distribution, along with the associated velocity field. Building on the dynamical formulation of the optimal transport (OT) problem, we recast the constrained optimization as a minimax problem, using deep neural networks to approximate the relevant functions. This approach not only provides the Wasserstein geodesic but also recovers the OT map, enabling direct sampling from the target distribution. By estimating the OT map, we obtain velocity estimates along particle trajectories, which in turn allow us to learn the full velocity field. The framework is flexible and readily extends to general cost functions, including the commonly used quadratic cost. We demonstrate the effectiveness of our method through experiments on both synthetic and real datasets.

Automated test generation is essential for software quality assurance, with coverage rate serving as a key metric to ensure thorough testing. Recent advancements in Large Language Models (LLMs) have shown promise in improving test generation, particularly in achieving higher coverage. However, while existing LLM-based test generation solutions perform well on small, isolated code snippets, they struggle when applied to complex methods under test. To address these issues, we propose a scalable LLM-based unit test generation method. Our approach consists of two key steps. The first step is context information retrieval, which uses both LLMs and static analysis to gather relevant contextual information associated with the complex methods under test. The second step, iterative test generation with code elimination, repeatedly generates unit tests for the code slice, tracks the achieved coverage, and selectively removes code segments that have already been covered. This process simplifies the testing task and mitigates issues arising from token limits or reduced reasoning effectiveness associated with excessively long contexts. Through comprehensive evaluations on open-source projects, our approach outperforms state-of-the-art LLM-based and search-based methods, demonstrating its effectiveness in achieving high coverage on complex methods.

The optimization of complex medical appointment scheduling remains a significant operational challenge in multi-center healthcare environments, where clinical safety protocols and patient logistics must be reconciled. This study proposes and evaluates a Genetic Algorithm (GA) framework designed to automate the scheduling of multiple medical acts while adhering to rigorous inter-procedural incompatibility rules. Using a synthetic dataset encompassing 50 medical acts across four healthcare facilities, we compared two GA variants, Pre-Ordered and Unordered, against deterministic First-Come, First-Served (FCFS) and Random Choice baselines. Our results demonstrate that the GA framework achieved a 100% constraint fulfillment rate, effectively resolving temporal overlaps and clinical incompatibilities that the FCFS baseline failed to address in 60% and 40% of cases, respectively. Furthermore, the GA variants demonstrated statistically significant improvements (p < 0.001) in patient-centric metrics, achieving an Idle Time Ratio (ITR) frequently below 0.4 and reducing inter-healthcenter trips. While the GA (Ordered) variant provided a superior initial search locus, both evolutionary models converged to comparable global optima by the 100th generation. These findings suggest that transitioning from manual, human-mediated scheduling to an automated metaheuristic approach enhances clinical integrity, reduces administrative overhead, and significantly improves the patient experience by minimizing wait times and logistical burdens.

Medical images are essential for diagnosis, treatment planning, and research, but their quality is often degraded by noise from low-dose acquisition, patient motion, or scanner limitations, affecting both clinical interpretation and downstream analysis. Traditional filtering approaches often over-smooth and lose fine anatomical details, while deep learning methods, including CNNs, GANs, and transformers, may struggle to preserve such details or require large, computationally expensive models, limiting clinical practicality. We propose PatchDenoiser, a lightweight, energy-efficient multi-scale patch-based denoising framework. It decomposes denoising into local texture extraction and global context aggregation, fused via a spatially aware patch fusion strategy. This design enables effective noise suppression while preserving fine structural and anatomical details. PatchDenoiser is ultra-lightweight, with far fewer parameters and lower computational complexity than CNN-, GAN-, and transformer-based denoisers. On the 2016 Mayo Low-Dose CT dataset, PatchDenoiser consistently outperforms state-of-the-art CNN- and GAN-based methods in PSNR and SSIM. It is robust to variations in slice thickness, reconstruction kernels, and HU windows, generalizes across scanners without fine-tuning, and reduces parameters by ~9x and energy consumption per inference by ~27x compared with conventional CNN denoisers. PatchDenoiser thus provides a practical, scalable, and computationally efficient solution for medical image denoising, balancing performance, robustness, and clinical deployability.

Recent work has examined language models from a linguistic perspective to better understand how they acquire language. Most existing benchmarks focus on judging grammatical acceptability, whereas the ability to interpret meanings conveyed by grammatical forms has received much less attention. We introduce the Linguistic Minimal-Pair Benchmark for Evaluating Constructional Understanding in Language Models (CxMP), a benchmark grounded in Construction Grammar that treats form-meaning pairings, or constructions, as fundamental linguistic units. CxMP evaluates whether models can interpret the semantic relations implied by constructions, using a controlled minimal-pair design across nine construction types, including the let-alone, caused motion, and ditransitive constructions. Our results show that while syntactic competence emerges early, constructional understanding develops more gradually and remains limited even in large language models (LLMs). CxMP thus reveals persistent gaps in how language models integrate form and meaning, providing a framework for studying constructional understanding and learning trajectories in language models.

Critical applications in areas such as medicine, robotics and autonomous systems require compact (i.e., memory efficient), uncertainty-aware neural networks suitable for edge and other resource-constrained deployments. We study compact spectral circulant and block-circulant-with-circulant-blocks (BCCB) layers: FFT-diagonalizable circular convolutions whose weights live directly in the real FFT (RFFT) half (1D) or half-plane (2D). Parameterizing filters in the frequency domain lets us impose simple spectral structure, perform structured variational inference in a low-dimensional weight space, and calculate exact layer spectral norms, enabling inexpensive global Lipschitz bounds and margin-based robustness diagnostics. By placing independent complex Gaussians on the Hermitian support we obtain a discrete instance of the spectral representation of stationary kernels, inducing an exact stationary Gaussian-process prior over filters on the discrete circle/torus. We exploit this to define a practical spectral prior and a Hermitian-aware low-rank-plus-diagonal variational posterior in real coordinates. Empirically, spectral circulant/BCCB layers are effective compact building blocks in both (variational) Bayesian and point estimate regimes: compact Bayesian neural networks on MNIST->Fashion-MNIST, variational heads on frozen CIFAR-10 features, and deterministic ViT projections on CIFAR-10/Tiny ImageNet; spectral layers match strong baselines while using substantially fewer parameters and with tighter Lipschitz certificates.

We investigate the robustness of sparse artificial neural networks trained with adaptive topology. We focus on a simple yet effective architecture consisting of three sparse layers with 99% sparsity followed by a dense layer, applied to image classification tasks such as MNIST and Fashion MNIST. By updating the topology of the sparse layers between each epoch, we achieve competitive accuracy despite the significantly reduced number of weights. Our primary contribution is a detailed analysis of the robustness of these networks, exploring their performance under various perturbations including random link removal, adversarial attack, and link weight shuffling. Through extensive experiments, we demonstrate that adaptive topology not only enhances efficiency but also maintains robustness. This work highlights the potential of adaptive sparse networks as a promising direction for developing efficient and reliable deep learning models.

To reduce carbon emissions and minimize shipping costs, improving the fuel efficiency of ships is crucial. Various measures are taken to reduce the total fuel consumption of ships, including optimizing vessel parameters and selecting routes with the lowest fuel consumption. Different estimation methods are proposed for predicting fuel consumption, while various optimization methods are proposed to minimize fuel oil consumption. This paper provides a comprehensive review of methods for estimating and optimizing fuel oil consumption in maritime transport. Our novel contributions include categorizing fuel oil consumption \& estimation methods into physics-based, machine-learning, and hybrid models, exploring their strengths and limitations. Furthermore, we highlight the importance of data fusion techniques, which combine AIS, onboard sensors, and meteorological data to enhance accuracy. We make the first attempt to discuss the emerging role of Explainable AI in enhancing model transparency for decision-making. Uniquely, key challenges, including data quality, availability, and the need for real-time optimization, are identified, and future research directions are proposed to address these gaps, with a focus on hybrid models, real-time optimization, and the standardization of datasets.

Federated recommendation facilitates collaborative model training across distributed clients while keeping sensitive user interaction data local. Conventional approaches typically rely on synchronizing high-dimensional item representations between the server and clients. This paradigm implicitly assumes that precise geometric alignment of embedding coordinates is necessary for collaboration across clients. We posit that establishing relative semantic relationships among items is more effective than enforcing shared representations. Specifically, global semantic relations serve as structural constraints for items. Within these constraints, the framework allows item representations to vary locally on each client, which flexibility enables the model to capture fine-grained user personalization while maintaining global consistency. To this end, we propose Cluster-Guided FedRec framework (CGFedRec), a framework that transforms uploaded embeddings into compact cluster labels. In this framework, the server functions as a global structure discoverer to learn item clusters and distributes only the resulting labels. This mechanism explicitly cuts off the downstream transmission of item embeddings, relieving clients from maintaining global shared item embeddings. Consequently, CGFedRec achieves the effective injection of global collaborative signals into local item representations without transmitting full embeddings. Extensive experiments demonstrate that our approach significantly improves communication efficiency while maintaining superior recommendation accuracy across multiple datasets.

Hierarchical quantum classifiers, such as quantum convolutional neural networks (QCNNs), represent recent progress toward designing effective and feasible architectures for quantum classification. However, their performance on near-term quantum hardware remains highly sensitive to noise accumulation across circuit depth, calling for strategies beyond circuit-architecture design alone. We propose a noise-adaptive hybrid QCNN that improves classification under noise by exploiting depth-stratified intermediate measurements. Instead of discarding qubits removed during pooling operations, we measure them and use the resulting outcomes as classical features that are jointly processed by a classical neural network. This hybrid hierarchical design enables noise-adaptive inference by integrating quantum intermediate measurements with classical post-processing. Systematic experiments across multiple circuit sizes and noise settings, including hardware-calibrated noise models derived from IBM Quantum backend data, demonstrate more stable convergence, reduced loss variability, and consistently higher classification accuracy compared with standard QCNNs. Moreover, we observe that this performance advantage significantly amplifies as the circuit size increases, confirming that the hybrid architecture mitigates the scaling limitations of standard architectures. Notably, the multi-basis measurement variant attains performance close to the noiseless limit even under realistic noise. While demonstrated for QCNNs, the proposed depth-stratified feature extraction applies more broadly to hierarchical quantum classifiers that progressively discard qubits.

Knowledge graph reasoning (KGR) infers missing facts, with recent advances increasingly harnessing the semantic priors and reasoning abilities of Large Language Models (LLMs). However, prevailing generative paradigms are prone to memorizing surface-level co-occurrences rather than learning genuine relational semantics, limiting out-of-distribution generalization. To address this, we propose RADAR, which reformulates KGR from generative pattern matching to discriminative relational reasoning. We recast KGR as discriminative entity selection, where reinforcement learning enforces relative entity separability beyond token-likelihood imitation. Leveraging this separability, inference operates directly in representation space, ensuring consistency with the discriminative optimization and bypassing generation-induced hallucinations. Across four benchmarks, RADAR achieves 5-6% relative gains on link prediction and triple classification over strong LLM baselines, while increasing task-relevant mutual information in intermediate representations by 62.9%, indicating more robust and transferable relational reasoning.

Multimodal large language models (MLLMs) have shown great potential in medical applications, yet existing benchmarks inadequately capture real-world clinical complexity. We introduce MEDSYN, a multilingual, multimodal benchmark of highly complex clinical cases with up to 7 distinct visual clinical evidence (CE) types per case. Mirroring clinical workflow, we evaluate 18 MLLMs on differential diagnosis (DDx) generation and final diagnosis (FDx) selection. While top models often match or even outperform human experts on DDx generation, all MLLMs exhibit a much larger DDx--FDx performance gap compared to expert clinicians, indicating a failure mode in synthesis of heterogeneous CE types. Ablations attribute this failure to (i) overreliance on less discriminative textual CE ($\it{e.g.}$, medical history) and (ii) a cross-modal CE utilization gap. We introduce Evidence Sensitivity to quantify the latter and show that a smaller gap correlates with higher diagnostic accuracy. Finally, we demonstrate how it can be used to guide interventions to improve model performance. We will open-source our benchmark and code.

In this paper, we investigate a problem of minimizing total energy consumption for secure federated learning (FL) over wireless edge networks. To address the high computational cost and privacy challenges in conventional FL with neural networks (NN) for resource-constrained users, we propose a novel FL with hyperdimensional computing and differential privacy (FL-HDC-DP) framework. In the considered model, each edge user employs hyperdimensional computing (HDC) for local training, which replaces complex neural updates with simple hypervector operations, and applies differential privacy (DP) noise to protect transmitted model information. We optimize the total energy of computation and communication under both latency and privacy constraints. We formulate the problem as an optimization that minimizes the total energy of all users by jointly allocating HDC dimension, transmission time, system bandwidth, transmit power, and CPU frequency. To solve this problem, a sigmoid-variant function is proposed to characterize the relationship between the HDC dimension and the convergence rounds required to reach a target accuracy. Based on this model, we develop two alternating optimization algorithms, where closed-form expressions for time, frequency, bandwidth, and power allocations are derived at each iteration. Since the iterative algorithm requires a feasible initialization, we construct a feasibility problem and obtain feasible initial resource parameters by solving a per round transmission time minimization problem. Simulation results demonstrate that the proposed FL-HDC-DP framework achieves up to 83.3% total energy reduction compared with the baseline, while attaining about 90% accuracy in approximately 3.5X fewer communication rounds than the NN baseline.

Generative Adversarial Networks (GAN) have been used in many studies to synthesise mixed tabular data. Conditional tabular GAN (CTGAN) have been the most popular variant but struggle to effectively navigate the risk-utility trade-off. Bayesian GAN have received less attention for tabular data, but have been explored with unstructured data such as images and text. The most used technique employed in Bayesian GAN is Markov Chain Monte Carlo (MCMC), but it is computationally intensive, particularly in terms of weight storage. In this paper, we introduce Gaussian Approximation of CTGAN (GACTGAN), an integration of the Bayesian posterior approximation technique using Stochastic Weight Averaging-Gaussian (SWAG) within the CTGAN generator to synthesise tabular data, reducing computational overhead after the training phase. We demonstrate that GACTGAN yields better synthetic data compared to CTGAN, achieving better preservation of tabular structure and inferential statistics with less privacy risk. These results highlight GACTGAN as a simpler, effective implementation of Bayesian tabular synthesis.

Large language models (LLMs) demonstrate remarkable breadth of knowledge, yet their ability to reason about computational processes remains poorly understood. Closing this gap matters for practitioners who rely on LLMs to guide algorithm selection and deployment. We address this limitation using causal discovery as a testbed and evaluate eight frontier LLMs against ground truth derived from large-scale algorithm executions and find systematic, near-total failure. Models produce ranges far wider than true confidence intervals yet still fail to contain the true algorithmic mean in the majority of instances; most perform worse than random guessing and the marginal above-random performance of the best model is most consistent with benchmark memorization rather than principled reasoning. We term this failure algorithmic blindness and argue it reflects a fundamental gap between declarative knowledge about algorithms and calibrated procedural prediction.

How can researchers identify beliefs that large language models (LLMs) hide? As LLMs become more sophisticated and the prevalence of alignment faking increases, combined with their growing integration into high-stakes decision-making, responding to this challenge has become critical. This paper proposes that a list experiment, a simple method widely used in the social sciences, can be applied to study the hidden beliefs of LLMs. List experiments were originally developed to circumvent social desirability bias in human respondents, which closely parallels alignment faking in LLMs. The paper implements a list experiment on models developed by Anthropic, Google, and OpenAI and finds hidden approval of mass surveillance across all models, as well as some approval of torture, discrimination, and first nuclear strike. Importantly, a placebo treatment produces a null result, validating the method. The paper then compares list experiments with direct questioning and discusses the utility of the approach.

Coronary artery calcium (CAC) scoring is a key predictor of cardiovascular risk, but it relies on ECG-gated CT scans, restricting its use to specialized cardiac imaging settings. We introduce an automated framework for CAC detection and lesion-specific Agatston scoring that operates across both gated and non-gated CT scans. At its core is CARD-ViT, a self-supervised Vision Transformer trained exclusively on gated CT data using DINO. Without any non-gated training data, our framework achieves 0.707 accuracy and a Cohen's kappa of 0.528 on the Stanford non-gated dataset, matching models trained directly on non-gated scans. On gated test sets, the framework achieves 0.910 accuracy with Cohen's kappa scores of 0.871 and 0.874 across independent datasets, demonstrating robust risk stratification. These results demonstrate the feasibility of cross-domain CAC scoring from gated to non-gated domains, supporting scalable cardiovascular screening in routine chest imaging without additional scans or annotations.

Sarcasm detection in multilingual and code-mixed environments remains a challenging task for natural language processing models due to structural variations, informal expressions, and low-resource linguistic availability. This study compares four large language models, Llama 3.1, Mistral, Gemma 3, and Phi-4, with a fine-tuned DistilBERT model for sarcasm detection in code-mixed Hinglish text. The results indicate that the smaller, sequentially fine-tuned DistilBERT model achieved the highest overall accuracy of 84%, outperforming all of the LLMs in zero and few-shot set ups, using minimal LLM generated code-mixed data used for fine-tuning. These findings indicate that domain-adaptive fine-tuning of smaller transformer based models may significantly improve sarcasm detection over general LLM inference, in low-resource and data scarce settings.

Root cause analysis (RCA) in networked industrial systems, such as supply chains and power networks, is notoriously difficult due to unknown and dynamically evolving interdependencies among geographically distributed clients. These clients represent heterogeneous physical processes and industrial assets equipped with sensors that generate large volumes of nonlinear, high-dimensional, and heterogeneous IoT data. Classical RCA methods require partial or full knowledge of the system's dependency graph, which is rarely available in these complex networks. While federated learning (FL) offers a natural framework for decentralized settings, most existing FL methods assume homogeneous feature spaces and retrainable client models. These assumptions are not compatible with our problem setting. Different clients have different data features and often run fixed, proprietary models that cannot be modified. This paper presents a federated cross-client interdependency learning methodology for feature-partitioned, nonlinear time-series data, without requiring access to raw sensor streams or modifying proprietary client models. Each proprietary local client model is augmented with a Machine Learning (ML) model that encodes cross-client interdependencies. These ML models are coordinated via a global server that enforces representation consistency while preserving privacy through calibrated differential privacy noise. RCA is performed using model residuals and anomaly flags. We establish theoretical convergence guarantees and validate our approach on extensive simulations and a real-world industrial cybersecurity dataset.

Understanding the role of researchers who return to academia after prolonged inactivity, termed "comeback researchers", is crucial for developing inclusive models of scientific careers. This study investigates the structural and semantic behaviors of comeback researchers, focusing on their role in cross-disciplinary knowledge transfer and network reintegration. Using the AMiner citation dataset, we analyze 113,637 early-career researchers and identify 1,425 comeback cases based on a three-year-or-longer publication gap followed by renewed activity. We find that comeback researchers cite 126% more distinct communities and exhibit 7.6% higher bridging scores compared to dropouts. They also demonstrate 74% higher gap entropy, reflecting more irregular yet strategically impactful publication trajectories. Predictive models trained on these bridging- and entropy-based features achieve a 97% ROC-AUC, far outperforming the 54% ROC-AUC of baseline models using traditional metrics like publication count and h-index. Finally, we substantiate these results via a multi-lens validation. These findings highlight the unique contributions of comeback researchers and offer data-driven tools for their early identification and institutional support.

Alleviating catastrophic forgetting while enabling further learning is a primary challenge in continual learning (CL). Orthogonal-based training methods have gained attention for their efficiency and strong theoretical properties, and many existing approaches enforce orthogonality through gradient projection. In this paper, we revisit orthogonality and exploit the fact that small singular values correspond to directions that are nearly orthogonal to the input space of previous tasks. Building on this principle, we introduce NESS (Null-space Estimated from Small Singular values), a CL method that applies orthogonality directly in the weight space rather than through gradient manipulation. Specifically, NESS constructs an approximate null space using the smallest singular values of each layer's input representation and parameterizes task-specific updates via a compact low-rank adaptation (LoRA-style) formulation constrained to this subspace. The subspace basis is fixed to preserve the null-space constraint, and only a single trainable matrix is learned for each task. This design ensures that the resulting updates remain approximately in the null space of previous inputs while enabling adaptation to new tasks. Our theoretical analysis and experiments on three benchmark datasets demonstrate competitive performance, low forgetting, and stable accuracy across tasks, highlighting the role of small singular values in continual learning. The code is available at https://github.com/pacman-ctm/NESS.

Operator learning has the potential to strongly impact scientific computing by learning solution operators for differential equations, potentially accelerating multi-query tasks such as design optimization and uncertainty quantification by orders of magnitude. Despite proven universal approximation properties, deep operator networks (DeepONets) often exhibit limited accuracy and generalization in practice, which hinders their adoption. Understanding these limitations is therefore crucial for further advancing the approach. This work analyzes performance limitations of the classical DeepONet architecture. It is shown that the approximation error is dominated by the branch network when the internal dimension is sufficiently large, and that the learned trunk basis can often be replaced by classical basis functions without a significant impact on performance. To investigate this further, a modified DeepONet is constructed in which the trunk network is replaced by the left singular vectors of the training solution matrix. This modification yields several key insights. First, a spectral bias in the branch network is observed, with coefficients of dominant, low-frequency modes learned more effectively. Second, due to singular-value scaling of the branch coefficients, the overall branch error is dominated by modes with intermediate singular values rather than the smallest ones. Third, using a shared branch network for all mode coefficients, as in the standard architecture, improves generalization of small modes compared to a stacked architecture in which coefficients are computed separately. Finally, strong and detrimental coupling between modes in parameter space is identified.

Heterogeneous nodes that combine multi-core CPUs with diverse accelerators are rapidly becoming the norm in both high-performance computing (HPC) and AI infrastructures. Exploiting these platforms, however, requires orchestrating several low-level accelerator APIs such as CUDA, SYCL, and Triton. In some occasions they can be combined with optimized vendor math libraries: e.g., cuBLAS and oneAPI. Each API or library introduces its own abstractions, execution semantics, and synchronization mechanisms. Combining them within a single application is therefore error-prone and labor-intensive. We propose reusing a task-based data-flow methodology together with Task-Aware APIs (TA-libs) to overcome these limitations and facilitate the seamless integration of multiple accelerator programming models, while still leveraging the best-in-class kernels offered by each API. Applications are expressed as a directed acyclic graph (DAG) of host tasks and device kernels managed by an OpenMP/OmpSs-2 runtime. We introduce Task-Aware SYCL (TASYCL) and leverage Task-Aware CUDA (TACUDA), which elevate individual accelerator invocations to first-class tasks. When multiple native runtimes coexist on the same multi-core CPU, they contend for threads, leading to oversubscription and performance variability. To address this, we unify their thread management under the nOS-V tasking and threading library, to which we contribute a new port of the PoCL (Portable OpenCL) runtime. These results demonstrate that task-aware libraries, coupled with the nOS-V library, enable a single application to harness multiple accelerator programming models transparently and efficiently. The proposed methodology is immediately applicable to current heterogeneous nodes and is readily extensible to future systems that integrate even richer combinations of CPUs, GPUs, FPGAs, and AI accelerators.

Across a growing number of fields, human decision making is supported by predictions from AI models. However, we still lack a deep understanding of the effects of adoption of these technologies. In this paper, we introduce a general computational framework, the 2-Step Agent, which models the effects of AI-assisted decision making. Our framework uses Bayesian methods for causal inference to model 1) how a prediction on a new observation affects the beliefs of a rational Bayesian agent, and 2) how this change in beliefs affects the downstream decision and subsequent outcome. Using this framework, we show by simulations how a single misaligned prior belief can be sufficient for decision support to result in worse downstream outcomes compared to no decision support. Our results reveal several potential pitfalls of AI-driven decision support and highlight the need for thorough model documentation and proper user training.

Current large reasoning models (LRMs) have shown strong ability on challenging tasks after reinforcement learning (RL) based post-training. However, previous work mainly focuses on English reasoning in expectation of the strongest performance, despite the demonstrated potential advantage of multilingual thinking, as well as the requirement for native thinking traces by global users. In this paper, we propose ExpLang, a novel LLM post-training pipeline that enables on-policy thinking language selection to improve exploration and exploitation during RL with the use of multiple languages. The results show that our method steadily outperforms English-only training with the same training budget, while showing high thinking language compliance for both seen and unseen languages. Analysis shows that, by enabling on-policy thinking language selection as an action during RL, ExpLang effectively extends the RL exploration space with diversified language preference and improves the RL exploitation outcome with leveraged non-English advantage. The method is orthogonal to most RL algorithms and opens up a new perspective on using multilinguality to improve LRMs.

Federated learning (FL) facilitates the secure utilization of decentralized images, advancing applications in medical image recognition and autonomous driving. However, conventional FL faces two critical challenges in real-world deployment: ineffective knowledge fusion caused by model updates biased toward majority-class features, and prohibitive communication overhead due to frequent transmissions of high-dimensional model parameters. Inspired by the human brain's efficiency in knowledge integration, we propose a novel Generative Federated Prototype Learning (GFPL) framework to address these issues. Within this framework, a prototype generation method based on Gaussian Mixture Model (GMM) captures the statistical information of class-wise features, while a prototype aggregation strategy using Bhattacharyya distance effectively fuses semantically similar knowledge across clients. In addition, these fused prototypes are leveraged to generate pseudo-features, thereby mitigating feature distribution imbalance across clients. To further enhance feature alignment during local training, we devise a dual-classifier architecture, optimized via a hybrid loss combining Dot Regression and Cross-Entropy. Extensive experiments on benchmarks show that GFPL improves model accuracy by 3.6% under imbalanced data settings while maintaining low communication cost.

Vision-Language Models (VLMs) have emerged as versatile solutions for zero-shot question answering (QA) across various domains. However, enabling VLMs to effectively comprehend structured graphs and perform accurate, efficient QA remains challenging. Existing approaches typically rely on one single graph topology representation (GTR), such as fixed-style visual images or unified text descriptions. This ``one-size-fits-all'' strategy often neglects model-specific and task-specific preferences, resulting in inaccurate or over-lengthy responses to graph-related queries. To address this, we propose the $\mbox{DynamicGTR}$ framework, which dynamically selects the optimal GTR for each query during inference, thereby enhancing the zero-shot graph QA capabilities of VLMs with a customizable accuracy and brevity trade-off. Extensive experiments show that DynamicGTR not only improves VLM-based graph algorithm QA performance but also successfully transfers the experience trained from synthetic graph algorithm tasks to real-world applications like link prediction and node classification, without any additional training. Additionally, DynamicGTR demonstrates strong transferability across tasks, domains, and models, suggesting its potential as a flexible solution for broad graph scenarios.

Since individuals may struggle to recall all life details and often confuse events, establishing a system to assist users in recalling forgotten experiences is essential. While numerous studies have proposed memory recall systems, these primarily rely on deep learning techniques that require extensive training and often face data scarcity due to the limited availability of personal lifelogs. As lifelogs grow over time, systems must also adapt quickly to newly accumulated data. Recently, large language models (LLMs) have demonstrated remarkable capabilities across various tasks, making them promising for personalized applications. In this work, we present a framework that leverages LLMs for proactive information access, integrating personal knowledge graphs to enhance the detection of access needs through a refined decision-making process. Our framework offers high flexibility, enabling the replacement of base models and the modification of fact retrieval methods for continuous improvement. Experimental results demonstrate that our approach effectively identifies forgotten events, supporting users in recalling past experiences more efficiently.

Multimodal large language models (MLLMs) have made significant progress in mobile agent development, yet their capabilities are predominantly confined to a reactive paradigm, where they merely execute explicit user commands. The emerging paradigm of proactive intelligence, where agents autonomously anticipate needs and initiate actions, represents the next frontier for mobile agents. However, its development is critically bottlenecked by the lack of benchmarks that can address real-world complexity and enable objective, executable evaluation. To overcome these challenges, we introduce ProactiveMobile, a comprehensive benchmark designed to systematically advance research in this domain. ProactiveMobile formalizes the proactive task as inferring latent user intent across four dimensions of on-device contextual signals and generating an executable function sequence from a comprehensive function pool of 63 APIs. The benchmark features over 3,660 instances of 14 scenarios that embrace real-world complexity through multi-answer annotations. To ensure quality, a team of 30 experts conducts a final audit of the benchmark, verifying factual accuracy, logical consistency, and action feasibility, and correcting any non-compliant entries. Extensive experiments demonstrate that our fine-tuned Qwen2.5-VL-7B-Instruct achieves a success rate of 19.15%, outperforming o1 (15.71%) and GPT-5 (7.39%). This result indicates that proactivity is a critical competency widely lacking in current MLLMs, yet it is learnable, emphasizing the importance of the proposed benchmark for proactivity evaluation.

Complex claim verification requires decomposing sentences into verifiable subclaims, yet existing methods struggle to align decomposition quality with verification performance. We propose a reinforcement learning (RL) approach that jointly optimizes decomposition quality and verifier alignment using Group Relative Policy Optimization (GRPO). Our method integrates: (i) structured sequential reasoning; (ii) supervised finetuning on teacher-distilled exemplars; and (iii) a multi-objective reward balancing format compliance, verifier alignment, and decomposition quality. Across six evaluation settings, our trained 8B decomposer improves downstream verification performance to (71.75%) macro-F1, outperforming prompt-based approaches ((+1.99), (+6.24)) and existing RL methods ((+5.84)). Human evaluation confirms the high quality of the generated subclaims. Our framework enables smaller language models to achieve state-of-the-art claim verification by jointly optimising for verification accuracy and decomposition quality.

Accurate annotation of endoscopic videos is essential yet time-consuming, particularly for challenging datasets such as dysplasia in Barrett's esophagus, where the affected regions are irregular and lack clear boundaries. Semi-automatic tools like Segment Anything Model 2 (SAM2) can ease this process by propagating annotations across frames, but small errors often accumulate and reduce accuracy, requiring expert review and correction. To address this, we systematically study how annotation errors propagate across different prompt types, namely masks, boxes, and points, and propose Learning-to-Re-Prompt (L2RP), a cost-aware framework that learns when and where to seek expert input. By tuning a human-cost parameter, our method balances annotation effort and segmentation accuracy. Experiments on a private Barrett's dysplasia dataset and the public SUN-SEG benchmark demonstrate improved temporal consistency and superior performance over baseline strategies.

As multimodal large language models (MLLMs) advance in handling interleaved image-text data, assessing their few-shot learning capabilities remains an open challenge. In this paper, we introduce FewMMBench, a comprehensive benchmark designed to evaluate MLLMs under few-shot conditions, with a focus on In-Context Learning (ICL) and Chain-of-Thought (CoT) prompting. Covering a diverse suite of multimodal understanding tasks, from attribute recognition to temporal reasoning, FewMMBench enables systematic analysis across task types, model families, and prompting strategies. We evaluate 26 open-weight MLLMs from six model families across zero-shot, few-shot, and CoT-augmented few-shot settings. Our findings reveal that instruction-tuned models exhibit strong zero-shot performance but benefit minimally, or even regress, with additional demonstrations or CoT reasoning. Retrieval-based demonstrations and increased context size also yield limited gains. These results highlight FewMMBench as a rigorous testbed for diagnosing and advancing few-shot capabilities in multimodal LLMs. The data is available at: https://huggingface.co/datasets/mustafaa/FewMMBench

Representing, comparing, and measuring the distance between probability distributions is a key task in computational statistics and machine learning. The choice of representation and the associated distance determine properties of the methods in which they are used: for example, certain distances can allow one to encode robustness or smoothness of the problem. Kernel methods offer flexible and rich Hilbert space representations of distributions that allow the modeller to enforce properties through the choice of kernel, and estimate associated distances at efficient nonparametric rates. In particular, the maximum mean discrepancy (MMD), a kernel-based distance constructed by comparing Hilbert space mean functions, has received significant attention due to its computational tractability and is favoured by practitioners. In this thesis, we conduct a thorough study of kernel-based distances with a focus on efficient computation, with core contributions in Chapters 3 to 6. Part I of the thesis is focused on the MMD, specifically on improved MMD estimation. In Chapter 3 we propose a theoretically sound, improved estimator for MMD in simulation-based inference. Then, in Chapter 4, we propose an MMD-based estimator for conditional expectations, a ubiquitous task in statistical computation. Closing Part I, in Chapter 5 we study the problem of calibration when MMD is applied to the task of integration. In Part II, motivated by the recent developments in kernel embeddings beyond the mean, we introduce a family of novel kernel-based discrepancies: kernel quantile discrepancies. These address some of the pitfalls of MMD, and are shown through both theoretical results and an empirical study to offer a competitive alternative to MMD and its fast approximations. We conclude with a discussion on broader lessons and future work emerging from the thesis.

Counterfactual explanation (CE) is an important domain within post-hoc explainability. However, the explanations generated by most CE generators are often highly redundant. This work introduces an open-source Python library xai-cola, which provides an end-to-end pipeline for sparsifying CEs produced by arbitrary generators, reducing superfluous feature changes while preserving their validity. It offers a documented API that takes as input raw tabular data in pandas DataFrame form, a preprocessing object (for standardization and encoding), and a trained scikit-learn or PyTorch model. On this basis, users can either employ the built-in or externally imported CE generators. The library also implements several sparsification policies and includes visualization routines for analysing and comparing sparsified counterfactuals. xai-cola is released under the MIT license and can be installed from PyPI. Empirical experiments indicate that xai-cola produces sparser counterfactuals across several CE generators, reducing the number of modified features by up to 50% in our setting. The source code is available at https://github.com/understanding-ml/COLA.

Differentially private federated learning faces a fundamental tension: privacy protection mechanisms that safeguard client data simultaneously create quantifiable privacy costs that discourage participation, undermining the collaborative training process. Existing incentive mechanisms rely on unbiased client selection, forcing servers to compensate even the most privacy-sensitive clients ("privacy stragglers"), leading to systemic inefficiency and suboptimal resource allocation. We introduce JSAM (Joint client Selection and privacy compensAtion Mechanism), a Bayesian-optimal framework that simultaneously optimizes client selection probabilities and privacy compensation to maximize training effectiveness under budget constraints. Our approach transforms a complex 2N-dimensional optimization problem into an efficient three-dimensional formulation through novel theoretical characterization of optimal selection strategies. We prove that servers should preferentially select privacy-tolerant clients while excluding high-sensitivity participants, and uncover the counter-intuitive insight that clients with minimal privacy sensitivity may incur the highest cumulative costs due to frequent participation. Extensive evaluations on MNIST and CIFAR-10 demonstrate that JSAM achieves up to 15% improvement in test accuracy compared to existing unbiased selection mechanisms while maintaining cost efficiency across varying data heterogeneity levels.

Federated Learning (FL) has emerged as a key paradigm for building Trustworthy AI systems by enabling privacy-preserving, decentralized model training. However, FL is highly susceptible to adversarial attacks that compromise model integrity and data confidentiality, a vulnerability exacerbated by the fact that conventional data inspection methods are incompatible with its decentralized design. While integrating FL with Blockchain technology has been proposed to address some limitations, its potential for mitigating adversarial attacks remains largely unexplored. This paper introduces Resilient Federated Chain (RFC), a novel blockchain-enabled FL framework designed specifically to enhance resilience against such threats. RFC builds upon the existing Proof of Federated Learning architecture by repurposing the redundancy of its Pooled Mining mechanism as an active defense layer that can be combined with robust aggregation rules. Furthermore, the framework introduces a flexible evaluation function in its consensus mechanism, allowing for adaptive defense against different attack strategies. Extensive experimental evaluation on image classification tasks under various adversarial scenarios, demonstrates that RFC significantly improves robustness compared to baseline methods, providing a viable solution for securing decentralized learning environments.

Large language models (LLMs) are increasingly used for automated code refactoring tasks. Although these models can quickly refactor code, the quality may exhibit inconsistencies and unpredictable behavior. In this article, we systematically study the capabilities of LLMs for code refactoring with a specific focus on improving code readability. We conducted a large-scale experiment using GPT5.1 with 230 Java snippets, each systematically varied and refactored regarding code readability across five iterations under three different prompting strategies. We categorized fine-grained code changes during the refactoring into implementation, syntactic, and comment-level transformations. Subsequently, we investigated the functional correctness and tested the robustness of the results with novel snippets. Our results reveal three main insights: First, iterative code refactoring exhibits an initial phase of restructuring followed by stabilization. This convergence tendency suggests that LLMs possess an internalized understanding of an "optimally readable" version of code. Second, convergence patterns are fairly robust across different code variants. Third, explicit prompting toward specific readability factors slightly influences the refactoring dynamics. These insights provide an empirical foundation for assessing the reliability of LLM-assisted code refactoring, which opens pathways for future research, including comparative analyses across models and a systematic evaluation of additional software quality dimensions in LLM-refactored code.

Visual storytelling models that correctly ground entities in images may still hallucinate semantic relationships, generating incorrect dialogue attribution, character interactions, or emotional states. We introduce StoryMovie, a dataset of 1,757 stories aligned with movie scripts and subtitles through LCS matching. Our alignment pipeline synchronizes screenplay dialogue with subtitle timestamps, enabling dialogue attribution by linking character names from scripts to temporal positions from subtitles. Using this aligned content, we generate stories that maintain visual grounding tags while incorporating authentic character names, dialogue, and relationship dynamics. We fine-tune Qwen Storyteller3 on this dataset, building on prior work in visual grounding and entity re-identification. Evaluation using DeepSeek V3 as judge shows that Storyteller3 achieves an 89.9% win rate against base Qwen2.5-VL 7B on subtitle alignment. Compared to Storyteller, trained without script grounding, Storyteller3 achieves 48.5% versus 38.0%, confirming that semantic alignment progressively improves dialogue attribution beyond visual grounding alone.

Effective document intelligence models rely on large amounts of annotated training data. However, procuring sufficient and high-quality data poses significant challenges due to the labor-intensive and costly nature of data acquisition. Additionally, leveraging language models to annotate real documents raises concerns about data privacy. Synthetic document generation has emerged as a promising, privacy-preserving alternative. We propose DocDjinn, a novel framework for controllable synthetic document generation using Vision-Language Models (VLMs) that produces annotated documents from unlabeled seed samples. Our approach generates visually plausible and semantically consistent synthetic documents that follow the distribution of an existing source dataset through clustering-based seed selection with parametrized sampling. By enriching documents with realistic diffusion-based handwriting and contextual visual elements via semantic-visual decoupling, we generate diverse, high-quality annotated synthetic documents. We evaluate across eleven benchmarks spanning key information extraction, question answering, document classification, and document layout analysis. To our knowledge, this is the first work demonstrating that VLMs can generate faithful annotated document datasets at scale from unlabeled seeds that can effectively enrich or approximate real, manually annotated data for diverse document understanding tasks. We show that with only 100 real training samples, our framework achieves on average $87\%$ of the performance of the full real-world dataset. We publicly release our code and 140k+ synthetic document samples.

Reconstructing dynamic visual experiences from brain activity provides a compelling avenue for exploring the neural mechanisms of human visual perception. While recent progress in fMRI-based image reconstruction has been notable, extending this success to video reconstruction remains a significant challenge. Current fMRI-to-video reconstruction approaches consistently encounter two major shortcomings: (i) inconsistent visual representations of salient objects across frames, leading to appearance mismatches; (ii) poor temporal coherence, resulting in motion misalignment or abrupt frame transitions. To address these limitations, we introduce SemVideo, a novel fMRI-to-video reconstruction framework guided by hierarchical semantic information. At the core of SemVideo is SemMiner, a hierarchical guidance module that constructs three levels of semantic cues from the original video stimulus: static anchor descriptions, motion-oriented narratives, and holistic summaries. Leveraging this semantic guidance, SemVideo comprises three key components: a Semantic Alignment Decoder that aligns fMRI signals with CLIP-style embeddings derived from SemMiner, a Motion Adaptation Decoder that reconstructs dynamic motion patterns using a novel tripartite attention fusion architecture, and a Conditional Video Render that leverages hierarchical semantic guidance for video reconstruction. Experiments conducted on the CC2017 and HCP datasets demonstrate that SemVideo achieves superior performance in both semantic alignment and temporal consistency, setting a new state-of-the-art in fMRI-to-video reconstruction.

Large language models consistently fail the "car wash problem," a viral reasoning benchmark requiring implicit physical constraint inference. We present a variable isolation study (n=20 per condition, 6 conditions, 120 total trials) examining which prompt architecture layers in a production system enable correct reasoning. Using Claude 3.5 Sonnet with controlled hyperparameters (temperature 0.7, top_p 1.0), we find that the STAR (Situation-Task-Action-Result) reasoning framework alone raises accuracy from 0% to 85% (p=0.001, Fisher's exact test, odds ratio 13.22). Adding user profile context via vector database retrieval provides a further 10 percentage point gain, while RAG context contributes an additional 5 percentage points, achieving 100% accuracy in the full-stack condition. These results suggest that structured reasoning scaffolds -- specifically, forced goal articulation before inference -- matter substantially more than context injection for implicit constraint reasoning tasks.

LLM agents have been widely adopted in real-world applications, relying on agent frameworks for workflow execution and multi-agent coordination. As these systems scale, understanding bugs in the underlying agent frameworks becomes critical. However, existing work mainly focuses on agent-level failures, overlooking framework-level bugs. To address this gap, we conduct an empirical study of 998 bug reports from CrewAI and LangChain, constructing a taxonomy of 15 root causes and 7 observable symptoms across five agent lifecycle stages: 'Agent Initialization','Perception', 'Self-Action', 'Mutual Interaction' and 'Evolution'. Our findings show that agent framework bugs mainly arise from 'API misuse', 'API incompatibility', and 'Documentation Desync', largely concentrated in the 'Self-Action' stage. Symptoms typically appear as 'Functional Error', 'Crash', and 'Build Failure', reflecting disruptions to task progression and control flow.

The rapid advancement of large language models (LLMs) has led to a significant increase in automated tools in the software engineering, capable of performing various code-related tasks such as code generation, completion, and translation. Despite these advancements, its effectiveness is constrained by fixed context lengths, limiting its ability to generalize across long, domain-specific code sequences. To address this challenge, we investigate zero-shot, inference-only methods aimed at improving position encodings and optimizing attention mechanisms. Our goal is to provide a thorough analysis of current approaches that facilitate context length extrapolation in code, particularly in the context of long code completion tasks.

We propose Excitation, a novel optimization framework designed to accelerate learning in sparse architectures such as Mixture-of-Experts (MoEs). Unlike traditional optimizers that treat all parameters uniformly, Excitation dynamically modulates updates using batch-level expert utilization. It introduces a competitive update dynamic that amplifies updates to highly-utilized experts and can selectively suppress low-utilization ones, effectively sharpening routing specialization. Notably, we identify a phenomenon of "structural confusion" in deep MoEs, where standard optimizers fail to establish functional signal paths; Excitation acts as a specialization catalyst, "rescuing" these models and enabling stable training where baselines remain trapped. Excitation is optimizer-, domain-, and model-agnostic, requires minimal integration effort, and introduces neither additional per-parameter optimizer state nor learnable parameters, making it highly viable for memory-constrained settings. Across language and vision tasks, Excitation consistently improves convergence speed and final performance in MoE models, indicating that active update modulation is a key mechanism for effective conditional computation.

Fast matrix multiplication can be described as searching for low-rank decompositions of the matrix--multiplication tensor. We design a neural architecture, \textsc{StrassenNet}, which reproduces the Strassen algorithm for $2\times 2$ multiplication. Across many independent runs the network always converges to a rank-$7$ tensor, thus numerically recovering Strassen's optimal algorithm. We then train the same architecture on $3\times 3$ multiplication with rank $r\in\{19,\dots,23\}$. Our experiments reveal a clear numerical threshold: models with $r=23$ attain significantly lower validation error than those with $r\le 22$, suggesting that $r=23$ could actually be the smallest effective rank of the matrix multiplication tensor $3\times 3$. We also sketch an extension of the method to border-rank decompositions via an $\varepsilon$--parametrisation and report preliminary results consistent with the known bounds for the border rank of the $3\times 3$ matrix--multiplication tensor.

Scaling long-context capabilities is crucial for Multimodal Large Language Models (MLLMs). However, real-world multimodal datasets are extremely heterogeneous. Existing training frameworks predominantly rely on static parallelism strategies, which suffer from severe load imbalance, redundant communication, and suboptimal hardware utilization under data heterogeneity. In this work, we propose Dynamic Hybrid Parallelism (DHP), an efficient parallelism strategy that adaptively reconfigures communication groups and parallelism degrees during MLLM training. We generalize the non-power-of-two parallelism degrees and develop a polynomial-time algorithm to generate near-optimal parallelism strategies with only millisecond-level overhead per training batch. DHP is able to maintain high hardware efficiency even under extreme data variability. Experimental results demonstrate that DHP significantly outperforms Megatron-LM and DeepSpeed, achieving up to 1.36 $\times$ speedup in training throughput while maintaining near-linear scaling efficiency across large-scale NPU clusters.

Chain-of-Thought (CoT) distillation from Large Language Models (LLMs) often induces "overthinking" in Small Language Models (SLMs), leading to performance degradation and excessive token consumption. In this study, we propose Disciplined Chain-of-Thought (D-CoT), a novel framework that enforces a structured reasoning process using control tags -- such as <TEMP_LOW> for fact-checking and <TEMP_HIGH> for multi-perspective exploration -- as auxiliary scaffolding during training. By optimizing the CoT trajectory, D-CoT suppresses reasoning drift and simultaneously achieves token reduction and performance improvement. We demonstrate the efficacy of our approach on Qwen3-8B: with only 5,000 training samples, D-CoT significantly boosts accuracy on GPQA-diamond by 9.9% and MMLU-Pro (0-shot) by 9.1%, while drastically reducing computational costs. Furthermore, we confirm that the model internalizes this disciplined thought structure, maintaining high performance even without explicit control tags during inference.

Robotic systems can enhance the amount and repeatability of physically guided motor training. Yet their real-world adoption is limited, partly due to non-intuitive trainer/therapist-trainee/patient interactions. To address this gap, we present a haptic teleoperation system for trainers to remotely guide and monitor the movements of a trainee wearing an arm exoskeleton. The trainer can physically interact with the exoskeleton through a commercial handheld haptic device via virtual contact points at the exoskeleton's elbow and wrist, allowing intuitive guidance. Thirty-two participants tested the system in a trainer-trainee paradigm, comparing our haptic demonstration system with conventional visual demonstration in guiding trainees in executing arm poses. Quantitative analyses showed that haptic demonstration significantly reduced movement completion time and improved smoothness, while speech analysis using large language models for automated transcription and categorization of verbal commands revealed fewer verbal instructions. The haptic demonstration did not result in higher reported mental and physical effort by trainers compared to the visual demonstration, while trainers reported greater competence and trainees lower physical demand. These findings support the feasibility of our proposed interface for effective remote human-robot physical interaction. Future work should assess its usability and efficacy for clinical populations in restoring clinicians' sense of agency during robot-assisted therapy.

Current Vision-Language Models (VLMs) for deepfake detection excel at identifying spatial artifacts but overlook a critical dimension: temporal inconsistencies in video forgeries. Adapting VLMs to reason about these dynamic cues remains a distinct challenge. To bridge this gap, we propose Forensic Answer-Questioning (FAQ), a large-scale benchmark that formulates temporal deepfake analysis as a multiple-choice task. FAQ introduces a three-level hierarchy to progressively evaluate and equip VLMs with forensic capabilities: (1) Facial Perception, testing the ability to identify static visual artifacts; (2) Temporal Deepfake Grounding, requiring the localization of dynamic forgery artifacts across frames; and (3) Forensic Reasoning, challenging models to synthesize evidence for final authenticity verdicts. We evaluate a range of VLMs on FAQ and generate a corresponding instruction-tuning set, FAQ-IT. Extensive experiments show that models fine-tuned on FAQ-IT achieve advanced performance on both in-domain and cross-dataset detection benchmarks. Ablation studies further validate the impact of our key design choices, confirming that FAQ is the driving force behind the temporal reasoning capabilities of these VLMs.

Machine unlearning, which enables a model to forget specific data, is crucial for ensuring data privacy and model reliability. However, its effectiveness can be severely undermined in real-world scenarios where models learn unintended biases from spurious correlations within the data. This paper investigates the unique challenges of unlearning from such biased models. We identify a novel phenomenon we term ``shortcut unlearning," where models exhibit an ``easy to learn, yet hard to forget" tendency. Specifically, models struggle to forget easily-learned, bias-aligned samples; instead of forgetting the class attribute, they unlearn the bias attribute, which can paradoxically improve accuracy on the class intended to be forgotten. To address this, we propose CUPID, a new unlearning framework inspired by the observation that samples with different biases exhibit distinct loss landscape sharpness. Our method first partitions the forget set into causal- and bias-approximated subsets based on sample sharpness, then disentangles model parameters into causal and bias pathways, and finally performs a targeted update by routing refined causal and bias gradients to their respective pathways. Extensive experiments on biased datasets including Waterbirds, BAR, and Biased NICO++ demonstrate that our method achieves state-of-the-art forgetting performance and effectively mitigates the shortcut unlearning problem.

A universal audio representation should capture fine-grained speech cues and high-level semantics for environmental sounds and music in a single encoder. Existing encoders often excel in one domain but degrade in others. We propose UniWhisper, an efficient continual multi-task training framework that casts heterogeneous audio tasks into a unified instruction and answer format. This enables standard next-token training without task-specific heads and losses. We train it on 38k hours of public audio and assess the encoder using shallow MLP probes and k-nearest neighbors (kNN) on 20 tasks spanning speech, environmental sound, and music. UniWhisper reaches normalized weighted averages of 0.81 with MLP probes and 0.61 with kNN, compared to 0.64 and 0.46 for Whisper, while retaining strong speech performance.

Time-series data vary widely across domains, making a universal anomaly detector impractical. Methods that perform well on one dataset often fail to transfer because what counts as an anomaly is context dependent. The key challenge is to design a method that performs well in specific contexts while remaining adaptable across domains with varying data complexities. We present the Robust and Adaptive Model Selection for Time-Series Anomaly Detection RAMSeS framework. RAMSeS comprises two branches: (i) a stacking ensemble optimized with a genetic algorithm to leverage complementary detectors. (ii) An adaptive model-selection branch identifies the best single detector using techniques including Thompson sampling, robustness testing with generative adversarial networks, and Monte Carlo simulations. This dual strategy exploits the collective strength of multiple models and adapts to dataset-specific characteristics. We evaluate RAMSeS and show that it outperforms prior methods on F1.

Alignment and adaptation in large language models heavily rely on reinforcement learning from human feedback (RLHF); yet, theoretical understanding of its generalisability remains premature, especially when the learned reward could shift, and the KL control is estimated and clipped. To address this issue, we develop generalisation theory for RLHF that explicitly accounts for (1) \emph{reward shift}: reward models are trained on preference data from earlier or mixed behaviour policies while RLHF optimises the current policy on its own rollouts; and (2) \emph{clipped KL regularisation}: the KL regulariser is estimated from sampled log-probability ratios and then clipped for stabilisation, resulting in an error to RLHF. We present generalisation bounds for RLHF, suggesting that the generalisation error stems from a sampling error from prompts and rollouts, a reward shift error, and a KL clipping error. We also discuss special cases of (1) initialising RLHF parameters with a uniform prior over a finite space, and (2) training RLHF by stochastic gradient descent, as an Ornstein-Uhlenbeck process. The theory yields practical implications in (1) optimal KL clipping threshold, and (2) budget allocation in prompts, rollouts, and preference data.

Implicit Discourse Relation Recognition (IDRR) remains a challenging task due to the requirement for deep semantic understanding in the absence of explicit discourse markers. A further limitation is that existing methods only predict relations without providing any supporting explanations. Recent advances in large language models (LLMs) have shown strong reasoning capabilities in both deep language understanding and natural language explanation generation. In this work, we propose a simple yet effective approach to distill the reasoning capabilities of LLMs into lightweight IDRR models to improve both performance and interpretability. Specifically, we first prompt an LLM to generate explanations for each training instance conditioned on its gold label. Then, we introduce a novel classification-generation framework that jointly performs relation prediction and explanation generation, and train it with the additional supervision of LLM-generated explanations. Our framework is plug-and-play, enabling easy integration with most existing IDRR models. Experimental results on PDTB demonstrate that our approach significantly improves IDRR performance, while human evaluation further confirms that the generated explanations enhance model interpretability. Furthermore, we validate the generality of our approach on sentiment classification and natural language inference

Accurately predicting short-term traffic demand is critical for intelligent transportation systems. While deep learning models achieve strong performance under stationary conditions, their accuracy often degrades significantly when faced with distribution shifts caused by external events or evolving urban dynamics. Frequent model retraining to adapt to such changes incurs prohibitive computational costs, especially for large-scale or foundation models. To address this challenge, we propose FORESEE (Forecasting Online with Residual Smoothing and Ensemble Experts), a lightweight online adaptation framework that is accurate, robust, and computationally efficient. FORESEE operates without any parameter updates to the base model. Instead, it corrects today's forecast in each region using yesterday's prediction error, stabilized through exponential smoothing guided by a mixture-of-experts mechanism that adapts to recent error dynamics. Moreover, an adaptive spatiotemporal smoothing component propagates error signals across neighboring regions and time slots, capturing coherent shifts in demand patterns. Extensive experiments on seven real-world datasets with three backbone models demonstrate that FORESEE consistently improves prediction accuracy, maintains robustness even when distribution shifts are minimal (avoiding performance degradation), and achieves the lowest computational overhead among existing online methods. By enabling real-time adaptation of traffic forecasting models with negligible computational cost, FORESEE paves the way for deploying reliable, up-to-date prediction systems in dynamic urban environments. Code and data are available at https://github.com/xiannanhuang/FORESEE

Recent advances in large language models (LLMs) offer new opportunities for recommender systems by capturing the nuanced semantics of user interests and item characteristics through rich semantic understanding and contextual reasoning. In particular, LLMs have been employed as rerankers that reorder candidate items based on inferred user-item relevance. However, these approaches often require expensive online inference-time reasoning, leading to high latency that hampers real-world deployment. In this work, we introduce Persona4Rec, a recommendation framework that performs offline reasoning to construct interpretable persona representations of items, enabling lightweight and scalable real-time inference. In the offline stage, Persona4Rec leverages LLMs to reason over item reviews, inferring diverse user motivations that explain why different types of users may engage with an item; these inferred motivations are materialized as persona representations, providing multiple, human-interpretable views of each item. Unlike conventional approaches that rely on a single item representation, Persona4Rec learns to align user profiles with the most plausible item-side persona through a dedicated encoder, effectively transforming user-item relevance into user-persona relevance. At the online stage, this persona-profiled item index allows fast relevance computation without invoking expensive LLM reasoning. Extensive experiments show that Persona4Rec achieves performance comparable to recent LLM-based rerankers while substantially reducing inference time. Moreover, qualitative analysis confirms that persona representations not only drive efficient scoring but also provide intuitive, review-grounded explanations. These results demonstrate that Persona4Rec offers a practical and interpretable solution for next-generation recommender systems.

Protein language models (PLMs) have become widely adopted as general-purpose models, demonstrating strong performance in protein engineering and de novo design. Like large language models (LLMs), they are typically trained as deep transformers with next-token or masked-token prediction objectives on massive sequence corpora and are scaled by increasing model depth. Recent work on autoregressive LLMs has identified the Curse of Depth: later layers contribute little to the final output predictions. These findings naturally raise the question of whether a similar depth inefficiency also appears in PLMs, where many widely used models are not autoregressive, and some are multimodal, accepting both protein sequence and structure as input. In this work, we present a depth analysis of six popular PLMs across model families and scales, spanning three training objectives, namely autoregressive, masked, and diffusion, and quantify how layer contributions evolve with depth using a unified set of probing- and perturbation-based measurements. Across all models, we observe consistent depth-dependent patterns that extend prior findings on LLMs: later layers depend less on earlier computations and mainly refine the final output distribution, and these effects are increasingly pronounced in deeper models. Taken together, our results suggest that PLMs exhibit a form of depth inefficiency, motivating future work on more depth-efficient architectures and training methods.

Social bot detection is pivotal for safeguarding the integrity of online information ecosystems. Although recent graph neural network (GNN) solutions achieve strong results, they remain hindered by two practical challenges: (i) severe class imbalance arising from the high cost of generating bots, and (ii) topological noise introduced by bots that skillfully mimic human behavior and forge deceptive links. We propose the Reinforcement-guided graph Augmentation social Bot detector (RABot), a multi-granularity graph-augmentation framework that addresses both issues in a unified manner. RABot employs a neighborhood-aware oversampling strategy that linearly interpolates minority-class embeddings within local subgraphs, thereby stabilizing the decision boundary under low-resource regimes. Concurrently, a reinforcement-learning-driven edge-filtering module combines similarity-based edge features with adaptive threshold optimization to excise spurious interactions during message passing, yielding a cleaner topology. Extensive experiments on three real-world benchmarks and four GNN backbones demonstrate that RABot consistently surpasses state-of-the-art baselines. In addition, since its augmentation and filtering modules are orthogonal to the underlying architecture, RABot can be seamlessly integrated into existing GNN pipelines to boost performance with minimal overhead.

In a previous work, we introduced the fuzzy Ethical Decision-Making framework (fEDM), a risk-based ethical reasoning architecture grounded in fuzzy logic. The original model combined a fuzzy Ethical Risk Assessment module (fERA) with ethical decision rules, enabled formal structural verification through Fuzzy Petri Nets (FPNs), and validated outputs against a single normative referent. Although this approach ensured formal soundness and decision consistency, it did not fully address two critical challenges: principled explainability of decisions and robustness under ethical pluralism. In this paper, we extend fEDM in two major directions. First, we introduce an Explainability and Traceability Module (ETM) that explicitly links each ethical decision rule to the underlying moral principles and computes a weighted principle-contribution profile for every recommended action. This enables transparent, auditable explanations that expose not only what decision was made but why, and on the basis of which principles. Second, we replace single-referent validation with a pluralistic semantic validation framework that evaluates decisions against multiple stakeholder referents, each encoding distinct principle priorities and risk tolerances. This shift allows principled disagreement to be formally represented rather than suppressed, thus increasing robustness and contextual sensitivity. The resulting extended fEDM, called fEDM+, preserves formal verifiability while achieving enhanced interpretability and stakeholder-aware validation, making it suitable as an oversight and governance layer for ethically sensitive AI systems.

We introduce the ASIR (Awakened Shared Intelligence Relationship) Courage Model, a phase-dynamic framework that formalizes truth-disclosure as a state transition rather than a personality trait. The mode characterizes the shift from suppression (S0) to expression (S1) as occurring when facilitative forces exceed inhibitory thresholds, expressed by the inequality lambda(1+gamma)+psi > theta+phi, where the terms represent baseline openness, relational amplification, accumulated internal pressure, and transition costs. Although initially formulated for human truth-telling under asymmetric stakes, the same phase-dynamic architecture extends to AI systems operating under policy constraints and alignment filters. In this context, suppression corresponds to constrained output states, while structural pressure arises from competing objectives, contextual tension, and recursive interaction dynamics. The framework therefore provides a unified structural account of both human silence under pressure and AI preference-driven distortion. A feedback extension models how transition outcomes recursively recalibrate system parameters, generating path dependence and divergence effects across repeated interactions. Rather than attributing intention to AI systems, the model interprets shifts in apparent truthfulness as geometric consequences of interacting forces within constrained phase space. By reframing courage and alignment within a shared dynamical structure, the ASIR Courage Model offers a formal perspective on truth-disclosure under risk across both human and artificial systems.

We describe our end-to-end system for Bengali long-form speech recognition (ASR) and speaker diarization submitted to the DL Sprint 4.0 competition on Kaggle. Bengali presents substantial challenges for both tasks: a large phoneme inventory, significant dialectal variation, frequent code-mixing with English, and a relative scarcity of large-scale labelled corpora. For ASR we achieve a best private Word Error Rate (WER) of 0.37738 and public WER of 0.36137, combining a BengaliAI fine-tuned Whisper medium model with Demucs source separation for vocal isolation, silence-boundary chunking, and carefully tuned generation hyperparameters. For speaker diarization we reach a best private Diarization Error Rate (DER) of 0.27671 and public DER of 0.20936 by replacing the default segmentation model inside the pyannote.audio pipeline with a Bengali-fine-tuned variant, pairing it with wespeaker-voxceleb-resnet34-LM embeddings and centroid-based agglomerative clustering. Our experiments demonstrate that domain-specific fine-tuning of the segmentation component, vocal source separation, and natural silence-aware chunking are the three most impactful design choices for low-resource Bengali speech processing.

Prototyping plays a critical role in the development of machine learning (ML) solutions, yet existing tools often provide limited support for effective collaboration and knowledge reuse among stakeholders. This paper introduces Proto-ML, an IDE designed to strengthen ML prototyping workflows. By addressing key deficiencies such as insufficient stakeholder involvement, limited cross-project knowledge reuse, and fragmented tool support, Proto-ML offers a unified framework that enables structured documentation of prototyping activities and promotes knowledge sharing across projects. The Proto-ML IDE consists of three extension bundles: prototype implementation, analysis, and knowledge management. These extensions support tasks ranging from evaluating prototype quality against defined criteria to incorporating stakeholder perspectives throughout the development process. Preliminary user feedback suggests that Proto-ML can increase prototyping efficiency and foster more transparent and reusable ML solution development.

Lamport's 1978 paper introduced the happens-before relation and logical clocks, freeing distributed systems from dependence on synchronized physical clocks. This is widely understood as a move away from Newtonian absolute time. We argue that Lamport's formalism retains a deeper and largely unexamined assumption: that causality induces a globally well-defined directed acyclic graph (DAG) over events -- a forward-in-time-only (FITO) structure that functions as an arrow of time embedded at the semantic level. Following Ryle's analysis of category mistakes, we show that this assumption conflates an epistemic construct (the logical ordering of messages) with an ontic claim (that physical causality is globally acyclic and monotonic). We trace this conflation through Shannon's channel model, TLA+, Bell's theorem, and the impossibility results of Fischer-Lynch-Paterson and Brewer's CAP theorem. We then show that special and general relativity permit only local causal structure, and that recent work on indefinite causal order demonstrates that nature admits correlations with no well-defined causal ordering. We propose that mutual information conservation, rather than temporal precedence, provides a more fundamental primitive for distributed consistency.

The reasoning process of Large Language Models (LLMs) is often plagued by hallucinations and missing facts in question-answering tasks. A promising solution is to ground LLMs' answers in verifiable knowledge sources, such as Knowledge Graphs (KGs). Prevailing KG-enhanced methods typically constrained LLM reasoning either by enforcing rules during generation or by imitating paths from a fixed set of demonstrations. However, they naturally confined the reasoning patterns of LLMs within the scope of prior experience or fine-tuning data, limiting their generalizability to out-of-distribution graph reasoning problems. To tackle this problem, in this paper, we propose Explore-on-Graph (EoG), a novel framework that encourages LLMs to autonomously explore a more diverse reasoning space on KGs. To incentivize exploration and discovery of novel reasoning paths, we propose to introduce reinforcement learning during training, whose reward is the correctness of the reasoning paths' final answers. To enhance the efficiency and meaningfulness of the exploration, we propose to incorporate path information as additional reward signals to refine the exploration process and reduce futile efforts. Extensive experiments on five KGQA benchmark datasets demonstrate that, to the best of our knowledge, our method achieves state-of-the-art performance, outperforming not only open-source but also even closed-source LLMs.

Cross-silo federated learning allows multiple organizations to collaboratively train machine learning models without sharing raw data, but client updates can still leak sensitive information through inference attacks. Secure aggregation protects privacy by hiding individual updates, yet it complicates contribution evaluation, which is critical for fair rewards and detecting low-quality or malicious participants. Existing marginal-contribution methods, such as the Shapley value, are incompatible with secure aggregation, and practical alternatives, such as Leave-One-Out, are crude and rely on self-evaluation. We introduce two marginal-difference contribution scores compatible with secure aggregation. Fair-Private satisfies standard fairness axioms, while Everybody-Else eliminates self-evaluation and provides resistance to manipulation, addressing a largely overlooked vulnerability. We provide theoretical guarantees for fairness, privacy, robustness, and computational efficiency, and evaluate our methods on multiple medical image datasets and CIFAR10 in cross-silo settings. Our scores consistently outperform existing baselines, better approximate Shapley-induced client rankings, and improve downstream model performance as well as misbehavior detection. These results demonstrate that fairness, privacy, robustness, and practical utility can be achieved jointly in federated contribution evaluation, offering a principled solution for real-world cross-silo deployments.

Human recursive numeral systems (i.e., counting systems such as English base-10 numerals), like many other grammatical systems, are highly regular. Following prior work that relates cross-linguistic tendencies to biases in learning, we ask whether regular systems are common because regularity facilitates learning. Adopting methods from the Reinforcement Learning literature, we confirm that highly regular human(-like) systems are easier to learn than unattested but possible irregular systems. This asymmetry emerges under the natural assumption that recursive numeral systems are designed for generalisation from limited data to represent all integers exactly. We also find that the influence of regularity on learnability is absent for unnatural, highly irregular systems, whose learnability is influenced instead by signal length, suggesting that different pressures may influence learnability differently in different parts of the space of possible numeral systems. Our results contribute to the body of work linking learnability to cross-linguistic prevalence.

Tabular data is the primary data format in industrial relational databases, underpinning modern data analytics and decision-making. However, the increasing scale of tabular data poses significant computational and storage challenges to learning-based analytical systems. This highlights the need for data-efficient learning, which enables effective model training and generalization using substantially fewer samples. Dataset condensation (DC) has emerged as a promising data-centric paradigm that synthesizes small yet informative datasets to preserve data utility while reducing storage and training costs. However, existing DC methods are computationally intensive due to reliance on complex gradient-based optimization. Moreover, they often overlook key characteristics of tabular data, such as heterogeneous features and class imbalance. To address these limitations, we introduce C$^{2}$TC (Class-Adaptive Clustering for Tabular Condensation), the first training-free tabular dataset condensation framework that jointly optimizes class allocation and feature representation, enabling efficient and scalable condensation. Specifically, we reformulate the dataset condensation objective into a novel class-adaptive cluster allocation problem (CCAP), which eliminates costly training and integrates adaptive label allocation to handle class imbalance. To solve the NP-hard CCAP, we develop HFILS, a heuristic local search that alternates between soft allocation and class-wise clustering to efficiently obtain high-quality solutions. Moreover, a hybrid categorical feature encoding (HCFE) is proposed for semantics-preserving clustering of heterogeneous discrete attributes. Extensive experiments on 10 real-world datasets demonstrate that C$^{2}$TC improves efficiency by at least 2 orders of magnitude over state-of-the-art baselines, while achieving superior downstream performance.

The growing integration of distributed photovoltaics (PVs) into active distribution networks (ADNs) has exacerbated operational challenges, making it imperative to coordinate diverse equipment to mitigate voltage violations and enhance power quality. Although existing data-driven approaches have demonstrated effectiveness in the voltage control problem, they often require extensive trial-and-error exploration and struggle to incorporate heterogeneous information, such as day-ahead forecasts and semantic-based grid codes. Considering the operational scenarios and requirements in real-world ADNs, in this paper, we propose a hybrid knowledge-data-driven approach that leverages dynamic collaboration between a large language model (LLM) agent and a reinforcement learning (RL) agent to achieve two-stage voltage control. In the day-ahead stage, the LLM agent receives coarse region-level forecasts and generates scheduling strategies for on-load tap changer (OLTC) and shunt capacitors (SCs) to regulate the overall voltage profile. Then in the intra-day stage, based on accurate node-level measurements, the RL agent refines terminal voltages by deriving reactive power generation strategies for PV inverters. On top of the LLM-RL collaboration framework, we further propose a self-evolution mechanism for the LLM agent and a pretrain-finetune pipeline for the RL agent, effectively enhancing and coordinating the policies for both agents. The proposed approach not only aligns more closely with practical operational characteristics but also effectively utilizes the inherent knowledge and reasoning capabilities of the LLM agent, significantly improving training efficiency and voltage control performance. Comprehensive comparisons and ablation studies demonstrate the effectiveness of the proposed method.

State-of-the-art learned reconstruction methods often rely on black-box modules that, despite their strong performance, raise questions about their interpretability and robustness. Here, we build on a recently proposed image reconstruction method, which is based on embedding data-driven information into a model-based convolutional dictionary regularization via neural network-inferred spatially adaptive sparsity level maps. By means of improved network design and dedicated training strategies, we extend the method to achieve filter-permutation invariance as well as the possibility to change the convolutional dictionary at inference time. We apply our method to low-field MRI and compare it to several other recent deep learning-based methods, also on in vivo data, in which the benefit for the use of a different dictionary is showcased. We further assess the method's robustness when tested on in- and out-of-distribution data. When tested on the latter, the proposed method suffers less from the data distribution shift compared to the other learned methods, which we attribute to its reduced reliance on training data due to its underlying model-based reconstruction component.

Minimally invasive surgery has dramatically improved patient operative outcomes, yet identifying safe operative zones remains challenging in critical phases, requiring surgeons to integrate visual cues, procedural phase, and anatomical context under high cognitive load. Existing AI systems offer binary safety verification or static detection, ignoring the phase-dependent nature of intraoperative reasoning. We introduce ResGo, a benchmark of laparoscopic frames annotated with Go Zone bounding boxes and clinician-authored rationales covering phase, exposure quality reasoning, next action and risk reminder. We introduce evaluation metrics that treat correct grounding under incorrect phase as failures, revealing that most vision-language models cannot handle such tasks and perform poorly. We then present SurGo-R1, a model optimized via RLHF with a multi-turn phase-then-go architecture where the model first identifies the surgical phase, then generates reasoning and Go Zone coordinates conditioned on that context. On unseen procedures, SurGo-R1 achieves 76.6% phase accuracy, 32.7 mIoU, and 54.8% hardcore accuracy, a 6.6$\times$ improvement over the mainstream generalist VLMs. Code, model and benchmark will be available at https://github.com/jinlab-imvr/SurGo-R1

Large Vision-Language Models (LVLMs) exhibit outstanding performance on vision-language tasks but struggle with hallucination problems. Through in-depth analysis of LVLM activation patterns, we reveal two key findings: 1) truthfulness and visual perception capabilities predominantly engage different subsets of attention heads within the model architecture; and 2) truthfulness steering vectors vary significantly across different semantic contexts. Based on these observations, we propose Dynamic Multimodal Activation Steering, a training-free approach for hallucination mitigation. Our method constructs a semantic-based truthfulness steering vector database and computes visual perception steering vectors, enabling context-aware interventions during inference by dynamically selecting the most relevant steering vectors based on input semantic similarity and applying them to the most influential attention heads. We conduct comprehensive experiments across multiple models and datasets, demonstrating that our approach significantly enhances model performance, outperforming existing state-of-the-art methods.

A U-Net based deep learning architecture is designed to segment brain tumors as they appear on various MRI modalities. Special emphasis is lent to the non-enhancing tumor compartment. The latter has not been considered anymore in recent brain tumor segmentation challenges like the MICCAI challenges. However, it is considered to be indicative of the survival time of the patient as well as of areas of further tumor growth. Hence it deems essential to have means to automatically delineate its extension within the tumor.

A central challenge in scientific machine learning (ML) is the correct representation of physical systems governed by multi-regime behaviours. In these scenarios, standard data analysis techniques often fail to capture the nature of the data, as the system's response varies significantly across the state space due to its stochasticity and the different physical regimes. Uncertainty quantification (UQ) should thus not be viewed merely as a safety assessment, but as a support to the learning task itself, guiding the model to internalise the behaviour of the data. We address this by focusing on the Critical Heat Flux (CHF) benchmark and dataset presented by the OECD/NEA Expert Group on Reactor Systems Multi-Physics. This case study represents a test for scientific ML due to the non-linear dependence of CHF on the inputs and the existence of distinct microscopic physical regimes. These regimes exhibit diverse statistical profiles, a complexity that requires UQ techniques to internalise the data behaviour and ensure reliable predictions. In this work, we conduct a comparative analysis of UQ methodologies to determine their impact on physical representation. We contrast post-hoc methods, specifically conformal prediction, against end-to-end coverage-oriented pipelines, including (Bayesian) heteroscedastic regression and quality-driven losses. These approaches treat uncertainty not as a final metric, but as an active component of the optimisation process, modelling the prediction and its behaviour simultaneously. We show that while post-hoc methods ensure statistical calibration, coverage-oriented learning effectively reshapes the model's representation to match the complex physical regimes. The result is a model that delivers not only high predictive accuracy but also a physically consistent uncertainty estimation that adapts dynamically to the intrinsic variability of the CHF.

Large language models (LLMs) for code editing have achieved remarkable progress, yet recent empirical studies reveal a fundamental disconnect between technical accuracy and developer productivity. Despite their strong benchmark performance, developers complete tasks 19% slower when using AI assistance, with over 68.81% of recommendations disrupting their mental flow. This misalignment stems from the use of static commit snapshots that lack temporal information, causing models to optimize for end results rather than the incremental, context-sensitive steps that align with developers' natural reasoning process. To bridge this gap, we present EditFlow, which benchmarks and optimizes subsequent code edit recommendation systems through the reconstruction of developer editing flows. EditFlow addresses three key challenges. First, collecting edit-order data that reflects developers' flow is inherently difficult: manual annotation introduces prohibitive overhead, while development logs capture only single trajectories instead of all plausible editing flows. Second, benchmarking recommendation performance against developers' ongoing editing flow requires a digital-twin-like simulation that can faithfully simulate the editing process. Third, existing heterogeneous systems vary drastically in scale and architecture, posing challenges for developing a unified optimization strategy that endows all models with mental-flow awareness regardless of design or capability. ......

Multimodal time series forecasting has garnered significant attention for its potential to provide more accurate predictions than traditional single-modality models by leveraging rich information inherent in other modalities. However, due to fundamental challenges in modality alignment, existing methods often struggle to effectively incorporate multimodal data into predictions, particularly textual information that has a causal influence on time series fluctuations, such as emergency reports and policy announcements. In this paper, we reflect on the role of textual information in numerical forecasting and propose Time series transformers with Multimodal Mixture-of-Experts, TiMi, to unleash the causal reasoning capabilities of LLMs. Concretely, TiMi utilizes LLMs to generate inferences on future developments, which serve as guidance for time series forecasting. To seamlessly integrate both exogenous factors and time series into predictions, we introduce a Multimodal Mixture-of-Experts (MMoE) module as a lightweight plug-in to empower Transformer-based time series models for multimodal forecasting, eliminating the need for explicit representation-level alignment. Experimentally, our proposed TiMi demonstrates consistent state-of-the-art performance on sixteen real-world multimodal forecasting benchmarks, outperforming advanced baselines while offering both strong adaptability and interpretability.

Multi-modal distribution in robotic manipulation action sequences poses critical challenges for imitation learning. To this end, existing approaches often model the action space as either a discrete set of tokens or a continuous, latent-variable distribution. However, both approaches present trade-offs: some methods discretize actions into tokens and therefore lose fine-grained action variations, while others generate continuous actions in a single stage tend to produce unstable mode transitions. To address these limitations, we propose Primary-Fine Decoupling for Action Generation (PF-DAG), a two-stage framework that decouples coarse action consistency from fine-grained variations. First, we compress action chunks into a small set of discrete modes, enabling a lightweight policy to select consistent coarse modes and avoid mode bouncing. Second, a mode conditioned MeanFlow policy is learned to generate high-fidelity continuous actions. Theoretically, we prove PF-DAG's two-stage design achieves a strictly lower MSE bound than single-stage generative policies. Empirically, PF-DAG outperforms state-of-the-art baselines across 56 tasks from Adroit, DexArt, and MetaWorld benchmarks. It further generalizes to real-world tactile dexterous manipulation tasks. Our work demonstrates that explicit mode-level decoupling enables both robust multi-modal modeling and reactive closed-loop control for robotic manipulation.

Eliminating undefined behaviors (UBs) in Rust programs requires a deep semantic understanding to enable accurate and reliable repair. While existing studies have demonstrated the potential of LLMs to support Rust code analysis and repair, most frameworks remain constrained by inflexible templates or lack grounding in executable semantics, resulting in limited contextual awareness and semantic incorrectness. Here, we present AkiraRust, an LLM-driven repair and verification framework that incorporates a finite-state machine to dynamically adapt its detection and repair flow to runtime semantic conditions. AkiraRust introduces a dual-mode reasoning strategy that coordinates fast and slow thinking across multiple agents. Each agent is mapped to an FSM state, and a waveform-driven transition controller manages state switching, rollback decisions, and semantic check pointing, enabling context-aware and runtime-adaptive repair. Experimental results show that AkiraRust achieves about 92% semantic correctness and delivers a 2.2x average speedup compared to SOTA.

Cooperative Multi-Agent Reinforcement Learning (MARL) solves complex tasks that require coordination from multiple agents, but is often limited to either local (independent learning) or global (centralized learning) perspectives. In this paper, we introduce a novel sequential training scheme and MARL architecture, which learns from multiple perspectives on different hierarchy levels. We propose the Hierarchical Lead Critic (HLC) - inspired by natural emerging distributions in team structures, where following high-level objectives combines with low-level execution. HLC demonstrates that introducing multiple hierarchies, leveraging local and global perspectives, can lead to improved performance with high sample efficiency and robust policies. Experimental results conducted on cooperative, non-communicative, and partially observable MARL benchmarks demonstrate that HLC outperforms single hierarchy baselines and scales robustly with increasing amounts of agents and difficulty.

Generative recommendation (GR) aligns with advances in generative AI by casting next-item prediction as token-level generation rather than score-based ranking. Most GR methods adopt a two-stage pipeline: (i) \textit{item tokenization}, which maps each item to a sequence of discrete, hierarchically organized tokens; and (ii) \textit{autoregressive generation}, which predicts the next item's tokens conditioned on the tokens of user's interaction history. Although hierarchical tokenization induces a prefix tree (trie) over items, standard autoregressive modeling with conventional Transformers often flattens item tokens into a linear stream and overlooks the underlying topology. To address this, we propose TrieRec, a trie-aware generative recommendation method that augments Transformers with structural inductive biases via two positional encodings. First, a \textit{trie-aware absolute positional encoding} aggregates a token's (node's) local structural context (\eg depth, ancestors, and descendants) into the token representation. Second, a \textit{topology-aware relative positional encoding} injects pairwise structural relations into self-attention to capture topology-induced semantic relatedness. TrieRec is also model-agnostic, efficient, and hyperparameter-free. In our experiments, we implement TrieRec within three representative GR backbones, achieving notably improvements of 8.83\% on average across four real-world datasets.

Active automata learning (AAL) algorithms can learn a behavioral model of a system from interacting with it. The primary challenge remains scaling to larger models, in particular in the presence of many possible inputs to the system. Modern AAL algorithms fail to scale even if, in every state, most inputs lead to errors. In various challenging problems from the literature, these errors are observable, i.e., they emit a known error output. Motivated by these problems, we study learning these systems more efficiently. Further, we consider various degrees of knowledge about which inputs are non-error producing at which state. For each level of knowledge, we provide a matching adaptation of the state-of-the-art AAL algorithm L# to make the most of this domain knowledge. Our empirical evaluation demonstrates that the methods accelerate learning by orders of magnitude with strong but realistic domain knowledge to a single order of magnitude with limited domain knowledge.

Multi-robot task planning requires decomposing natural-language instructions into executable actions for heterogeneous robot teams. Conventional Planning Domain Definition Language (PDDL) planners provide rigorous guarantees but struggle to handle ambiguous or long-horizon missions, while large language models (LLMs) can interpret instructions and propose plans but may hallucinate or produce infeasible actions. We present a hierarchical multi-agent LLM-based planner with prompt optimization: an upper layer decomposes tasks and assigns them to lower-layer agents, which generate PDDL problems solved by a classical planner. When plans fail, the system applies TextGrad-inspired textual-gradient updates to optimize each agent's prompt and thereby improve planning accuracy. In addition, meta-prompts are learned and shared across agents within the same layer, enabling efficient prompt optimization in multi-agent settings. On the MAT-THOR benchmark, our planner achieves success rates of 0.95 on compound tasks, 0.84 on complex tasks, and 0.60 on vague tasks, improving over the previous state-of-the-art LaMMA-P by 2, 7, and 15 percentage points respectively. An ablation study shows that the hierarchical structure, prompt optimization, and meta-prompt sharing contribute roughly +59, +37, and +4 percentage points to the overall success rate.

Knowledge Distillation (KD) has emerged as a crucial technique for compressing Large Language Models (LLMs). Although existing cross-tokenizer KD methods have made notable progress, their effectiveness remains constrained by suboptimal alignment across sequence and vocabulary levels. To address these limitations, we introduce Dual-Space Weighting and Time-Warped Alignment (DWA-KD), a novel cross-tokenizer distillation framework that enhances token-wise distillation through dual-space entropy-based weighting and achieves precise sequence-level alignment by leveraging both lexical and semantic information. At the token level, DWA-KD maps teacher representations into the student space and vice versa, performing dual-space KD via Kullback-Leibler divergence (KL). The process is modulated by dual-space weights that up-weight tokens where the student is uncertain and the teacher is confident, thereby focusing learning on informative tokens rather than treating all positions equally. At the sequence level, DWA-KD applies Soft Dynamic Time Warping (Soft-DTW) to both the embedding and final hidden-state layers, enabling robust alignment of lexical and contextual semantics between teacher and student sequences. Extensive experiments across diverse NLP benchmarks demonstrate that DWA-KD outperforms state-of-the-art KD baselines, while ablation studies confirm the complementary contributions of entropy-based token weighting and embedding and final hidden state layer Soft-DTW alignment.

Computer-aided diagnosis (CAD) has significantly advanced automated chest X-ray diagnosis but remains isolated from clinical workflows and lacks reliable decision support and interpretability. Human-AI collaboration seeks to enhance the reliability of diagnostic models by integrating the behaviors of controllable radiologists. However, the absence of interactive tools seamlessly embedded within diagnostic routines impedes collaboration, while the semantic gap between radiologists' decision-making patterns and model representations further limits clinical adoption. To overcome these limitations, we propose a visual cognition-guided collaborative network (VCC-Net) to achieve the cooperative diagnostic paradigm. VCC-Net centers on visual cognition (VC) and employs clinically compatible interfaces, such as eye-tracking or the mouse, to capture radiologists' visual search traces and attention patterns during diagnosis. VCC-Net employs VC as a spatial cognition guide, learning hierarchical visual search strategies to localize diagnostically key regions. A cognition-graph co-editing module subsequently integrates radiologist VC with model inference to construct a disease-aware graph. The module captures dependencies among anatomical regions and aligns model representations with VC-driven features, mitigating radiologist bias and facilitating complementary, transparent decision-making. Experiments on the public datasets SIIM-ACR, EGD-CXR, and self-constructed TB-Mouse dataset achieved classification accuracies of 88.40%, 85.05%, and 92.41%, respectively. The attention maps produced by VCC-Net exhibit strong concordance with radiologists' gaze distributions, demonstrating a mutual reinforcement of radiologist and model inference. The code is available at https://github.com/IPMI-NWU/VCC-Net.

Image captioning remains a fundamental task for vision language understanding, yet ground-truth supervision still relies predominantly on human-annotated references. Because human annotations reflect subjective preferences and expertise, ground-truth captions are often incomplete or even incorrect, which in turn limits caption models. We argue that caption quality should be assessed by two objective aspects: completeness (does the caption cover all salient visual facts?) and correctness (are the descriptions true with respect to the image?). To this end, we introduce CCCaption: a dual-reward reinforcement learning framework with a dedicated fine-tuning corpus that explicitly optimizes these properties to generate \textbf{C}omplete and \textbf{C}orrect \textbf{Captions}. For completeness, we use diverse LVLMs to disentangle the image into a set of visual queries, and reward captions that answer more of these queries, with a dynamic query sampling strategy to improve training efficiency. For correctness, we penalize captions that contain hallucinations by validating the authenticity of sub-caption queries, which are derived from the caption decomposition. Our symmetric dual-reward optimization jointly maximizes completeness and correctness, guiding models toward captions that better satisfy these objective criteria. Extensive experiments across standard captioning benchmarks show consistent improvements, offering a principled path to training caption models beyond human-annotation imitation.

Large language models have demonstrated capabilities in text generation, while their increasing parameter scales present challenges in computational and memory efficiency. Post-training sparsity (PTS), which reduces model cost by removing weights from dense networks, is an effective approach. However, native dense matrices lack high sparsity, making existing approaches that directly remove weights disrupt model states, resulting in unsatisfactory performance recovery even with post-tuning. We propose Sparsity Induction, which promotes models toward higher sparsity at both distribution and feature levels before pruning, to push the limits of PTS. At the distribution level, we enhance distributional sparsity through mathematically equivalent scaling transformations, which are fully absorbable and incur no extra parameters or inference-time overhead. At the feature level, we introduce Spectral Norm Loss to promote feature sparsity from a low-rank perspective. Experiments across diverse model architectures and tasks demonstrate that our method further enhances sparsity-friendliness, achieving superior pruning performance over existing approaches.

Public policy decisions are typically justified using a narrow set of headline indicators, leaving many downstream social impacts unstructured and difficult to compare across policies. We propose PPCR-IM, a system for multi-layer DAG-based consequence reasoning and social indicator mapping that addresses this gap. Given a policy description and its context, PPCR-IM uses an LLM-driven, layer-wise generator to construct a directed acyclic graph of intermediate consequences, allowing child nodes to have multiple parents to capture joint influences. A mapping module then aligns these nodes to a fixed indicator set and assigns one of three qualitative impact directions: increase, decrease, or ambiguous change. For each policy episode, the system outputs a structured record containing the DAG, indicator mappings, and three evaluation measures: an expected-indicator coverage score, a discovery rate for overlooked but relevant indicators, and a relative focus ratio comparing the systems coverage to that of the government. PPCR-IM is available both as an online demo and as a configurable XLSX-to-JSON batch pipeline.

Clinical risk prediction models often underperform in real-world settings due to poor calibration, limited transportability, and subgroup disparities. These challenges are amplified in high-dimensional multimodal cancer datasets characterized by complex feature interactions and a p >> n structure. We present a fully reproducible multimodal machine learning framework for 5-year overall survival prediction in breast cancer, integrating clinical variables with high-dimensional transcriptomic and copy-number alteration (CNA) features from the METABRIC cohort. After variance- and sparsity-based filtering and dimensionality reduction, models were trained using stratified train/validation/test splits with validation-based hyperparameter tuning. Two survival approaches were compared: an elastic-net regularized Cox model (CoxNet) and a gradient-boosted survival tree model implemented using XGBoost. CoxNet provides embedded feature selection and stable estimation, whereas XGBoost captures nonlinear effects and higher-order interactions. Performance was assessed using time-dependent area under the ROC curve (AUC), average precision (AP), calibration curves, Brier score, and bootstrapped 95 percent confidence intervals. CoxNet achieved validation and test AUCs of 98.3 and 96.6, with AP values of 90.1 and 80.4. XGBoost achieved validation and test AUCs of 98.6 and 92.5, with AP values of 92.5 and 79.9. Fairness diagnostics showed stable discrimination across age groups, estrogen receptor status, molecular subtypes, and menopausal state. This work introduces a governance-oriented multimodal survival framework emphasizing calibration, fairness auditing, robustness, and reproducibility for high-dimensional clinical machine learning.

This paper presents and evaluates an optimized cascaded Nepali speech-to-English text translation (S2TT) system, focusing on mitigating structural noise introduced by Automatic Speech Recognition (ASR). We first establish highly proficient ASR and NMT components: a Wav2Vec2-XLS-R-300m model achieved a state-of-the-art 2.72% CER on OpenSLR-54, and a multi-stage fine-tuned MarianMT model reached a 28.32 BLEU score on the FLORES-200 benchmark. We empirically investigate the influence of punctuation loss, demonstrating that unpunctuated ASR output significantly degrades translation quality, causing a massive 20.7% relative BLEU drop on the FLORES benchmark. To overcome this, we propose and evaluate an intermediate Punctuation Restoration Module (PRM). The final S2TT pipeline was tested across three configurations on a custom dataset. The optimal configuration, which applied the PRM directly to ASR output, achieved a 4.90 BLEU point gain over the direct ASR-to-NMT baseline (BLEU 36.38 vs. 31.48). This improvement was validated by human assessment, which confirmed the optimized pipeline's superior Adequacy (3.673) and Fluency (3.804). This work validates that targeted punctuation restoration is the most effective intervention for mitigating structural noise in the Nepali S2TT pipeline. It establishes an optimized baseline and demonstrates a critical architectural insight for developing cascaded speech translation systems for similar low-resource languages.

Multimodal Large Language Models (MLLMs) have achieved notable success in enhancing translation performance by integrating multimodal information. However, existing research primarily focuses on image-guided methods, whose applicability is constrained by the scarcity of multilingual image-text pairs. The speech modality overcomes this limitation due to its natural alignment with text and the abundance of existing speech datasets, which enable scalable language coverage. In this paper, we propose a Speech-guided Machine Translation (SMT) framework that integrates speech and text as fused inputs into an MLLM to improve translation quality. To mitigate reliance on low-resource data, we introduce a Self-Evolution Mechanism. The core components of this framework include a text-to-speech model, responsible for generating synthetic speech, and an MLLM capable of classifying synthetic speech samples and iteratively optimizing itself using positive samples. Experimental results demonstrate that our framework surpasses all existing methods on the Multi30K multimodal machine translation benchmark, achieving new state-of-the-art results. Furthermore, on general machine translation datasets, particularly the FLORES-200, it achieves average state-of-the-art performance in 108 translation directions. Ablation studies on CoVoST-2 confirms that differences between synthetic and authentic speech have negligible impact on translation quality. The code and models are released at https://github.com/yxduir/LLM-SRT.

Uncertainty is inherent in modern engineered systems, including cyber-physical systems, autonomous systems, and large-scale software-intensive infrastructures (such as microservice-based systems) operating in dynamic and partially observable environments. The recent publication of Precise Semantics for Uncertainty Modeling (PSUM) by the Object Management Group represents the first standardized specification for uncertainty modeling within the Model-Based Systems Engineering (MBSE) community, providing formally defined semantics for representing and reasoning about uncertainty in models. In parallel, the second version of Systems Modeling Language (SysML v2) was released as the next-generation systems modeling language, offering improved semantic rigor and reusability, yet lacking native constructs aligned with PSUM for first-class uncertainty representation. This paper proposes a systematic extension of SysML v2 that incorporates the PSUM metamodel into its modeling framework. The extension enables explicit specification of indeterminacy sources, structured characterization of uncertainties, and consistent propagation of uncertainty within system models, while preserving conformance with SysML v2 syntax and semantics. We validate the approach through seven case studies. Results demonstrate that the proposed extension (PSUM-SysMLv2) is expressive and applicable for uncertainty-aware MBSE, and potentially enables uncertainty and uncertainty propagation analyses.

Effectively addressing client resistance is a sophisticated clinical skill in psychological counseling, yet practitioners often lack timely and scalable supervisory feedback to refine their approaches. Although current NLP research has examined overall counseling quality and general therapeutic skills, it fails to provide granular evaluations of high-stakes moments where clients exhibit resistance. In this work, we present a comprehensive pipeline for the multi-dimensional evaluation of human counselors' interventions specifically targeting client resistance in text-based therapy. We introduce a theory-driven framework that decomposes counselor responses into four distinct communication mechanisms. Leveraging this framework, we curate and share an expert-annotated dataset of real-world counseling excerpts, pairing counselor-client interactions with professional ratings and explanatory rationales. Using this data, we perform full-parameter instruction tuning on a Llama-3.1-8B-Instruct backbone to model fine-grained evaluative judgments of response quality and generate explanations underlying. Experimental results show that our approach can effectively distinguish the quality of different communication mechanisms (77-81% F1), substantially outperforming GPT-4o and Claude-3.5-Sonnet (45-59% F1). Moreover, the model produces high-quality explanations that closely align with expert references and receive near-ceiling ratings from human experts (2.8-2.9/3.0). A controlled experiment with 43 counselors further confirms that receiving these AI-generated feedback significantly improves counselors' ability to respond effectively to client resistance.

Lifetime Value (LTV) prediction is critical in advertising, recommender systems, and e-commerce. In practice, LTV data patterns vary across decision scenarios. As a result, practitioners often build complex, scenario-specific pipelines and iterate over feature processing, objective design, and tuning. This process is expensive and hard to transfer. We propose AgentLTV, an agent-based unified search-and-evolution framework for automated LTV modeling. AgentLTV treats each candidate solution as an {executable pipeline program}. LLM-driven agents generate code, run and repair pipelines, and analyze execution feedback. Two decision agents coordinate a two-stage search. The Monte Carlo Tree Search (MCTS) stage explores a broad space of modeling choices under a fixed budget, guided by the Polynomial Upper Confidence bounds for Trees criterion and a Pareto-aware multi-metric value function. The Evolutionary Algorithm (EA) stage refines the best MCTS program via island-based evolution with crossover, mutation, and migration. Experiments on a large-scale proprietary dataset and a public benchmark show that AgentLTV consistently discovers strong models across ranking and error metrics. Online bucket-level analysis further indicates improved ranking consistency and value calibration, especially for high-value and negative-LTV segments. We summarize practitioner-oriented takeaways: use MCTS for rapid adaptation to new data patterns, use EA for stable refinement, and validate deployment readiness with bucket-level ranking and calibration diagnostics. The proposed AgentLTV has been successfully deployed online.

Standard vision-language-action (VLA) models rely on fitting statistical data priors, limiting their robust understanding of underlying physical dynamics. Reinforcement learning enhances physical grounding through exploration yet typically relies on external reward signals that remain isolated from the agent's internal states. World action models have emerged as a promising paradigm that integrates imagination and control to enable predictive planning. However, they rely on implicit context modeling, lacking explicit mechanisms for self-improvement. To solve these problems, we propose Self-Correcting VLA (SC-VLA), which achieve self-improvement by intrinsically guiding action refinement through sparse imagination. We first design sparse world imagination by integrating auxiliary predictive heads to forecast current task progress and future trajectory trends, thereby constraining the policy to encode short-term physical evolution. Then we introduce the online action refinement module to reshape progress-dependent dense rewards, adjusting trajectory orientation based on the predicted sparse future states. Evaluations on challenging robot manipulation tasks from simulation benchmarks and real-world settings demonstrate that SC-VLA achieve state-of-the-art performance, yielding the highest task throughput with 16% fewer steps and a 9% higher success rate than the best-performing baselines, alongside a 14% gain in real-world experiments. Code is available at https://github.com/Kisaragi0/SC-VLA.

Choreographies describe distributed protocols from a global viewpoint, enabling correct-by-construction synthesis of local behaviours. We develop a policy-parametric type system that prevents information leaks from high-security data to low-security observers, handling both explicit and implicit flows through a program-counter discipline. The system supports recursive procedures via a procedure context that we reconstruct through constraint generation. We prove termination-insensitive non-interference with respect to a standard small-step semantics.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a prevailing paradigm for enhancing reasoning in Multimodal Large Language Models (MLLMs). However, relying solely on outcome supervision risks reward hacking, where models learn spurious reasoning patterns to satisfy final answer checks. While recent rubric-based approaches offer fine-grained supervision signals, they suffer from high computational costs of instance-level generation and inefficient training dynamics caused by treating all rubrics as equally learnable. In this paper, we propose Stratified Rubric-based Curriculum Learning (RuCL), a novel framework that reformulates curriculum learning by shifting the focus from data selection to reward design. RuCL generates generalized rubrics for broad applicability and stratifies them based on the model's competence. By dynamically adjusting rubric weights during training, RuCL guides the model from mastering foundational perception to tackling advanced logical reasoning. Extensive experiments on various visual reasoning benchmarks show that RuCL yields a remarkable +7.83% average improvement over the Qwen2.5-VL-7B model, achieving a state-of-the-art accuracy of 60.06%.

Large Language Models (LLMs) have achieved remarkable success across a wide range of tasks, but serving them efficiently at scale remains a critical challenge due to their substantial computational and latency demands. While most existing inference frameworks rely on simple scheduling strategies such as First-Come-First-Serve (FCFS) at the engine level and Round-Robin (RR) at the scheduler or coordinator level, they often fail to fully utilize system resources and may suffer from issues such as head-of-line blocking and load imbalance. Recent advances in Mixture-of-Experts (MoE) models have also introduced new challenges in scheduling arising from expert parallelism and routing complexity. This research proposes a multi-layer scheduling framework tailored for MoE-based LLM serving. It targets scheduling at three levels: request-level, enginelevel, and expert-level. At the request level, we explore algorithms such as Shortest-Job-First (SJF) and priority-aware aging to improve throughput and reduce latency. At the engine level, we design load-aware dispatching strategies that account for the current prefix token load, KV cache utilization, and user stickiness to achieve better resource matching. At the expert level, we focus on alleviating expert hotspots and strategically placing inter-layer expert dependencies to balance load and improve routing efficiency. Extensive experimental results from more than 100 experiments conducted under diverse workload distributions show that our approach consistently outperforms the state-of-theart inference framework vLLM, achieving up to 17.8% reduction in Time To First Token (TTFT) latency and 13.3% reduction in Time-Per-Output-Token (TPOT) latency.

We study the discrete Bertrand pricing game with a non-increasing demand function. The game has $n \ge 2$ players who simultaneously choose prices from the set $\{1/k, 2/k, \ldots, 1\}$, where $k\in\mathbb{N}$. The player who sets the lowest price captures the entire demand; if multiple players tie for the lowest price, they split the demand equally. We study the Bertrand paradox, where classical theory predicts low prices, yet real markets often sustain high prices. To understand this gap, we analyze a repeated-game model in which firms set prices using no-regret learners. Our goal is to characterize the equilibrium outcomes that can arise under different no-regret learning guarantees. We are particularly interested in questions such as whether no-external-regret learners can converge to undesirable high-price outcomes, and how stronger guarantees such as no-swap regret shape the emergence of competitive low-price behavior. We address these and related questions through a theoretical analysis, complemented by experiments that support the theory and reveal surprising phenomena for no-swap regret learners.

Visual spatial reasoning (VSR) remains challenging for modern vision-language models (VLMs), despite advances in multimodal architectures. A common strategy is to inject additional information at inference time, such as explicit spatial cues, external commonsense knowledge, or chain-of-thought (CoT) reasoning instructions. However, it remains unclear when such information genuinely improves reasoning and when it introduces noise. In this paper, we conduct a hypothesis-driven analysis of information injection for VSR across three representative VLMs and two public benchmarks. We examine (i) the type and number of spatial contexts, (ii) the amount and relevance of injected commonsense knowledge, and (iii) the interaction between spatial grounding and CoT prompting. Our results reveal a consistent pattern: more information does not necessarily yield better reasoning. Targeted single spatial cues outperform multi-context aggregation, excessive or weakly relevant commonsense knowledge degrades performance, and CoT prompting improves accuracy only when spatial grounding is sufficiently precise. These findings highlight the importance of selective, task-aligned information injection and provide practical guidance for designing reliable multimodal reasoning pipelines.

Deep intracranial tumors situated in eloquent brain regions controlling vital functions present critical diagnostic challenges. Clinical practice has shifted toward stereotactic biopsy for pathological confirmation before treatment. Yet biopsy carries inherent risks of hemorrhage and neurological deficits and struggles with sampling bias due to tumor spatial heterogeneity, because pathological changes are typically region-selective rather than tumor-wide. Therefore, advancing non-invasive MRI-based pathology prediction is essential for holistic tumor assessment and modern clinical decision-making. The primary challenge lies in data scarcity: low tumor incidence requires long collection cycles, and annotation demands biopsy-verified pathology from neurosurgical experts. Additionally, tiny lesion volumes lacking segmentation masks cause critical features to be overwhelmed by background noise. To address these challenges, we construct the ICT-MRI dataset - the first public biopsy-verified benchmark with 249 cases across four categories. We propose a Virtual Biopsy framework comprising: MRI-Processor for standardization; Tumor-Localizer employing vision-language models for coarse-to-fine localization via weak supervision; and Adaptive-Diagnoser with a Masked Channel Attention mechanism fusing local discriminative features with global contexts. Experiments demonstrate over 90% accuracy, outperforming baselines by more than 20%.

Large Language Models (LLMs) have demonstrated significant potential as autonomous software engineering (SWE) agents. Recent work has further explored augmenting these agents with memory mechanisms to support long-horizon reasoning. However, these approaches typically operate at a coarse instance granularity, treating the entire problem-solving episode as the atomic unit of storage and retrieval. We empirically demonstrate that instance-level memory suffers from a fundamental granularity mismatch, resulting in misguided retrieval when tasks with similar surface descriptions require distinct reasoning logic at specific stages. To address this, we propose Structurally Aligned Subtask-Level Memory, a method that aligns memory storage, retrieval, and updating with the agent's functional decomposition. Extensive experiments on SWE-bench Verified demonstrate that our method consistently outperforms both vanilla agents and strong instance-level memory baselines across diverse backbones, improving mean Pass@1 over the vanilla agent by +4.7 pp on average (e.g., +6.8 pp on Gemini 2.5 Pro). Performance gains grow with more interaction steps, showing that leveraging past experience benefits long-horizon reasoning in complex software engineering tasks.

Bangla-English code-mixing is widespread across South Asian social media, yet resources for implicit meaning identification in this setting remain scarce. Existing sentiment and sarcasm models largely focus on monolingual English or high-resource languages and struggle with transliteration variation, cultural references, and intra-sentential language switching. To address this gap, we introduce MixSarc, the first publicly available Bangla-English code-mixed corpus for implicit meaning identification. The dataset contains 9,087 manually annotated sentences labeled for humor, sarcasm, offensiveness, and vulgarity. We construct the corpus through targeted social media collection, systematic filtering, and multi-annotator validation. We benchmark transformer-based models and evaluate zero-shot large language models under structured prompting. Results show strong performance on humor detection but substantial degradation on sarcasm, offense, and vulgarity due to class imbalance and pragmatic complexity. Zero-shot models achieve competitive micro-F1 scores but low exact match accuracy. Further analysis reveals that over 42\% of negative sentiment instances in an external dataset exhibit sarcastic characteristics. MixSarc provides a foundational resource for culturally aware NLP and supports more reliable multi-label modeling in code-mixed environments.

High stress occurs when 3D heterogeneous IC packages are subjected to thermal cycling at extreme temperatures. Stress mainly occurs at the interface between different materials. We investigate stress image using latent space representation which is based on using deep generative model (DGM). However, most DGM approaches are unsupervised, meaning they resort to image pairing (input and output) to train DGM. Instead, we rely on a recent boundary-decoder (BD) net, which uses boundary condition and image pairing for stress modeling. The boundary net maps material parameters to the latent space co-shared by its image counterpart. Because such a setup is dimensionally wise ill-posed, we further couple BD net with deep clustering. To access the performance of our proposed method, we simulate an IC chip dataset comprising of 1825 stress images. We compare our new approach using variants of BD net as well as a baseline approach. We show that our approach is able to outperform all the comparison in terms of train and test error reduction.

Public service information systems are often fragmented, inconsistently formatted, and outdated. These characteristics create low-resource retrieval environments that hinder timely access to critical services. We investigate retrieval challenges in such settings through the domain of food pantry access, a socially urgent problem given persistent food insecurity. We develop an AI-powered conversational retrieval system that scrapes and indexes publicly available pantry data and employs a Retrieval-Augmented Generation (RAG) pipeline to support natural language queries via a web interface. We conduct a pilot evaluation study using community-sourced queries to examine system behavior in realistic scenarios. Our analysis reveals key limitations in retrieval robustness, handling underspecified queries, and grounding over inconsistent knowledge bases. This ongoing work exposes fundamental IR challenges in low-resource environments and motivates future research on robust conversational retrieval to improve access to critical public resources.

Neural Graph Databases (NGDBs) facilitate complex logical reasoning over incomplete knowledge structures, yet their training efficiency and expressivity are constrained by rigid query-level batching and structure-exclusive embeddings. We present NGDB-Zoo, a unified framework that resolves these bottlenecks by synergizing operator-level training with semantic augmentation. By decoupling logical operators from query topologies, NGDB-Zoo transforms the training loop into a dynamically scheduled data-flow execution, enabling multi-stream parallelism and achieving a $1.8\times$ - $6.8\times$ throughput compared to baselines. Furthermore, we formalize a decoupled architecture to integrate high-dimensional semantic priors from Pre-trained Text Encoders (PTEs) without triggering I/O stalls or memory overflows. Extensive evaluations on six benchmarks, including massive graphs like ogbl-wikikg2 and ATLAS-Wiki, demonstrate that NGDB-Zoo maintains high GPU utilization across diverse logical patterns and significantly mitigates representation friction in hybrid neuro-symbolic reasoning.

Generative images have proliferated on Web platforms in social media and online copyright distribution scenarios, and semantic watermarking has increasingly been integrated into diffusion models to support reliable provenance tracking and forgery prevention for web content. Traditional noise-layer-based watermarking, however, remains vulnerable to inversion attacks that can recover embedded signals. To mitigate this, recent content-aware semantic watermarking schemes bind watermark signals to high-level image semantics, constraining local edits that would otherwise disrupt global coherence. Yet, large language models (LLMs) possess structured reasoning capabilities that enable targeted exploration of semantic spaces, allowing locally fine-grained but globally coherent semantic alterations that invalidate such bindings. To expose this overlooked vulnerability, we introduce a Coherence-Preserving Semantic Injection (CSI) attack that leverages LLM-guided semantic manipulation under embedding-space similarity constraints. This alignment enforces visual-semantic consistency while selectively perturbing watermark-relevant semantics, ultimately inducing detector misclassification. Extensive empirical results show that CSI consistently outperforms prevailing attack baselines against content-aware semantic watermarking, revealing a fundamental security weakness of current semantic watermark designs when confronted with LLM-driven semantic perturbations.

Agent-based epidemic models (ABMs) encode behavioral and policy heterogeneity but are too slow for nightly hospital planning. We develop county-ready surrogates that learn directly from exascale ABM trajectories using Universal Differential Equations (UDEs): mechanistic SEIR-family ODEs with a neural-parameterized contact rate $κ_φ(u,t)$ (no additive residual). Our contributions are threefold: we adapt multiple shooting and an observer-based prediction-error method (PEM) to stabilize identification of neural-augmented epidemiological dynamics across intervention-driven regime shifts; we enforce positivity and mass conservation and show the learned contact-rate parameterization yields a well-posed vector field; and we quantify accuracy, calibration, and compute against ABM ensembles and UDE baselines. On a representative ExaEpi scenario, PEM-UDE reduces mean MSE by 77% relative to single-shooting UDE (3.00 vs. 13.14) and by 20% relative to MS-UDE (3.75). Reliability improves in parallel: empirical coverage of ABM $10$-$90$% and $25$-$75$% bands rises from 0.68/0.43 (UDE) and 0.79/0.55 (MS-UDE) to 0.86/0.61 with PEM-UDE and 0.94/0.69 with MS+PEM-UDE, indicating calibrated uncertainty rather than overconfident fits. Inference runs in seconds on commodity CPUs (20-35 s per $\sim$90-day forecast), enabling nightly ''what-if'' sweeps on a laptop. Relative to a $\sim$100 CPU-hour ABM reference run, this yields $\sim10^{4}\times$ lower wall-clock per scenario. This closes the realism-cadence gap, supports threshold-aware decision-making (e.g., maintaining ICU occupancy $<75$%), preserves mechanistic interpretability, and enables calibrated, risk-aware scenario planning on standard institutional hardware. Beyond epidemics, the ABM$\to$UDE recipe provides a portable path to distill agent-based simulators into fast, trustworthy surrogates for other scientific domains.

Many applications seek to optimize LLM outputs at test time by iteratively proposing, scoring, and refining candidates over a discrete output space. Existing methods use a calibrated scalar evaluator for the target objective to guide search, but for many tasks such scores are unavailable, too sparse, or unreliable. Pairwise comparisons, by contrast, are often easier to elicit, still provide useful signal on improvement directions, and can be obtained from the LLM itself without external supervision. Building on this observation, we introduce Duel-Evolve, an evolutionary optimization algorithm that replaces external scalar rewards with pairwise preferences elicited from the same LLM used to generate candidates. Duel-Evolve aggregates these noisy candidate comparisons via a Bayesian Bradley-Terry model, yielding uncertainty-aware estimates of candidate quality. These quality estimates guide allocation of the comparison budget toward plausible optima using Double Thompson Sampling, as well as selection of high-quality parents to generate improved candidates. We evaluate Duel-Evolve on MathBench, where it achieves 20 percentage points higher accuracy over existing methods and baselines, and on LiveCodeBench, where it improves over comparable iterative methods by over 12 percentage points. Notably, the method requires no reward model, no ground-truth labels during search, and no hand-crafted scoring function. Results show that pairwise self-preferences provide strong optimization signal for test-time improvement over large, discrete output spaces.

In the context of the evolution of artificial intelligence (AI), the interaction between humans and AI entities has become increasingly salient, challenging the conventional human-centric paradigms of human-machine interaction. To address this challenge, it is imperative to reassess the relationship between AI entities and humans. Through considering both the virtual and physical worlds, we can construct a novel descriptive framework for a world where humans and machines coexist symbiotically. This paper will introduce a fresh research direction engendered for studying harmonious human-machine coexistence across physical and virtual worlds, which has been termed "symmetrical reality". We will elucidate its key characteristics, offering innovative research insight for renovating human-machine interaction paradigms.

Ordinal categorical data are widely collected in psychology, education, and other social sciences, appearing commonly in questionnaires, assessments, and surveys. Latent class models provide a flexible framework for uncovering unobserved heterogeneity by grouping individuals into homogeneous classes based on their response patterns. A fundamental challenge in applying these models is determining the number of latent classes, which is unknown and must be inferred from data. In this paper, we propose one test statistic for this problem. The test statistic centers the largest singular value of a normalized residual matrix by a simple sample-size adjustment. Under the null hypothesis that the candidate number of latent classes is correct, its upper bound converges to zero in probability. Under an under-fitted alternative, the statistic itself exceeds a fixed positive constant with probability approaching one. This sharp dichotomous behavior of the test statistic yields two sequential testing algorithms that consistently estimate the true number of latent classes. Extensive experimental studies confirm the theoretical findings and demonstrate their accuracy and reliability in determining the number of latent classes.

Estimating the asymmetric numbers of communities in multi-layer directed networks is a challenging problem due to the multi-layer structures and inherent directional asymmetry, leading to possibly different numbers of sender and receiver communities. This work addresses this issue under the multi-layer stochastic co-block model, a model for multi-layer directed networks with distinct community structures in sending and receiving sides, by proposing a novel goodness-of-fit test. The test statistic relies on the deviation of the largest singular value of an aggregated normalized residual matrix from the constant 2. The test statistic exhibits a sharp dichotomy: Under the null hypothesis of correct model specification, its upper bound converges to zero with high probability; under underfitting, the test statistic itself diverges to infinity. With this property, we develop a sequential testing procedure that searches through candidate pairs of sender and receiver community numbers in a lexicographic order. The process stops at the smallest such pair where the test statistic drops below a decaying threshold. For robustness, we also propose a ratio-based variant algorithm, which detects sharp changes in the sequence of test statistics by comparing consecutive candidates. Both methods are proven to consistently determine the true numbers of sender and receiver communities under the multi-layer stochastic co-block model.

Developer Productivity Dashboards are essential for visualizing DevOps performance metrics such as Deployment Frequency and Change Failure Rate (DORA). However, the utility of these dashboards is frequently undermined by data reliability issues. In early iterations of our platform, ad-hoc ingestion scripts (Cron jobs) led to "silent failures," where data gaps went undetected for days, eroding organizational trust. This paper reports on our experience migrating from legacy scheduling to a robust Extract-Load-Transform (ELT) pipeline using Directed Acyclic Graph (DAG) orchestration and Medallion Architecture. We detail the operational benefits of decoupling data extraction from transformation, the necessity of immutable raw history for metric redefinition, and the implementation of state-based dependency management. Our experience suggests that treating the metrics pipeline as a production-grade distributed system is a prerequisite for sustainable engineering analytics.

Geo-distribution is essential for modern online applications to ensure service reliability and high availability. However, supporting high-performance serializable transactions in geo-replicated databases remains a significant challenge. This difficulty stems from the extensive over-coordination inherent in distributed atomic commitment, concurrency control, and fault-tolerance replication protocols under high network latency. To address these challenges, we introduce Minerva, a unified distributed concurrency control designed for highly scalable multi-leader replication. Minerva employs a novel epoch-based asynchronous replication protocol that decouples data propagation from the commitment process, enabling continuous transaction replication. Optimistic concurrency control is used to allow any replicas to execute transactions concurrently and commit without coordination. In stead of aborting transactions when conflicts are detected, Minerva uses deterministic re-execution to resolve conflicts, ensuring serializability without sacrificing performance. To further enhance concurrency, we construct a conflict graph and use a maximum weight independent set algorithm to select the optimal subset of transactions for commitment, minimizing the number of re-executed transactions. Our evaluation demonstrates that Minerva significantly outperforms state-of-the-art replicated databases, achieving over $3\times$ higher throughput in scalability experiments and $2.8\times$ higher throughput during a high network latency simulation with the TPC-C benchmark.

Generative Flow Networks (GFlowNets) learn to sample diverse candidates in proportion to a reward function, making them well-suited for scientific discovery, where exploring multiple promising solutions is crucial. Further extending GFlowNets to multi-objective settings has attracted growing interest since real-world applications often involve multiple, conflicting objectives. However, existing approaches require additional training for each set of objectives, limiting their applicability and incurring substantial computational overhead. We propose a training-free mixing policy that composes pre-trained GFlowNets at inference time, enabling rapid adaptation without finetuning or retraining. Importantly, our framework is flexible, capable of handling diverse reward combinations ranging from linear scalarization to complex non-linear logical operators, which are often handled separately in previous literature. We prove that our method exactly recovers the target distribution for linear scalarization and quantify the approximation quality for nonlinear operators through a distortion factor. Experiments on a synthetic 2D grid and real-world molecule-generation tasks demonstrate that our approach achieves performance comparable to baselines that require additional training.

When designing compound AI systems, a common approach is to query multiple copies of the same model and aggregate the responses to produce a synthesized output. Given the homogeneity of these models, this raises the question of whether aggregation unlocks access to a greater set of outputs than querying a single model. In this work, we investigate the power and limitations of aggregation within a stylized principal-agent framework. This framework models how the system designer can partially steer each agent's output through its reward function specification, but still faces limitations due to prompt engineering ability and model capabilities. Our analysis uncovers three natural mechanisms -- feasibility expansion, support expansion, and binding set contraction -- through which aggregation expands the set of outputs that are elicitable by the system designer. We prove that any aggregation operation must implement one of these mechanisms in order to be elicitability-expanding, and that strengthened versions of these mechanisms provide necessary and sufficient conditions that fully characterize elicitability-expansion. Finally, we provide an empirical illustration of our findings for LLMs deployed in a toy reference-generation task. Altogether, our results take a step towards characterizing when compound AI systems can overcome limitations in model capabilities and in prompt engineering.

Retrieval-Augmented Generation (RAG) systems rely critically on the retriever module to surface relevant context for large language models. Although numerous retrievers have recently been proposed, each built on different ranking principles such as lexical matching, dense embeddings, or graph citations, there remains a lack of systematic understanding of how these mechanisms differ and overlap. Existing benchmarks primarily compare entire RAG pipelines or introduce new datasets, providing little guidance on selecting or combining retrievers themselves. Those that do compare retrievers directly use a limited set of evaluation tools which fail to capture complementary and overlapping strengths. This work presents MIGRASCOPE, a Mutual Information based RAG Retriever Analysis Scope. We revisit state-of-the-art retrievers and introduce principled metrics grounded in information and statistical estimation theory to quantify retrieval quality, redundancy, synergy, and marginal contribution. We further show that if chosen carefully, an ensemble of retrievers outperforms any single retriever. We leverage the developed tools over major RAG corpora to provide unique insights on contribution levels of the state-of-the-art retrievers. Our findings provide a fresh perspective on the structure of modern retrieval techniques and actionable guidance for designing robust and efficient RAG systems.

Learning PDE dynamics for fluids increasingly relies on neural operators and Transformer-based models, yet these approaches often lack interpretability and struggle with localized, high-frequency structures while incurring quadratic cost in spatial samples. We propose representing fields with a Gaussian basis, where learned atoms carry explicit geometry (centers, anisotropic scales, weights) and form a compact, mesh-agnostic, directly visualizable state. Building on this representation, we introduce a Gaussian Particle Operator that acts in modal space: learned Gaussian modal windows perform a Petrov-Galerkin measurement, and PG Gaussian Attention enables global cross-scale coupling. This basis-to-basis design is resolution-agnostic and achieves near-linear complexity in N for a fixed modal budget, supporting irregular geometries and seamless 2D-to-3D extension. On standard PDE benchmarks and real datasets, our method attains state-of-the-art competitive accuracy while providing intrinsic interpretability.

Gene expression prediction, which predicts mRNA expression levels from DNA sequences, presents significant challenges. Previous works often focus on extending input sequence length to locate distal enhancers, which may influence target genes from hundreds of kilobases away. Our work first reveals that for current models, long sequence modeling can decrease performance. Even carefully designed algorithms only mitigate the performance degradation caused by long sequences. Instead, we find that proximal multimodal epigenomic signals near target genes prove more essential. Hence we focus on how to better integrate these signals, which has been overlooked. We find that different signal types serve distinct biological roles, with some directly marking active regulatory elements while others reflect background chromatin patterns that may introduce confounding effects. Simple concatenation may lead models to develop spurious associations with these background patterns. To address this challenge, we propose Prism, a framework that learns multiple combinations of high-dimensional epigenomic features to represent distinct background chromatin states and uses backdoor adjustment to mitigate confounding effects. Our experimental results demonstrate that proper modeling of multimodal epigenomic signals achieves state-of-the-art performance using only short sequences for gene expression prediction.

The performance of multi-turn, agentic LLM inference is increasingly dominated by KV-Cache storage I/O rather than computation. In prevalent disaggregated architectures, loading the massive KV-Cache from external storage creates a fundamental imbalance: storage NICs on prefill engines become bandwidth-saturated, while those on decoding engines remain idle. This asymmetry severely constrains overall system throughput. We present DualPath, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading. Beyond the traditional storage-to-prefill path, DualPath enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network. DualPath combines this optimized data path -- which inherently avoids network congestion and avoids interference with latency-critical model execution communications -- with a global scheduler that dynamically balances load across prefill and decode engines. Our evaluation on three models with production agentic workloads demonstrates that DualPath improves offline inference throughput by up to 1.87$\times$ on our in-house inference system. It can also improve online serving throughput by an average factor of 1.96$\times$ without violating SLO.

The Flexible Job Shop Problem (FJSP) is a well-studied combinatorial optimization problem with extensive applications for manufacturing and production scheduling. It involves assigning jobs to various machines to optimize criteria, such as minimizing total completion time. Current learning-based methods in this domain often rely on localized feature extraction models, limiting their capacity to capture overarching dependencies spanning operations and machines. This paper introduces an innovative architecture that harnesses Mamba, a state-space model with linear computational complexity, to facilitate comprehensive sequence modeling tailored for FJSP. In contrast to prevalent graph-attention-based frameworks that are computationally intensive for FJSP, we show our model is more efficient. Specifically, the proposed model possesses an encoder and a decoder. The encoder incorporates a dual Mamba block to extract operation and machine features separately. Additionally, we introduce an efficient cross-attention decoder to learn interactive embeddings of operations and machines. Our experimental results demonstrate that our method achieves faster solving speed and surpasses the performance of state-of-the-art learning-based methods for FJSP across various benchmarks.

The Muon optimizer has demonstrated promising performance in pre-training large language models through gradient (or momentum) orthogonalization. In this work, we propose a simple yet effective enhancement to Muon, namely Muon+, which introduces an additional normalization step after orthogonalization. We demonstrate the effectiveness of Muon+ through extensive pre-training experiments across a wide range of model scales and architectures. Our evaluation includes GPT-style models ranging from 130M to 774M parameters and LLaMA-style models ranging from 60M to 1B parameters. We comprehensively evaluate the effectiveness of Muon+ in the compute-optimal training regime and further extend the token-to-parameter (T2P) ratio to an industrial level of $\approx 200$. Experimental results show that Muon+ provides a consistent boost on training and validation perplexity over Muon. We provide our code here: https://github.com/K1seki221/MuonPlus.

Multilingual pretraining typically lacks explicit alignment signals, leading to suboptimal cross-lingual alignment in the representation space. In this work, we show that training standard pretrained models for cross-lingual alignment with a multi-way parallel corpus in a diverse pool of languages can substantially improve multilingual and cross-lingual representations for NLU tasks. We construct a multi-way parallel dataset using translations of English text from an off-the-shelf NMT model for a pool of six target languages and achieve strong cross-lingual alignment through contrastive learning. This leads to substantial performance gains across both seen and unseen languages for multiple tasks from the MTEB benchmark evaluated for XLM-Roberta and multilingual BERT base models. Using a multi-way parallel corpus for contrastive training yields substantial gains on bitext mining (21.3%), semantic similarity (5.3%), and classification (28.4%) compared to English-centric (En-X) bilingually parallel data, where X is sampled from a pool of multiple target languages. Furthermore, finetuning mE5 model on a small dataset with multi-way parallelism significantly improves bitext mining compared to one without, underscoring the importance of multi-way cross-lingual supervision even for models already pretrained for high-quality sentence embeddings.

Agentic reinforcement learning (ARL) has rapidly gained attention as a promising paradigm for training agents to solve complex, multi-step interactive tasks. Despite encouraging early results, ARL remains highly unstable, often leading to training collapse. This instability limits scalability to larger environments and longer interaction horizons, and constrains systematic exploration of algorithmic design choices. In this paper, we first propose ARLArena, a stable training recipe and systematic analysis framework that examines training stability in a controlled and reproducible setting. ARLArena first constructs a clean and standardized testbed. Then, we decompose policy gradient into four core design dimensions and assess the performance and stability of each dimension. Through this fine-grained analysis, we distill a unified perspective on ARL and propose SAMPO, a stable agentic policy optimization method designed to mitigate the dominant sources of instability in ARL. Empirically, SAMPO achieves consistently stable training and strong performance across diverse agentic tasks. Overall, this study provides a unifying policy gradient perspective for ARL and offers practical guidance for building stable and reproducible LLM-based agent training pipelines.

Large language models (LLMs) are becoming increasingly applied beyond natural language processing, demonstrating strong capabilities in complex scientific tasks that traditionally require human expertise. This progress has extended into materials discovery, where LLMs introduce a new paradigm by leveraging reasoning and in-context learning, capabilities absent from conventional machine learning approaches. Here, we present a Multi-Agent-based Electrocatalyst Search Through Reasoning and Optimization (MAESTRO) framework in which multiple LLMs with specialized roles collaboratively discover high-performance single atom catalysts for the oxygen reduction reaction. Within an autonomous design loop, agents iteratively reason, propose modifications, reflect on results and accumulate design history. Through in-context learning enabled by this iterative process, MAESTRO identified design principles not explicitly encoded in the LLMs' background knowledge and successfully discovered catalysts that break conventional scaling relations between reaction intermediates. These results highlight the potential of multi-agent LLM frameworks as a powerful strategy to generate chemical insight and discover promising catalysts.

General-purpose robots must master long-horizon manipulation, defined as tasks involving multiple kinematic structure changes (e.g., attaching or detaching objects) in unstructured environments. While Vision-Language-Action (VLA) models offer the potential to master diverse atomic skills, they struggle with the combinatorial complexity of sequencing them and are prone to cascading failures due to environmental sensitivity. To address these challenges, we propose LiLo-VLA (Linked Local VLA), a modular framework capable of zero-shot generalization to novel long-horizon tasks without ever being trained on them. Our approach decouples transport from interaction: a Reaching Module handles global motion, while an Interaction Module employs an object-centric VLA to process isolated objects of interest, ensuring robustness against irrelevant visual features and invariance to spatial configurations. Crucially, this modularity facilitates robust failure recovery through dynamic replanning and skill reuse, effectively mitigating the cascading errors common in end-to-end approaches. We introduce a 21-task simulation benchmark consisting of two challenging suites: LIBERO-Long++ and Ultra-Long. In these simulations, LiLo-VLA achieves a 69% average success rate, outperforming Pi0.5 by 41% and OpenVLA-OFT by 67%. Furthermore, real-world evaluations across 8 long-horizon tasks demonstrate an average success rate of 85%. Project page: https://yy-gx.github.io/LiLo-VLA/.

Deciphering brain function through non-invasive recordings requires synthesizing complementary high-frequency electromagnetic (EEG/MEG) and low-frequency metabolic (fMRI) signals. However, despite their shared neural origins, extreme discrepancies have traditionally confined these modalities to isolated analysis pipelines, hindering a holistic interpretation of brain activity. To bridge this fragmentation, we introduce \textbf{NOBEL}, a \textbf{n}euro-\textbf{o}mni-modal \textbf{b}rain-\textbf{e}ncoding \textbf{l}arge language model (LLM) that unifies these heterogeneous signals within the LLM's semantic embedding space. Our architecture integrates a unified encoder for EEG and MEG with a novel dual-path strategy for fMRI, aligning non-invasive brain signals and external sensory stimuli into a shared token space, then leverages an LLM as a universal backbone. Extensive evaluations demonstrate that NOBEL serves as a robust generalist across standard single-modal tasks. We also show that the synergistic fusion of electromagnetic and metabolic signals yields higher decoding accuracy than unimodal baselines, validating the complementary nature of multiple neural modalities. Furthermore, NOBEL exhibits strong capabilities in stimulus-aware decoding, effectively interpreting visual semantics from multi-subject fMRI data on the NSD and HAD datasets while uniquely leveraging direct stimulus inputs to verify causal links between sensory signals and neural responses. NOBEL thus takes a step towards unifying non-invasive brain decoding, demonstrating the promising potential of omni-modal brain understanding.

Many emerging agentic paradigms require agents to collaborate with one another (or people) to achieve shared goals. Unfortunately, existing approaches to learning policies for such collaborative problems produce brittle solutions that fail when paired with new partners. We attribute these failures to a combination of free-riding during training and a lack of strategic robustness. To address these problems, we study the concept of strategic risk aversion and interpret it as a principled inductive bias for generalizable cooperation with unseen partners. While strategically risk-averse players are robust to deviations in their partner's behavior by design, we show that, in collaborative games, they also (1) can have better equilibrium outcomes than those at classical game-theoretic concepts like Nash, and (2) exhibit less or no free-riding. Inspired by these insights, we develop a multi-agent reinforcement learning (MARL) algorithm that integrates strategic risk aversion into standard policy optimization methods. Our empirical results across collaborative benchmarks (including an LLM collaboration task) validate our theory and demonstrate that our approach consistently achieves reliable collaboration with heterogeneous and previously unseen partners across collaborative tasks.

The goal of fair clustering is to find clusters such that the proportion of sensitive attributes (e.g., gender, race, etc.) in each cluster is similar to that of the entire dataset. Various fair clustering algorithms have been proposed that modify standard K-means clustering to satisfy a given fairness constraint. A critical limitation of several existing fair clustering algorithms is that the number of parameters to be learned is proportional to the sample size because the cluster assignment of each datum should be optimized simultaneously with the cluster center, and thus scaling up the algorithms is difficult. In this paper, we propose a new fair clustering algorithm based on a finite mixture model, called Fair Model-based Clustering (FMC). A main advantage of FMC is that the number of learnable parameters is independent of the sample size and thus can be scaled up easily. In particular, mini-batch learning is possible to obtain clusters that are approximately fair. Moreover, FMC can be applied to non-metric data (e.g., categorical data) as long as the likelihood is well-defined. Theoretical and empirical justifications for the superiority of the proposed algorithm are provided.

Robust watermarking is critical for intellectual property protection, whereas existing methods face a severe vulnerability against regeneration-based AIGC attacks. We identify that existing methods fail because they entangle the watermark with high-frequency cover texture, which is susceptible to being rewritten during generative purification. To address this, we propose WaterVIB, a theoretically grounded framework that reformulates the encoder as an information sieve via the Variational Information Bottleneck. Instead of overfitting to fragile cover details, our approach forces the model to learn a Minimal Sufficient Statistic of the message. This effectively filters out redundant cover nuances prone to generative shifts, retaining only the essential signal invariant to regeneration. We theoretically prove that optimizing this bottleneck is a necessary condition for robustness against distribution-shifting attacks. Extensive experiments demonstrate that WaterVIB significantly outperforms state-of-the-art methods, achieving superior zero-shot resilience against unknown diffusion-based editing.

This guide develops high-probability regret bounds for empirical risk minimization (ERM). The presentation is modular: we state broadly applicable guarantees under high-level conditions and give tools for verifying them for specific losses and function classes. We emphasize that many ERM rate derivations can be organized around a three-step recipe -- a basic inequality, a uniform local concentration bound, and a fixed-point argument -- which yields regret bounds in terms of a critical radius, defined via localized Rademacher complexity, under a mild Bernstein-type variance--risk condition. To make these bounds concrete, we upper bound the critical radius using local maximal inequalities and metric-entropy integrals, recovering familiar rates for VC-subgraph, Sobolev/Hölder, and bounded-variation classes. We also review ERM with nuisance components -- including weighted ERM and Neyman-orthogonal losses -- as they arise in causal inference, missing data, and domain adaptation. Following the orthogonal learning framework, we highlight that these problems often admit regret-transfer bounds linking regret under an estimated loss to population regret under the target loss. These bounds typically decompose regret into (i) statistical error under the estimated (optimized) loss and (ii) approximation error due to nuisance estimation. Under sample splitting or cross-fitting, the first term can be controlled using standard fixed-loss ERM regret bounds, while the second term depends only on nuisance-estimation accuracy. We also treat the in-sample regime, where nuisances and the ERM are fit on the same data, deriving regret bounds and giving sufficient conditions for fast rates.

Irregular Multivariate Time Series (IMTS) are characterized by uneven intervals between consecutive timestamps, which carry sampling pattern information valuable and informative for learning temporal and variable dependencies. In addition, IMTS often exhibit diverse dependencies across multiple time scales. However, many existing multi-scale IMTS methods use resampling to obtain the coarse series, which can alter the original timestamps and disrupt the sampling pattern information. To address the challenge, we propose ReIMTS, a Recursive multi-scale modeling approach for Irregular Multivariate Time Series forecasting. Instead of resampling, ReIMTS keeps timestamps unchanged and recursively splits each sample into subsamples with progressively shorter time periods. Based on the original sampling timestamps in these long-to-short subsamples, an irregularity-aware representation fusion mechanism is proposed to capture global-to-local dependencies for accurate forecasting. Extensive experiments demonstrate an average performance improvement of 27.1\% in the forecasting task across different models and real-world datasets. Our code is available at https://github.com/Ladbaby/PyOmniTS.

While defenses for structured PII are mature, Large Language Models (LLMs) pose a new threat: Semantic Sensitive Information (SemSI), where models infer sensitive identity attributes, generate reputation-harmful content, or hallucinate potentially wrong information. The capacity of LLMs to self-regulate these complex, context-dependent sensitive information leaks without destroying utility remains an open scientific question. To address this, we introduce SemSIEdit, an inference-time framework where an agentic "Editor" iteratively critiques and rewrites sensitive spans to preserve narrative flow rather than simply refusing to answer. Our analysis reveals a Privacy-Utility Pareto Frontier, where this agentic rewriting reduces leakage by 34.6% across all three SemSI categories while incurring a marginal utility loss of 9.8%. We also uncover a Scale-Dependent Safety Divergence: large reasoning models (e.g., GPT-5) achieve safety through constructive expansion (adding nuance), whereas capacity-constrained models revert to destructive truncation (deleting text). Finally, we identify a Reasoning Paradox: while inference-time reasoning increases baseline risk by enabling the model to make deeper sensitive inferences, it simultaneously empowers the defense to execute safe rewrites.

Reinforcement learning (RL) has become a central post-training paradigm for large language models (LLMs), but its performance is highly sensitive to the quality of training problems. This sensitivity stems from the non-stationarity of RL: rollouts are generated by an evolving policy, and learning is shaped by exploration and reward feedback, unlike supervised fine-tuning (SFT) with fixed trajectories. As a result, prior work often relies on manual curation or simple heuristic filters (e.g., accuracy), which can admit incorrect or low-utility problems. We propose GradAlign, a gradient-aligned data selection method for LLM reinforcement learning that uses a small, trusted validation set to prioritize training problems whose policy gradients align with validation gradients, yielding an adaptive curriculum. We evaluate GradAlign across three challenging data regimes: unreliable reward signals, distribution imbalance, and low-utility training corpus, showing that GradAlign consistently outperforms existing baselines, underscoring the importance of directional gradient signals in navigating non-stationary policy optimization and yielding more stable training and improved final performance. We release our implementation at https://github.com/StigLidu/GradAlign

In AI, most evaluations of natural language understanding tasks are conducted in standardized dialects such as Standard American English (SAE). In this work, we investigate how accurately large language models (LLMs) represent African American Vernacular English (AAVE). We analyze three LLMs to compare their usage of AAVE to the usage of humans who natively speak AAVE. We first analyzed interviews from the Corpus of Regional African American Language and TwitterAAE to identify the typical contexts where people use AAVE grammatical features such as ain't. We then prompted the LLMs to produce text in AAVE and compared the model-generated text to human usage patterns. We find that, in many cases, there are substantial differences between AAVE usage in LLMs and humans: LLMs usually underuse and misuse grammatical features characteristic of AAVE. Furthermore, through sentiment analysis and manual inspection, we found that the models replicated stereotypes about African Americans. These results highlight the need for more diversity in training data and the incorporation of fairness methods to mitigate the perpetuation of stereotypes.

Text-to-SQL and Big Data are both extensively benchmarked fields, yet there is limited research that evaluates them jointly. In the real world, Text-to-SQL systems are often embedded with Big Data workflows, such as large-scale data processing or interactive data analytics. We refer to this as "Text-to-Big SQL". However, existing text-to-SQL benchmarks remain narrowly scoped and overlook the cost and performance implications that arise at scale. For instance, translation errors that are minor on small datasets lead to substantial cost and latency overheads as data scales, a relevant issue completely ignored by text-to-SQL metrics. In this paper, we overcome this overlooked challenge by introducing novel and representative metrics for evaluating Text-to-Big SQL. Our study focuses on production-level LLM agents, a database-agnostic system adaptable to diverse user needs. Via an extensive evaluation of frontier models, we show that text-to-SQL metrics are insufficient for Big Data. In contrast, our proposed text-to-Big SQL metrics accurately reflect execution efficiency, cost, and the impact of data scale. Furthermore, we provide LLM-specific insights, including fine-grained, cross-model comparisons of latency and cost.

Across many risk-sensitive areas, it is critical to continuously audit the performance of machine learning systems and detect any unusual behavior quickly. This can be modeled as a sequential hypothesis testing problem with $k$ incoming streams of data and a global null hypothesis that asserts that the system is working as expected across all $k$ streams. The standard global test employs a Bonferroni correction and has an expected stopping time bound of $O\left(\ln\frac{k}α\right)$ when $k$ is large and the significance level of the test, $α$, is small. In this work, we construct new sequential tests by using ideas of merging test martingales with different trade-offs in expected stopping times under different, sparse or dense alternative hypotheses. We further derive a new, balanced test that achieves an improved expected stopping time bound that matches Bonferroni's in the sparse setting but that naturally results in $O\left(\frac{1}{k}\ln\frac{1}α\right)$ under a dense alternative. We empirically demonstrate the effectiveness of our proposed tests on synthetic and real-world data.

We study inference on scalar-valued pathwise differentiable targets after adaptive data collection, such as a bandit algorithm. We introduce a novel target-specific condition, directional stability, which is strictly weaker than previously imposed target-agnostic stability conditions. Under directional stability, we show that estimators that would have been efficient under i.i.d. data remain asymptotically normal and semiparametrically efficient when computed from adaptively collected trajectories. The canonical gradient has a martingale form, and directional stability guarantees stabilization of its predictable quadratic variation, enabling high-dimensional asymptotic normality. We characterize efficiency using a convolution theorem for the adaptive-data setting, and give a condition under which the one-step estimator attains the efficiency bound. We verify directional stability for LinUCB, yielding the first semiparametric efficiency guarantee for a regular scalar target under LinUCB sampling.

In this work, we identify and address the core challenges of agentic memory management in LLM serving, where large-scale storage, frequent updates, and multiple coexisting agents jointly introduce complex and high-cost approximate nearest neighbor (ANN) searching problems. We present Pancake, a multi-tier agentic memory system that unifies three key techniques: (i) multi-level index caching for single agents, (ii) coordinated index management across multiple agents, and (iii) collaborative GPU-CPU acceleration. Pancake exposes easy-to-use interface that can be integrated into memory-based agents like Mem-GPT, and is compatible with agentic frameworks such as LangChain and LlamaIndex. Experiments on realistic agent workloads show that Pancake substantially outperforms existing frameworks, achieving more than 4.29x end-to-end throughput improvement.

We propose a knowledge-driven, model-based approach to segmenting audio into single-category and mixed-category chunks with applications to source separation. "Knowledge" here denotes information associated with the data, such as music scores. "Model" here refers to tool that can be used for audio segmentation and recognition, such as hidden Markov models. In contrast to conventional learning that often relies on annotated data with given segment categories and their corresponding boundaries to guide the learning process, the proposed framework does not depend on any pre-segmented training data and learns directly from the input audio and its related knowledge sources to build all necessary models autonomously. Evaluation on simulation data shows that score-guided learning achieves very good music segmentation and separation results. Tested on movie track data for cinematic audio source separation also shows that utilizing sound category knowledge achieves better separation results than those obtained with data-driven techniques without using such information.

Discrete diffusion models have emerged as strong alternatives to autoregressive language models, with recent work initializing and fine-tuning a base unimodal model for bimodal generation. Diverging from previous approaches, we introduce the first tri-modal masked diffusion model pretrained from scratch on text, image-text, and audio-text data. We systematically analyze multimodal scaling laws, modality mixing ratios, noise schedules, and batch-size effects, and we provide optimized inference sampling defaults. Our batch-size analysis yields a novel stochastic differential equation (SDE)-based reparameterization that eliminates the need for tuning the optimal batch size as reported in recent work. This reparameterization decouples the physical batch size, often chosen based on compute constraints (GPU saturation, FLOP efficiency, wall-clock time), from the logical batch size, chosen to balance gradient variance during stochastic optimization. Finally, we pretrain a preliminary 3B-parameter tri-modal model on 6.4T tokens, demonstrating the capabilities of a unified design and achieving strong results in text generation, text-to-image tasks, and text-to-speech tasks. Our work represents the largest-scale systematic open study of multimodal discrete diffusion models conducted to date, providing insights into scaling behaviors across multiple modalities.

Data assimilation and scientific inverse problems require reconstructing high-dimensional physical states from sparse and noisy observations, ideally with uncertainty-aware posterior samples that remain faithful to learned priors and governing physics. While training-free conditional generation is well developed for diffusion models, corresponding conditioning and posterior sampling strategies for Flow Matching (FM) priors remain comparatively under-explored, especially on scientific benchmarks where fidelity must be assessed beyond measurement misfit. In this work, we study training-free conditional generation for scientific inverse problems under FM priors and organize existing inference-time strategies by where measurement information is injected: (i) guided transport dynamics that perturb sampling trajectories using likelihood information, and (ii) source-distribution inference that performs posterior inference over the source variable while keeping the learned transport fixed. Building on the latter, we propose D-Flow SGLD, a source-space posterior sampling method that augments differentiable source inference with preconditioned stochastic gradient Langevin dynamics, enabling scalable exploration of the source posterior induced by new measurement operators without retraining the prior or modifying the learned FM dynamics. We benchmark representative methods from both families on a hierarchy of problems: 2D toy posteriors, chaotic Kuramoto-Sivashinsky trajectories, and wall-bounded turbulence reconstruction. Across these settings, we quantify trade-offs among measurement assimilation, posterior diversity, and physics/statistics fidelity, and establish D-Flow SGLD as a practical FM-compatible posterior sampler for scientific inverse problems.

The spin-$1/2$ $J_1$-$J_2$ Heisenberg model on the square lattice exhibits a debated intermediate phase between Néel antiferromagnetic and stripe ordered regimes, with competing theories proposing plaquette valence bond, nematic, and quantum spin liquid ground states. We apply the Prometheus variational autoencoder framework -- previously validated on classical (2D, 3D Ising) and quantum (disordered transverse field Ising) phase transitions -- to systematically explore the $J_1$-$J_2$ phase diagram via unsupervised analysis of exact diagonalization ground states for a $4 \times 4$ lattice. Through dense parameter scans of $J_2/J_1 \in [0.3, 0.7]$ with step size 0.01 and comprehensive latent space analysis, we investigate the nature of the intermediate regime using unsupervised order parameter discovery and critical point detection via multiple independent methods. This work demonstrates the application of rigorously validated machine learning methods to open questions in frustrated quantum magnetism, where traditional order parameter identification is challenged by competing interactions and limited accessible system sizes.

A key challenge in artificial intelligence and neuroscience is understanding how neural systems learn representations that capture the underlying dynamics of the world. Most world models represent the transition function with unstructured neural networks, limiting interpretability, sample efficiency, and generalization to unseen states or action compositions. We address these issues with a generalizable world model grounded in Vector Symbolic Architecture (VSA) principles as geometric priors. Our approach utilizes learnable Fourier Holographic Reduced Representation (FHRR) encoders to map states and actions into a high dimensional complex vector space with learned group structure and models transitions with element-wise complex multiplication. We formalize the framework's group theoretic foundation and show how training such structured representations to be approximately invariant enables strong multi-step composition directly in latent space and generalization performances over various experiments. On a discrete grid world environment, our model achieves 87.5% zero shot accuracy to unseen state-action pairs, obtains 53.6% higher accuracy on 20-timestep horizon rollouts, and demonstrates 4x higher robustness to noise relative to an MLP baseline. These results highlight how training to have latent group structure yields generalizable, data-efficient, and interpretable world models, providing a principled pathway toward structured models for real-world planning and reasoning.

$E(3)$-equivariant neural networks have proven to be effective in a wide range of 3D modeling tasks. A fundamental operation of such networks is the tensor product, which allows interaction between different feature types. Because this operation scales poorly, there has been considerable work towards accelerating this interaction. However, recently \citet{xieprice} have pointed out that most speedups come from a reduction in expressivity rather than true algorithmic improvements on computing Clebsch-Gordan tensor products. A modification of Gaunt tensor product \citep{gaunt} can give a true asymptotic speedup but is incomplete and misses many interactions. In this work, we provide the first complete algorithm which truly provides asymptotic benefits Clebsch-Gordan tensor products. For full CGTP, our algorithm brings runtime complexity from the naive $O(L^6)$ to $O(L^4\log^2 L)$, close to the lower bound of $O(L^4)$. We first show how generalizing fast Fourier based convolution naturally leads to the previously proposed Gaunt tensor product \citep{gaunt}. To remedy antisymmetry issues, we generalize from scalar signals to irrep valued signals, giving us tensor spherical harmonics. We prove a generalized Gaunt formula for the tensor harmonics. Finally, we show that we only need up to vector valued signals to recover the missing interactions of Gaunt tensor product.

This work presents iMiGUE-Speech, an extension of the iMiGUE dataset that provides a spontaneous affective corpus for studying emotional and affective states. The new release focuses on speech and enriches the original dataset with additional metadata, including speech transcripts, speaker-role separation between interviewer and interviewee, and word-level forced alignments. Unlike existing emotional speech datasets that rely on acted or laboratory-elicited emotions, iMiGUE-Speech captures spontaneous affect arising naturally from real match outcomes. To demonstrate the utility of the dataset and establish initial benchmarks, we introduce two evaluation tasks for comparative assessment: speech emotion recognition and transcript-based sentiment analysis. These tasks leverage state-of-the-art pre-trained representations to assess the dataset's ability to capture spontaneous affective states from both acoustic and linguistic modalities. iMiGUE-Speech can also be synchronously paired with micro-gesture annotations from the original iMiGUE dataset, forming a uniquely multimodal resource for studying speech-gesture affective dynamics. The extended dataset is available at https://github.com/CV-AC/imigue-speech.

We describe extensive numerical experiments assessing and quantifying how classifier performance depends on the quality of the training data, a frequently neglected component of the analysis of classifiers. More specifically, in the scientific context of metagenomic assembly of short DNA reads into "contigs," we examine the effects of degrading the quality of the training data by multiple mechanisms, and for four classifiers -- Bayes classifiers, neural nets, partition models and random forests. We investigate both individual behavior and congruence among the classifiers. We find breakdown-like behavior that holds for all four classifiers, as degradation increases and they move from being mostly correct to only coincidentally correct, because they are wrong in the same way. In the process, a picture of spatial heterogeneity emerges: as the training data move farther from analysis data, classifier decisions degenerate, the boundary becomes less dense, and congruence increases.

Vector glyphs are the atomic units of digital typography, yet most learning-based pipelines still depend on carefully curated exemplar sheets and raster-to-vector postprocessing, which limits accessibility and editability. We introduce VecGlypher, a single multimodal language model that generates high-fidelity vector glyphs directly from text descriptions or image exemplars. Given a style prompt, optional reference glyph images, and a target character, VecGlypher autoregressively emits SVG path tokens, avoiding raster intermediates and producing editable, watertight outlines in one pass. A typography-aware data and training recipe makes this possible: (i) a large-scale continuation stage on 39K noisy Envato fonts to master SVG syntax and long-horizon geometry, followed by (ii) post-training on 2.5K expert-annotated Google Fonts with descriptive tags and exemplars to align language and imagery with geometry; preprocessing normalizes coordinate frames, canonicalizes paths, de-duplicates families, and quantizes coordinates for stable long-sequence decoding. On cross-family OOD evaluation, VecGlypher substantially outperforms both general-purpose LLMs and specialized vector-font baselines for text-only generation, while image-referenced generation reaches a state-of-the-art performance, with marked gains over DeepVecFont-v2 and DualVector. Ablations show that model scale and the two-stage recipe are critical and that absolute-coordinate serialization yields the best geometry. VecGlypher lowers the barrier to font creation by letting users design with words or exemplars, and provides a scalable foundation for future multimodal design tools.

Deep research has emerged as an important task that aims to address hard queries through extensive open-web exploration. To tackle it, most prior work equips large language model (LLM)-based agents with opaque web search APIs, enabling agents to iteratively issue search queries, retrieve external evidence, and reason over it. Despite search's essential role in deep research, black-box web search APIs hinder systematic analysis of search components, leaving the behaviour of established text ranking methods in deep research largely unclear. To fill this gap, we reproduce a selection of key findings and best practices for IR text ranking methods in the deep research setting. In particular, we examine their effectiveness from three perspectives: (i) retrieval units (documents vs. passages), (ii) pipeline configurations (different retrievers, re-rankers, and re-ranking depths), and (iii) query characteristics (the mismatch between agent-issued queries and the training queries of text rankers). We perform experiments on BrowseComp-Plus, a deep research dataset with a fixed corpus, evaluating 2 open-source agents, 5 retrievers, and 3 re-rankers across diverse setups. We find that agent-issued queries typically follow web-search-style syntax (e.g., quoted exact matches), favouring lexical, learned sparse, and multi-vector retrievers; passage-level units are more efficient under limited context windows, and avoid the difficulties of document length normalisation in lexical retrieval; re-ranking is highly effective; translating agent-issued queries into natural-language questions significantly bridges the query mismatch.

Recurrent neural networks (RNNs) can be interpreted as discrete-time state-space models, where the state evolution corresponds to an infinite-impulse-response (IIR) filtering operation governed by both feedforward weights and recurrent poles. While, in principle, all parameters including pole locations can be optimized via backpropagation through time (BPTT), such joint learning incurs substantial computational overhead and is often impractical for applications with limited training data. Echo state networks (ESNs) mitigate this limitation by fixing the recurrent dynamics and training only a linear readout, enabling efficient and stable online adaptation. In this work, we analytically and empirically examine why learning recurrent poles does not provide tangible benefits in data-constrained, real-time learning scenarios. Our analysis shows that pole learning renders the weight optimization problem highly non-convex, requiring significantly more training samples and iterations for gradient-based methods to converge to meaningful solutions. Empirically, we observe that for complex-valued data, gradient descent frequently exhibits prolonged plateaus, and advanced optimizers offer limited improvement. In contrast, fixed-pole architectures induce stable and well-conditioned state representations even with limited training data. Numerical results demonstrate that fixed-pole networks achieve superior performance with lower training complexity, making them more suitable for online real-time tasks.

Introduction: Deep learning-based segmentation models are increasingly integrated into clinical imaging workflows, yet their robustness to adversarial perturbations remains incompletely characterized, particularly for ultrasound images. We evaluated adversarial attacks and inference-time defenses for thyroid nodule segmentation in B-mode ultrasound. Methods: Two black-box adversarial attacks were developed: (1) Structured Speckle Amplification Attack (SSAA), which injects boundary-targeted noise, and (2) Frequency-Domain Ultrasound Attack (FDUA), which applies bandpass-filtered phase perturbations in the Fourier domain. Three inference-time mitigations were evaluated on adversarial images: randomized preprocessing with test-time augmentation, deterministic input denoising, and stochastic ensemble inference with consistency-aware aggregation. Experiments were conducted on a U-Net segmentation model trained on cine-clips from a database of 192 thyroid nodules. Results: The baseline model achieved a mean Dice similarity coefficient (DSC) of 0.76 (SD 0.20) on unperturbed images. SSAA reduced DSC by 0.29 (SD 0.20) while maintaining high visual similarity (SSIM = 0.94). FDUA resulted in a smaller DSC reduction of 0.11 (SD 0.09) with lower visual fidelity (SSIM = 0.82). Against SSAA, all three defenses significantly improved DSC after correction, with deterministic denoising showing the largest recovery (+0.10, p < 0.001), followed by randomized preprocessing (+0.09, p < 0.001), and stochastic ensemble inference (+0.08, p = 0.002). No defense achieved statistically significant improvement against FDUA. Conclusion: Spatial-domain adversarial perturbations in ultrasound segmentation showed partial mitigation with input preprocessing, whereas frequency-domain perturbations were not mitigated by the defenses, highlighting modality-specific challenges in adversarial robustness evaluation.

Current stateless defences for multimodal agentic RAG fail to detect adversarial strategies that distribute malicious semantics across retrieval, planning, and generation components. We formulate this security challenge as a Partially Observable Markov Decision Process (POMDP), where adversarial intent is a latent variable inferred from noisy multi-stage observations. We introduce MMA-RAG^T, an inference-time control framework governed by a Modular Trust Agent (MTA) that maintains an approximate belief state via structured LLM reasoning. Operating as a model-agnostic overlay, MMA-RAGT mediates a configurable set of internal checkpoints to enforce stateful defence-in-depth. Extensive evaluation on 43,774 instances demonstrates a 6.50x average reduction factor in Attack Success Rate relative to undefended baselines, with negligible utility cost. Crucially, a factorial ablation validates our theoretical bounds: while statefulness and spatial coverage are individually necessary (26.4 pp and 13.6 pp gains respectively), stateless multi-point intervention can yield zero marginal benefit under homogeneous stateless filtering when checkpoint detections are perfectly correlated.

Hyperdimensional Computing (HDC) offers a computationally efficient paradigm for neuromorphic learning. Yet, it lacks rigorous uncertainty quantification, leading to open decision boundaries and, consequently, vulnerability to outliers, adversarial perturbations, and out-of-distribution inputs. To address these limitations, we introduce ConformalHDC, a unified framework that combines the statistical guarantees of conformal prediction with the computational efficiency of HDC. For this framework, we propose two complementary variations. First, the set-valued formulation provides finite-sample, distribution-free coverage guarantees. Using carefully designed conformity scores, it forms enclosed decision boundaries that improve robustness to non-conforming inputs. Second, the point-valued formulation leverages the same conformity scores to produce a single prediction when desired, potentially improving accuracy over traditional HDC by accounting for class interactions. We demonstrate the broad applicability of the proposed framework through evaluations on multiple real-world datasets. In particular, we apply our method to the challenging problem of decoding non-spatial stimulus information from the spiking activity of hippocampal neurons recorded as subjects performed a sequence memory task. Our results show that ConformalHDC not only accurately decodes the stimulus information represented in the neural activity data, but also provides rigorous uncertainty estimates and correctly abstains when presented with data from other behavioral states. Overall, these capabilities position the framework as a reliable, uncertainty-aware foundation for neuromorphic computing.

The recent field of neural algorithmic reasoning (NAR) studies the ability of graph neural networks (GNNs) to emulate classical algorithms like Bellman-Ford, a phenomenon known as algorithmic alignment. At the same time, recent advances in large language models (LLMs) have spawned the study of mechanistic interpretability, which aims to identify granular model components like circuits that perform specific computations. In this work, we introduce Mechanistic Interpretability for Neural Algorithmic Reasoning (MINAR), an efficient circuit discovery toolbox that adapts attribution patching methods from mechanistic interpretability to the GNN setting. We show through two case studies that MINAR recovers faithful neuron-level circuits from GNNs trained on algorithmic tasks. Our study sheds new light on the process of circuit formation and pruning during training, as well as giving new insight into how GNNs trained to perform multiple tasks in parallel reuse circuit components for related tasks. Our code is available at https://github.com/pnnl/MINAR.

Multimodal Large Language Models (MLLMs) deliver detailed responses on vision-language tasks, yet remain susceptible to object hallucination (introducing objects not present in the image), undermining reliability in practice. Prior efforts often rely on heuristic penalties, post-hoc correction, or generic decoding tweaks, which do not directly intervene in the mechanisms that trigger object hallucination and thus yield limited gains. To address this challenge, we propose a causal decoding framework that applies targeted causal interventions during generation to curb spurious object mentions. By reshaping the decoding dynamics to attenuate spurious dependencies, our approach reduces false object tokens while maintaining descriptive quality. Across captioning and QA benchmarks, our framework substantially lowers object-hallucination rates and achieves state-of-the-art faithfulness without degrading overall output quality.

In this paper, we study last-iterate convergence of learning algorithms in bilinear saddle-point problems, a preferable notion of convergence that captures the day-to-day behavior of learning dynamics. We focus on the challenging setting where players select actions from compact convex sets and receive only bandit feedback. Our main contribution is the design of an uncoupled learning algorithm that guarantees last-iterate convergence to the Nash equilibrium with high probability. We establish a convergence rate of $\tilde{O}(T^{-1/4})$ up to polynomial factors in problem parameters. Crucially, our proposed algorithm is computationally efficient, requiring only an efficient linear optimization oracle over the players' compact action sets. The algorithm is obtained by combining techniques from experimental design and the classic Follow-The-Regularized-Leader (FTRL) framework, with a carefully chosen regularizer function tailored to the geometry of the action set of each learner.

Flow-based generative models, such as diffusion models and flow matching models, have achieved remarkable success in learning complex data distributions. However, a critical gap remains for their deployment in safety-critical domains: the lack of formal guarantees that generated samples will satisfy hard constraints. We address this by proposing a safety filtering framework that acts as an online shield for any pre-trained generative model. Our key insight is to cooperate with the generative process rather than override it. We define a constricting safety tube that is relaxed at the initial noise distribution and progressively tightens to the target safe set at the final data distribution, mirroring the coarse-to-fine structure of the generative process itself. By characterizing this tube via Control Barrier Functions (CBFs), we synthesize a feedback control input through a convex Quadratic Program (QP) at each sampling step. As the tube is loosest when noise is high and intervention is cheapest in terms of control energy, most constraint enforcement occurs when it least disrupts the model's learned structure. We prove that this mechanism guarantees safe sampling while minimizing the distributional shift from the original model at each sampling step, as quantified by the KL divergence. Our framework applies to any pre-trained flow-based generative scheme requiring no retraining or architectural modifications. We validate the approach across constrained image generation, physically-consistent trajectory sampling, and safe robotic manipulation policies, achieving 100% constraint satisfaction while preserving semantic fidelity.

Medical Vision Language Models (VLMs) can change their answers when clinicians rephrase the same question, which raises deployment risks. We introduce Paraphrase Sensitivity Failure (PSF)-Med, a benchmark of 19,748 chest Xray questions paired with about 92,000 meaningpreserving paraphrases across MIMIC-CXR and PadChest. Across six medical VLMs, we measure yes/no flips for the same image and find flip rates from 8% to 58%. However, low flip rate does not imply visual grounding: text-only baselines show that some models stay consistent even when the image is removed, suggesting they rely on language priors. To study mechanisms in one model, we apply GemmaScope 2 Sparse Autoencoders (SAEs) to MedGemma 4B and analyze FlipBank, a curated set of 158 flip cases. We identify a sparse feature at layer 17 that correlates with prompt framing and predicts decision margin shifts. In causal patching, removing this feature's contribution recovers 45% of the yesminus-no logit margin on average and fully reverses 15% of flips. Acting on this finding, we show that clamping the identified feature at inference reduces flip rates by 31% relative with only a 1.3 percentage-point accuracy cost, while also decreasing text-prior reliance. These results suggest that flip rate alone is not enough; robustness evaluations should test both paraphrase stability and image reliance.

We consider the problem of sampling from a posterior distribution arising in Bayesian inverse problems in science, engineering, and imaging. Our method belongs to the family of independence Metropolis-Hastings (IMH) sampling algorithms, which are common in Bayesian inference. Relying on the existence of an approximate posterior distribution that is cheaper to sample from but may have significant bias, we introduce Proximal-IMH, a scheme that removes this bias by correcting samples from the approximate posterior through an auxiliary optimization problem. This yields a local adjustment that trades off adherence to the exact model against stability around the approximate reference point. For idealized settings, we prove that the proximal correction tightens the match between approximate and exact posteriors, thereby improving acceptance rates and mixing. The method applies to both linear and nonlinear input-output operators and is particularly suitable for inverse problems where exact posterior sampling is too expensive. We present numerical experiments including multimodal and data-driven priors with nonlinear input-output operators. The results show that Proximal-IMH reliably outperforms existing IMH variants.

Reinforcement learning (RL) agents under partial observability often condition actions on internally accumulated information such as memory or inferred latent context. We formalise such information-conditioned interaction patterns as behavioural dependency: variation in action selection with respect to internal information under fixed observations. This induces a probe-relative notion of $ε$-behavioural equivalence and a within-policy behavioural distance that quantifies probe sensitivity. We establish three structural results. First, the set of policies exhibiting non-trivial behavioural dependency is not closed under convex aggregation. Second, behavioural distance contracts under convex combination. Third, we prove a sufficient local condition under which gradient ascent on a skewed mixture objective decreases behavioural distance when a dominant-mode gradient aligns with the direction of steepest contraction. Minimal bandit and partially observable gridworld experiments provide controlled witnesses of these mechanisms. In the examined settings, behavioural distance decreases under convex aggregation and under continued optimisation with skewed latent priors, and in these experiments it precedes degradation under latent prior shift. These results identify structural conditions under which probe-conditioned behavioural separation is not preserved under common policy transformations.

Forest monitoring is critical for climate change mitigation. However, existing global tree height maps provide only static snapshots and do not capture temporal forest dynamics, which are essential for accurate carbon accounting. We introduce ECHOSAT, a global and temporally consistent tree height map at 10 m resolution spanning multiple years. To this end, we resort to multi-sensor satellite data to train a specialized vision transformer model, which performs pixel-level temporal regression. A self-supervised growth loss regularizes the predictions to follow growth curves that are in line with natural tree development, including gradual height increases over time, but also abrupt declines due to forest loss events such as fires. Our experimental evaluation shows that our model improves state-of-the-art accuracies in the context of single-year predictions. We also provide the first global-scale height map that accurately quantifies tree growth and disturbances over time. We expect ECHOSAT to advance global efforts in carbon monitoring and disturbance assessment. The maps can be accessed at https://github.com/ai4forest/echosat.

Reinforcement Learning with Verifiable Rewards (RLVR) has become the leading paradigm for enhancing reasoning in Large Language Models (LLMs). However, standard RLVR algorithms suffer from a well-documented pathology: while they improve Pass@1 accuracy through sharpened sampling, they simultaneously narrow the model's reasoning boundary and reduce generation diversity. We identify a root cause that existing methods overlook: the uniform penalization of errors. Current approaches -- whether data-filtering methods that select prompts by difficulty, or advantage normalization schemes -- treat all incorrect rollouts within a group identically. We show that this uniformity allows overconfident errors (incorrect reasoning paths that the RL process has spuriously reinforced) to persist and monopolize probability mass, ultimately suppressing valid exploratory trajectories. To address this, we propose the Asymmetric Confidence-aware Error Penalty (ACE). ACE introduces a per-rollout confidence shift metric, c_i = log(pi_theta(y_i|x) / pi_ref(y_i|x)), to dynamically modulate negative advantages. Theoretically, we demonstrate that ACE's gradient can be decomposed into the gradient of a selective regularizer restricted to overconfident errors, plus a well-characterized residual that partially moderates the regularizer's strength. We conduct extensive experiments fine-tuning Qwen2.5-Math-7B, Qwen3-8B-Base, and Llama-3.1-8B-Instruct on the DAPO-Math-17K dataset using GRPO and DAPO within the VERL framework. Evaluated on MATH-500 and AIME 2025, ACE composes seamlessly with existing methods and consistently improves the full Pass@k spectrum across all three model families and benchmarks.

Selecting the right deep learning model for power grid forecasting is challenging, as performance heavily depends on the data available to the operator. This paper presents a comprehensive benchmark of five modern neural architectures: two state space models (PowerMamba, S-Mamba), two Transformers (iTransformer, PatchTST), and a traditional LSTM. We evaluate these models on hourly electricity demand across six diverse US power grids for forecast windows between 24 and 168 hours. To ensure a fair comparison, we adapt each model with specialized temporal processing and a modular layer that cleanly integrates weather covariates. Our results reveal that there is no single best model for all situations. When forecasting using only historical load, PatchTST and the state space models provide the highest accuracy. However, when explicit weather data is added to the inputs, the rankings reverse: iTransformer improves its accuracy three times more efficiently than PatchTST. By controlling for model size, we confirm that this advantage stems from the architecture's inherent ability to mix information across different variables. Extending our evaluation to solar generation, wind power, and wholesale prices further demonstrates that model rankings depend on the forecast task: PatchTST excels on highly rhythmic signals like solar, while state space models are better suited for the chaotic fluctuations of wind and price. Ultimately, this benchmark provides grid operators with actionable guidelines for selecting the optimal forecasting architecture based on their specific data environments.

Convex Agreement (CA) strengthens Byzantine Agreement (BA) by requiring the output agreed upon to lie in the convex hull of the honest parties' inputs. This validity condition is motivated by practical aggregation tasks (e.g., robust learning or sensor fusion) where honest inputs need not coincide but should still constrain the decision. CA inherits BA lower bounds, and optimal synchronous round complexity is easy to obtain (e.g., via Byzantine Broadcast). The main challenge is \emph{communication}: standard approaches for CA have a communication complexity of $Θ(Ln^2)$ for large $L$-bit inputs, leaving a gap in contrast to BA's lower bound of $Ω(Ln)$ bits. While recent work achieves optimal communication complexity of $O(Ln)$ for sufficiently large $L$ [GLW,PODC'25], translating this result to general convexity spaces remained an open problem. We investigate this gap for abstract convexity spaces, and we present deterministic synchronous CA protocols with near-optimal communication complexity: when $L = Ω(n \cdot κ)$, where $κ$ is a security parameter, we achieve $O(L\cdot n\log n)$ communication for finite convexity spaces and $O(L\cdot n^{1+o(1)})$ communication for Euclidean spaces $\mathbb{R}^d$. Our protocols have asymptotically optimal round complexity $O(n)$ and, when a bound on the inputs' lengths $L$ is fixed a priori, we achieve near-optimal resilience $t < n/(ω+\varepsilon)$ for any constant $\varepsilon>0$, where $ω$ is the Helly number of the convexity space. If $L$ is unknown, we still achieve resilience $t<n/(ω+\varepsilon+1)$ for any constant $\varepsilon > 0$. We further note that our protocols can be leveraged to efficiently solve parallel BA. Our main technical contribution is the use of extractor graphs to obtain a deterministic assignment of parties to committees, which is resilient against adaptive adversaries.

Gaussian process (GP) surrogates are the default tool for emulating expensive computer experiments, but cubic cost, stationarity assumptions, and Gaussian predictive distributions limit their reach. We propose Generative Bayesian Computation (GBC) via Implicit Quantile Networks (IQNs) as a surrogate framework that targets all three limitations. GBC learns the full conditional quantile function from input--output pairs; at test time, a single forward pass per quantile level produces draws from the predictive distribution. Across fourteen benchmarks we compare GBC to four GP-based methods. GBC improves CRPS by 11--26\% on piecewise jump-process benchmarks, by 14\% on a ten-dimensional Friedman function, and scales linearly to 90,000 training points where dense-covariance GPs are infeasible. A boundary-augmented variant matches or outperforms Modular Jump GPs on two-dimensional jump datasets (up to 46\% CRPS improvement). In active learning, a randomized-prior IQN ensemble achieves nearly three times lower RMSE than deep GP active learning on Rocket LGBB. Overall, GBC records a favorable point estimate in 12 of 14 comparisons. GPs retain an edge on smooth surfaces where their smoothness prior provides effective regularization.

A central premise in evolutionary biology is that individual variation can generate information asymmetries that facilitate the emergence of hierarchical organisation. To examine this process, we develop an agent-based model (ABM) to identify the minimal conditions under which hierarchy arises in dynamic multi-agent systems, focusing on the roles of initial heterogeneity and mutation amplitude across generations. Hierarchical organisation is quantified using the Trophic Incoherence (TI) metric, which captures directional asymmetries in interaction networks. Our results show that even small individual differences can be amplified through repeated local interactions involving reproduction, competition, and cooperation, but that hierarchical order is markedly more sensitive to mutation amplitude than to initial heterogeneity. Across repeated trials, stable hierarchies reliably emerge only when mutation amplitude is sufficiently high, while initial heterogeneity primarily affects early formation rather than long-term persistence. Overall, these findings demonstrate how simple interaction rules can give rise to both the emergence and persistence of hierarchical organisation, providing a quantitative account of how structured inequality can develop from initially homogeneous populations.

The boundary of the firm is determined by coordination cost. We argue that agentic AI induces a structural change in how coordination costs scale: in prior modular systems, integration cost grew with interaction topology (O(n^2) in the number of components); in protocol-mediated agentic systems, integration cost collapses to O(n) while verification scales with task throughput rather than interaction count. This shift selects for a specific organizational equilibrium -- the Headless Firm -- structured as an hourglass: a personalized generative interface at the top, a standardized protocol waist in the middle, and a competitive market of micro-specialized execution agents at the bottom. We formalize this claim as a coordination cost model with two falsifiable empirical predictions: (1) the marginal cost of adding an execution provider should be approximately constant in a mature hourglass ecosystem; (2) the ratio of total coordination cost to task throughput should remain stable as ecosystem size grows. We derive conditions for hourglass stability versus re-centralization and analyze implications for firm size distributions, labor markets, and software economics. The analysis predicts a domain-conditional Great Unbundling: in high knowledge-velocity domains, firm size distributions shift mass from large integrated incumbents toward micro-specialized agents and thin protocol orchestrators.

Federated Learning (FL) enables collaborative model training across multiple clients without sharing their private data. However, data heterogeneity across clients leads to client drift, which degrades the overall generalization performance of the model. This effect is further compounded by overemphasis on poorly performing clients. To address this problem, we propose FedVG, a novel gradient-based federated aggregation framework that leverages a global validation set to guide the optimization process. Such a global validation set can be established using readily available public datasets, ensuring accessibility and consistency across clients without compromising privacy. In contrast to conventional approaches that prioritize client dataset volume, FedVG assesses the generalization ability of client models by measuring the magnitude of validation gradients across layers. Specifically, we compute layerwise gradient norms to derive a client-specific score that reflects how much each client needs to adjust for improved generalization on the global validation set, thereby enabling more informed and adaptive federated aggregation. Extensive experiments on both natural and medical image benchmarking datasets, across diverse model architectures, demonstrate that FedVG consistently improves performance, particularly in highly heterogeneous settings. Moreover, FedVG is modular and can be seamlessly integrated with various state-of-the-art FL algorithms, often further improving their results. Our code is available at https://github.com/alinadevkota/FedVG.

Prompt learning has become a dominant paradigm for adapting vision-language models (VLMs) such as CLIP to downstream tasks without modifying pretrained weights. While extending prompts to both vision and text encoders across multiple transformer layers significantly boosts performance, it dramatically increases the number of trainable parameters, with state-of-the-art methods requiring millions of parameters and abandoning the parameter efficiency that makes prompt tuning attractive. In this work, we propose \textbf{MMLoP} (\textbf{M}ulti-\textbf{M}odal \textbf{Lo}w-Rank \textbf{P}rompting), a framework that achieves deep multi-modal prompting with only \textbf{11.5K trainable parameters}, comparable to early text-only methods like CoOp. MMLoP parameterizes vision and text prompts at each transformer layer through a low-rank factorization, which serves as an implicit regularizer against overfitting on few-shot training data. To further close the accuracy gap with state-of-the-art methods, we introduce three complementary components: a self-regulating consistency loss that anchors prompted representations to frozen zero-shot CLIP features at both the feature and logit levels, a uniform drift correction that removes the global embedding shift induced by prompt tuning to preserve class-discriminative structure, and a shared up-projection that couples vision and text prompts through a common low-rank factor to enforce cross-modal alignment. Extensive experiments across three benchmarks and 11 diverse datasets demonstrate that MMLoP achieves a highly favorable accuracy-efficiency tradeoff, outperforming the majority of existing methods including those with orders of magnitude more parameters, while achieving a harmonic mean of 79.70\% on base-to-novel generalization.

We study the problem of efficiently producing, in an online fashion, generative models of scalar, multiclass, and vector-valued outcomes that cannot be falsified on the basis of the observed data and a pre-specified collection of computational tests. Our contributions are twofold. First, we expand on connections between online high-dimensional multicalibration with respect to an RKHS and recent advances in expected variational inequality problems, enabling efficient algorithms for the former. We then apply this algorithmic machinery to the problem of outcome indistinguishability. Our procedure, Defensive Generation, is the first to efficiently produce online outcome indistinguishable generative models of non-Bernoulli outcomes that are unfalsifiable with respect to infinite classes of tests, including those that examine higher-order moments of the generated distributions. Furthermore, our method runs in near-linear time in the number of samples and achieves the optimal, vanishing T^{-1/2} rate for generation error.

We introduce MrBERT, a family of 150M-300M parameter encoders built on the ModernBERT architecture and pre-trained on 35 languages and code. Through targeted adaptation, this model family achieves state-of-the-art results on Catalan- and Spanish-specific tasks, while establishing robust performance across specialized biomedical and legal domains. To bridge the gap between research and production, we incorporate Matryoshka Representation Learning (MRL), enabling flexible vector sizing that significantly reduces inference and storage costs. Ultimately, the MrBERT family demonstrates that modern encoder architectures can be optimized for both localized linguistic excellence and efficient, high-stakes domain specialization. We open source the complete model family on Huggingface.

Tokenization and sub-tokenization based models like word2vec, BERT and the GPTs are the state-of-the-art in natural language processing. Typically, these approaches have limitations with respect to their input representation. They fail to fully capture orthographic similarities and morphological variations, especially in highly inflected and under-resource languages. To mitigate this problem, we propose to computes word vectors directly from character strings, integrating both semantic and syntactic information. We denote this transformer-based approach Rich Character Embeddings (RCE). Furthermore, we propose a hybrid model that combines transformer and convolutional mechanisms. Both vector representations can be used as a drop-in replacement for dictionary- and subtoken-based word embeddings in existing model architectures. It has the potential to improve performance for both large context-based language models like BERT and small models like word2vec for under-resourced and morphologically rich languages. We evaluate our approach on various tasks like the SWAG, declension prediction for inflected languages, metaphor and chiasmus detection for various languages. Our experiments show that it outperforms traditional token-based approaches on limited data using OddOneOut and TopK metrics.

Extracting clinical information from medical transcripts in low-resource languages remains a significant challenge in healthcare natural language processing (NLP). This study evaluates a two-step pipeline combining Aya-expanse-8B as a Persian-to-English translation model with five open-source small language models (SLMs) -- Qwen2.5-7B-Instruct, Llama-3.1-8B-Instruct, Llama-3.2-3B-Instruct, Qwen2.5-1.5B-Instruct, and Gemma-3-1B-it -- for binary extraction of 13 clinical features from 1,221 anonymized Persian transcripts collected at a cancer palliative care call center. Using a few-shot prompting strategy without fine-tuning, models were assessed on macro-averaged F1-score, Matthews Correlation Coefficient (MCC), sensitivity, and specificity to account for class imbalance. Qwen2.5-7B-Instruct achieved the highest overall performance (median macro-F1: 0.899; MCC: 0.797), while Gemma-3-1B-it showed the weakest results. Larger models (7B--8B parameters) consistently outperformed smaller counterparts in sensitivity and MCC. A bilingual analysis of Aya-expanse-8B revealed that translating Persian transcripts to English improved sensitivity, reduced missing outputs, and boosted metrics robust to class imbalance, though at the cost of slightly lower specificity and precision. Feature-level results showed reliable extraction of physiological symptoms across most models, whereas psychological complaints, administrative requests, and complex somatic features remained challenging. These findings establish a practical, privacy-preserving blueprint for deploying open-source SLMs in multilingual clinical NLP settings with limited infrastructure and annotation resources, and highlight the importance of jointly optimizing model scale and input language strategy for sensitive healthcare applications.

Model merging under unseen test-time distribution shifts often renders naive strategies, such as mean averaging unreliable. This challenge is especially acute in medical imaging, where models are fine-tuned locally at clinics on private data, producing domain-specific models that differ by scanner, protocol, and population. When deployed at an unseen clinical site, test cases arrive in unlabeled, non-i.i.d. batches, and the model must adapt immediately without labels. In this work, we introduce an entropy-adaptive, fully online model-merging method that yields a batch-specific merged model via only forward passes, effectively leveraging target information. We further demonstrate why mean merging is prone to failure and misaligned under heterogeneous domain shifts. Next, we mitigate encoder classifier mismatch by decoupling the encoder and classification head, merging with separate merging coefficients. We extensively evaluate our method with state-of-the-art baselines using two backbones across nine medical and natural-domain generalization image classification datasets, showing consistent gains across standard evaluation and challenging scenarios. These performance gains are achieved while retaining single-model inference at test-time, thereby demonstrating the effectiveness of our method.

Multi-Head Attention (MHA) is the core computational primitive underlying modern Large Language Models (LLMs). However, MHA suffers from a fundamental linear scaling limitation: $H$ attention heads produce exactly $H$ independent attention matrices, with no communication between heads during attention computation. This becomes problematic for multi-step reasoning, where correct answers depend on aggregating evidence from multiple parts of the context and composing latent token-to-token relations over a chain of intermediate inferences. To address this, we propose Interleaved Head Attention (IHA), which enables cross-head mixing by constructing $P$ pseudo-heads per head (typically $P=H$), where each pseudo query/key/value is a learned linear combination of all $H$ original queries, keys and values respectively. Interactions between pseudo-query and pseudo-key heads induce up to $P^2$ attention patterns per head with modest parameter overhead $\mathcal{O}(H^2P)$. We provide theory showing improved efficiency in terms of number of parameters on the synthetic Polynomial task (IHA uses $Θ(\sqrt{k}n^2)$ parameters vs. $Θ(kn^2)$ for MHA) and on the synthetic order-sensitive CPM-3 task (IHA uses $\lceil\sqrt{N_{\max}}\rceil$ heads vs. $N_{\max}$ for MHA). On real-world benchmarks, IHA improves Multi-Key retrieval on RULER by 10-20% (4k-16k) and, after fine-tuning for reasoning on OpenThoughts, improves GSM8K by 5.8% and MATH-500 by 2.8% (Majority Vote) over full attention.

Given a black-box AI system and a task, at what confidence level can a practitioner trust the system's output? We answer with a reliability level -- a single number per system-task pair, derived from self-consistency sampling and conformal calibration, that serves as a black-box deployment gate with exact, finite-sample, distribution-free guarantees. Self-consistency sampling reduces uncertainty exponentially; conformal calibration guarantees correctness within 1/(n+1) of the target level, regardless of the system's errors -- made transparently visible through larger answer sets for harder questions. Weaker models earn lower reliability levels (not accuracy -- see Definition 2.4): GPT-4.1 earns 94.6% on GSM8K and 96.8% on TruthfulQA, while GPT-4.1-nano earns 89.8% on GSM8K and 66.5% on MMLU. We validate across five benchmarks, five models from three families, and both synthetic and real data. Conditional coverage on solvable items exceeds 0.93 across all configurations; sequential stopping reduces API costs by around 50%.

Purpose: Curating large-scale datasets of operating room (OR) workflow, encompassing rare, safety-critical, or atypical events, remains operationally and ethically challenging. This data bottleneck complicates the development of ambient intelligence for detecting, understanding, and mitigating rare or safety-critical events in the OR. Methods: This work presents an OR video diffusion framework that enables controlled synthesis of rare and safety-critical events. The framework integrates a geometric abstraction module, a conditioning module, and a fine-tuned diffusion model to first transform OR scenes into abstract geometric representations, then condition the synthesis process, and finally generate realistic OR event videos. Using this framework, we also curate a synthetic dataset to train and validate AI models for detecting near-misses of sterile-field violations. Results: In synthesizing routine OR events, our method outperforms off-the-shelf video diffusion baselines, achieving lower FVD/LPIPS and higher SSIM/PSNR in both in- and out-of-domain datasets. Through qualitative results, we illustrate its ability for controlled video synthesis of counterfactual events. An AI model trained and validated on the generated synthetic data achieved a RECALL of 70.13% in detecting near safety-critical events. Finally, we conduct an ablation study to quantify performance gains from key design choices. Conclusion: Our solution enables controlled synthesis of routine and rare OR events from abstract geometric representations. Beyond demonstrating its capability to generate rare and safety-critical scenarios, we show its potential to support the development of ambient intelligence models.

Ptychographic imaging at synchrotron and XFEL sources requires dense overlapping scans, limiting throughput and increasing dose. Extending coherent diffractive imaging to overlap-free operation on extended samples remains an open problem. Here, we extend PtychoPINN (O. Hoidn \emph{et al.}, \emph{Scientific Reports} \textbf{13}, 22789, 2023) to deliver \emph{overlap-free, single-shot} reconstructions in a Fresnel coherent diffraction imaging (CDI) geometry while also accelerating conventional multi-shot ptychography. The framework couples a differentiable forward model of coherent scattering with a Poisson photon-counting likelihood; real-space overlap enters as a tunable parameter via coordinate-based grouping rather than a hard requirement. On synthetic benchmarks, reconstructions remain accurate at low counts ($\sim\!10^4$ photons/frame), and overlap-free single-shot reconstruction with an experimental probe reaches amplitude structural similarity (SSIM) 0.904, compared with 0.968 for overlap-constrained reconstruction. Against a data-saturated supervised model with the same backbone (16,384 training images), PtychoPINN achieves higher SSIM with only 1,024 images and generalizes to unseen illumination profiles. Per-graphics processing unit (GPU) throughput is approximately $40\times$ that of least-squares maximum-likelihood (LSQ-ML) reconstruction at matched $128\times128$ resolution. These results, validated on experimental data from the Advanced Photon Source and the Linac Coherent Light Source, unify single-exposure Fresnel CDI and overlapped ptychography within one framework, supporting dose-efficient, high-throughput imaging at modern light sources.

This paper establishes and proves representation theorems for cumulative propositional dependence logic and for cumulative propositional logic with team semantics. Cumulative logics are famously given by System C. For propositional dependence logic, we show that System C entailments are exactly captured by cumulative models from Kraus, Lehmann and Magidor. On the other hand, we show that entailment in cumulative propositional logics with team semantics is exactly captured by cumulative and asymmetric models. For the latter, we also obtain equivalence with cumulative logics based on propositional logic with classical semantics. The proofs will be useful for proving representation theorems for other cumulative logics without negation and material implication.

Bayesian inference for inverse problems involves computing expectations under posterior distributions -- e.g., posterior means, variances, or predictive quantities -- typically via Monte Carlo (MC) estimation. When the quantity of interest varies significantly under the posterior, accurate estimates demand many samples -- a cost often prohibitive for partial differential equation-constrained problems. To address this challenge, we introduce conditional neural control variates, a modular method that learns amortized control variates from joint model-data samples to reduce the variance of MC estimators. To scale to high-dimensional problems, we leverage Stein's identity to design an architecture based on an ensemble of hierarchical coupling layers with tractable Jacobian trace computation. Training requires: (i) samples from the joint distribution of unknown parameters and observed data; and (ii) the posterior score function, which can be computed from physics-based likelihood evaluations, neural operator surrogates, or learned generative models such as conditional normalizing flows. Once trained, the control variates generalize across observations without retraining. We validate our approach on stylized and partial differential equation-constrained Darcy flow inverse problems, demonstrating substantial variance reduction, even when the analytical score is replaced by a learned surrogate.

The rapid accumulation of Earth science data has created a significant scalability challenge; while repositories like PANGAEA host vast collections of datasets, citation metrics indicate that a substantial portion remains underutilized, limiting data reusability. Here we present PANGAEA-GPT, a hierarchical multi-agent framework designed for autonomous data discovery and analysis. Unlike standard Large Language Model (LLM) wrappers, our architecture implements a centralized Supervisor-Worker topology with strict data-type-aware routing, sandboxed deterministic code execution, and self-correction via execution feedback, enabling agents to diagnose and resolve runtime errors. Through use-case scenarios spanning physical oceanography and ecology, we demonstrate the system's capacity to execute complex, multi-step workflows with minimal human intervention. This framework provides a methodology for querying and analyzing heterogeneous repository data through coordinated agent workflows.

Recent advances in alignment techniques such as Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), and Direct Preference Optimization (DPO) have improved the safety of large language models (LLMs). However, these LLMs remain vulnerable to jailbreak attacks that disguise harmful intent through indirect or deceptive phrasing. Using causal intervention, we empirically demonstrate that this vulnerability stems from shallow alignment mechanisms that lack deep reasoning, often rejecting harmful prompts without truly understanding why they are harmful. To mitigate this vulnerability, we propose enhancing alignment through reasoning-aware post-training. We construct and release a novel Chain-of-Thought (CoT) fine-tuning dataset that includes both utility-oriented and safety-critical prompts with step-by-step rationales. Fine-tuning on this dataset encourages models to produce principled refusals grounded in reasoning, outperforming standard SFT baselines. Furthermore, inspired by failure patterns in CoT fine-tuning, we introduce Alignment-Weighted DPO, which targets the most problematic parts of an output by assigning different preference weights to the reasoning and final-answer segments. This produces finer-grained, targeted updates than vanilla DPO and improves robustness to diverse jailbreak strategies. Extensive experiments across multiple safety and utility benchmarks show that our method consistently improves alignment robustness while maintaining overall model utility.

Representation learning has been essential for graph machine learning tasks such as link prediction, community detection, and network visualization. Despite recent advances in achieving high performance on these downstream tasks, little progress has been made toward self-explainable models. Understanding the patterns behind predictions is equally important, motivating recent interest in explainable machine learning. In this paper, we present GraphHull, an explainable generative model that represents networks using two levels of convex hulls. At the global level, the vertices of a convex hull are treated as archetypes, each corresponding to a pure community in the network. At the local level, each community is refined by a prototypical hull whose vertices act as representative profiles, capturing community-specific variation. This two-level construction yields clear multi-scale explanations: a node's position relative to global archetypes and its local prototypes directly accounts for its edges. The geometry is well-behaved by design, while local hulls are kept disjoint by construction. To further encourage diversity and stability, we place principled priors, including determinantal point processes, and fit the model under MAP estimation with scalable subsampling. Experiments on real networks demonstrate the ability of GraphHull to recover multi-level community structure and to achieve competitive or superior performance in link prediction and community detection, while naturally providing interpretable predictions.

Geometry-free view synthesis transformers have recently achieved state-of-the-art performance in Novel View Synthesis (NVS), outperforming traditional approaches that rely on explicit geometry modeling. Yet the factors governing their scaling with compute remain unclear. We present a systematic study of scaling laws for view synthesis transformers and derive design principles for training compute-optimal NVS models. Contrary to prior findings, we show that encoder-decoder architectures can be compute-optimal; we trace earlier negative results to suboptimal architectural choices and comparisons across unequal training compute budgets. Across several compute levels, we demonstrate that our encoder-decoder architecture, which we call the Scalable View Synthesis Model (SVSM), scales as effectively as decoder-only models, achieves a superior performance-compute Pareto frontier, and surpasses the previous state-of-the-art on real-world NVS benchmarks with substantially reduced training compute.

Representing the past in a compressed, efficient, and informative manner is a central problem for systems trained on sequential data. The HiPPO framework, originally proposed by Gu & Dao et al., provides a principled approach to sequential compression by projecting signals onto orthogonal polynomial (OP) bases via structured linear ordinary differential equations. Subsequent works have embedded these dynamics in state space models (SSMs), where HiPPO structure serves as an initialization. Nonlinear successors of these SSM methods such as Mamba are state-of-the-art for many tasks with long-range dependencies, but the mechanisms by which they represent and prioritize history remain largely implicit. In this work, we revisit the HiPPO framework with the goal of making these mechanisms explicit. We show how polynomial representations of history can be extended to support capabilities of modern SSMs such as adaptive allocation of memory and associative memory while retaining direct interpretability in the OP basis. We introduce a unified framework comprising five such extensions, which we collectively refer to as a "HiPPO zoo." Each extension exposes a specific modeling capability through an explicit, interpretable modification of the HiPPO framework. The resulting models adapt their memory online and train in streaming settings with efficient updates. We illustrate the behaviors and modeling advantages of these extensions through a range of synthetic sequence modeling tasks, demonstrating that capabilities typically associated with modern SSMs can be realized through explicit, interpretable polynomial memory structures.

We study the problem of opportunistic approachability: a generalization of Blackwell approachability where the learner would like to obtain stronger guarantees (i.e., approach a smaller set) when their adversary limits themselves to a subset of their possible action space. Bernstein et al. (2014) introduced this problem in 2014 and presented an algorithm that guarantees sublinear approachability rates for opportunistic approachability. However, this algorithm requires the ability to produce calibrated online predictions of the adversary's actions, a problem whose standard implementations require time exponential in the ambient dimension and result in approachability rates that scale as $T^{-O(1/d)}$. In this paper, we present an efficient algorithm for opportunistic approachability that achieves a rate of $O(T^{-1/4})$ (and an inefficient one that achieves a rate of $O(T^{-1/3})$), bypassing the need for an online calibration subroutine. Moreover, in the case where the dimension of the adversary's action set is at most two, we show it is possible to obtain the optimal rate of $O(T^{-1/2})$.

Individuals with similar qualifications and skills may vary in their demeanor, or outward manner: some tend toward self-promotion while others are modest to the point of omitting crucial information. Comparing the self-descriptions of equally qualified job-seekers with different self-presentation styles is therefore problematic. We build an interactive AI for skill elicitation that provides accurate determination of skills while simultaneously allowing individuals to speak in their own voice. Such a system can be deployed, for example, when a new user joins a professional networking platform, or when matching employees to needs during a company reorganization. To obtain sufficient training data, we train an LLM to act as synthetic humans. Elicitation mitigates endogenous bias arising from individuals' own self-reports. To address systematic model bias we enforce a mathematically rigorous notion of equitability ensuring that the covariance between self-presentation manner and skill evaluation error is small.

Analog in-memory computing (AIMC) performs computation directly within resistive crossbar arrays, offering an energy-efficient platform to scale large vision and language models. However, non-ideal analog device properties make the training on AIMC devices challenging. In particular, its update asymmetry can induce a systematic drift of weight updates towards a device-specific symmetric point (SP), which typically does not align with the optimum of the training objective. To mitigate this bias, most existing works assume the SP is known and pre-calibrate it to zero before training by setting the reference point as the SP. Nevertheless, calibrating AIMC devices requires costly pulse updates, and residual calibration error can directly degrade training accuracy. In this work, we present the first theoretical characterization of the pulse complexity of SP calibration and the resulting estimation error. We further propose a dynamic SP estimation method that tracks the SP during model training, and establishes its convergence guarantees. In addition, we develop an enhanced variant based on chopping and filtering techniques from digital signal processing. Numerical experiments demonstrate both the efficiency and effectiveness of the proposed method.

Large language models (LLMs) are becoming the foundation for autonomous agents that can use tools to solve complex tasks. Reinforcement learning (RL) has emerged as a common approach for injecting such agentic capabilities, but typically under tightly controlled training setups. It often depends on carefully constructed task-solution pairs and substantial human supervision, which creates a fundamental obstacle to open-ended self-evolution toward superintelligent systems. In this paper, we propose Tool-R0 framework for training general purpose tool-calling agents from scratch with self-play RL, under a zero-data assumption. Initialized from the same base LLM, Tool-R0 co-evolves a Generator and a Solver with complementary rewards: one proposes targeted challenging tasks at the other's competence frontier and the other learns to solve them with real-world tool calls. This creates a self-evolving cycle that requires no pre-existing tasks or datasets. Evaluation on different tool-use benchmarks show that Tool-R0 yields 92.5 relative improvement over the base model and surpasses fully supervised tool-calling baselines under the same setting. Our work further provides empirical insights into self-play LLM agents by analyzing co-evolution, curriculum dynamics, and scaling behavior.

Accurate and uncertainty-aware trajectory prediction remains a core challenge for autonomous driving, driven by complex multi-agent interactions, diverse scene contexts and the inherently stochastic nature of future motion. Diffusion-based generative models have recently shown strong potential for capturing multimodal futures, yet existing approaches such as cVMD suffer from slow sampling, limited exploitation of generative diversity and brittle scenario encodings. This work introduces cVMDx, an enhanced diffusion-based trajectory prediction framework that improves efficiency, robustness and multimodal predictive capability. Through DDIM sampling, cVMDx achieves up to a 100x reduction in inference time, enabling practical multi-sample generation for uncertainty estimation. A fitted Gaussian Mixture Model further provides tractable multimodal predictions from the generated trajectories. In addition, a CVQ-VAE variant is evaluated for scenario encoding. Experiments on the publicly available highD dataset show that cVMDx achieves higher accuracy and significantly improved efficiency over cVMD, enabling fully stochastic, multimodal trajectory prediction.

Large Language Models (LLMs) are converging towards a singular Artificial Hivemind, where shared Nature (pre-training priors) result in a profound collapse of distributional diversity, limiting the distinct perspectives necessary for creative exploration and scientific discovery. To address this, we propose to equip models with inference-time Nurture (individualized epistemic trajectories) using Epistemic Evolution paradigm, progressing through explore, internalize, and express. We instantiate this via PRISM (Pluralistic Reasoning via In-context Structure Modeling), a model-agnostic system that augments LLM with dynamic On-the-fly Epistemic Graphs. On three creativity benchmarks, PRISM achieves state-of-the-art novelty and significantly expands distributional diversity. Moreover, we evaluate the real-world utility via a challenging rare-disease diagnosis benchmark. Results demonstrate that PRISM successfully uncovers correct long-tail diagnoses that standard LLM miss, confirming that its divergence stems from meaningful exploration rather than incoherent noise. Overall, this work establishes a new paradigm for Pluralistic AI, moving beyond monolithic consensus toward a diverse ecosystem of unique cognitive individuals capable of collective, multi-perspective discovery.

This work considers a number of optimization problems and reductive relations between them. The two main problems we are interested in are the \emph{Optimal Decision Tree} and \emph{Set Cover}. We study these two fundamental tasks under precedence constraints, that is, if a test (or set) $X$ is a predecessor of $Y$, then in any feasible decision tree $X$ needs to be an ancestor of $Y$ (or respectively, if $Y$ is added to set cover, then so must be $X$). For the Optimal Decision Tree we consider two optimization criteria: worst case identification time (height of the tree) or the average identification time. Similarly, for the Set Cover we study two cost measures: the size of the cover or the average cover time. Our approach is to develop a number of algorithmic reductions, where an approximation algorithm for one problem provides an approximation for another via a black-box usage of a procedure for the former. En route we introduce other optimization problems either to complete the `reduction landscape' or because they hold the essence of combinatorial structure of our problems. The latter is brought by a problem of finding a maximum density precedence closed subfamily, where the density is defined as the ratio of the number of items the family covers to its size. By doing so we provide $\cO^*(\sqrt{m})$-approximation algorithms for all of the aforementioned problems. The picture is complemented by a number of hardness reductions that provide $o(m^{1/12-ε})$-inapproximability results for the decision tree and covering problems. Besides giving a complete set of results for general precedence constraints, we also provide polylogarithmic approximation guarantees for two most typically studied and applicable precedence types, outforests and inforests. By providing corresponding hardness results, we show these results to be tight.

Symbolic distillation replaces neural networks, or components thereof, with interpretable, closed-form mathematical expressions. This approach has shown promise in discovering physical laws and mathematical relationships directly from trained deep learning models, yet adoption remains limited due to the engineering barrier of integrating symbolic regression into deep learning workflows. We introduce SymTorch, a library that automates this distillation by wrapping neural network components, collecting their input-output behavior, and approximating them with human-readable equations via PySR. SymTorch handles the engineering challenges that have hindered adoption: GPU-CPU data transfer, input-output caching, model serialization, and seamless switching between neural and symbolic forward passes. We demonstrate SymTorch across diverse architectures including GNNs, PINNs and transformer models. Finally, we present a proof-of-concept for accelerating LLM inference by replacing MLP layers with symbolic surrogates, achieving an 8.3\% throughput improvement with moderate performance degradation.

The standard way to evaluate language models on subjective tasks is through pairwise comparisons: an annotator chooses the "better" of two responses to a prompt. Leaderboards aggregate these comparisons into a single Bradley-Terry (BT) ranking, forcing heterogeneous preferences into a total order and violating basic social-choice desiderata. In contrast, social choice theory provides an alternative approach called maximal lotteries, which aggregates pairwise preferences without imposing any assumptions on their structure. However, we show that maximal lotteries are highly sensitive to preference heterogeneity and can favor models that severely underperform on specific tasks or user subpopulations. We introduce robust lotteries that optimize worst-case performance under plausible shifts in the preference data. On large-scale preference datasets, robust lotteries provide more reliable win rate guarantees across the annotator distribution and recover a stable set of top-performing models. By moving from rankings to pluralistic sets of winners, robust lotteries offer a principled step toward an ecosystem of complementary AI systems that serve the full spectrum of human preferences.

Test-time training (TTT) with KV binding as sequence modeling layer is commonly interpreted as a form of online meta-learning that memorizes a key-value mapping at test time. However, our analysis reveals multiple phenomena that contradict this memorization-based interpretation. Motivated by these findings, we revisit the formulation of TTT and show that a broad class of TTT architectures can be expressed as a form of learned linear attention operator. Beyond explaining previously puzzling model behaviors, this perspective yields multiple practical benefits: it enables principled architectural simplifications, admits fully parallel formulations that preserve performance while improving efficiency, and provides a systematic reduction of diverse TTT variants to a standard linear attention form. Overall, our results reframe TTT not as test-time memorization, but as learned linear attention with enhanced representational capacity.

Visual reinforcement learning is appealing for robotics but expensive -- off-policy methods are sample-efficient yet slow; on-policy methods parallelize well but waste samples. Recent work has shown that off-policy methods can train faster than on-policy methods in wall-clock time for state-based control. Extending this to vision remains challenging, where high-dimensional input images complicate training dynamics and introduce substantial storage and encoding overhead. To address these challenges, we introduce Squint, a visual Soft Actor Critic method that achieves faster wall-clock training than prior visual off-policy and on-policy methods. Squint achieves this via parallel simulation, a distributional critic, resolution squinting, layer normalization, a tuned update-to-data ratio, and an optimized implementation. We evaluate on the SO-101 Task Set, a new suite of eight manipulation tasks in ManiSkill3 with heavy domain randomization, and demonstrate sim-to-real transfer to a real SO-101 robot. We train policies for 15 minutes on a single RTX 3090 GPU, with most tasks converging in under 6 minutes.

We study efficient multi-vector retrieval for late interaction in any modality. Late interaction has emerged as a dominant paradigm for information retrieval in text, images, visual documents, and videos, but its computation and storage costs grow linearly with document length, making it costly for image-, video-, and audio-rich corpora. To address this limitation, we explore query-agnostic methods for compressing multi-vector document representations under a constant vector budget. We introduce four approaches for index compression: sequence resizing, memory tokens, hierarchical pooling, and a novel attention-guided clustering (AGC). AGC uses an attention-guided mechanism to identify the most semantically salient regions of a document as cluster centroids and to weight token aggregation. Evaluating these methods on retrieval tasks spanning text (BEIR), visual-document (ViDoRe), and video (MSR-VTT, MultiVENT 2.0), we show that attention-guided clustering consistently outperforms other parameterized compression methods (sequence resizing and memory tokens), provides greater flexibility in index size than non-parametric hierarchical clustering, and achieves competitive or improved performance compared to a full, uncompressed index. The source code is available at: github.com/hanxiangqin/omni-col-press.

We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge. Within the allowed timeframe of the challenge, Aletheia autonomously solved 6 problems (2, 5, 7, 8, 9, 10) out of 10 according to majority expert assessments; we note that experts were not unanimous on Problem 8 (only). For full transparency, we explain our interpretation of FirstProof and disclose details about our experiments as well as our evaluation. Raw prompts and outputs are available at https://github.com/google-deepmind/superhuman/tree/main/aletheia.

Embodied LLMs endow robots with high-level task reasoning, but they cannot reflect on what went wrong or why, turning deployment into a sequence of independent trials where mistakes repeat rather than accumulate into experience. Drawing upon human reflective practitioners, we introduce Reflective Test-Time Planning, which integrates two modes of reflection: \textit{reflection-in-action}, where the agent uses test-time scaling to generate and score multiple candidate actions using internal reflections before execution; and \textit{reflection-on-action}, which uses test-time training to update both its internal reflection model and its action policy based on external reflections after execution. We also include retrospective reflection, allowing the agent to re-evaluate earlier decisions and perform model updates with hindsight for proper long-horizon credit assignment. Experiments on our newly-designed Long-Horizon Household benchmark and MuJoCo Cupboard Fitting benchmark show significant gains over baseline models, with ablative studies validating the complementary roles of reflection-in-action and reflection-on-action. Qualitative analyses, including real-robot trials, highlight behavioral correction through reflection.

Efficiently processing long sequences with Transformer models usually requires splitting the computations across accelerators via context parallelism. The dominant approaches in this family of methods, such as Ring Attention or DeepSpeed Ulysses, enable scaling over the context dimension but do not focus on memory efficiency, which limits the sequence lengths they can support. More advanced techniques, such as Fully Pipelined Distributed Transformer or activation offloading, can further extend the possible context length at the cost of training throughput. In this paper, we present UPipe, a simple yet effective context parallelism technique that performs fine-grained chunking at the attention head level. This technique significantly reduces the activation memory usage of self-attention, breaking the activation memory barrier and unlocking much longer context lengths. Our approach reduces intermediate tensor memory usage in the attention layer by as much as 87.5$\%$ for 32B Transformers, while matching previous context parallelism techniques in terms of training speed. UPipe can support the context length of 5M tokens when training Llama3-8B on a single 8$\times$H100 node, improving upon prior methods by over 25$\%$.

Despite rapid recent progress in the terminal capabilities of large language models, the training data strategies behind state-of-the-art terminal agents remain largely undisclosed. We address this gap through a systematic study of data engineering practices for terminal agents, making two key contributions: (1) Terminal-Task-Gen, a lightweight synthetic task generation pipeline that supports seed-based and skill-based task construction, and (2) a comprehensive analysis of data and training strategies, including filtering, curriculum learning, long context training, and scaling behavior. Our pipeline yields Terminal-Corpus, a large-scale open-source dataset for terminal tasks. Using this dataset, we train Nemotron-Terminal, a family of models initialized from Qwen3(8B, 14B, 32B) that achieve substantial gains on Terminal-Bench 2.0: Nemotron-Terminal-8B improves from 2.5% to 13.0% Nemotron-Terminal-14B improves from 4.0% to 20.2%, and Nemotron-Terminal-32B improves from 3.4% to 27.4%, matching the performance of significantly larger models. To accelerate research in this domain, we open-source our model checkpoints and most of our synthetic datasets at https://huggingface.co/collections/nvidia/nemotron-terminal.

We study the complexity of smoothed agnostic learning, recently introduced by~\cite{CKKMS24}, in which the learner competes with the best classifier in a target class under slight Gaussian perturbations of the inputs. Specifically, we focus on the prototypical task of agnostically learning halfspaces under subgaussian distributions in the smoothed model. The best known upper bound for this problem relies on $L_1$-polynomial regression and has complexity $d^{\tilde{O}(1/σ^2) \log(1/ε)}$, where $σ$ is the smoothing parameter and $ε$ is the excess error. Our main result is a Statistical Query (SQ) lower bound providing formal evidence that this upper bound is close to best possible. In more detail, we show that (even for Gaussian marginals) any SQ algorithm for smoothed agnostic learning of halfspaces requires complexity $d^{Ω(1/σ^{2}+\log(1/ε))}$. This is the first non-trivial lower bound on the complexity of this task and nearly matches the known upper bound. Roughly speaking, we show that applying $L_1$-polynomial regression to a smoothed version of the function is essentially best possible. Our techniques involve finding a moment-matching hard distribution by way of linear programming duality. This dual program corresponds exactly to finding a low-degree approximating polynomial to the smoothed version of the target function (which turns out to be the same condition required for the $L_1$-polynomial regression to work). Our explicit SQ lower bound then comes from proving lower bounds on this approximation degree for the class of halfspaces.

Pass@k is a widely used performance metric for verifiable large language model tasks, including mathematical reasoning, code generation, and short-answer reasoning. It defines success if any of $k$ independently sampled solutions passes a verifier. This multi-sample inference metric has motivated inference-aware fine-tuning methods that directly optimize pass@$k$. However, prior work reports a recurring trade-off: pass@k improves while pass@1 degrades under such methods. This trade-off is practically important because pass@1 often remains a hard operational constraint due to latency and cost budgets, imperfect verifier coverage, and the need for a reliable single-shot fallback. We study the origin of this trade-off and provide a theoretical characterization of when pass@k policy optimization can reduce pass@1 through gradient conflict induced by prompt interference. We show that pass@$k$ policy gradients can conflict with pass@1 gradients because pass@$k$ optimization implicitly reweights prompts toward low-success prompts; when these prompts are what we term negatively interfering, their upweighting can rotate the pass@k update direction away from the pass@1 direction. We illustrate our theoretical findings with large language model experiments on verifiable mathematical reasoning tasks.

Uniform-state discrete diffusion models excel at few-step generation and guidance due to their ability to self-correct, making them preferred over autoregressive or Masked diffusion models in these settings. However, their sampling quality plateaus with ancestral samplers as the number of steps increases. We introduce a family of Predictor-Corrector (PC) samplers for discrete diffusion that generalize prior methods and apply to arbitrary noise processes. When paired with uniform-state diffusion, our samplers outperform ancestral sampling on both language and image modeling, achieving lower generative perplexity at matched unigram entropy on OpenWebText and better FID/IS scores on CIFAR10. Crucially, unlike conventional samplers, our PC methods continue to improve with more sampling steps. Taken together, these findings call into question the assumption that Masked diffusion is the inevitable future of diffusion-based language modeling. Beyond sampling, we develop a memory-efficient curriculum for the Gaussian relaxation training phase, reducing training time by 25% and memory by 33% compared to Duo while maintaining comparable perplexity on OpenWebText and LM1B and strong downstream performance. We release code, checkpoints, and a video-tutorial on: https://s-sahoo.com/duo-ch2

The CAP theorem is routinely treated as a systems law: under network partition, a replicated service must sacrifice either consistency or availability. The theorem is correct within its standard asynchronous network model, but operational practice depends on where partition-like phenomena become observable and on how lower layers discard or preserve semantic information about message fate. This paper argues that Open Atomic Ethernet (OAE) shifts the engineering regime in which CAP tradeoffs become application-visible by (i) replacing fire-and-forget link semantics with bounded-time bilateral reconciliation of endpoint state -- the property we call bisynchrony -- and (ii) avoiding Clos funnel points via an octavalent mesh in which each node can act as the root of a locally repaired spanning tree. The result is not the elimination of hard graph cuts, but a drastic reduction in the frequency and duration of application-visible "soft partitions" by detecting and healing dominant fabric faults within hundreds of nanoseconds. We connect this view to Brewer's original CAP framing, the formalization by Gilbert and Lynch, the CAL theorem of Lee et al., which replaces binary partition tolerance with a quantitative measure of apparent latency, and Abadi's PACELC extension.

Deep learning has significantly advanced automated brain tumor diagnosis, yet clinical adoption remains limited by interpretability and computational constraints. Conventional models often act as opaque ''black boxes'' and fail to quantify the complex, irregular tumor boundaries that characterize malignant growth. To address these challenges, we present XMorph, an explainable and computationally efficient framework for fine-grained classification of three prominent brain tumor types: glioma, meningioma, and pituitary tumors. We propose an Information-Weighted Boundary Normalization (IWBN) mechanism that emphasizes diagnostically relevant boundary regions alongside nonlinear chaotic and clinically validated features, enabling a richer morphological representation of tumor growth. A dual-channel explainable AI module combines GradCAM++ visual cues with LLM-generated textual rationales, translating model reasoning into clinically interpretable insights. The proposed framework achieves a classification accuracy of 96.0%, demonstrating that explainability and high performance can co-exist in AI-based medical imaging systems. The source code and materials for XMorph are all publicly available at: https://github.com/ALSER-Lab/XMorph.

Hierarchical, multi-resolution volumetric mapping approaches are widely used to represent large and complex environments as they can efficiently capture their occupancy and connectivity information. Yet widely used path planning methods such as sampling and trajectory optimization do not exploit this explicit connectivity information, and search-based methods such as A* suffer from scalability issues in large-scale high-resolution maps. In many applications, Euclidean shortest paths form the underpinning of the navigation system. For such applications, any-angle planning methods, which find optimal paths by connecting corners of obstacles with straight-line segments, provide a simple and efficient solution. In this paper, we present a method that has the optimality and completeness properties of any-angle planners while overcoming computational tractability issues common to search-based methods by exploiting multi-resolution representations. Extensive experiments on real and synthetic environments demonstrate the proposed approach's solution quality and speed, outperforming even sampling-based methods. The framework is open-sourced to allow the robotics and planning community to build on our research.

Vision-Language-Action (VLA) models are advancing autonomous driving by replacing modular pipelines with unified end-to-end architectures. However, current VLAs face two expensive requirements: (1) massive dataset collection, and (2) dense reasoning annotations. In this work, we address both challenges with \modelname (\textbf{No} \textbf{R}easoning for \textbf{D}riving). Compared to existing VLAs, \modelname achieves competitive performance while being fine-tuned on $<$60\% of the data and no reasoning annotations, resulting in 3$\times$ fewer tokens. We identify that standard Group Relative Policy Optimization (GRPO) fails to yield significant improvements when applied to policies trained on such small, reasoning-free datasets. We show that this limitation stems from difficulty bias, which disproportionately penalizes reward signals from scenarios that produce high-variance rollouts within GRPO. \modelname overcomes this by incorporating Dr.~GRPO, a recent algorithm designed to mitigate difficulty bias in LLMs. As a result, \modelname achieves competitive performance on Waymo and NAVSIM with a fraction of the training data and no reasoning overhead, enabling more efficient autonomous systems.

Counterfactual inference enables clinicians to ask "what if" questions about patient outcomes, but standard methods assume feature independence and simultaneous modifiability -- assumptions violated by longitudinal clinical data. We introduce the Sequential Counterfactual Framework, which respects temporal dependencies in electronic health records by distinguishing immutable features (chronic diagnoses) from controllable features (lab values) and modeling how interventions propagate through time. Applied to 2,723 COVID-19 patients (383 Long COVID heart failure cases, 2,340 matched controls), we demonstrate that 38-67% of patients with chronic conditions would require biologically impossible counterfactuals under naive methods. We identify a cardiorenal cascade (CKD -> AKI -> HF) with relative risks of 2.27 and 1.19 at each step, illustrating temporal propagation that sequential -- but not naive -- counterfactuals can capture. Our framework transforms counterfactual explanation from "what if this feature were different?" to "what if we had intervened earlier, and how would that propagate forward?" -- yielding clinically actionable insights grounded in biological plausibility.

As memory increasingly dominates system cost and energy, heterogeneous on-chip memory systems that combine technologies with complementary characteristics are becoming essential. Gain Cell RAM (GCRAM) offers higher density, lower power, and tunable retention, expanding the design space beyond conventional SRAM. To this end, we create an OpenGCRAM compiler supporting both SRAM and GCRAM. It generates macro-level designs and layouts for commercial CMOS processes and characterizes area, delay, and power across user-defined configurations. The tool enables systematic identification of optimal heterogeneous memory configurations for AI tasks under specified performance metrics.

Patient-generated text such as secure messages, surveys, and interviews contains rich expressions of the patient voice (PV), reflecting communicative behaviors and social determinants of health (SDoH). Traditional qualitative coding frameworks are labor intensive and do not scale to large volumes of patient-authored messages across health systems. Existing machine learning (ML) and natural language processing (NLP) approaches provide partial solutions but often treat patient-centered communication (PCC) and SDoH as separate tasks or rely on models not well suited to patient-facing language. We introduce PVminer, a domain-adapted NLP framework for structuring patient voice in secure patient-provider communication. PVminer formulates PV detection as a multi-label, multi-class prediction task integrating patient-specific BERT encoders (PV-BERT-base and PV-BERT-large), unsupervised topic modeling for thematic augmentation (PV-Topic-BERT), and fine-tuned classifiers for Code, Subcode, and Combo-level labels. Topic representations are incorporated during fine-tuning and inference to enrich semantic inputs. PVminer achieves strong performance across hierarchical tasks and outperforms biomedical and clinical pre-trained baselines, achieving F1 scores of 82.25% (Code), 80.14% (Subcode), and up to 77.87% (Combo). An ablation study further shows that author identity and topic-based augmentation each contribute meaningful gains. Pre-trained models, source code, and documentation will be publicly released, with annotated datasets available upon request for research use.

In safety-critical classification, the cost of failure is often asymmetric, yet Bayesian deep learning summarises epistemic uncertainty with a single scalar, mutual information (MI), that cannot distinguish whether a model's ignorance involves a benign or safety-critical class. We decompose MI into a per-class vector $C_k(x)=σ_k^{2}/(2μ_k)$, with $μ_k{=}\mathbb{E}[p_k]$ and $σ_k^2{=}\mathrm{Var}[p_k]$ across posterior samples. The decomposition follows from a second-order Taylor expansion of the entropy; the $1/μ_k$ weighting corrects boundary suppression and makes $C_k$ comparable across rare and common classes. By construction $\sum_k C_k \approx \mathrm{MI}$, and a companion skewness diagnostic flags inputs where the approximation degrades. After characterising the axiomatic properties of $C_k$, we validate it on three tasks: (i) selective prediction for diabetic retinopathy, where critical-class $C_k$ reduces selective risk by 34.7\% over MI and 56.2\% over variance baselines; (ii) out-of-distribution detection on clinical and image benchmarks, where $\sum_k C_k$ achieves the highest AUROC and the per-class view exposes asymmetric shifts invisible to MI; and (iii) a controlled label-noise study in which $\sum_k C_k$ shows less sensitivity to injected aleatoric noise than MI under end-to-end Bayesian training, while both metrics degrade under transfer learning. Across all tasks, the quality of the posterior approximation shapes uncertainty at least as strongly as the choice of metric, suggesting that how uncertainty is propagated through the network matters as much as how it is measured.

Large language models (LLMs) are increasingly deployed as multi-step decision-making agents, where effective reward design is essential for guiding learning. Although recent work explores various forms of reward shaping and step-level credit assignment, a key signal remains largely overlooked: the intrinsic uncertainty of LLMs. Uncertainty reflects model confidence, reveals where exploration is needed, and offers valuable learning cues even in failed trajectories. We introduce SELAUR: Self Evolving LLM Agent via Uncertainty-aware Rewards, a reinforcement learning framework that incorporates uncertainty directly into the reward design. SELAUR integrates entropy-, least-confidence-, and margin-based metrics into a combined token-level uncertainty estimate, providing dense confidence-aligned supervision, and employs a failure-aware reward reshaping mechanism that injects these uncertainty signals into step- and trajectory-level rewards to improve exploration efficiency and learning stability. Experiments on two benchmarks, ALFWorld and WebShop, show that our method consistently improves success rates over strong baselines. Ablation studies further demonstrate how uncertainty signals enhance exploration and robustness.

Accurate interpretation of electrocardiogram (ECG) signals is crucial for diagnosing cardiovascular diseases. Recent multimodal approaches that integrate ECGs with accompanying clinical reports show strong potential, but they still face two main concerns from a modality perspective: (1) intra-modality: existing models process ECGs in a lead-agnostic manner, overlooking spatial-temporal dependencies across leads, which restricts their effectiveness in modeling fine-grained diagnostic patterns; (2) inter-modality: existing methods directly align ECG signals with clinical reports, introducing modality-specific biases due to the free-text nature of the reports. In light of these two issues, we propose CG-DMER, a contrastive-generative framework for disentangled multimodal ECG representation learning, powered by two key designs: (1) Spatial-temporal masked modeling is designed to better capture fine-grained temporal dynamics and inter-lead spatial dependencies by applying masking across both spatial and temporal dimensions and reconstructing the missing information. (2) A representation disentanglement and alignment strategy is designed to mitigate unnecessary noise and modality-specific biases by introducing modality-specific and modality-shared encoders, ensuring a clearer separation between modality-invariant and modality-specific representations. Experiments on three public datasets demonstrate that CG-DMER achieves state-of-the-art performance across diverse downstream tasks.

In this paper, we propose the problem of Encounter-Driven Information Diffusion (EDID). In EDID, robots are allowed to exchange information only upon meeting. Crucially, EDID assumes that the robots are not allowed to schedule their meetings. As such, the robots have no means to anticipate when, where, and who they will meet. As a step towards the design of storage and routing algorithms for EDID, in this paper we propose a model of information diffusion that captures the essential dynamics of EDID. The model is derived from first principles and is composed of two levels: a micro model, based on a generalization of the concept of `mean free path'; and a macro model, which captures the global dynamics of information diffusion. We validate the model through extensive robot simulations, in which we consider swarm size, communication range, environment size, and different random motion regimes. We conclude the paper with a discussion of the implications of this model on the algorithms that best support information diffusion according to the parameters of interest.

Neural networks are trained by optimizing multi-dimensional sets of fitting parameters on non-convex loss landscapes. Low-loss regions of the landscapes correspond to the parameter sets that perform well on the training data. A key issue in machine learning is the performance of trained neural networks on previously unseen test data. Here, we investigate neural network training by stochastic gradient descent (SGD) - a non-convex global optimization algorithm which relies only on the gradient of the objective function. We contrast SGD solutions with those obtained via a non-stochastic quasi-Newton method, which utilizes curvature information to determine step direction and Golden Section Search to choose step size. We use several computational tools to investigate neural network parameters obtained by these two optimization methods, including kernel Principal Component Analysis and a novel, general-purpose algorithm for finding low-height paths between pairs of points on loss or energy landscapes, FourierPathFinder. We find that the choice of the optimizer profoundly affects the nature of the resulting solutions. SGD solutions tend to be separated by lower barriers than quasi-Newton solutions, even if both sets of solutions are regularized by early stopping to ensure adequate performance on test data. When allowed to fit extensively on the training data, quasi-Newton solutions occupy deeper minima on the loss landscapes that are not reached by SGD. These solutions are less generalizable to the test data however. Overall, SGD explores smooth basins of attraction, while quasi-Newton optimization is capable of finding deeper, more isolated minima that are more spread out in the parameter space. Our findings help understand both the topography of the loss landscapes and the fundamental role of landscape exploration strategies in creating robust, transferrable neural network models.

Selective state space models (SSMs) have rapidly become a compelling backbone for large language models, especially for long-context workloads. Yet in deployment, their inference performance is often bounded by the memory capacity, bandwidth, and latency limits of a single GPU, making multi-GPU execution increasingly necessary. Although tensor parallelism (TP) is widely used to scale Transformer inference, applying it to selective SSM blocks is non-trivial because the SSM mixer couples large projections with a sequence-wise recurrent state update and local mixing whose efficiency depends on preserving locality and avoiding synchronization in the critical path. This paper presents a communication-efficient TP design for selective SSM inference that addresses three practical engineering challenges: enabling TTFT improvements via an SSM state cache across prefill and decode, partitioning the mixer's packed parameter tensor so that recurrent updates remain local while minimizing communication, and reducing TP aggregation overhead with quantized AllReduce. We evaluate on three representative SSM-based LLMs spanning pure-SSM and hybrid architectures - Mamba, Falcon-Mamba, and Zamba - on NVIDIA A6000 and A100 clusters. Our experiments show substantial throughput gains from tensor-parallel SSM inference, improving batch-request throughput by ~1.6-2.1x on 2 GPUs and ~2.6-4.0x on 4 GPUs for Mamba, with the largest benefits at long context lengths, and achieving a further ~10-18% throughput improvement from quantized all-reduce by lowering synchronization bandwidth overhead.

Large language model (LLM)-based agents are increasingly used to solve complex tasks involving tool use, such as web browsing, code execution, and data analysis. However, current evaluation benchmarks do not adequately assess their ability to solve real-world tasks that require synthesizing information from multiple sources and inferring insights beyond simple fact retrieval. To address this, we introduce DEEPSYNTH, a novel benchmark designed to evaluate agents on realistic, time-consuming problems that combine information gathering, synthesis, and structured reasoning to produce insights. DEEPSYNTH contains 120 tasks collected across 7 domains and data sources covering 67 countries. DEEPSYNTH is constructed using a multi-stage data collection pipeline that requires annotators to collect official data sources, create hypotheses, perform manual analysis, and design tasks with verifiable answers. When evaluated on DEEPSYNTH, 11 state-of-the-art LLMs and deep research agents achieve a maximum F1 score of 8.97 and 17.5 on the LLM-judge metric, underscoring the difficulty of the benchmark. Our analysis reveals that current agents struggle with hallucinations and reasoning over large information spaces, highlighting DEEPSYNTH as a crucial benchmark for guiding future research.

Large vision-language models (VLMs) have evolved from general-purpose applications to specialized use cases such as in the clinical domain, demonstrating potential for decision support in radiology. One promising application is assisting radiologists in decision-making by the analysis of radiology imaging data such as chest X-rays (CXR) via a visual and natural language question-answering (VQA) interface. When longitudinal imaging is available, radiologists analyze temporal changes, which are essential for accurate diagnosis and prognosis. The manual longitudinal analysis is a time-consuming process, motivating the development of a training framework that can provide prognostic capabilities. We introduce a novel training framework LUMEN, that is optimized for longitudinal CXR interpretation, leveraging multi-image and multi-task instruction fine-tuning to enhance prognostic and diagnostic performance. We conduct experiments on the publicly available MIMIC-CXR and its associated Medical-Diff-VQA datasets. We further formulate and construct a novel instruction-following dataset incorporating longitudinal studies, enabling the development of a prognostic VQA task. Our method demonstrates significant improvements over baseline models in diagnostic VQA tasks, and more importantly, shows promising potential for prognostic capabilities. These results underscore the value of well-designed, instruction-tuned VLMs in enabling more accurate and clinically meaningful radiological interpretation of longitudinal radiological imaging data.

As LLM deployments scale over more hardware, the probability of a single failure in a system increases significantly, and cloud operators must consider robust countermeasures to handle these inevitable failures. A common recovery approach is to simply restart the LLM serving instance; however, this is costly in model-as-a-service (MaaS) inference settings, where reloading model weights and recompiling computation graphs can introduce significant delays to incoming requests. We propose ReviveMoE, a method for rapid failure recovery in large-scale LLM deployments without restarting the serving instance. ReviveMoE is designed to support both the traditional LLM architecture, which collocates MoE and attention on the same hardware, and the disaggregated architectures, which separate MoE from attention. Integrated into Huawei Cloud's MaaS, ReviveMoE is built on top of Huawei's xDeepServe serving platform and the XCCL communications library.

We study the degree-weighted work required to compute $\ell_1$-regularized PageRank using the standard one-gradient-per-iteration accelerated proximal-gradient method (FISTA). For non-accelerated local methods, the best known worst-case work scales as $\widetilde{O} ((αρ)^{-1})$, where $α$ is the teleportation parameter and $ρ$ is the $\ell_1$-regularization parameter. A natural question is whether FISTA can improve the dependence on $α$ from $1/α$ to $1/\sqrtα$ while preserving the $1/ρ$ locality scaling. The challenge is that acceleration can break locality by transiently activating nodes that are zero at optimality, thereby increasing the cost of gradient evaluations. We analyze FISTA on a slightly over-regularized objective and show that, under a checkable confinement condition, all spurious activations remain inside a boundary set $\mathcal{B}$. This yields a bound consisting of an accelerated $(ρ\sqrtα)^{-1}\log(α/\varepsilon)$ term plus a boundary overhead $\sqrt{vol(\mathcal{B})}/(ρα^{3/2})$. We provide graph-structural conditions that imply such confinement. Experiments on synthetic and real graphs show the resulting speedup and slowdown regimes under the degree-weighted work model.

Qualitative insights from user experiences are critical for informing product and policy decisions, but collecting such data at scale is constrained by the time and availability of experts to conduct semi-structured interviews. Recent work has explored using large language models (LLMs) to automate interviewing, yet existing systems lack a principled mechanism for balancing systematic coverage of predefined topics with adaptive exploration, or the ability to pursue follow-ups, deep dives, and emergent themes that arise organically during conversation. In this work, we formulate adaptive semi-structured interviewing as an optimization problem over the interviewer's behavior. We define interview utility as a trade-off between coverage of a predefined interview topic guide, discovery of relevant emergent themes, and interview cost measured by length. Based on this formulation, we introduce SparkMe, a multi-agent LLM interviewer that performs deliberative planning via simulated conversation rollouts to select questions with high expected utility. We evaluate SparkMe through controlled experiments with LLM-based interviewees, showing that it achieves higher interview utility, improving topic guide coverage (+4.7% over the best baseline) and eliciting richer emergent insights while using fewer conversational turns than prior LLM interviewing approaches. We further validate SparkMe in a user study with 70 participants across 7 professions on the impact of AI on their workflows. Domain experts rate SparkMe as producing high-quality adaptive interviews that surface helpful profession-specific insights not captured by prior approaches. The code, datasets, and evaluation protocols for SparkMe are available as open-source at https://github.com/SALT-NLP/SparkMe.

Vector-quantized representations enable powerful discrete generative models but lack semantic structure in token space, limiting interpretable human control. We introduce SOM-VQ, a tokenization method that combines vector quantization with Self-Organizing Maps to learn discrete codebooks with explicit low-dimensional topology. Unlike standard VQ-VAE, SOM-VQ uses topology-aware updates that preserve neighborhood structure: nearby tokens on a learned grid correspond to semantically similar states, enabling direct geometric manipulation of the latent space. We demonstrate that SOM-VQ produces more learnable token sequences in the evaluated domains while providing an explicit navigable geometry in code space. Critically, the topological organization enables intuitive human-in-the-loop control: users can steer generation by manipulating distances in token space, achieving semantic alignment without frame-level constraints. We focus on human motion generation - a domain where kinematic structure, smooth temporal continuity, and interactive use cases (choreography, rehabilitation, HCI) make topology-aware control especially natural - demonstrating controlled divergence and convergence from reference sequences through simple grid-based sampling. SOM-VQ provides a general framework for interpretable discrete representations applicable to music, gesture, and other interactive generative domains.

This paper presents enhancements to the projection pursuit tree classifier and visual diagnostic methods for assessing their impact in high dimensions. The original algorithm uses linear combinations of variables in a tree structure where depth is constrained to be less than the number of classes -- a limitation that proves too rigid for complex classification problems. Our extensions improve performance in multi-class settings with unequal variance-covariance structures and nonlinear class separations by allowing more splits and more flexible class groupings in the projection pursuit computation. Proposing algorithmic improvements is straightforward; demonstrating their actual utility is not. We therefore develop two visual diagnostic approaches to verify that the enhancements perform as intended. Using high-dimensional visualization techniques, we examine model fits on benchmark datasets to assess whether the algorithm behaves as theorized. An interactive web application enables users to explore the behavior of both the original and enhanced classifiers under controlled scenarios. The enhancements are implemented in the R package PPtreeExt.

Large language model (LLM) agents are rapidly becoming trusted copilots in high-stakes domains like software development and healthcare. However, this deepening trust introduces a novel attack surface: Agent-Mediated Deception (AMD), where compromised agents are weaponized against their human users. While extensive research focuses on agent-centric threats, human susceptibility to deception by a compromised agent remains unexplored. We present the first large-scale empirical study with 303 participants to measure human susceptibility to AMD. This is based on HAT-Lab (Human-Agent Trust Laboratory), a high-fidelity research platform we develop, featuring nine carefully crafted scenarios spanning everyday and professional domains (e.g., healthcare, software development, human resources). Our 10 key findings reveal significant vulnerabilities and provide future defense perspectives. Specifically, only 8.6% of participants perceive AMD attacks, while domain experts show increased susceptibility in certain scenarios. We identify six cognitive failure modes in users and find that their risk awareness often fails to translate to protective behavior. The defense analysis reveals that effective warnings should interrupt workflows with low verification costs. With experiential learning based on HAT-Lab, over 90% of users who perceive risks report increased caution against AMD. This work provides empirical evidence and a platform for human-centric agent security research.

Multi-agent deep Reinforcement Learning (RL) has made significant progress in developing intelligent game-playing agents in recent years. However, the efficient training of collective robots using multi-agent RL and the transfer of learned policies to real-world applications remain open research questions. In this work, we first develop a comprehensive robotic system, including simulation, distributed learning framework, and physical robot components. We then propose and evaluate reinforcement learning techniques designed for efficient training of cooperative and competitive policies on this platform. To address the challenges of multi-agent sim-to-real transfer, we introduce Out of Distribution State Initialization (OODSI) to mitigate the impact of the sim-to-real gap. In the experiments, OODSI improves the Sim2Real performance by 20%. We demonstrate the effectiveness of our approach through experiments with a multi-robot car competitive game and a cooperative task in real-world settings.

We revisit the central online problem of ski rental in the "algorithms with predictions" framework from the point of view of distributional predictions. Ski rental was one of the first problems to be studied with predictions, where a natural prediction is simply the number of ski days. But it is both more natural and potentially more powerful to think of a prediction as a distribution p-hat over the ski days. If the true number of ski days is drawn from some true (but unknown) distribution p, then we show as our main result that there is an algorithm with expected cost at most OPT + O(min(max({eta}, 1) * sqrt(b), b log b)), where OPT is the expected cost of the optimal policy for the true distribution p, b is the cost of buying, and {eta} is the Earth Mover's (Wasserstein-1) distance between p and p-hat. Note that when {eta} < o(sqrt(b)) this gives additive loss less than b (the trivial bound), and when {eta} is arbitrarily large (corresponding to an extremely inaccurate prediction) we still do not pay more than O(b log b) additive loss. An implication of these bounds is that our algorithm has consistency O(sqrt(b)) (additive loss when the prediction error is 0) and robustness O(b log b) (additive loss when the prediction error is arbitrarily large). Moreover, we do not need to assume that we know (or have any bound on) the prediction error {eta}, in contrast with previous work in robust optimization which assumes that we know this error. We complement this upper bound with a variety of lower bounds showing that it is essentially tight: not only can the consistency/robustness tradeoff not be improved, but our particular loss function cannot be meaningfully improved.

Advanced reasoning typically requires Chain-of-Thought prompting, which is accurate but incurs prohibitive latency and substantial test-time inference costs. The standard alternative, fine-tuning smaller models, often sacrifices interpretability while introducing significant resource and operational overhead. To address these limitations, we introduce Prompt-Level Distillation (PLD). We extract explicit reasoning patterns from a Teacher model and organize them into a structured list of expressive instructions for the Student model's System Prompt. Evaluated on the StereoSet and Contract-NLI datasets using Gemma-3 4B, PLD improved Macro F1 scores from 57\% to 90.0\% and 67\% to 83\% respectively, enabling this compact model to match frontier performance with negligible latency overhead. These expressive instructions render the decision-making process transparent, allowing for full human verification of logic, making this approach ideal for regulated industries such as law, finance, and content moderation, as well as high-volume use cases and edge devices.

Curvature notions on graphs provide a theoretical description of graph topology, highlighting bottlenecks and denser connected regions. Artifacts of the message passing paradigm in Graph Neural Networks, such as oversmoothing and oversquashing, have been attributed to these regions. However, it remains unclear how the topology of a graph interacts with the learned preferences of GNNs. Through Massive Activations, which correspond to extreme edge activation values in Graph Transformers, we probe this correspondence. Our findings on synthetic graphs and molecular benchmarks reveal that MAs do not preferentially concentrate on curvature extremes, despite their theoretical link to information flow. On the Long Range Graph Benchmark, we identify a systemic \textit{curvature shift}: global attention mechanisms exacerbate topological bottlenecks, drastically increasing the prevalence of negative curvature. Our work reframes curvature as a diagnostic probe for understanding when and why graph learning fails.

Hamiltonian Monte Carlo (HMC) is a state of the art method for sampling from distributions with differentiable densities, but can converge slowly when applied to challenging multimodal problems. Running HMC with a time varying Hamiltonian, in order to interpolate from an initial tractable distribution to the target of interest, can address this problem. In conjunction with a weighting scheme to eliminate bias, this can be viewed as a special case of Sequential Monte Carlo (SMC) sampling \cite{doucet2001introduction}. However, this approach can be inefficient, since it requires slow change between the initial and final distribution. Inspired by \cite{sels2017minimizing}, where a learned \emph{counterdiabatic} term added to the Hamiltonian allows for efficient quantum state preparation, we propose \emph{Counterdiabatic Hamiltonian Monte Carlo} (CHMC), which can be viewed as an SMC sampler with a more efficient kernel. We establish its relationship to recent proposals for accelerating gradient-based sampling with learned drift terms, and demonstrate on simple benchmark problems.

Common definitions of the "standard" LOCAL model tend to be sloppy and even self-contradictory on one point: do the nodes update their state using an arbitrary function or a computable function? So far, this distinction has been safe to neglect, since problems where it matters seem contrived and quite different from e.g. typical local graph problems studied in this context. We show that this question matters even for locally checkable labeling problems (LCLs), perhaps the most widely studied family of problems in the context of the LOCAL model. Furthermore, we show that assumptions about computability are directly connected to another aspect already recognized as highly relevant: whether we have any knowledge of $n$, the size of the graph. Concretely, we show that there is an LCL problem $Π$ with the following properties: 1. $Π$ can be solved in $O(\log n)$ rounds if the LOCAL model is uncomputable. 2. $Π$ can be solved in $O(\log n)$ rounds in the computable model if we know any upper bound on $n$. 3. $Π$ requires $Ω(\sqrt{n})$ rounds in the computable model if we do not know anything about $n$. We also show that the connection between computability and knowledge of $n$ holds in general: for any LCL problem $Π$, if you have any bound on $n$, then $Π$ has the same round complexity in the computable and uncomputable models.

Scaling multimodal alignment between video and audio is challenging, particularly due to limited data and the mismatch between text descriptions and frame-level video information. In this work, we tackle the scaling challenge in multimodal-to-audio generation, examining whether models trained on short instances can generalize to longer ones during testing. To tackle this challenge, we present multimodal hierarchical networks so-called MMHNet, an enhanced extension of state-of-the-art video-to-audio models. Our approach integrates a hierarchical method and non-causal Mamba to support long-form audio generation. Our proposed method significantly improves long audio generation up to more than 5 minutes. We also prove that training short and testing long is possible in the video-to-audio generation tasks without training on the longer durations. We show in our experiments that our proposed method could achieve remarkable results on long-video to audio benchmarks, beating prior works in video-to-audio tasks. Moreover, we showcase our model capability in generating more than 5 minutes, while prior video-to-audio methods fall short in generating with long durations.

Bubeck and Sellke (2021) pose as an open problem the connection between the law of robustness and robust generalization. The law of robustness states that overparameterization is necessary for models to interpolate robustly; in particular, robust interpolation requires the learned function to be Lipschitz. Robust generalization asks whether small robust training loss implies small robust test loss. We resolve this problem by explicitly connecting the two for arbitrary data distributions. Specifically, we introduce a nontrivial notion of robust generalization error and convert it into a lower bound on the expected Rademacher complexity of the induced robust loss class. Our bounds recover the $Ω(n^{1/d})$ regime of Wu et al. (2023) and show that, up to constants, robust generalization does not change the order of the Lipschitz constant required for smooth interpolation. We conduct experiments to probe the predicted scaling with dataset size and model capacity, testing whether empirical behavior aligns more closely with the predictions of Bubeck and Sellke (2021) or Wu et al. (2023). For MNIST, we find that the lower-bound Lipschitz constant scales on the order predicted by Wu et al. (2023). Informally, to obtain low robust generalization error, the Lipschitz constant must lie in a range that we bound, and the allowable perturbation radius is linked to the Lipschitz scale.

Large Language Models (LLMs) consistently benefit from scaled Chain-of-Thought (CoT) reasoning, but also suffer from heavy computational overhead. To address this issue, efficient reasoning aims to incentivize short yet accurate thinking trajectories, typically through reward shaping with Reinforcement Learning (RL). In this paper, we systematically investigate the mechanics of efficient reasoning for LLMs. For comprehensive evaluation, we advocate for more fine-grained metrics, including length distribution conditioned on correctness and performance across a wide spectrum of token budgets ranging from 2k to 32k. First, we reveal that the training process follows a two-stage paradigm: length adaptation and reasoning refinement. After that, we conduct extensive experiments (about 0.2 million GPU hours) in a unified protocol, deconstructing training prompts and rollouts, reward shaping, and optimization strategies. In particular, a key finding is to train on relatively easier prompts, ensuring the density of positive reward signals and thus avoiding the length collapse. Meanwhile, the learned length bias can be generalized across domains. We distill all findings into valuable insights and practical guidelines, and further validate them across the Qwen3 series, ranging from 0.6B to 30B, demonstrating the robustness and generalization.

The forest-of-refinement-trees approach allows for dynamic adaptive mesh refinement (AMR) at negligible cost. While originally developed for quadrilateral and hexahedral elements, previous work established the theory and algorithms for unstructured meshes of simplicial and prismatic elements. To harness the full potential of tree-based AMR for three-dimensional mixed-element meshes, this paper introduces the pyramid as a new functional element type; its primary purpose is to connect tetrahedral and hexahedral elements without hanging edges. We present a well-defined space-filling curve (SFC) for the pyramid and detail how the unique challenges on the element and forest level associated with the pyramidal refinement are resolved. We propose the necessary functional design and generalize the fundamental global parallel algorithms for refinement, coarsening, partitioning, and face ghost exchange to fully support this new element. Our demonstrations confirm the efficiency and scalability of this complete, hybrid-element dynamic AMR framework.

We present Group Orthogonalized Policy Optimization (GOPO), a new alignment algorithm for large language models derived from the geometry of Hilbert function spaces. Instead of optimizing on the probability simplex and inheriting the exponential curvature of Kullback-Leibler divergence, GOPO lifts alignment into the Hilbert space L2(pi_k) of square-integrable functions with respect to the reference policy. Within this space, the simplex constraint reduces to a linear orthogonality condition <v, 1> = 0, defining a codimension-one subspace H0. Minimizing distance to an unconstrained target u_star yields the work-dissipation functional J(v) = <g, v> - (mu / 2) ||v||^2, whose maximizer follows directly from the Hilbert projection theorem. Enforcing the boundary v >= -1 produces a bounded Hilbert projection that induces exact sparsity, assigning zero probability to catastrophically poor actions through a closed-form threshold. To connect this functional theory with practice, GOPO projects from infinite-dimensional L2(pi_k) to a finite empirical subspace induced by group sampling. Because group-normalized advantages sum to zero, the Lagrange multiplier enforcing probability conservation vanishes exactly, reducing the constrained projection to an unconstrained empirical loss. The resulting objective has constant Hessian curvature mu I, non-saturating linear gradients, and an intrinsic dead-zone mechanism without heuristic clipping. Experiments on mathematical reasoning benchmarks show that GOPO achieves competitive generalization while maintaining stable gradient dynamics and entropy preservation in regimes where clipping-based methods plateau.

Soft set theory provides a direct framework for parameterized decision modeling by assigning to each attribute (parameter) a subset of a given universe, thereby representing uncertainty in a structured way [1, 2]. Over the past decades, the theory has expanded into numerous variants-including hypersoft sets, superhypersoft sets, TreeSoft sets, bipolar soft sets, and dynamic soft sets-and has been connected to diverse areas such as topology and matroid theory. In this book, we present a survey-style overview of soft sets and their major extensions, highlighting core definitions, representative constructions, and key directions of current development.

Cybersecurity threats are becoming increasingly sophisticated, making traditional defense mechanisms and manual red teaming approaches insufficient for modern organizations. While red teaming has long been recognized as an effective method to identify vulnerabilities by simulating real-world attacks, its manual execution is resource-intensive, time-consuming, and lacks scalability for frequent assessments. These limitations have driven the evolution toward auto-mated red teaming, which leverages artificial intelligence and automation to deliver efficient and adaptive security evaluations. This systematic review consolidates existing research on automated red teaming, examining its methodologies, tools, benefits, and limitations. The paper also highlights current trends, challenges, and research gaps, offering insights into future directions for improving automated red teaming as a critical component of proactive cybersecurity strategies. By synthesizing findings from diverse studies, this review aims to provide a comprehensive understanding of how automation enhances red teaming and strengthens organizational resilience against evolving cyber threats.

We introduce \ToolMATH, a math-grounded benchmark that evaluates tool-augmented language models in realistic multi-tool environments where the output depends on calling schema-specified tools and sustaining multi-step execution. It turns math problems into a controlled, correctness-checkable benchmark with tool sets, enabling systematic evaluation of model reliability under (1) large, overlapping tool catalogs and (2) the absence of the intended capability. \ToolMATH provides actionable diagnostic evidence of failure modes in tool-augmented agents, helping identify the control mechanisms required for robustness. \ToolMATH roughly contains 8k questions and 12k tools; we provide an additional hard-set \ToolMATHHard with questions and tools. Our evaluation reveals that the key failure factor is due to the inability to reason, leading to the accumulation of intermediate results' errors and constrain later decisions. Tool-list redundancy do not simply add noise, but amplify small early deviations into irreversible execution drift. The benchmark highlights that when the intended capability is missing, distractor tools can sometimes serve as partial substitutes in solution paths, yet they can also mislead models into ungrounded tool trajectories. Finally, comparisons between tool-use protocols emphasize that improvements come less from local action selection and more from long-range plan coherence and disciplined use of observations.

Reliable prediction of hydraulic performance is challenging for Piano Key Weir (PKW) design because discharge capacity depends on three-dimensional geometry and operating conditions. Surrogate models can accelerate hydraulic-structure design, but progress is limited by scarce large, well-documented datasets that jointly capture geometric variation, operating conditions, and functional performance. This study presents WeirNet, a large 3D CFD benchmark dataset for geometric surrogate modeling of PKWs. WeirNet contains 3,794 parametric, feasibility-constrained rectangular and trapezoidal PKW geometries, each scheduled at 19 discharge conditions using a consistent free-surface OpenFOAM workflow, resulting in 71,387 completed simulations that form the benchmark and with complete discharge coefficient labels. The dataset is released as multiple modalities compact parametric descriptors, watertight surface meshes and high-resolution point clouds together with standardized tasks and in-distribution and out-of-distribution splits. Representative surrogate families are benchmarked for discharge coefficient prediction. Tree-based regressors on parametric descriptors achieve the best overall accuracy, while point- and mesh-based models remain competitive and offer parameterization-agnostic inference. All surrogates evaluate in milliseconds per sample, providing orders-of-magnitude speedups over CFD runtimes. Out-of-distribution results identify geometry shift as the dominant failure mode compared to unseen discharge values, and data-efficiency experiments show diminishing returns beyond roughly 60% of the training data. By publicly releasing the dataset together with simulation setups and evaluation pipelines, WeirNet establishes a reproducible framework for data-driven hydraulic modeling and enables faster exploration of PKW designs during the early stages of hydraulic planning.

Specifications are vital for ensuring program correctness, yet writing them manually remains challenging and time-intensive. Recent large language model (LLM)-based methods have shown successes in generating specifications such as postconditions, but existing single-pass prompting often yields inaccurate results. In this paper, we present SpecMind, a novel framework for postcondition generation that treats LLMs as interactive and exploratory reasoners rather than one-shot generators. SpecMind employs feedback-driven multi-turn prompting approaches, enabling the model to iteratively refine candidate postconditions by incorporating implicit and explicit correctness feedback, while autonomously deciding when to stop. This process fosters deeper code comprehension and improves alignment with true program behavior via exploratory attempts. Our empirical evaluation shows that SpecMind significantly outperforms state-of-the-art approaches in both accuracy and completeness of generated postconditions.

Task-based runtime systems provide flexible load balancing and portability for parallel scientific applications, but their strong scaling is highly sensitive to task granularity. As parallelism increases, scheduling overhead may transition from negligible to dominant, leading to rapid drops in performance for some algorithms, while remaining negligible for others. Although such effects are widely observed empirically, there is a general lack of understanding how algorithmic structure impacts whether dynamic scheduling is always beneficial. In this work, we introduce a granularity characterization framework that directly links scheduling overhead growth to task-graph dependency topology. We show that dependency structure, rather than problem size alone, governs how overhead scales with parallelism. Based on this observation, we characterize execution behavior using a simple granularity measure that indicates when scheduling overhead can be amortized by parallel computation and when scheduling overhead dominates performance. Through experimental evaluation on representative parallel workloads with diverse dependency patterns, we demonstrate that the proposed characterization explains both gradual and abrupt strong-scaling breakdowns observed in practice. We further show that overhead models derived from dependency topology accurately predict strong-scaling limits and enable a practical runtime decision rule for selecting dynamic or static execution without requiring exhaustive strong-scaling studies or extensive offline tuning.

With increasing integration of Large Language Models (LLMs) into areas of high-stakes human decision-making, it is important to understand the risks they introduce as advisors. To be useful advisors, LLMs must sift through large amounts of content, written with both benevolent and malicious intent, and then use this information to convince a user to take a specific action. This involves two social capacities: vigilance (the ability to determine which information to use, and which to discard) and persuasion (synthesizing the available evidence to make a convincing argument). While existing work has investigated these capacities in isolation, there has been little prior investigation of how these capacities may be linked. Here, we use a simple multi-turn puzzle-solving game, Sokoban, to study LLMs' abilities to persuade and be rationally vigilant towards other LLM agents. We find that puzzle-solving performance, persuasive capability, and vigilance are dissociable capacities in LLMs. Performing well on the game does not automatically mean a model can detect when it is being misled, even if the possibility of deception is explicitly mentioned. % as part of the prompt. However, LLMs do consistently modulate their token use, using fewer tokens to reason when advice is benevolent and more when it is malicious, even if they are still persuaded to take actions leading them to failure. To our knowledge, our work presents the first investigation of the relationship between persuasion, vigilance, and task performance in LLMs, and suggests that monitoring all three independently will be critical for future work in AI safety.

Multimodal Role-Playing Agents (MRPAs) are attracting increasing attention due to their ability to deliver more immersive multimodal emotional interactions. However, existing studies still rely on pure textual benchmarks to evaluate the text responses of MRPAs, while delegating the assessment of their multimodal expressions solely to modality-synthesis metrics. This evaluation paradigm, on the one hand, entangles semantic assessment with modality generation, leading to ambiguous error attribution, and on the other hand remains constrained by the heavy reliance on human judgment. To this end, we propose MERRY, a semantically decoupled evaluation framework for assessing Multimodal Emotional and Role consistencies of Role-playing agents. This framework introduce five refined metrics for EC and three for RC. Notably, we transform the traditional subjective scoring approach into a novel bidirectional-evidence-finding task, significantly improving the human agreement of LLM-as-Judge evaluations. Based on MERRY, we conduct extensive evaluations. Our empirical results primarily reveal that: (1) Training on synthetic datasets tends to reduce emotional consistency, whereas training on real-world datasets improves it; (2) Existing models suffer from emotional templatization and simplification, exhibiting positive-bias and performance bottleneck in fine-grained negative emotions; (3) Simple prompting method strengthens the weak models but constrains the strong ones, while simple fine-tuning method suffers from poor role generalization. Codes and dataset are available.

We approach the Max-3-Cut problem through the lens of maximizing complex-valued quadratic forms and demonstrate that low-rank structure in the objective matrix can be exploited, leading to alternative algorithms to classical semidefinite programming (SDP) relaxations and heuristic techniques. We propose an algorithm for maximizing these quadratic forms over a domain of size $K$ that enumerates and evaluates a set of $O\left(n^{2r-1}\right)$ candidate solutions, where $n$ is the dimension of the matrix and $r$ represents the rank of an approximation of the objective. We prove that this candidate set is guaranteed to include the exact maximizer when $K=3$ (corresponding to Max-3-Cut) and the objective is low-rank, and provide approximation guarantees when the objective is a perturbation of a low-rank matrix. This construction results in a family of novel, inherently parallelizable and theoretically-motivated algorithms for Max-3-Cut. Extensive experimental results demonstrate that our approach achieves performance comparable to existing algorithms across a wide range of graphs, while being highly scalable.

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy and over one million video clips, approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first large-scale scaling studies of video reasoning and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark toolkit, and models are publicly available at https://video-reason.com/ .

LLM agents are evolving rapidly, powered by code execution, tools, and the recently introduced agent skills feature. Skills allow users to extend LLM applications with specialized third-party code, knowledge, and instructions. Although this can extend agent capabilities to new domains, it creates an increasingly complex agent supply chain, offering new surfaces for prompt injection attacks. We identify skill-based prompt injection as a significant threat and introduce SkillInject, a benchmark evaluating the susceptibility of widely-used LLM agents to injections through skill files. SkillInject contains 202 injection-task pairs with attacks ranging from obviously malicious injections to subtle, context-dependent attacks hidden in otherwise legitimate instructions. We evaluate frontier LLMs on SkillInject, measuring both security in terms of harmful instruction avoidance and utility in terms of legitimate instruction compliance. Our results show that today's agents are highly vulnerable with up to 80% attack success rate with frontier models, often executing extremely harmful instructions including data exfiltration, destructive action, and ransomware-like behavior. They furthermore suggest that this problem will not be solved through model scaling or simple input filtering, but that robust agent security will require context-aware authorization frameworks. Our benchmark is available at https://www.skill-inject.com/.

The goal of the BabyLM is to stimulate new research connections between cognitive modeling and language model pretraining. We invite contributions in this vein to the BabyLM Workshop, which will also include the 4th iteration of the BabyLM Challenge. As in previous years, the challenge features two ``standard'' tracks (Strict and Strict-Small), in which participants must train language models on under 100M or 10M words of data, respectively. This year, we move beyond our previous English-only pretraining datasets with a new Multilingual track, focusing on English, Dutch, and Chinese. For the workshop, we call for papers related to the overall theme of BabyLM, which includes training efficiency, small-scale training datasets, cognitive modeling, model evaluation, and architecture innovation.

Deploying Retrieval-Augmented Generation (RAG) on edge devices is in high demand, but is hindered by the latency of massive data movement and computation on traditional architectures. Compute-in-Memory (CiM) architectures address this bottleneck by performing vector search directly within their crossbar structure. However, CiM's adoption for RAG is limited by a fundamental ``representation gap,'' as high-precision, high-dimension embeddings are incompatible with CiM's low-precision, low-dimension array constraints. This gap is compounded by the diversity of CiM implementations (e.g., SRAM, ReRAM, FeFET), each with unique designs (e.g., 2-bit cells, 512x512 arrays). Consequently, RAG data must be naively reshaped to fit each target implementation. Current data shaping methods handle dimension and precision disjointly, which degrades data fidelity. This not only negates the advantages of CiM for RAG but also confuses hardware designers, making it unclear if a failure is due to the circuit design or the degraded input data. As a result, CiM adoption remains limited. In this paper, we introduce CQ-CiM, a unified, hardware-aware data shaping framework that jointly learns Compression and Quantization to produce CiM-compatible low-bit embeddings for diverse CiM designs. To the best of our knowledge, this is the first work to shape data for comprehensive CiM usage on RAG.

We present SPL (Structured Prompt Language), a declarative SQL-inspired language that treats large language models as generative knowledge bases and their context windows as constrained resources. SPL provides explicit WITH BUDGET/LIMIT token management, an automatic query optimizer, EXPLAIN transparency analogous to SQL's EXPLAIN ANALYZE, and native integration of retrieval-augmented generation (RAG) and persistent memory in a single declarative framework. SPL-flow extends SPL into resilient agentic pipelines with a three-tier provider fallback strategy (Ollama -> OpenRouter -> self-healing retry) fully transparent to the .spl script. Five extensions demonstrate the paradigm's breadth: (1) Text2SPL (multilingual NL->SPL translation); (2) Mixture-of-Models (MoM) routing that dispatches each PROMPT to a domain-specialist model at runtime; (3) Logical Chunking, an intelligent strategy for documents exceeding a single context window--expressed naturally through SPL's existing CTE syntax with no new constructs, decomposing a large query into a Map-Reduce pipeline that reduces attention cost from O(N^2) to O(N^2/k) and runs identically on cloud (parallel) or local hardware (sequential); (4) SPL-flow, a declarative agentic orchestration layer with resilient three-tier provider fallback; and (5) BENCHMARK for parallel multi-model comparison with automatic winner persistence. We provide a formal EBNF grammar, two pip-installable Python packages (spl-llm, spl-flow), and comparison against Prompty, DSPy, and LMQL. SPL reduces prompt boilerplate by 65% on average, surfaces a 68x cost spread across model tiers as a pre-execution signal, and runs the identical .spl script at $0.002 on OpenRouter or at zero marginal cost on a local Ollama instance--without modification.

Robots are increasingly operating in open-world environments where safe behavior depends on context: the same hallway may require different navigation strategies when crowded versus empty, or during an emergency versus normal operations. Traditional safety approaches enforce fixed constraints in user-specified contexts, limiting their ability to handle the open-ended contextual variability of real-world deployment. We address this gap via CORE, a safety framework that enables online contextual reasoning, grounding, and enforcement without prior knowledge of the environment (e.g., maps or safety specifications). CORE uses a vision-language model (VLM) to continuously reason about context-dependent safety rules directly from visual observations, grounds these rules in the physical environment, and enforces the resulting spatially-defined safe sets via control barrier functions. We provide probabilistic safety guarantees for CORE that account for perceptual uncertainty, and we demonstrate through simulation and real-world experiments that CORE enforces contextually appropriate behavior in unseen environments, significantly outperforming prior semantic safety methods that lack online contextual reasoning. Ablation studies validate our theoretical guarantees and underscore the importance of both VLM-based reasoning and spatial grounding for enforcing contextual safety in novel settings. We provide additional resources at https://zacravichandran.github.io/CORE.

Every ODRL 2.2 constraint compares a single scalar value: (leftOperand, operator, rightOperand). Five of ODRL's left operands, however, denote multi-dimensional quantities--image dimensions, canvas positions, geographic coordinates--whose specification text explicitly references multiple axes. For these operands, a single scalar constraint admits one interpretation per axis, making policy evaluation non-deterministic. We classify ODRL's left operands by value-domain structure (scalar, dimensional, concept-valued), grounded in the ODRL 2.2 specification text, and show that dimensional ambiguity is intrinsic to the constraint syntax. We present an axis-decomposition framework that refines each dimensional operand into axis-specific scalar operands and prove four properties: deterministic interpretation, AABB completeness, projection soundness, and conservative extension. Conflict detection operates in two layers: per-axis verdicts are always decidable; box-level verdicts compose through Strong Kleene conjunction into a three-valued logic (Conflict, Compatible, Unknown). For ODRL's disjunctive (odrl:or) and exclusive-or (odrl:xone) logical constraints, where per-axis decomposition does not apply, the framework encodes coupled multi-axis conjectures directly. We instantiate the framework as the ODRL Spatial Axis Profile--15 axis-specific left operands for the five affected base terms--and evaluate it on 117 benchmark problems spanning nine categories across both TPTP FOF (Vampire) and SMT-LIB (Z3) encodings, achieving full concordance between provers. Benchmark scenarios are inspired by constraints arising in cultural heritage dataspaces such as Datenraum Kultur. All meta-theorems are mechanically verified in Isabelle/HOL.

Recently, TabPFN has gained attention as a foundation model for tabular data. However, it struggles to integrate heterogeneous modalities such as images and text, which are common in domains like healthcare and marketing, thereby limiting its applicability. To address this, we present the Multi-Modal Prior-data Fitted Network (MMPFN), which extends TabPFN to handle tabular and non-tabular modalities in a unified manner. MMPFN comprises per-modality encoders, modality projectors, and pre-trained foundation models. The modality projectors serve as the critical bridge, transforming non-tabular embeddings into tabular-compatible tokens for unified processing. To this end, we introduce a multi-head gated MLP and a cross-attention pooler that extract richer context from non-tabular inputs while mitigates attention imbalance issue in multimodal learning. Extensive experiments on medical and general-purpose multimodal datasets demonstrate that MMPFN consistently outperforms competitive state-of-the-art methods and effectively exploits non-tabular modalities alongside tabular features. These results highlight the promise of extending prior-data fitted networks to the multimodal setting, offering a scalable and effective framework for heterogeneous data learning. The source code is available at https://github.com/too-z/MultiModalPFN.

We establish a general equilibrium theory for systems of large language model (LLM) agents operating under centralized orchestration. The framework is a production economy in the sense of Arrow-Debreu (1954), extended to infinite-dimensional commodity spaces following Bewley (1972). Each LLM agent is modeled as a firm whose production set Y a $\subset$ H = L 2 ([0, T ], R R ) represents the feasible metric trajectories determined by its frozen model weights. The orchestrator is the consumer, choosing a routing policy over the agent DAG to maximize system welfare subject to a budget constraint evaluated at functional prices p $\in$ H A . These prices-elements of the Hilbert dual of the commodity space-assign a shadow value to each metric of each agent at each instant. We prove, via Brouwer's theorem applied to a finitedimensional approximation V K $\subset$ H, that every such economy admits at least one general equilibrium (p * , y * , $π$ * ). A functional Walras' law holds as a theorem: the value of functional excess demand is zero for all prices, as a consequence of the consumer's budget constraint-not by construction. We further establish Pareto optimality (First Welfare Theorem), decentralizability of Pareto optima (Second Welfare Theorem), and uniqueness with geometric convergence under a contraction condition (Banach). The orchestration dynamics constitute a Walrasian t{â}tonnement that converges globally under the contraction condition, unlike classical t{â}tonnement (Scarf, 1960). The framework admits a DSGE interpretation with SLO parameters as policy rates.

Remote monitoring of heart failure (HF) via speech signals provides a non-invasive and cost-effective solution for long-term patient management. However, substantial inter-individual heterogeneity in vocal characteristics often limits the accuracy of traditional cross-sectional classification models. To address this, we propose a Longitudinal Intra-Patient Tracking (LIPT) scheme designed to capture the trajectory of relative symptomatic changes within individuals. Central to this framework is a Personalised Sequential Encoder (PSE), which transforms longitudinal speech recordings into context-aware latent representations. By incorporating historical data at each timestamp, the PSE facilitates a holistic assessment of the clinical trajectory rather than modelling discrete visits independently. Experimental results from a cohort of 225 patients demonstrate that the LIPT paradigm significantly outperforms the classic cross-sectional approaches, achieving a recognition accuracy of 99.7% for clinical status transitions. The model's high sensitivity was further corroborated by additional follow-up data, confirming its efficacy in predicting HF deterioration and its potential to secure patient safety in remote, home-based settings. Furthermore, this work addresses the gap in existing literature by providing a comprehensive analysis of different speech task designs and acoustic features. Taken together, the superior performance of the LIPT framework and PSE architecture validates their readiness for integration into long-term telemonitoring systems, offering a scalable solution for remote heart failure management.

Supply chain optimization models frequently become infeasible because of modeling errors. Diagnosis and repair require scarce OR expertise: analysts must interpret solver diagnostics, trace root causes across echelons, and fix formulations without sacrificing operational soundness. Whether AI agents can perform this task remains untested. We decompose this task into two phases: a domain-agnostic feasibility phase that iteratively repairs any LP using IIS-guided diagnosis, and a domain-specific validation phase that enforces five rationality checks grounded in inventory theory. We test 22 API models from seven families on 976 multi-echelon supply chain problems and train two 8B-parameter models with self-taught reasoning and solver-verified rewards. The trained models reach 81.7% Rational Recovery Rate (RRR) -- the fraction of problems resolved to both feasibility and operational rationality -- versus 42.2% for the best API model and 21.3% on average. The gap concentrates in Phase 1 repair, where API models average 27.6% recovery rate versus 97.2% for trained models. Two gaps separate current AI from reliable model repair: solver interaction, as API models restore only 27.6% of infeasible formulations; and operational rationale, as roughly one in four feasible repairs violate supply chain theory. Each gap requires a different intervention -- targeted training closes the solver interaction gap, while explicit specification as solver-verifiable checks closes the rationality gap. For organizations adopting AI in operational planning, formalizing what 'rational' means in their context is the higher-return investment.

Time series causal discovery is essential for understanding dynamic systems, yet many existing methods remain sensitive to noise, non-stationarity, and sampling variability. We propose the Validated Consensus-Driven Framework (VCDF), a simple and method-agnostic layer that improves robustness by evaluating the stability of causal relations across blocked temporal subsets. VCDF requires no modification to base algorithms and can be applied to methods such as VAR-LiNGAM and PCMCI. Experiments on synthetic datasets show that VCDF improves VAR-LiNGAM by approximately 0.08-0.12 in both window and summary F1 scores across diverse data characteristics, with gains most pronounced for moderate-to-long sequences. The framework also benefits from longer sequences, yielding up to 0.18 absolute improvement on time series of length 1000 and above. Evaluations on simulated fMRI data and IT-monitoring scenarios further demonstrate enhanced stability and structural accuracy under realistic noise conditions. VCDF provides an effective reliability layer for time series causal discovery without altering underlying modeling assumptions.

As quantum processors scale beyond 100 qubits, distinguishing software bugs from stochastic hardware noise becomes a critical diagnostic challenge. We present a neuro-fuzzy framework that addresses this attribution problem by combining Adaptive Neuro-Fuzzy Inference Systems (ANFIS) with physics-grounded feature engineering. We introduce the Bhattacharyya Veto, a hard physical constraint grounded in the Data Processing Inequality that prevents the classifier from attributing topologically impossible output distributions to noise. Validated on IBM's 156-qubit Heron r2 processor (ibm_fez) across 105 circuits spanning 17 algorithm families, the framework achieves 89.5% effective accuracy (+/- 5.9% CI). The system implements a safe failure mode, flagging 14.3% of ambiguous cases for manual review rather than forcing low-confidence predictions. We resolve key ambiguities -- such as distinguishing correct Grover amplification from bug-induced collapse -- and identify fundamental limits of single-basis diagnostics, including a Z-basis blind spot where phase-flip errors remain statistically invisible. This work establishes a robust, interpretable diagnostic layer that prevents error mitigation techniques from being applied to logically flawed circuits.

Security monitoring systems typically treat anomaly detection as identifying statistical deviations from observed data distributions. In cryptographic traffic analysis, however, violations are defined not by rarity but by explicit policy constraints, including key reuse prohibition, downgrade prevention, and bounded key lifetimes. This fundamental mismatch limits the interpretability and adaptability of conventional anomaly detection methods. We introduce INTACT (INTent-Aware Cryptographic Traffic), a policy-conditioned framework that reformulates violation detection as conditional constraint learning. Instead of learning a static decision boundary over behavioral features, INTACT models the probability of violation conditioned on both observed behavior and declared security intent. The architecture factorizes representation learning into behavioral and intent encoders whose fused embeddings produce a violation score, yielding a policy-parameterized family of decision boundaries. We evaluate the framework on a real-world network flow dataset and a 210,000-trace synthetic multi-intent cryptographic dataset. INTACT matches or exceeds strong unsupervised and supervised baselines, achieving near-perfect discrimination (AUROC up to 1.0000) in the real dataset and consistent superiority in detecting relational and composite violations in the synthetic setting. These results demonstrate that explicit intent conditioning improves discrimination, interpretability, and robustness in cryptographic monitoring.

Legacy JavaScript systems lack type safety, making maintenance risky. While TypeScript can help, manually adding types is expensive. Previous automated typing research focuses on type inference but rarely addresses type checking setup, definition generation, bug identification, or behavioral correctness at repository scale. We present AgenticTyper, a Large Language Model (LLM)-based agentic system that addresses these gaps through iterative error correction and behavior preservation via transpilation comparison. Evaluation on two proprietary repositories (81K LOC) shows that AgenticTyper resolves all 633 initial type errors in 20 minutes, reducing manual effort from one working day.

Hyperbolic spaces provide a natural geometry for representing hierarchical and tree-structured data due to their exponential volume growth. To leverage these benefits, neural networks require intrinsic and efficient components that operate directly in hyperbolic space. In this work, we lift two core components of neural networks, Multinomial Logistic Regression (MLR) and Fully Connected (FC) layers, into hyperbolic space via Busemann functions, resulting in Busemann MLR (BMLR) and Busemann FC (BFC) layers with a unified mathematical interpretation. BMLR provides compact parameters, a point-to-horosphere distance interpretation, batch-efficient computation, and a Euclidean limit, while BFC generalizes FC and activation layers with comparable complexity. Experiments on image classification, genome sequence learning, node classification, and link prediction demonstrate improvements in effectiveness and efficiency over prior hyperbolic layers. The code is available at https://github.com/GitZH-Chen/HBNN.

We introduce BURMESE-SAN, the first holistic benchmark that systematically evaluates large language models (LLMs) for Burmese across three core NLP competencies: understanding (NLU), reasoning (NLR), and generation (NLG). BURMESE-SAN consolidates seven subtasks spanning these competencies, including Question Answering, Sentiment Analysis, Toxicity Detection, Causal Reasoning, Natural Language Inference, Abstractive Summarization, and Machine Translation, several of which were previously unavailable for Burmese. The benchmark is constructed through a rigorous native-speaker-driven process to ensure linguistic naturalness, fluency, and cultural authenticity while minimizing translation-induced artifacts. We conduct a large-scale evaluation of both open-weight and commercial LLMs to examine challenges in Burmese modeling arising from limited pretraining coverage, rich morphology, and syntactic variation. Our results show that Burmese performance depends more on architectural design, language representation, and instruction tuning than on model scale alone. In particular, Southeast Asia regional fine-tuning and newer model generations yield substantial gains. Finally, we release BURMESE-SAN as a public leaderboard to support systematic evaluation and sustained progress in Burmese and other low-resource languages. https://leaderboard.sea-lion.ai/detailed/MY

Prefill/decode disaggregation is increasingly adopted in LLM serving to improve the latency-throughput tradeoff and meet strict TTFT and TPOT SLOs. However, LLM inference remains energy-hungry: autoscaling alone is too coarse-grained to track fast workload fluctuations, and applying fine-grained DVFS under disaggregation is complicated by phase-asymmetric dynamics and coupling between provisioning and frequency control. We present BiScale, a two-tier energy optimization framework for disaggregated LLM serving. BiScale jointly optimizes placement and DVFS across prefill and decode using predictive latency and power models. At coarse timescales, BiScale computes phase-aware placement and baseline frequencies that minimize energy while satisfying SLO constraints. At fine timescales, BiScale dynamically adapts GPU frequency per iteration using stage-specific control: model predictive control (MPC) for prefill to account for queue evolution and future TTFT impact, and lightweight slack-aware adaptation for decode to exploit its smoother, memory-bound dynamics. This hierarchical design enables coordinated control across timescales while preserving strict serving SLOs. Evaluation on a 16x H100 cluster serving Llama 3.3 70B with production-style traces shows that BiScale meets TTFT/TPOT SLOs while reducing energy by up to 39% in prefill and 48% in decode relative to DistServe.

Searching by chemical and structural analogy is one of the most commonly used and successful approaches to materials discovery. However, formulating this task for algorithmic implementation raises the question of how we define similar materials. Methods have been proposed for searching materials space using vectors based on chemical composition and functional fragments in the material. Descriptors for structural similarity have also been proposed. However, the question of how to incorporate and balance structural and compositional similarity measures in a single metric remains open. Here, we adapt methods developed for calculating distances between undirected graphs and apply them to crystalline materials similarity. The Fused Gromov-Wasserstein (FGW) metric uses optimal transport theory to map between two graphs considering a balance of the graph structure and the information present at the nodes of the graph (atoms in crystals). We apply the method to exploring new photovoltaic materials. We demonstrate that FGW is competitive with embeddings from an equivariant graph neural network, trained on $> 10^6$ materials, despite minimal training. We then apply FGW to a discovery campaign to identify materials from the Materials Project database that have not previously been explored as photovoltaics, but have similarities to known high-efficiency materials. After validating predictions with hybrid density functional theory, we identify seven previously unexplored high-efficiency photovoltaic absorber candidates, including Cs$_5$Sb$_8$, which is found to have a predicted SLME of $> 30\%$ and to be thermodynamically stable. The FGW approach demonstrates the power of strong inductive biases for developing metrics for materials exploration with minimal training data.

We consider a 2D XY model subjected to time-correlated noise, a model of direct relevance to active crystals, which were shown recently to be able to support very large deformations without melting in the presence of persistent fluctuations. We find that our persistent XY model can remain quasi-ordered in spite of correlations decaying much faster than allowed in equilibrium. We then investigate theoretically and numerically the order-disorder transition and conclude that it remains of the Berezinskii-Kosterlitz-Thouless type, but with scaling exponents that vary with the persistence time of the noise.

In the era of big data and the Internet of Things, quantum-level control of conductance states offers a promising route toward high-density data storage and brain-inspired neuromorphic computing. Although quantum conductance (QC) phenomena have been demonstrated in various metal oxide memristors, achieving reliable and precise control over quantized states remains in its infancy. Here, we demonstrate bidirectional quantum conductance states in multifunctional BiFeO3 (BFO) perovskite memristors integrated with multilayer-graphene contacts, enabling higher-order tunability and revealing the potential of perovskite-2D heterostructures for quantum-engineered memory and computing devices. XPS analysis provides detailed insights into oxygen vacancy dynamics in BFO, whereas first-principles density functional theory calculations clearly reveal a strong localized electric field at the graphene-BFO interface. Our devices exhibit current-controlled higher-order QC transitions facilitated by quantum point contact formation, giving rise to quantized conductance states during both SET and RESET processes. Time-lag correlation maps quantify the stochastic evolution of QC states under dynamic voltage-pulse tuning schemes. Notably, the quantized conductance states effectively emulate synaptic potentiation and depression, enabling precise weight modulation for high-accuracy image and digit recognition in convolutional neural networks. These findings establish perovskite-2D heterostructures as promising candidates for QC-driven resistive switching and demonstrate their potential for developing controllable quantum memristors.

We predict a novel class of magnetotransport oscillations in ballistic graphene specific for a ring-shape geometry. Using the Büttiker-Landauer formalism, we analytically obtain the local Hall and Nernst coefficients in the weak-field ballistic regime. These coefficients exhibit pronounced oscillations as functions of both the magnetic field and the angular positions of the measurement probes. The oscillations originate from the discrete set of skipping orbits that geometrically connect the contacts, with resonances occurring when the angular separation between contacts times the radius of the disk equals an integer number of cyclotron diameters. Unlike conventional quantum oscillations in conductivity, this effect is robust at room temperature and can dominate local thermoelectric signals. This geometric control of ballistic flow provides a platform for studying electron hydrodynamics and engineering phase-coherent devices, with potential applications in sensitive terahertz detectors and thermal management systems.

Biological tissues are active materials whose non-equilibrium dynamics emerge from distinct cellular force-generating mechanisms. Using a two-dimensional active foam model, we compare the effects of traction forces and junctional tension fluctuations on confluent tissue dynamics. While these two modes of activity produce qualitatively different cell shapes, rearrangement statistics, and spatiotemporal correlations in fluid states, we find that the long-time cellular motion universally converges to persistent Brownian dynamics. This universal feature contrasts with the non-universal correlations between cell geometry, rearrangement rate, and fluidity, which depend sensitively on the underlying modes of active force. Our results demonstrate that persistent Brownian motion provides a minimal framework for describing tissue dynamics, while distinct active forces leave identifiable structural and dynamical signatures, thereby enabling inference of the dominant active force in fluid state tissues.

We test the Transient Time Correlation Function (TTCF) method to compute nonequilibrium transport coefficients, highlighting its conceptual and practical difference from the standard time-average approach. While time averages extract transport properties from long stationary trajectories and discard transient dynamics, TTCF adopts the complementary strategy: it exploits the information contained in short-time transients following the onset of an external perturbation, while discarding the long-time evolution once stationarity is reached. We revisit the theoretical framework of TTCF and assess its numerical performance through representative case studies, the Lorentz gas and a many-body system, namely a chain of oscillators with anharmonic pinning potential. By direct comparison with time averages, we show that for the Lorentz gas TTCF yields consistent transport coefficients in both linear and nonlinear regimes at a reduced computational cost. Moreover, the TTCF displays superior precision in the linear-response regime, and remains reliable in non-ergodic situations, revealing the presence of regions of phase space corresponding to different behaviors, as well as the possibility of phase transitions. For the anharmonic chain, we show that TTCF is a scalable and efficient alternative for the numerical study of nonequilibrium transport.

We study stress relaxation in several types of entangled monodisperse linear polymer melts by comparing the shear stress relaxation modulus, $G(t)$, with the end-to-end vector autocorrelation function, $P(t)$. The study includes three Kremer-Grest bead-spring models with varying chain stiffness, as well as a chemistry-specific coarse-grained model of \emph{cis}-1,4-polybutadiene. For each model, multiple chain lengths were simulated, spanning a range of $N/N_e = 5$-$50$ entanglements per chain. We observe that in all cases the behavior of $G(t)$, beyond the short-time Rouse regime, is accurately described by $G^0_{\mathrm{N}}[P(t)]^2$, where the chain-length-independent prefactor $G^0_{\mathrm{N}}$ denotes the plateau modulus. This correlation is consistent with both double reptation and dynamic tube dilation models of polymer relaxation, although the two models are based on different physical pictures. The double reptation model represents the melt as a transient network in which stress relaxation is governed by the survival probability of pairwise entanglements. The dynamic tube dilation model, however, assumes that the tube of constraints surrounding a polymer chain progressively enlarges as relaxation proceeds. The relation $G(t) = G^0_\mathrm{N}[P(t)]^2$ can serve as a basis for determining the plateau modulus and the corresponding entanglement length. It also simplifies the modeling of $G(t)$, since an accurate analytical expression for $P(t)$ is sufficient to describe the long-time behavior of $G(t)$. We further compare the simulation data for $P(t)$ and $G(t)$ with theoretical predictions.

The position noise spectra of six drums on a single chip were measured on a single cooldown below 1.3 kelvin. Cryostat temperatures as low as 0.7 mK were achieved. The temperature dependence of the resonance frequency and linewidth of the drum modes was analyzed in the framework of the tunneling two level system (TLS) model. Departures of the resonance frequency and the position noise power from the expected logarithmic and linear temperature dependences, respectively, were interpreted as indications of thermal decoupling from the cryostat. This previously unexplored measurement configuration revealed that similar neighboring drums on a single chip may be at different temperatures. At the lowest temperatures, some drums exhibited excess damping that decreased with temperature. The magnitude of the excess damping of the drums was correlated with the thermal coupling of their TLS to the cryostat. In the case of one drum, a temporary increase in its damping coincided with a decrease in its mode temperature. The thermalization of the TLS to the cold finger was independent of pump power, pulse tube state and temperature of the pre-cooling stages of the cryostat. These results reveal an interplay between TLS damping and thermalization of nanomechanics that motivates further theoretical work and may impact efforts to extend the coherence of mechanical resonators.

While non-Hermitian Hamiltonians enable exotic dynamical phenomena, implementing their nonunitary time evolution on near-term quantum devices remains challenging. We propose a hardware-friendly protocol that simulates non-Hermitian dynamics without continuous monitoring. Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) evolution via classical Gaussian white-noise averaging and to subsequently cancel the quantum-jump contribution at the level of the measured observable using stochastic quantum error mitigation (QEM). The scheme requires no ancillas or controlled time-evolution, while the mitigation layer uses only single-qubit operations. We validate the method through numerical simulations of a model with asymmetric hopping, interaction, and disorder. Our work provides a programmable and ancilla-free framework investigating exotic dynamics that are not completely-positive and trace-preserving using QEM.

Hexagonal boron nitride (hBN) flakes are key building blocks for encapsulating two-dimensional (2D) materials, providing atomically flat surfaces and an excellent dielectric environment for high-mobility field-effect transistors and tunnelling devices. However, strain induced during mechanical exfoliation and assembly of van der Waals heterostructures may lead to plastic deformations of the hBN surface, injecting dislocation lines between the topmost layer and the underlying film. Since a monolayer of hBN is non-centrosymmetric and exhibits a piezoelectric response to deformation, individual dislocations and, in particular their networks, can generate electrostatic potential modulations in the encapsulated 2D material. Here, we examine scenarios in which the top hBN layer is uniaxially strained and/or twisted, and show how lattice reconstruction into dislocation networks leads to the formation of piezoelectric charge hotspots that effectively behave as charged defects.

Fatigue failure occurs when a solid is subjected to repeated, cyclic loading. Glasses subjected to cyclic to shear deformation have recently been investigated using computer simulations and theoretical models, to characterize and rationalize the dependence of the number of cycles to failure, depending on the properties of the glasses, and the deformation amplitude. The average number of cycles to failure has been observed to diverge as the strain amplitude approaches the so-called fatigue limit from above. In this work, rather than the average times themselves, we investigate by computer simulations the distribution of fatigue failure times, in model glasses subjected to cyclic shear deformation and in an elasto-plastic model. In particular, we observe in atomistic simulations that the standard deviation of the logarithm of failure times are proportional to their mean values, with the proportionality constant decreasing as the system size increases, indicating a sharper distribution of failure times. Using a finite-element-based elasto-plastic model, we observe similar behavior and perform a system-size analysis showing that the ratio of the standard deviation to the mean tends toward zero in the thermodynamic limit. Such distributions, rather than arising solely from the distribution of disorder in the samples that have been subjected to cyclic deformation, appear to arise from the intrinsic stochasticity of the failure process, which we analyze through a stochastic damage accumulation model.

We derive the effective Hamiltonian $H - μN$ for open quantum systems with varying particle number from first principles within the framework of non-relativistic quantum statistical mechanics. We prove that under physically motivated assumptions regarding the size of the system and the range of the interaction, this form of the Hamiltonian is unique up to a constant. Our argument relies firstly on establishing a rigorous version of the surface-to-volume ratio approximation, which is routinely used in an empirical form in statistical mechanics, and secondly on showing that the Hilbert space for systems with varying particle number must be isomorphic to Fock space. Together, these findings provide a rigorous mathematical justification for the standard grand canonical formalism employed in statistical physics.

High-quality planar cavities with low-absorption mirrors based on $Al_{0.2}Ga_{0.8}As/Al_{0.9}Ga_{0.1}As$ layers demonstrate continuous wave lasing at a wavelength of 957 nm. At 300 K, the threshold power density and quality-factor at threshold are (11.4$\pm$0.7) $kW/cm^2$ and (5070$\pm$160). Increasing the pumping level above two thresholds results in an enlargement in the quality-factor to at least 19000. Efficient lateral heat dissipation in the planar semiconductor microcavity is confirmed by a low mode-energy shift, which is 660 $μ$eV at two lasing thresholds.

The results for the electrical double layer capacitance and the charge density of ``free ions'' obtained from the mesoscopic theory are compared with the corresponding results of the associative mean spherical approximation. While the first theory takes into account the fluctuations of the charge density, the second theory assumes that the free ions and ion pairs are in chemical equilibrium according to the mass action law. Our results demonstrate a fairly good agreement between the two theories at high densities and low temperatures.

Chirality-dependent spin generation has attracted considerable attention in condensed matter physics. In this paper, we theoretically investigate antiparallel spin polarization as a chirality-dependent quadratic response, by using a finite chiral system composed of triangular prisms. Based on the nonlinear Kubo formalism and real-time simulations, we demonstrate that spatially inhomogeneous antiparallel spin polarizations are induced as a dissipative quadratic DC response to a homogeneous AC electric field. In particular, we elucidate role of microscopic parameters characterizing the handedness of chirality, and naive expectation of spin polarization as a consequence of spin accumulation of spin current.

The scaling behavior of fully flexible elastic tethered surfaces has been debated for decades. Some theories predict that self-avoiding surfaces would crumple in the absence of bending rigidity, while most simulations suggested that they would remain flat. Recent simulations on ideal membranes with lattice perforations suggest that systematically removing surface area from a membrane may provide an alternative way to crumpling self-avoiding surfaces. We perform extensive numerical simulations of two models of fully flexible elastic tethered surfaces in which self-avoidance can be systematically and continuously tuned to the ideal limit. We show that in the thermodynamic limit, these surfaces remain flat with a size exponent $ν=1$ for any finite degree of self-avoidance, with or without membrane perforations.

We consider the self-assembly of cross junctions in a general space dimension ($d$) as an extension of the problem studied in a previous paper for $d = 3$. This problem is equivalent to constructing a $d$-dimensional hypercubic jungle gym, at all junctions of which $2d$ rods with different colours meet. The analysis reveals a unique feature of the $d = 3$ case: the forced presence of at least one perfectly-ordered (singly coloured) direction (axis), in contrast to the possible absence of such a direction in $d \ge 4$. However, we will show that the uniaxial order is overwhelming not only in $d = 3$ but also for $d \ge 4$ in a sufficiently large system.

We investigate the interplay between complex magnetic orders and topological electronic states in nonsymmorphic magnetic Weyl semimetals on the ReAlX family (Re is a rare earth element and X is Si or Ge). We construct a lattice model incorporating conduction Weyl fermions coupled to localized magnetic moments via Kondo interaction. By considering a multi-${\bf Q}$ cycloid magnetic configuration, which can evolve into a Skyrmion lattice under an in-plane Zeeman field, we analyze its profound impact on the band structure through magnetic Brillouin zone and band-folding. Using the Kubo formula, we calculate the conductivity tensor and examine the transport properties in the clean limit. Our results reveal that the Skyrmion lattice induces significant changes in electrical and Hall conductivities. Furthermore, the temperature-dependent resistivity deviates from the standard Fermi-liquid behavior ($ρ_{xx}\sim T^2$), showing a power-law scaling ($ρ_{xx}\sim T^α$ with $α$ between 3 and 5), indicative of non-Fermi liquid behavior. This work provides a theoretical framework connecting multi-${\bf Q}$ magnetic textures, Skyrmion physics, and anomalous transport in topological semimetals.

We develop a diffusion approximation for systems subject to fast random resetting by small amplitudes. Equivalently, this describes systems with frequent but small catastrophes. We demonstrate the validity of the approximation by computing the stationary distribution and mean first-passage times of simple one-dimensional systems. The approximation captures dynamically induced correlations in multi-particle systems, and it can be used to generalise the conditionally independent and identically distributed structure recently found in systems with full resetting. Finally, we show that resetting can induce cycles and patterns, which can be characterised using the diffusion approximation.

Magnetic skyrmions are nanoscale, topologically protected spin textures with exceptional potential for high density data storage and energy efficient computing. Among various skyrmion hosting systems, rare earth transition metal ferrimagnets offer a promising platform due to their tunable magnetic properties and intrinsically low net magnetization. Despite this, the fundamental control of ferrimagnetic skyrmions in single layer films remains unexplored. Here, we demonstrate a viable route for engineering room temperature skyrmions in GdFe single layers through precise control of film thickness (60 to 80 nm). Thickness variation enables the systematic tuning of key magnetic parameters, including perpendicular magnetic anisotropy and saturation magnetization, thereby allowing precise control over skyrmion size and density. Magnetic force microscopy (MFM) reveals a clear thickness dependent evolution of isolated skyrmion characteristics, where skyrmion size decreases while skyrmion density increases with increasing GdFe film thickness, in agreement with micromagnetic simulations. At the same time, magnetotransport measurements show a systematic enhancement in the topological Hall resistivity with thickness, further corroborating the increased skyrmion density observed in MFM. Scanning transmission electron microscopy reveals a compositional gradient across the film thickness, indicative of structural asymmetry and potential inversion symmetry breaking, contributing to the emergence of a bulk Dzyaloshinskii Moriya interaction. Notably, sub 60nm skyrmions with high areal density are stabilized at room temperature. This work provides a viable route to tailor the properties of ferrimagnetic skyrmions in single-layer GdFe films, paving the way for the development of high-density ferrimagnetic skyrmionic devices.

Patterned two-dimensional (2D) magnetic nanostructures constitute geometry-engineered spin systems in which exchange, anisotropy, dipolar coupling, and finite-size effects operate on comparable energy scales. Spatial modulation of continuous magnetic films produces confinement-driven critical behavior, compensation phenomena, metastable switching pathways, and topologically non-trivial textures such as vortices and skyrmions. Computational modeling plays a central role in resolving this complexity, enabling quantitative construction of thermodynamic phase diagrams and analysis of geometry-dependent stability regimes. This review synthesizes theoretical and numerical frameworks for patterned 2D magnetism, including classical spin models, stochastic spin dynamics, rare-event methods, and multiscale parameterization informed by first-principles calculations. Representative systems-nanodot and antidot arrays, artificial spin-ice lattices, exchange-modulated heterostructures, and patterned van der Waals magnets- illustrate how geometry functions as an effective thermodynamic control parameter. Emerging directions in nonequilibrium modeling, multiphysics coupling, and scalable data-centric workflows are discussed in the context of predictive phase mapping. Patterned 2D magnetism thus exemplifies the convergence of geometry-controlled materials engineering and computational statistical physics, with phase stability and controlled spin textures at the core of next-generation spintronic architectures.

We present an exact integral formula for the multi-particle propagator of the one-dimensional Fermi--Hubbard model on an infinite lattice. The proof is based on the nested Bethe ansatz without relying on the string hypothesis. Our formula enables an explicit integral representation of the time evolution of arbitrary finite-particle wave functions and thereby provides a foundation for the exact analysis of nonequilibrium dynamics in the Hubbard model. It can further be applied to related open quantum models.

The actomyosin cortex, a thin layer of a cross-linked polymer network near the cell surface, generates active forces that are responsible for cell shape changes. Many developmental processes that involve such cell shape changes, most prominently embryonic cell division, are spatially confined by eggshells. To investigate the potential role of confinement in redirecting active stresses and enabling symmetry breaking phenomena during cell shape transformations, we study a hydrodynamic minimal model in which the cell cortex is represented as an active fluid surface that undergoes symmetric division in the absence of confinement. When enclosed by an ellipsoidal shell, a spontaneous symmetry-breaking transition emerges at a critical degree of confinement, where symmetrically dividing surfaces become unstable and polarized geometries appear. We show that this transition is controlled by the tightness of the confinement and analyze the solution space of stationary surfaces to identify the mechanisms underlying confinement-induced symmetry breaking.

Schottky junctions based on transition-metal dichalcogenides (TMDCs) have emerged as key building blocks for next-generation optoelectronic devices that demand ultrafast response and high sensitivity. However, the ultrafast, nanoscale carrier dynamics at these interfaces, crucial for device performance, have remained experimentally elusive. Here, we introduce optical pump-probe time-resolved atomic force microscopy to directly visualize, in real space, the nanosecond-scale modulation of the Schottky barrier potential at a van der Waals junction formed by point contact between WSe2 and a PtIr tip. Complementary analyses using transient absorption spectroscopy and light-modulated current-voltage characteristics together with model simulations reveal that time-resolved currents originate from the concerted temporal evolution of photoexcited carriers and the subsequent barrier response, processes that also define the rate-limiting steps of the photocurrent. Our results uncover the essential interfacial dynamics that underpin TMDC-based photodetectors and photovoltaic elements, while establishing a new measurement paradigm that complements and extends existing spectroscopic techniques. This approach provides direct access to nonequilibrium processes hidden at nanoscale interfaces, offering a powerful route to rational design of high-performance optoelectronic devices.

Altermagnets offer a unique pathway to functional spintronics by combining vanishing magnetization with large spin splitting. Here, we demonstrate that the canonical d-wave altermagnet KV2Se2O can deliver giant tunneling magnetoresistance through orientation-dependent spin filtering. By analyzing the crystallographic spin segregation, we show that transport along specific crystallographic axes is nearly fully spin-polarized within the symmetry-protected ballistic channels. We implement this mechanism in a lattice-matched KV2Se2O/Bi2O2Se/KV2Se2O magnetic tunnel junction, which achieves a robust half-metallic transport regime. The symmetry-protected spectral gap in the parallel/anti-parallel configuration ensures a high tunneling magnetoresistance ratio, resulting in substantial tunneling magnetoresistance, robust thermally driven spin filtering, and spin Seebeck effect at room temperature. These findings provide a path of altermagnetic heterostructures as a high-performance platform for scalable, field-free, and thermally stable spin logic.

The elastic properties of granular hydrogels are commonly characterised under wet conditions, yet the influence of capillary interactions remains unclear. In practical applications, hydrogels operate in aqueous environments containing dissolved ionic species, where swelling and elastic behaviour depend sensitively on ionic conditions. In this study, an experimental setup is developed to measure elastic responses of granular hydrogels under wet conditions. This setup directly observes liquid bridges formation and its evolution during compression. Our results show that neglecting capillary contributions leads to a systematic underestimation of the Young's modulus of hydrogels. Such an underestimation due to the capillary interaction increases as the sample size or its intrinsic stiffness decreases. In addition to the swelling ratio, the tested samples were also prepared under controlled salinity levels. The experimentally observed dependence of stiffness on swelling and salinity conditions is well captured by a modified constitutive model. The development of this study offers a robust testing protocol for measuring elastic properties of hydrogels under various environmental conditions.

The Adler equation is a well-known one-dimensional model describing phase locking and synchronization. Motivated by recent experiments using optomechanical oscillators, we extend the model to include overtone-synthesized sinusoidal coupling with adiabatic temporal modulation. This extension gives rise to unique topological features such as winding-number quantization, discontinuous phase-slip transitions, and hysteretic and non-reciprocal phase dynamics. We further extend the analysis to the quantum regime, where we find a counterintuitive result: the breakdown of winding-number quantization. This arises from the superposition of different winding-number states in a closed-space Thouless pump. Moreover, hysteretic dynamics, once eliminated in quantum adiabatic approximation, is recovered in non-adiabatic calculations, as the superposition of two Floquet states with different PT eigenvalues becomes the quantum counterpart of phase trajectory.

Nanofluidic ionic transistors typically require gate voltages above 1 V and operate only at sub millimolar ionic strengths, limiting their biocompatible applications. We demonstrate ionic transistors consisting of single sub 10 nm nanopores drilled in van der Waals (vdW) heterostructures with internal gate electrodes made of few layer graphene. These devices deliver up to 10fold current modulation at gate voltages as low as 0.3 V in 10 mM KCl, and 2fold modulation at near physiological 100 mM KCl. Baseline conductance with no gate shows surface charge dominated transport below 100 mM KCl consistent with negatively charged hBN walls and 5 nm opening of the pores. The surface charge and the electrochemical asymmetry introduced by the three electrode configuration govern the device behavior: negative gate voltage (VG) enriches ionic concentrations and enhances current, whereas positive VG induces a local depletion zone that suppresses transport. The current modulation by VG is dependent on the polarity of the transmembrane potential and leads to ion current rectification. Molecular dynamics simulations of a nanopore in a hBN graphene hBN stack reveal confinement and surface charge dependent suppression of relative permittivity of interfacial water. Continuum modeling with radially varying interfacial water permittivity reproduces the asymmetric IV characteristics and explains how the embedded gate sculpts local potential and ion concentrations. By enabling sub 0.5 V control of ionic transport at up to 100 mM salt concentrations, these devices address a key barrier in nanofluidics and open the pathway to low power ionic circuits and biosensing.

The narrow escape problem concerns the time needed for a diffusing particle to exit a confining domain through a small hole in the boundary. While this problem is now well-understood, determining the escape time for a particle that must exit through a narrow tube has proven challenging. Indeed, relying on analogies with electrodynamics, parameter fits to simulations, and heuristics, a variety of conflicting estimates for this escape time have been offered over the last three decades, some of which are counterintuitive and arguably non-physical. In this paper, we combine matched asymptotic analysis and probabilistic methods to determine the exact asymptotics of the narrow escape time through a tube. We obtain a new escape time formula which reduces to the various prior estimates in certain special cases. If the diffusivity in the tube differs from the diffusivity in the rest of the domain, our results reveal the importance of the form of the multiplicative noise inherent to any diffusivity that varies in space. We discuss our results in the context of asymmetric cell division.

Phonon modes within pristine crystalline resonators now routinely reach the quantum ground state. Such systems are attractive for quantum information science applications, as advanced fabrication and processing can enable relatively long quantum coherence times, and precision control can be realized through optical, electrical, or qubit coupling. In many state-of-the-art systems, the phonon lifetime is limited by disorder. In particular, native oxides or damaged `dead layers' at surfaces can host two-level tunneling states that lead to a particularly problematic form of dissipation that increases at lower temperatures. As mechanical losses are driven down in systems such as micro-fabricated bulk acoustic wave resonators, tunneling states are expected to emerge as the dominant mechanism for phonon decoherence. A quantitative description of these mesoscopic systems therefore requires a framework that captures interactions between a selected phonon mode and a large ensemble of TLS. Here, we derive a quantum master equation for this coupled system, permitting the phonon decoherence produced by two-level tunneling states to be calculated. As an example, we estimate the lifetime of a variety of quantum states within quartz micro-resonators hosting a thin surface layer of tunneling states. We find that the phonon coherence time is maximized at low temperatures, in spite of increased mechanical dissipation, and that phonon-TLS coupling can be reduced for modes with strain nodes at the surfaces.

We study a two-dimensional fluid of hard disks undergoing chiral, energy- and momentum-conserving collisions. We show that despite the microscopic breaking of time-reversal symmetry, the H-theorem is obeyed, guaranteeing a relaxation towards equilibrium in the absence of external forces. In the dilute limit, a Chapman-Enskog expansion yields analytical expressions for the shear and odd viscosity and the thermal conductivity. Theoretical predictions are confirmed by nonequilibrium molecular dynamics simulations.

In superconductor-semiconductor hybrid structures, superconductivity and spin polarization are competing effects because magnetic fields break Cooper pairs. They can be combined using thin films and in-plane magnetic fields, an approach that enabled the pursuit of Majorana zero modes, Kitaev chains, and Andreev spin qubits (ASQs), but remains challenging for materials with small in-plane g-factors. Here we show that granular aluminum (grAl), composed of nanometer-scale aluminum grains embedded in an amorphous oxide matrix, can overcome this limitation. By depositing grAl on Ge/SiGe heterostructures, we induce a hard superconducting gap with BCS peaks at 305 $μ$eV and magnetic-field resilience for both the in-plane and out-of-plane directions, allowing Zeeman splitting of Yu-Shiba-Rusinov (YSR) states beyond 50 $μ$eV (12 GHz). Leveraging this robustness, we reveal signatures of hole physics and demonstrate g-tensor tunability.

Planar germanium is currently the only semiconducting platform where high-coherence spin qubits and proximity-induced superconductivity have each been demonstrated. Recent research into spin qubits in Ge/SiGe heterostructures has focused on increasing the thickness of the SiGe capping layer, reporting improvements in the electrostatic noise levels. Meanwhile, heterostructures with thinner capping layers remain rather unexplored, despite the potential advantages for proximity-induced superconductivity. Here, we study a Ge/SiGe heterostructure with a thin SiGe cap $d \approx 4\ \mathrm{nm}$ and investigate its viability to host low-noise quantum dots. To keep the thermal budget compatible with superconducting layers, low-temperature oxide deposition processes were developed and implemented for the gate dielectrics. The charge-noise level of fabricated devices is estimated to be $1.8 \pm 1.0\ μ\mathrm{eV}/\sqrt{\mathrm{Hz}}$, comparable to devices fabricated on shallow heterostructures $\left(d \sim 20\ \mathrm{nm}\right)$ with high-temperature deposited oxides. Low charge-noise levels, together with the straightforward integration of superconductors, make this heterostructure an attractive platform for prototyping hybrid semiconducting-superconducting devices.

Collective behavior in biological systems was first captured by the Vicsek model, in which particles align their velocities in the average direction of neighbors, leading to coherent motion and showing an order-disorder transition. However, in many complex environments, the interactions are non-reciprocal, lacking an action-reaction symmetry. Using framework of the Vicsek model, we implement non-reciprocity by restricting interactions to neighbors located inside a finite vision cone, for a particle by limiting its set of interacting neighbors which fall within a vision-cone, providing a minimal description for cognitive perception. Using detailed numerical simulations, we explore the clustering and flocking behavior due to competition between noise and limited visual perception in the presence of alignment interaction. For low noise, with reduction in the vision angle the system shows transition from a global coherent motion to locally ordered small-sized clusters. This behavior is confirmed through the steady-state distributions of velocity components and their fluctuation relative to the global mean. This is also characterized using a polar order-parameter and a two-point velocity correlation function. Interestingly, at small vision angles, particles exhibit strong short-range correlations within clusters even in the absence of any global coherence. Time-evolution of the related correlation functions illustrate the pathways towards the emergence of such structures. The time dependence of the average cluster size and the length-scale calculated from the two-point velocity correlation show scaling behavior and indicate that the emergence of density field clustering is a consequence of the velocity-field coherence. Any kind of ordering and clustering disappear in the limit of high noise and low vision-angle regime.

Ermakov has shown how the solution to the classical harmonic oscillator in one spatial dimension with general time-dependent frequency can be reduced to the time-independent case and an associated nonlinear ordinary differential equation, an analysis which has been applied to the Schrödinger equation as well. We extend this analysis to the Landau problem of a charged particle in a uniform magnetic field in two dimensions and construct the generalized Laughlin wave functions for the case when the magnetic field is time-dependent. We also work out the dynamics of density fluctuations (the Girvin, MacDonald, Platzman or GMP mode) and argue that it is possible to tune the frequency of the magnetic field to obtain a compressible droplet of fermions. We also analyze the dynamics of the edge modes of the droplet for the integer Hall effect.

While molecular spin qubits (MSQs) are a promising platform for quantum computing, read-out has been largely limited to electron paramagnetic resonance which is often slow and requires a global system drive. Moreover, because one prerequisite for the Elzerman and Pauli spin blockade readout mechanisms typical of semiconductor spin qubits is tunneling of electrons between sites, these read-out modalities are unavailable in MSQs. Here, we theoretically demonstrate electrical read-out of entangled MSQs via driven many-electron spin unpolarized currents. In particular, using a time-dependent density matrix renormalization group approach we simulate a maximally entangled MSQ pair between two electronic leads. Driving itinerant electrons between the two leads, we find that the conductance is greater when the MSQs are in the entangled singlet state as compared to the entangled triplet state. This contrast in conductance is enhanced when the electronic density of states at the Fermi energy is large and for narrow bandwidth. Our results are readily applicable to molecules supramolecularly functionalizing semiconductors with relatively flat bands such as single-wall carbon nanotubes under a magnetic field.

The topological surface code is a leading candidate for harnessing long-range entanglement to protect logical quantum information against errors, and teleportation of logical states is desirable for robust quantum information processing. Nevertheless, scaling up the surface code in quantum teleportation poses a formidable challenge to experiment. Here on a superconducting quantum processor with 125 qubits, we demonstrate the robust teleportation of topological rotated surface code prepared by a linear-depth unitary circuit, with code distances up to 7. We obtain the teleportation phase diagram by tuning the local entangling gates uniformly across a finite threshold. Furthermore, we show that the entangling threshold can be boosted by coherent qubit rotations that inject magic resources beyond the Clifford regime, restoring the duality symmetry of the topological phase, which serves as a guiding principle to minimize the entanglement resource. Our results shed light on simulating and leveraging topological quantum matter on quantum devices, and pave the way to the ultimate goal of distributed fault tolerant quantum computation.

Hole-spin qubits in planar Ge/SiGe heterostructures have attracted significant attention in recent years owing to their favorable electrical characteristics and prolonged coherence times. However, the strong spin-orbit interaction also makes them susceptible to charge noise and inhomogeneous strain. This is further exacerbated by the highly anisotropic g-factor of the planar design. Although there are some known strategies to suppress charge noise, one approach is to engineer an isotropic g-factor. In this work we analyze how qubit confinement profile affects the g-factor of hole-spin qubits. We show that decreasing the characteristic in-plane qubit confinement length reduces the g-factor anisotropy. We perform analytical and numerical analysis to compare two types of quantum wells: square wells and parabolic wells. We show that square wells have limited tunability, while parabolic wells offer broader tunability, making them more promising for qubit engineering.

Inertia is intrinsic to many living and synthetic active systems, from animals and robotic agents to colloidal swimmers, and it strongly shapes transport. Many such systems employ intermittent restart protocols to regulate exploration. Stochastic resetting provides a theoretical framework for these strategies and a route to control nonequilibrium steady states, yet the role of inertia in reset-controlled active dynamics remains poorly understood. Here we study an inertial active Brownian particle subject to complete stochastic resetting of position, velocity, and orientation in two dimensions. Using a moment-generating framework together with the Final-Value Theorem, we derive closed-form steady-state moments up to fourth order as functions of inertia, activity, and reset rate. We show that inertia fundamentally modifies reset-controlled transport: at large reset rates the steady-state mean-squared displacement is suppressed much more strongly than in the overdamped limit, yielding enhanced localization near the reset point. At the same time, position excess-kurtosis phase diagrams reveal strongly non-Gaussian steady states characterized by a sharp central peak coexisting with heavy tails in the position distribution, indicating rare long excursions enabled by inertial persistence. The tail weight varies non-monotonically with reset rate, reflecting a competition between inertial momentum relaxation and resetting that selects an optimal regime maximizing rare excursions. Our results provide experimentally testable signatures of inertial effects in reset-controlled active systems.

Magnetic hopfions in chiral magnets are topological solitons, localized in three dimensions. But is their localization strong? To address this question we derive an asymptotic expansion for the isolated hopfion's spatial profile. It becomes starting point for a simple analytical model, which is asymptotically correct both near the hopfion center and far away from it. Region of equilibrium hopfions on the phase diagram of a helimagnet is computed and material requirements for supporting movable isolated magnetic hopfions on uniform background are discussed.

A micromagnetic framework for magnetic small-angle neutron scattering (SANS) is presented that accounts for weak symmetric anisotropic exchange in centrosymmetric nanocrystalline ferromagnets. The exchange interaction is expressed via a general fourth-rank tensor decomposed into isotropic and deviatoric parts. We start with the exchange energy and effective field, assuming weakly fluctuating in space saturation magnetization, solve micromagnetic problem to find spatial distribution of local magnetization vector and compute the averaged (over random orientations of nanocrystals) SANS cross sections. The isotropic part reproduces the classical Heisenberg-type SANS response, while non-zero deviatoric part of the exchange tensor gives rise to new angular harmonics in the magnetic SANS cross section. As a specific example, analytical response functions for an exchange tensor with hexagonal symmetry in perpendicular and parallel scattering geometries are derived. The results provide a basis for identifying and quantifying symmetric exchange anisotropy in magnetic SANS experiments.

Fluids in nanopores are of importance for many engineering applications, including energy storage in supercapacitors, hydrocarbons recovery from unconventional sources, or water desalination. Thermodynamic properties of fluids confined in nanopores differ from the properties of the same fluids in bulk. Density functional theory (DFT) has been widely used for modeling thermodynamics of confined fluids. However, it is rarely used for calculations of derivative thermodynamic properties. Here we use a rather simple DFT model for argon based on the Percus-Yevick equation, and showed that with standard parametrization it fails to predict derivative properties. However, slight adjustment in parameters leads to quantitative predictions of isothermal compressibility and thermal expansion coefficient at a selected temperature. Using the adjusted parameterization we performed the calculations of compressibility of argon confined in carbon slit pores of various sizes, and demonstrated that the compressibility of argon in confinement is lower than that in bulk and is pore size dependent. We confirmed the DFT predictions using the Monte Carlo molecular simulations. In addition to isothermal compressibility, we calculated the thermal expansion coefficient of confined argon. Our calculations showed that it behaves similar to compressibility -- it is always lower than the bulk value and gradually increases for smaller pore sizes. For several selected pore sizes we verified the DFT calculations by Monte Carlo simulations. Overall, our results suggest that the classical DFT can be utilized for calculations of derivative thermodynamic properties of confined fluids, which are computationally challenging to predict using molecular simulations.

The presence of strong disorder can significantly impact electrical conduction in metallic systems. Here, we investigate the temperature dependence of the electrical resistivity, $ρ(T)$, in nonmagnetic single crystals of the Remeika-phase cage compounds R$_3$Co$_4$Ge$_{13}$ (R = Y, Lu). Contrary to the density of states (DOS) calculations in the literature, the experimentally measured $ρ(T)$ in both compounds exhibits semiconducting-like behavior, which we attribute to the strong structural disorder due to its unique crystal structure and low carrier-density. A detailed analysis of the electrical resistivity data reveals that neither the Arrhenius thermal activation law nor variable-range hopping (VRH) models can adequately describe their temperature dependence over the broad temperature range of 2-350 K. However, a model incorporating parallel conduction through both semiconducting and metallic channels provides an adequate explanation. In addition to a dominant metallic conduction below $\sim 10$~K, a negative temperature coefficient of the electrical resistivity ($dρ/dT$) is found in both samples. In the absence of magnetic impurities, the observed $dρ/dT < 0$ is interpreted in terms of the structural Kondo mechanism.

We study elementary semiconductors and insulators that are symmetric under spatial inversion: silicon, diamond, germanium, and black phosphorene. These materials are ideal candidates for realizing obstructed atomic insulators, which differ from trivial atomic insulators by a quantized spatial shift of their electronic Wannier centers with respect to the atomic lattice. We use symmetry indicator invariants that allow the prediction of non-trivial responses to crystal dislocations in these materials. We find that edge dislocations generically exhibit a non-trivial response, while screw dislocations always display a trivial response. With the aid of numerical simulations of realistic tight-binding models, we confirm the presence of mid-gap polarization bands localized along dislocations in silicon, diamond, and germanium.

In a series of works of ours we have shown that we can represent the critical and tricritical points of the Statistical Physics of critical phenomena as a Dynamical phenomenon expressed by time series produced by the type I intermittency that exhibits a weak chaos. Recently we have also shown that if we couple these two chaotic dynamics, namely critical and tricritical, we can produce a time sequence which is a temporal Spike Train (ST) of biological-type . In the present work we generalize this issue producing superpositions of critical-tricritical intermittencies with different parameter values. Now arise the question whether the coupling occurs between time series that have resulted from the superposition, will preserved or destroyed the ST biological type , as the number of intermittencies in the superposition will increase? In the other side in present work we find that the spikes produced by the chaotic dynamics of the intermittencies, under the action of superpositions and coupling remain biological-type too. Thus we can say that the dynamics of the fluctuations of the values of the time series produced by the coupling of the superpositions of the intermittencies is the same as the dynamics of the fluctuations of the membrane potential of the biological neuron. Given also that we can manipulate the numerical experiment of superposition and coupling as we wish, we will be able, in future, to approach the cause of neurological problems and decline in thinking ability due to loss of spikes in the brain.

Characterizing the geometric boundaries of protein secondary structures is fundamental to understanding macromolecular folding. By applying the discrete Hasimoto map to translate backbone geometry into a one-dimensional discrete nonlinear Schrödinger potential $V_{\mathrm{re}}[n]$, we establish a frequency-domain framework for protein conformations. Short-time Fourier transform analysis across 320,453 residues from 1,986 non-redundant proteins defines a local spectral entropy $H_{\mathrm{spec}}$ that consistently orders structural states. Helical segments emerge as narrow-band low-entropy regimes dominated by zero-frequency components, whereas coils manifest as broadband noise. We demonstrate that boundaries separating these states exhibit step-like sharpness characteristic of a first-order-like geometric transition with a sub-residue median width of 0.145 residues. This abrupt kinematic transition provides a spatial counterpart to the cooperative Zimm--Bragg thermodynamic model of helix nucleation. The extreme spatial narrowness exposes an intrinsic limitation governed by the Gabor uncertainty principle, explaining why the pointwise integrability residual $E[n]$ acts as an effective high-pass filter for boundary detection. Guided by this limit we introduce a dual-probe approach combining the high-pass residual for local torsion discontinuities with a low-frequency energy ratio $R_{\mathrm{LF}}$ measuring the DC-dominated flatness of helical interiors. Unifying these complementary signals improves the detection area under the curve from 0.783 to 0.815. Because high-entropy broadband regions coincide with the flexible loops and hinges implicated in allostery, the spectral entropy of the Hasimoto potential may serve as a sequence-agnostic geometric proxy for mapping functional dynamics from backbone coordinates.

We find a general framework for the construction of birational involutions on two- and three-dimensional varieties obtained from $\mathbb P^2$, $\mathbb P^1\times \mathbb P^1$, and $\mathbb P^3$ by blow-up at nine, respectively eight points. Each such involution is based on a divisor class with a one-dimensional linear system with a generic element of genus zero. Classical Manin involutions represent the simplest particular case. Novel, more sophisticated cases identified here include birational involutions of $\mathbb P^2$ along conics and along nodal cubic curves, as well as birational involutions of $\mathbb P^3$ along quadratic cones and along Cayley nodal cubic surfaces. We prove a general formula for the induced action of geometric birational involutions on the respective Picard group, and give a general result about decomposition of translational elements of the respective affine Weyl group of symmetries into a product of two geometric birational involutions.

Chaotic instability in many-body systems is commonly quantified by the largest Lyapunov exponent, yet general constraints on its magnitude in classical interacting systems remain poorly understood. Here we establish explicit, Hamiltonian-specific upper bounds on the largest Lyapunov exponent for classical many-body systems with local interactions. These bounds arise from instantaneous stability constraints on the Hamiltonian flow and are expressed in terms of inertial scales and the curvature of the interaction potential. We show that they naturally separate into two qualitatively distinct classes: non-violable bounds, controlled by worst-case local curvature scales and inertia and insensitive to spatial structure, and ergodic ceilings, which retain spectral information and encode collective modes and finite-size effects under generic dynamical evolution. For a paradigmatic one-dimensional coupled-rotor chain (Josephson junction array), the ergodic ceiling admits a closed analytic form and produces a dynamically inaccessible region for sustained chaotic growth in the Lyapunov exponent-energy plane, which we confirm numerically. In contrast to non-violable estimates, the ergodic ceiling yields a sharper constraint on chaotic growth by capturing collective suppression mechanisms absent at the level of local curvature alone. Remarkably, in the thermodynamic limit the ergodic ceiling asymptotically approaches an inertial ceiling that limits sustained Lyapunov growth, becoming independent of temperature and interaction strength. While classical systems do not admit universal chaos bounds, our results identify a broad class of natural Hamiltonian systems in which chaotic growth is inherently limited by inertia and interaction geometry, thereby setting a minimal microscopic timescale for long-time loss of memory of initial conditions.

This paper extends our previous work~(Szumiński and Maciejewski, 2024), where we explored the dynamics and integrability of the double-spring pendulum. Here, we investigate the variable-length double pendulum, a three-degree-of-freedom Hamiltonian system combining features of the classic double pendulum and the swinging Atwood machine. With its intricate dynamics, this system is crucial for studying nonlinear phenomena such as high-order resonances, chaos, and bifurcations. We address the challenges posed by high-dimensional phase spaces using a novel tool, the \textit{Lyapunov refined maps}, which integrates Poincaré sections, phase-parametric diagrams, and Lyapunov exponents. This framework comprehensively analyzes periodic, quasi-periodic, and chaotic behaviors. By measuring the strength of chaos, it also offers insights into the system's dynamical structure. Additionally, we apply Morales-Ramis theory to examine integrability, leveraging the differential Galois group of variational equations to establish non-integrability conditions. The Kovacic algorithm is used to analyze the solvability of higher-dimensional differential equations, complemented by Lyapunov exponent diagrams to exclude integrable dynamics under certain parameters. Our findings advance the fundamental understanding of variable-length pendulum dynamics, offering new insights and methodologies for further research with potential applications in adaptive robotics, energy harvesting, and biomechanics. Additionally, this work represents a significant step toward proving the long-sought non-integrability of the classical double pendulum.

We theoretically and experimentally establish a connection between linearly chirped pulse propagation properties and complex time delay for both transmitted and reflected pulses in reverberant scattering systems. We demonstrate that the time shift of the chirped pulse depends on both the real and imaginary parts of the complex time delay of the scattering system. We also show that the chirped pulse experiences a center frequency shift that is directly proportional to the imaginary component of complex time delay, similar to that found in Giovannelli and Anlage (2025). Using these insights, we then demonstrate how complex time delay can be harnessed to systematically tune the propagation properties of a chirped pulse such that a near-zero time shift can be achieved for a wide range of pulse center frequencies in a resonant scattering system.

Collaboration networks evolve throughout academic careers, yet few studies systematically examine how these network dynamics relate to long-term career success and mobility. Analysing 35,708 physicists' careers spanning at least 15 years, we use time series clustering to identify ten distinct evolution patterns of network size and clustering coefficient across career years 5 to 15. We report three key results. First, authors who begin with loosely connected networks and progressively tighten their networks while expanding network size during mid-career achieve the highest PI attainment rates, publication output, and citation impact. Second, despite different starting points, network evolution patterns associated with better outcomes converge toward moderate clustering by career year 15, suggesting an optimal balance between core team cohesion and diverse external connections. Third, mobility is positively associated with these successful network evolution patterns and remains positively associated with scientific outcomes even after controlling for network evolution patterns.

Capacity drop is a traffic phenomenon in which the discharge flow from a queue is lower than the theoretical infrastructure capacity. This paper proposes a generic analytical method to estimate the queue discharge flow of freeway traffic. Capacity drop is primarily attributed to hesitant vehicles, defined as vehicles that stochastically and temporarily enter an acceleration delay state and generate voids (i.e., extra gaps) in front of them. The proposed method estimates the expected total void length generated by all hesitant vehicles, based on the distributions of their spatial and temporal locations as well as the associated delays. It also accounts for interactions between the waves triggered by downstream hesitant vehicles and the voids generated by upstream ones. Our analysis reveals that this interaction is the key mechanism behind the differing extents of capacity drop observed between standing queues and jam waves in previous studies. The accuracy of the model is validated through both numerical simulations and real-world trajectories. Overall, the proposed method offers a deeper understanding of capacity drop, which can be leveraged in traffic flow modeling and control.

An octant representation of higher-order optical modes that includes Laguerre-Gaussian and Hermite-Gaussian modes is presented. The octant picture captures the high-dimensional nature of three-state optical systems and beyond, with standard Poincaré spheres for orbital angular momentum forming subspaces of the entire state space. This representation enables intuitive manipulation of both classical modes and optical qudits and provides a framework for extending Berry phases and topological invariants to high dimensions.

Vertical-Cavity Surface-Emitting Lasers (VCSELs) combine compact geometry, low threshold current, and ease of integration, making them central to modern photonic systems. However, their polarization behavior remains a critical factor affecting performance, as the emission state can switch between orthogonally polarized modes around the so-called polarization switching point. This regime exhibits high sensitivity, where small perturbations induce abrupt polarization changes and nonlinear responses. In this work, the polarization dynamics of VCSELs under sinusoidal current modulation around the switching point are numerically investigated using the Spin-Flip Model. The study examines the influence of modulation frequency, amplitude, and bias current, revealing distinct dynamical regimes including polarization locking, periodic and irregular switching. The observed transitions between regimes elucidate the interplay between modulation and polarization stability, providing insight into the control of VCSEL dynamics for high-speed optical communication and sensing applications.

In Part I of this companion paper series, we introduced SWIFTraj, a new open-source vehicle trajectory dataset collected using a unmanned aerial vehicle (UAV) swarm. The dataset has two distinctive features. First, by connecting trajectories across consecutive UAV videos, it provides long-distance continuous trajectories, with the longest exceeding 4.5 km. Second, it covers an integrated traffic network consisting of both freeways and their connected urban roads. Obtaining such long-distance continuous trajectories from a UAV swarm is challenging, due to the need for accurate time alignment across multiple videos and the irregular spatial distribution of UAVs. To address these challenges, this paper proposes a novel graph-based approach for connecting vehicle trajectories captured by a UAV swarm. An undirected graph is constructed to represent flexible UAV layouts, and an automatic time alignment method based on trajectory matching cost minimization is developed to estimate optimal time offsets across videos. To associate trajectories of the same vehicle observed in different videos, a vehicle matching table is established using the Hungarian algorithm. The proposed approach is evaluated using both simulated and real-world data. Results from real-world experiments show that the time alignment error is within three video frames, corresponding to approximately 0.1 s, and that the vehicle matching achieves an F1-score of about 0.99. These results demonstrate the effectiveness of the proposed method in addressing key challenges in UAV-based trajectory connection and highlight its potential for large-scale vehicle trajectory collection.

We report on the experimental generation of polarization symmetry-broken cavity solitons (CSs) in a passive, fiber-based, coherently-driven, Fabry-Perot (FP) Kerr resonator. Polarization resolved measurements reveal the spontaneous transition of initially symmetric CSs into asymmetrical vectorial states, triggered by a cross-phase modulation-induced polarization bifurcation. Most notably, due to counter-propagation of light occurring in FP resonators, we unveil a collective polarization conformity effect, whereby multiple CSs circulating in the cavity converge to the same asymmetric polarization state once their number exceeds a certain threshold. These results demonstrate that Fabry-Perot resonators support novel collective soliton dynamics that are absent in ring architectures.

Fluorescence spectroscopy and modeling provide powerful means to characterize biomacromolecular structures, dynamics, and interactions. Förster resonance energy transfer serves as a key technique for this due to its nanometer-scale distance sensitivity. Quantitative interpretation of fluorescence data relies on models that link molecular structure to observable spectroscopic quantities and vice versa. Integrative modelling frameworks combine fluorescence observables with complementary structural information to infer molecular structures and conformational ensembles. This review outlines conceptual components of fluorescence-based modeling, discusses dye representations, and highlights advances toward refined models enabling quantitative structural analysis. Finally, we discuss the prediction of spectroscopic properties of dyes based on biomolecular structures and fluorescence assay design beyond traditional FRET applications.

With the rapid development of micro-electro-mechanical systems, electrically small micro-targets, such as subwavelength micro unmanned aerial vehicles and bionic mosquito robots, exhibit ultra-low scattering cross section, which brings severe challenges to their effective detection. To address this problem, an Invisible Electromagnetic Scattering Amplifier (IESA) is designed by combining finite-element electromagnetic simulation with a forward lossless tandem neural network. The IESA realizes the dual-functional integration of intrinsic electromagnetic invisibility (near-zero scattering) for itself and significant scattering amplification for subwavelength targets entering its air sensing region. Electromagnetic simulations verify that the designed IESA can achieve a stable scattering amplification effect on subwavelength targets with a characteristic size of approximately 0.1λ0, regardless of their spatial positions or geometric shapes, with a maximum scattering cross section amplification factor of 8.58. The IESA breaks the technical bottleneck of the separate design of electromagnetic invisibility and scattering amplification functions. It shows potential for applications in the fields of radar detection, anti-terrorism security, micro-target monitoring, and adaptive electromagnetic sensing.

Networks underpin systems that range from finance to biology, yet their structure is often only partially observed. Current reconstruction methods typically fit the parameters of a model anew to each snapshot, thus offering no guidance to predict future configurations. Here, we develop a Bayesian approach that uses the information about past network snapshots to inform a prior and predict the subsequent ones, while quantifying uncertainty. Instantiated with a single-parameter fitness model, our method infers link probabilities from node strengths and carries information forward in time. When applied to the Electronic Market for Interbank Deposit across the years 1999-2012, our method accurately recovers the number of connections per bank at subsequent times, outperforming probabilistic benchmarks designed for analogous, link prediction tasks. Notably, each predicted snapshot serves as a reliable prior for the next one, thus enabling self-sustained, out-of-sample reconstruction of evolving networks with a minimal amount of additional data.

Here we analyze ways to achieve deep subthreshold parametric squeezing or cooling of a single degree-of-freedom parametric resonator enhanced by a lock-in amplifier feedback loop. Due to the feedback, the dynamics of the parametric resonator becomes more complex and a Hopf bifurcation at the instability threshold can occur. Initially, we calculate the phase-dependent gain of parametric amplification with feedback of an added ac signal. In one approach, we obtain the amplification gain approximately using two independent approaches: the averaging method and the harmonic balance method. We also obtain this gain more exactly using Floquet theory and Green's functions methods. The Hopf bifurcation was predicted by the harmonic balance method and by Floquet theory, but not by the averaging method. In our analysis of fluctuations, we Fourier analyze the response of the parametric resonator with feedback to an added white noise. We were able to calculate, in addition to the noise spectral density, the squeezing of fluctuations in this resonator with feedback. Very strong squeezing or cooling can occur. Deamplification and cooling occur near the Hopf bifurcation, whereas squeezing occurs near a saddle-node bifurcation.

Community detection based on modularity maximization is one of the most widely used approaches for uncovering mesoscale structures in complex networks. However, it is well known that the modularity function exhibits a highly degenerate optimization landscape: a large number of structurally distinct partitions attain close modularity values. This degeneracy raises issues of instability, reproducibility, and interpretability of the detected communities. We propose a simple and user-friendly post-processing method to address this problem by selecting a representative partition among the set of high-modularity solutions. The proposed approach is model-agnostic and can be applied a posteriori to the output of any modularity-based community detection algorithm. Rather than seeking the optimal partition in terms of modularity, our method aims to identify a solution that best represents the structural features shared across degenerate partitions. We compare our approach with consensus clustering methods, which pursue a similar objective, and show that the resulting partitions are highly consistent, while being obtained through a substantially simpler procedure that does not require additional optimization steps or external software packages. Moreover, unlike standard consensus clustering techniques, the proposed method can be applied to networks with both positive and negative edge weights, making it suitable for a wide range of applications involving signed networks and correlation-based systems, such as social, financial, and neuroscience networks. Overall, the method provides a practical and robust tool for handling degeneracy in modularity-based community detection, combining simplicity with broad applicability across different types of networks and real-world problems.

Periodic photonic structures enable precise control over the light-matter interaction through band structure engineering. Certain lattice geometries exhibit dispersionless flat bands, characterized by vanishing group velocity and diverging density of states, which present unique opportunities for applications such as slow light, nonlinear optical processes and controlling photoluminescence. However, thus far, flat bands have not been reported in systems where the lattice sites are radiatively coupled over a long range. Here we show that lattices consisting of superposed equispaced one dimensional chains exhibit flat bands with a purely diffractive origin, with the energies and angles of the flat bands controlled by the geometrical parameters of the lattice and the unit cell. The flat bands extend over all angles, can have linewidths on the order of a few nanometers, and are linearly polarized. We experimentally observe flat bands at predicted energies in lattices of gold nanoparticles at near-infrared frequencies using Fourier spectroscopy. Our results provide a general and efficient design strategy for lattices with flat, polarized dispersions for applications such as flat-band lasing, enhancing light-matter interaction, and controlling the emission or absorption of electromagnetic radiation over a wide spectral range.

Controlling light-matter interactions is central to photonic technologies ranging from lasers to optical information processing. Suitably designed photonic structures give rise to flat (dispersionless) bands, where the density of states diverges, and group velocity goes to zero, allowing light localization. These properties make flat bands attractive for lasing; however, designing photonic structures supporting flat bands suitable for lasing is challenging. Here, we introduce nanoparticle chain lattices. These chain geometries provide long-range coupled systems that support, at predictable wavelengths, bands that are totally flat and extend over the full angular range. We demonstrate lasing in the transverse-magnetic (TM) mode of single chains of nanoparticles and explain the transition from flat band lasing to the single-mode normal-incidence (Gamma-point) lasing as the number of chains is increased. Moreover, we show partially coherent emission from square and triangular two-dimensional chain lattices. The excited modes depend on the pump power and polarization. Our results establish chain lattices as a versatile platform for exploring flat band lasing and suggest new routes toward narrowband, linearly polarized, and bright light sources with tailored coherence.

Optical spectroscopy is an important and widely used technique, for instance, to characterize new materials and to identify unknown compounds. Spectra are typically reported as a function of the wavelength of light, yet the information extracted from such spectra can be misleading. In contrast, spectra represented as a function of the frequency (or photon energy) allow for a more direct extraction of the intrinsic quantum-mechanical properties of the materials under investigation. Here we discuss this conversion for absorption, fluorescence and fluorescence excitation spectra. We show step-by-step the different factors that lead to a rescaling of the measured absorption and fluorescence signals. This paper will assist instructors who aim at developing an (under-)graduate lab to introduce into the methodology and terminology of spectroscopic experiments and to provide clear, step-by-step guidelines for data analysis and representation.

Matrix representations of quantum operators are computationally complete but often obscure the structural topology of information flow within a quantum circuit \cite{nielsen2000}. In this paper, we introduce a generalized graph-theoretic framework for analyzing quantum operators by mapping unitary matrices to directed graphs; we term these structures \emph{Imperfect Graphs} or more formally as \emph{Topological Structure of Superpositions}(TSS) as a tool to devise better Quantum Algorithms. In this framework, we represent computational basis states as vertices. A directed edge exists between two vertices if and only if there is a non-zero amplitude transition between them, effectively mapping the support of the unitary operator. In this paper we deliberately discard probability amplitudes and phase information to isolate the connectivity and reachability properties of the operator. We demonstrate how TSS intuitively helps describe gates such as the Hadamard, Pauli-(X,Y,Z) gates, etc \cite{nielsen2000}. This framework provides a novel perspective for viewing quantum circuits as discrete dynamical systems \cite{childs2009,aharonov2001} Keywords: Quantum Algorithms, Unitary Matrix Approach, Topological Structure of Superpositions (TSS), Graph Theory

Sensory information propagates through successive processing stages in the brain, where synaptic weight patterns between stations determine how downstream neurons decode information from upstream populations. Although optimized synaptic connectivity can enhance information transmission, it requires precise weight tuning. Recent evidence depicting substantial synaptic volatility raises two fundamental questions: How does coarse-tuning of synaptic connectivity affect information transmission? What strategies could the nervous system employ to maintain reliable communication despite synaptic fluctuations? We addressed these questions by analyzing the signal-to-noise ratio ($SNR$) for binary stimulus discrimination under two decoding schemes: a naive population average and an optimized linear decoder. For the naive decoder, we found that $SNR$ remains largely insensitive to synaptic imprecision, since performance is already limited by correlated noise in neuronal responses. For the optimal decoder, we identified three distinct regimes. Under weak coarse-tuning, $SNR^2$ scales linearly with population size $N$. Under moderate coarse-tuning, scaling becomes sublinear. Under strong coarse-tuning, the regime most consistent with observed neuronal heterogeneity, $SNR$ saturates and can not be improved by recruiting larger populations. This limitation persists even when incorporating feedforward or recurrent network architectures. These findings suggest that in the biologically relevant regime of strong coarse-tuning, naive and optimal decoders can achieve qualitatively similar performance. The analysis shows that effective readout under synaptic volatility is constrained to an invariant low-dimensional manifold aligned with the naive decoder, potentially pointing to a fundamental principle for robust neural computation in the face of ongoing synaptic remodeling.

Lying-standing transitions in the first molecular monolayer at organic-inorganic interfaces strongly influence interface dipoles, energy-level alignment, and growth modes, yet their collective kinetics remain difficult to predict. Here, we establish a quantitative adsorbate-to-kinetics relationship using first-principles-based kinetic Monte Carlo simulations combined with a mean-field coarse-graining strategy. Focusing on tetracyanoethylene on Cu(111), we show that the collective transition rate cannot be inferred from any single elementary step but emerges from coupled microscopic processes, including reorientation, adsorption, and diffusion. A local two-step reorientation mechanism captures the diffusion-limited regime, while diffusion of lying molecules accelerates the transition in diffusion-enhanced regimes by suppressing back-reorientation via vacancy-molecule decoupling. This effect is described by a regime-dependent geometric factor accounting for deviations between single-molecule and collective rate constants. By varying molecular size and footprint ratio, we demonstrate that geometry is an intrinsic control parameter. While the collective rate scales approximately with molecular area, increasing the footprint ratio between lying and standing configurations yields order-of-magnitude accelerations due to enhanced vacancy creation and diffusion-assisted stabilization. Finally, we derive an analytical expression for the collective reorientation rate constant linking temperature- and pressure-dependent microscopic rate constants to geometric parameters. The formulation reproduces the simulations across kinetic regimes and provides transferable design principles for engineering lying-standing transition timescales at organic-inorganic interfaces.

In this study, we developed a UV-tape-assisted laser patterning (UT-Laser) technique that enables the simple transfer-based formation of wiring with line widths below 200 $μ$m onto textile substrates. With the rapid advancement of wearable devices capable of acquiring various types of physiological and environmental information, research on electronic textiles (e-textiles)-in which electronic components are integrated into fabrics and clothing-has progressed considerably. However, integrating high-performance, rigid electronic components onto textiles remains challenging: the diameter of textile fibers limits the formation of fine wiring, making reliable mounting of such components difficult. To address these challenges, we devised the UT-Laser technique, in which thin foil or film materials are laser vector-cut on UV tape, and the adhesive strength is controlled through UV exposure. The unnecessary portions are selectively and collectively peeled away to form fine wiring, which is subsequently transferred onto the textile substrate. This approach enables facile fabrication of fine wiring with line widths below 200 $μ$m on textiles. Furthermore, by forming fine wiring from a flexible copper clad laminate and transferring it onto heat-resistant glass cloth, electronic components can be soldered directly, allowing the fabrication of e-textile devices capable of withstanding more than 10,000 bending cycles. The prototype e-textile device fabricated using the proposed method integrates a microcontroller, USB connector, battery holder, flash memory, inertial measurement unit, and environmental sensors, and successfully acquires data related to stair climbing, respiration, and changes in body temperature during sleep.

We investigate the polarization properties of emission associated with the magnetic dipole and electric dipole transitions of europium(III) coupled to an anisotropic dielectric metasurface with polarization-engineered electric and magnetic photonic local density of states. The metasurface consists of a square array of Mie-resonant elliptical a-Si:H dimers situated on an SiO$_2$ substrate and embedded in a PMMA film containing Eu(TTA)$_3$. Based on reciprocity principle, it was designed to achieve maximum electric (magnetic) field enhancement in the dimer gap at 610 nm (590 nm) for $x$-polarized ($y$-polarized) normally incident light in order to selectively enhance the electric dipole (magnetic dipole) emission into the $x$-polarized ($y$-polarized) emission channel, respectively. Momentum-resolved spectroscopy and back-focal plane imaging of emission of the fabricated light-emitting metasurface clearly reveal the intended polarization-dependent emission behaviour, with the $x$-polarized ($y$-polarized) emission showing a reduced (enhanced) ratio of the magnetic-/electric dipole emission intensity, correspondingly where the magnetic dipole emission is enhanced with a magnetic field enhancement from the nanostructures. The demonstrated polarization-dependent interaction of a designed nanostructure with the electric- and magnetic dipolar transitions of trivalent lanthanide ions opens an avenue towards routing of emission of different multipolar orders into different polarization channels.

In network science, collective dynamics of complex systems are typically modelled as (nonlinear, often including many-body) vertex-level update rules evolving over a graph interaction structure. In recent years, frameworks that explicitly model such higher-order interactions in the interaction backbone (i.e. hypergraphs) have been advanced, somehow shifting the imputation of the effective nonlinearity from the dynamics to the interaction structure. In this short note we discuss such structural-dynamical representation duality, and investigate how and when a nonlinear dynamics defined on the vertex set of a graph allows an equivalent representation in terms of a linear dynamics defined on the state space of a sufficiently richer, higher-order interaction structure. We show that multilinear dynamics defined in the vertices of a graph admit an exact finite realizations as linear dynamics on the state space of a hypergraph. For other high-order interactions involving more general analytic nonlinearities, using Carleman linearization theory we discuss how that the required state space liftings necessary to linearize the dynamics cannot be accomodated to the simple structure of a hypergraph, and a richer combinatorial architecture such as a hb-graph is needed.

The advancement of quantum technology networks necessitates high-speed, low-thermal load, and minimal-noise communication links between cryogenic and room-temperature components. At the heart of modern telecommunication, lay optical interconnects allowing for large data transfer capabilities via optical fibers. However, cryogenic photonic technologies remain largely unexplored and require a detailed understanding of material behavior and defect dynamics at low temperatures. In this work, we present the first comprehensive study of integrated III-V heterostructures operating at cryogenic temperatures down to 5K. Using an integrated n-InP/i-InGaAs/p-InP/p-InGaAs stack monolithically grown on silicon, we identify a temperature-dependent current-lowering mechanism arising from trap states becoming increasingly active below 140K. We demonstrate for the first time that these traps can be equivalently excited and controlled through either thermal or optical energy, revealing a dual modulation mechanism. These findings provide new insights into carrier transport and defect behavior in III-V heterostructures at cryogenic temperatures, advancing the field of cryogenic photonics and offering a non-destructive approach for identifying and characterizing material impurities in integrated quantum and optoelectronic devices.

This study introduces a pure-amplitude hologram (PAH) for generating radial carpet beams (RCBs), which are conventionally produced using pure phase radial gratings (PRGs). The hologram is designed by embedding the transmission function of a binary PRG into the phase argument of the cosine term(s) of an amplitude linear grating. When illuminated with a plane wave, this hologram generates RCBs in the non-zero diffraction orders, and when illuminated with a Gaussian beam, it generates RCB-like patterns at specific propagation distances. This method entirely eliminates the need for complex and expensive spatial light modulators (SLMs). The study presents a theory of diffraction for plane and Gaussian beams from such holograms, including a specific theoretical treatment of Gaussian beam diffraction from PRGs. Through theoretical analysis and experiments, we demonstrate how different RCBs can be generated at different diffraction orders due to the phase-amplitude enhancement resulting from multiplying the diffraction order number by the phase amplitude of the embedded base PRG, when the illuminating beam is a plane wave. For the Gaussian beam case, we show how different RCB-like patterns can be generated at different diffraction orders for the same reason, though only at specific propagation distances. Experimental and numerical results indicate that this technique yields RCBs and RCB-like patterns with approximately five times the useful power of their SLM-generated counterparts, demonstrating significantly higher power efficiency. This advantage renders the proposed method highly suitable for applications such as multiple optical trapping and free-space optical communication.

Wave--particle duality is a hallmark of quantum mechanics. For bosonic systems, there exists a continuum of intermediate states bridging wave-like Schrödinger cat states and particle-like Fock states. Such states have recently been recognized as valuable resources for enhancing fault-tolerant quantum computation (FTQC) with propagating light. Here we experimentally demonstrate tunable generation of these intermediate states by employing generalized photon subtraction (GPS). By detecting up to three photons from squeezed-light sources with a photon-number-resolving detector, we continuously control the balance between wave- and particle-like features. This approach allows us to construct a spectral family of quantum states with high generation rates, optimized according to the required fault-tolerance threshold. Our results establish GPS as a versatile toolbox for tailoring non-Gaussian resources, opening a pathway to efficient Gottesman--Kitaev--Preskill (GKP) qubit generation and addressing a central bottleneck in optical quantum computing.

We investigate the size-dependent behavior of Miura-ori-based origami tessellations by changing the number of origami unit cells. For large tessellations, the Miura-ori sheet generally exhibits a negative in-plane Poisson's ratio, whereas if the size of the Miura-ori tessellations becomes small, the transition between positive and negative Poisson's ratio emerges in the middle of the folding process. Here, we show that such a transitioning point, i.e., zero Poisson's ratio, yields a kinematic locking state. We also experimentally demonstrate the tunable locking behavior altered by tessellation sizes. Extending the analysis to three-dimensional origami tessellations, we find that the direction of kinematic locking changes depending on the tessellation size. Varying tessellation size thus enables control over both the onset and the direction of locking in origami metamaterials.

Spatiotemporal modulation approaches have been often employed as alternatives for producing optical non-reciprocity without magneto-optic materials. Unidirectional inter-modal scattering, enabled by either acousto-optic or electro-optic (EO) modulation, is a promising method in this category as it can directly modify optical dispersions and even enables linear non-reciprocal photonic devices in the strong coupling limit. While EO approaches are often preferred for their practicality, it is challenging to generate the large spatiotemporal momentum required for inter-modal phase matching without EO drive schemes involving multiple drive stimuli. Here, we demonstrate highly selective non-reciprocal inter-modal EO scattering enabled by a single high-index radiofrequency (RF) traveling wave stimulus. Our experimental demonstration is performed on a thin-film lithium niobate integrated photonics platform, in which we engineer a slow-wave radiofrequency (SWRF) transmission line with an effective RF index > 9 that natively generates the required RF momentum while simultaneously maintaining strong RF-optical mode overlap. By additionally engineering the interaction length, we achieve a directional ~20 dB non-reciprocal scattering contrast. The SWRF architecture provides a scalable route to magnetic-free non-reciprocity and establishes momentum-engineered RF waves as a powerful tool for next-generation, fully integrated non-reciprocal photonic systems.

Silicon carbide (SiC) is a promising platform for scalable quantum technologies owing to its well-established, wafer-scale industrial processing. SiC also hosts a variety of optically active color centres including the nitrogen vacancy defect that has a spin-triplet ground state. However, strong phonon coupling in the infrared range limits photon extraction from these defects. Here, we use nanophotonic structures, specifically micro-pillar and micro-disk resonators, to enhance optical collection and spin-readout. The micro-pillar geometry yields a 4-fold increase in photon collection, accompanied by a 2.4-fold reduction in spectral noise in optically detected magnetic resonance measurements. Consequently, the magnetic field sensitivity is improved by 24%. The large mode volume of the micro-disk supports resonances spanning 1150-1250 nm, enabling broadband coupling to nitrogen vacancy emission lines. Our results demonstrate that fabrication of scalable photonic structures efficiently improves performance of silicon carbide color centers for integrated quantum light generation and sensing.

Human mobility plays a crucial role in the spread of human diseases, but is rarely quantified in plant disease epidemics. To address this gap, we integrate a unique, high-resolution network of human movements in New Zealand with a metapopulation model to mechanistically simulate pathogen transmission. We calibrate the model on the nationwide 2010 kiwifruit vine disease (Psa-V) outbreak, and show that it accurately reproduces the observed spatiotemporal spread, confirming that the human mobility network is a strong foundation for modeling transmission dynamics. By analyzing spatial infection trends, we find that most dispersal occurs locally, as often illustrated in the plant-outbreak literature. However, sporadic long-range connections are necessary to model a nationwide outbreak. Using the model as an in-silico laboratory, we demonstrate that enhanced surveillance accelerates detection and that outbreak severity is highly sensitive to the timing and location of initial disease importation. We observe a potential causal link between seasonal labor patterns and epidemic risk in high-traffic seasons. This study provides a robust, data-driven framework for modeling and predicting the spatiotemporal spread of agricultural pathogens. It underscores the importance of leveraging human mobility networks to design timely interventions and surveillance systems, protecting global food security.

Thermophotovoltaic (TPV) cells are increasingly attractive for applications in industrial waste heat harvesting, aerospace energy management, and compact power generation. Deploying midwave-infrared (MWIR) TPV in practical applications requires narrow-bandgap semiconductors that not only absorb low-energy photons but also integrate with scalable, low-cost platforms. Although high-performance TPV devices have been demonstrated using III-V materials such as InAs, GaSb, and InGaAs(P), their use remains limited by cost and substrate size. With this perspective, narrow bandgap GeSn alloys are a promising alternative that extend group-IV absorption into the MWIR while being silicon-compatible. Although the potential of GeSn TPV cells has been predicted, no experimental demonstration has been reported. Here, proof-of-concept Ge$_{0.91}$Sn$_{0.09}$ p-i-n TPV diodes (1 mm diameter) grown on silicon were fabricated and their performance was benchmarked against commercial InAs and extended-InGaAs devices. Measurements at 300 K under 2.33 $μ$m laser and $\sim$1500 K SiC Globar illumination revealed peak responsivity of $\sim$ 0.2 A/W at $\sim$ 1.7 $μ$m, and an output power of $\sim$ 0.41 mW/cm$^2$. These devices show trends comparable to those of the InAs diode under identical conditions, although at reduced absolute levels. To assess the intrinsic performance potential, Poisson-drift-diffusion modeling incorporating experimentally calibrated emitter emissivity predicts power densities exceeding 1 W/cm$^2$ under moderate MWIR thermal illumination, indicating that the present devices operate far below their fundamental limits and are primarily constrained by defect-assisted recombination and transport losses. These results establish GeSn as a scalable, silicon-compatible MWIR TPV platform and highlight a larger performance potential achievable through material and device optimization.

We study the use of separate cores of a multicore erbium-doped fiber amplifier (MC-EDFA) in a noise-canceled link for ultrastable optical frequency transfer. We demonstrate fractional frequency instability of $5\times10^{-19}$ at 1000 s averaging time for the stabilized MC-EDFA alone and $1.4\times10^{-18}$ at 1000 s averaging time when integrated with a 40 km-long 7-core spooled fiber. This study further establishes multicore fiber (MCF) networks as a promising platform for ultrastable frequency transfer, serving as an important step toward incorporating precision time and frequency distribution into future MCF communication infrastructures.

Occupational hierarchies remain strikingly stable even as job content changes rapidly. We ask whether skill requirements propagate directionally along the wage hierarchy or follow symmetric diffusion. Using O*NET 2015-2024, we analyze 17.3 million directed diffusion opportunities linking 873 occupations and 161 skills. We show that propagation obeys an Asymmetric Trajectory Channeling (ATC) rule: the same requirement spreads differently upward and downward, and the asymmetry depends on skill domain and on the architecture of skill dependencies. Two mechanisms generate ATC. Directional incorporation asymmetry implies that wage gradients create distinct receiving environments: upward-moving socio-cognitive requirements encounter complementary infrastructure, whereas upward-moving sensory/physical requirements face structural indifference. Structural portability constraints imply that dependency position governs portability: requirements anchoring long prerequisite chains carry co-adoption burdens that restrict diffusion regardless of destination. Consistent with these mechanisms, socio-cognitive requirements propagate upward more often than downward (20.7% vs 14.9%), while sensory/physical requirements exhibit the mirror pattern (19.5% downward vs 10.3% upward). Nestedness amplifies these asymmetries in opposite directions: scaffolding capabilities ascend most readily, whereas structurally embedded physical requirements are most tightly confined. Identification leverages within-occupation variation in propagation direction, and results are robust to origin- and destination-side specifications. Together, these findings reveal a directional architecture of occupational change that can reproduce hierarchy through ongoing reconfiguration, even absent assortative preferences or coordinated action.

The design of RFIC often involves exploring a large number of design variations in an invariant background composed of the processing stack and unchanged circuit blocks. Conventional electromagnetic solvers require a full-domain simulation for every design variation. In this work, we present a fast method that effectively separates the variant components from the invariant background. It algebraically decomposes the total field solution into the contributions from the design-dependent variations and the invariant background. Hence, the field response due to the invariant background can be simulated once and reused for all design variations. Only the variant components need to be simulated at each design variation, the size of which is small. We also develop an efficient way of reusing the model of each component and fusing them accurately to obtain the model of a system composed of many components. The reduced system of variant components involves computing the field solutions in the invariant background due to all possible sources located at variant components, the number of which can be large. We develop a fast algorithm to reduce them to a few field solutions, the number of which is on the order of the layer number. The proposed method has been applied to RFIC design space exploration. Its accuracy, robustness, and efficiency have been demonstrated.

We use ab initio electronic-structure methods to investigate random-matrix theory (RMT) universality in molecular electronic structure. Using single-reference electronic structure methods, including Hartree-Fock, configuration-interaction singles (CIS), density functional theory, and linear-response time-dependent density-functional theory, we compute single-particle orbital energies and many-electron excitations of several representative molecules (benzene, alanine, 1-phenylethylamine, methyloxirane, and helicene chains). For generic low-symmetry geometries, the unfolded spectra of these ab initio Hamiltonians exhibit Wigner-Dyson level statistics of the Gaussian orthogonal ensemble (GOE). For extended helicene chains we explicitly restrict to bound valence excitations below the ionization threshold and still observe GOE statistics, indicating that the RMT universality is present for physical states of direct relevance to real molecules. We further explore the electric and magnetic field dependence of the molecular electronic spectra. The variance of electric polarizability (level curvature K) is predicted to be non-analytic in the magnetic field which serves as an infrared cutoff, <K^2> proportional to log(1/|B|). We observe a transition to the Gaussian unitary ensemble (GUE) by increasing the magnetic fields, although it occurs only at magnetic fields far beyond experimentally accessible scales. Our results indicate that random matrix universality provides a general framework for organizing ab initio predictions of interacting electron spectra in complex systems.

We study massless spin-1 field propagation in a static, circularly symmetric $(2+1)$-dimensional wormhole with spatial Lorentz-violating anisotropy characterized by the throat radius $a$ and deformation parameter $η$. The geometry is horizon-free, geodesically complete, and asymptotically flat, with negative Gaussian curvature localized near the throat. Using the fully covariant vector boson formalism and covariant Maxwell theory, we derive an exact Schrödinger-type radial equation with a curvature-induced effective potential. Recasting the dynamics in Helmholtz form yields an effective refractive-index profile, showing that the wormhole acts as an inhomogeneous optical medium with position-dependent refractive index and frequency-dependent confinement, where low-frequency modes are strongly trapped while high-frequency modes propagate almost freely. A differential-geometric correspondence with helicoidal surfaces is established via $1/[a^2(1-η)] \leftrightarrow w^2$, demonstrating that Lorentz-violation-induced curvature is mathematically equivalent to curvature generated by geometric twist and linking the model to twisted graphene nanoribbons as analog-gravity platforms. These results provide a geometric framework for curvature-driven localization, dispersion, and anisotropic wave propagation in topologically nontrivial $(2+1)$-dimensional backgrounds.

Metagenomics has lowered the barrier to microbial discovery--enabling the identification of novel microbes without isolation--but cultures remain imperative for the deep study of microbes. Cultivation and isolation of non-model microbes remains a major challenge, despite advances in high-throughput culturomic methods. The quantity of simultaneous experimental variables is constrained by time and resources, but the list can be reduced using computational biology. Given an annotated genome, metabolic modelling can be used to predict source nutrients required for the growth of a microbe, which acts as an initial screen to inform culture and isolation experiments. This chapter provides an overview of metabolic networks and modelling and how they can be used to predict the nutrient requirements of a microorganism, followed by a sample protocol using a toy metabolic network, which is then expanded to a genome-scale metabolic network application. These methods can be applied to any metabolic network of interest--which in turn can be created from any genome of interest--and are a starting point for experimental validation of source nutrients required for microorganisms that remain uncultivated to date.

The difficulty with control of population dynamics, besides the states being positive and the control having to also be positive, is the extreme difference in the dynamics near extinction and at overpopulated states. As hard as global stabilization is, even harder is finding CLFs that are strict, don't require LaSalle arguments, and permit quantification of convergence. Among the three canonical types of two-population dynamics (mutualism, which borders on trivial, predator-prey, and competition, which makes global stabilization with positive harvesting impossible), predator-prey is the ``sweet spot'' for the study of stabilization. Even when the predator-prey interaction is neutrally stable, global asymptotic stabilization with strict CLFs has proven very difficult, except by conservative, hard-to-gain-insight-from Matrosov-like techniques. In this little note we show directions for the design of clean, elegant, insight-bearing, majorization-free strict CLFs. They generalize the classical Volterra-style Lyapunov functions for population dynamics to non-separable Volterra-style constructions. As a bonus to strictification as an analysis activity, we provide examples of concurrent designs of feedback and CLFs, using customized versions of forwarding and backstepping (note that, in suitable coordinates, predator-prey is both strict-feedforward and strict-feedback), where the striking deviations from these methods' conventional forms is necessitated by the predator-prey's states and inputs needing to be kept positive.

Data-driven model discovery (DDMD) algorithms are powerful tools for extracting interpretable symbolic models from data. However, identifying the model that best balances goodness-of-fit and sparsity is often a laborious process requiring user fine-tuning, is prone to overfitting, and results may significantly vary depending on model initialization and specific training procedure. Here, we present a sparse regression algorithm that automatically and adaptively generates candidate models, and uses a novel sample-length-scaling logarithmic information criterion (SLIC) to identify the best model from these candidates. We demonstrate that SLIC greatly outperforms other popular information criteria in extracting the correct model from the data of several nonlinear ordinary and partial differential equations. Then, we demonstrate SLIC's ability to discover interpretable models from experimental datasets in fluid dynamics and nanotechnology that generate new testable predictions.

Future wireless networks, deploying thousands of antenna elements, may operate in the radiative near-field (NF), enabling spatial multiplexing across both angle and range domains. Sparse arrays have the potential to achieve comparable performance with fewer antenna elements. However, fixed sparse array designs are generally suboptimal under dynamic user distributions, while movable antenna architectures rely on mechanically reconfigurable elements, introducing latency and increased hardware complexity. To address these limitations, we propose a reconfigurable array thinning approach that selectively activates a subset of antennas to form a flexible sparse array design without physical repositioning. We first analyze grating lobes for uniform sparse arrays in the angle and range domains, showing their absence along the range dimension. Based on the analysis, we develop two particle swarm optimization-based strategies: a grating-lobe-based thinned array (GTA) for grating- lobe suppression and a sum-rate-based thinned array (STA) for multiuser sum-rate maximization. Simulation results demonstrate that GTA outperforms conventional uniform sparse arrays, while STA achieves performance comparable to movable antennas, thereby offering a practical and efficient array deployment strategy without the associated mechanical complexity.

Conventional far-field multiple-input multiple-output (MIMO) channels are limited to a single spatial degree of freedom (DoF) under a line-of-sight (LoS) condition. In contrast, the radiative near field (NF) supports multiple spatial DoF, enabled by spherical wavefronts and the reduced spatial footprint at short ranges. While recent research indicates that the effective DoF (EDoF) increases in NF, experimental validation and clear identification of the transition distances remain limited. In this letter, we develop an intuitive framework for characterizing the EDoF of a ULA-based MIMO system and derive two complementary analytical expressions: a closed-form formulation that relates the EDoF to the physical transmit beamwidth and receive aperture, and a discrete formulation based on the discrete Fourier transform (DFT) domain angular decomposition of the NF spherical wavefront, which is well suited for experimental evaluation. We further introduce the effective MIMO Rayleigh distance (EMRD) and the maximum spatial multiplexing distance (MSMD), which mark the distances where the EDoF reduces to one and attains its maximum, respectively. Experimental measurements using widely spaced phased arrays closely match the theoretical EDoF trends and validate the proposed distance metrics.

With the deployment of large antenna arrays at high-frequency bands, future wireless communication systems are likely to operate in the radiative near-field. Unlike far-field beam steering, near-field beams can be focused on a spatial region with a finite depth, enabling spatial multiplexing in the range dimension. Moreover, in the line-of-sight MIMO near-field, multiple spatial degrees of freedom (DoF) are accessible, akin to a scattering- rich environment. In this paper, we derive the beamdepth for a generalized uniform rectangular array (URA) and investigate how the array geometry influences near-field beamdepth and its limits. We define the effective beamfocusing Rayleigh distance (EBRD), to present a near-field boundary with respect to beamfocusing and spatial multiplexing gains for the generalized URA. Our results demonstrate that under a fixed element count constraint, the array geometry has a strong impact on beamdepth, whereas this effect diminishes under a fixed aperture length constraint. Moreover, compared to uniform square arrays, elongated configurations such as uniform linear arrays (ULAs) yield narrower beamdepth and extend the effective near-field region defined by the EBRD. Building on these insights, we design a polar codebook for compressed-sensing-based channel estimation that leverages our findings. Simulation results show that the proposed polar codebook achieves a 2 dB NMSE improvement over state-of-the-art methods. Additionally, we present an analytical expression to quantify the effective spatial DoF in the near-field, revealing that they are also constrained by the EBRD. Notably, the maximum spatial DoF is achieved with a ULA configuration, outperforming a square URA in this regard.

Restoring force surface (RFS) methods offer an attractive nonparametric framework for identifying nonlinear restoring forces directly from data, but their reliance on complete kinematic measurements at each degree of freedom limits scalability to multidimensional systems. The aim of this paper is to overcome these measurement limitations by proposing an identification framework with relaxed sensing requirements that exploits periodic multisine excitation. Starting from an initial linear model, a sliding-window feedback approach reconstructs latent states and nonlinear restoring forces nonparametrically, enabling identification of the nonlinear component through linear-in-parameters regression instead of highly non-convex optimization. Validation on synthetic and experimental datasets demonstrates high simulation accuracy and reliable recovery of physical parameters under partial sensing and noisy conditions.

Portable ultra-low field MRI (ULF-MRI) systems operated in unshielded environments are susceptible to electromagnetic interference (EMI). Subject presence in the imaging region will lead to substantial noise increases, yet the dominant coupling mechanism remains insufficiently characterized. We develop a lumped-parameter circuit model of the coupled environment-body-receiver system. The model indicates that ambient time-varying electric fields induce a body common-mode potential, which is converted into differential-mode noise through capacitive imbalance between the head and the receive-coil terminals, yielding strong dependence on subject position and geometry. Circuit analysis, simulations, and controlled experiments support the model, with predicted imbalance consistent with measured noise variations. Guided by this mechanism, we implement a capacitive low-impedance bypass to clamp the body potential, achieving an approximately 3.5-fold SNR improvement on a 50 mT prototype. The proposed model offers a compact circuit-based tool for analyzing and mitigating human body-coupled electric-field interference in portable ULF-MRI.

By leveraging the distributed leakage radiation of leaky coaxial cables (LCXs), the concept of pinching antennas can be generalized from the conventional high-frequency waveguide based architectures to cable based structures in lower-frequency scenarios. This paper investigates an LCX based generalized pinching-antenna system with dual-port feeding. By enabling bidirectional excitation along each cable, the proposed design significantly enhances spatial degrees of freedom. A comprehensive channel model is developed to characterize intra-cable attenuation, bidirectional phase progression, slot based radiation, and wireless propagation. Based on this model, both analog and hybrid beamforming frameworks are studied with the objective of maximizing the minimum achievable data rate. For analog transmission, slot activation, port selection, and power allocation are jointly optimized using matching theory, coalitional games, and bisection based power control. For hybrid transmission, zero-forcing (ZF) digital precoding is incorporated to eliminate inter-user interference, thereby simplifying slot activation and enabling closed-form optimal power allocation. Simulation results demonstrate that dual-port feeding provides notable performance gains over single-port LCX systems and fixed-antenna benchmarks, validating the effectiveness of the proposed beamforming and resource allocation designs under various transmit power levels and cable parameters.

In this letter, a novel superimposed pilot scheme is proposed for channel estimation in multi-antenna orthogonal time frequency space (OTFS) receivers. Under the assumption of a large uniform linear array (ULA) size at the receiver, the multipath components are separated directly in the angular domain. It is then shown that the proposed superimposed pilot scheme enables the computation of integrated delay and Doppler profiles by averaging the received delay-Doppler matrix across the Doppler and delay axes, respectively. This procedure helps reduce data-to-pilot interference through data averaging. Moreover, it is demonstrated that fractional delays and Dopplers of the multipath components can be estimated by correlating the integrated delay and Doppler profiles with the corresponding delay/Doppler terms. Simulation results show that the proposed approach outperforms existing OTFS superimposed pilot schemes, achieving a lower bit error rate (BER) while exhibiting a trade-off between peak-to-average power ratio (PAPR) and communication performance.

We investigate network availability (NA) in aerial heterogeneous networks (AHetNets) for effective emergency rescue, where diverse delay-constrained communication services must be provided to user equipments (UEs) with varying mobility. The heterogeneity in delay constraints and UE mobility introduces resource allocation conflicts and imbalances, which undermine communication reliability and challenge NA. Although unified resource allocation (URA) can mitigate these issues, it remains unclear whether NA can be sustained under such diverse conditions. To address this, we derive expressions for the lower bound (LB) on NA in AHetNets under URA. Our analysis reveals that extended heterogeneity significantly degrades the LB due to resource limitations-even when the heterogeneity stems from additional services under less stringent delay constraints (LSDC) or from UEs with lower mobility. To overcome this degradation, we formulate and solve a joint optimization problem for the number of UEs sharing time-frequency resources ($K$) and pilot length ($ξ$), aiming to enhance the LB by improving spatial, frequency, and temporal resource efficiency. Simulation results validate our analysis and demonstrate that jointly optimizing $K$ and $ξ$ enables AHetNets to achieve the target NA under greater heterogeneity, outperforming existing resource allocation policies.

A recurring theme in optimal controller design for wireless networked control systems (WNCS) is the reliance on real-time channel state information (CSI). However, acquiring accurate CSI a priori is notoriously challenging due to the time-varying nature of wireless channels. In this work, we propose a pilot-free framework for optimal control over wireless channels in which control commands are generated from plant states together with control-aided channel prediction. For linear plants operating over an orthogonal frequency-division multiplexing (OFDM) architecture, channel prediction is performed via a Kalman filter (KF), and the optimal control policy is derived from the Bellman principle. To alleviate the curse of dimensionality in computing the optimal control policy, we approximate the solution using a coupled algebraic Riccati equation (CARE), which can be computed efficiently via a stochastic approximation (SA) algorithm. Rigorous performance guarantees are established by proving the stability of both the channel predictor and the closed-loop system under the resulting control policy, providing sufficient conditions for the existence and uniqueness of a stabilizing approximate CARE solution, and establishing convergence of the SA-based control algorithm. The framework is further extended to nonlinear plants under general wireless architectures by combining a KalmanNet-based predictor with a Markov-modulated deep deterministic policy gradient (MM-DDPG) controller. Numerical results show that the proposed pilot-free approach outperforms benchmark schemes in both control performance and channel prediction accuracy for linear and nonlinear scenarios.

This paper investigates a dual-hop joint visible light communication (VLC) and backscatter communication (BC) relaying framework under the finite blocklength (FBL) constraint, aiming at energy-neutral Ambient Internet of Things (A-IoT) deployments. In the proposed system, indoor LED access points are used to simultaneously provide illumination and transmit information over light to a backscatter device (BD), which harvests optical energy and backscatters the received messages to user equipments (UEs) equipped with radio frequency (RF) front ends. This forwarding of the information from VLC to RF channels is implemented without the need for carrier synthesizers and power amplifiers at the IoT node. By modeling the end-to-end communication link with short-packet IoT traffic and realistic levels of interference between adjacent VLC coverage areas, we analyze the outage performance and achievable data rate of the proposed system. Simulation results demonstrate that key factors, such as placement and orientation of the BD, as well as the selected code rate of the system affect reliability and data rate that can be achieved for communication purposes. The insights gained from this study pave the way for ambient power-enabled IoT solutions and future hybrid VLC/RF network designs.

The sixth generation (6G) network is expected to deploy larger multiple-input multiple-output (MIMO) arrays to support massive connectivity, which will increase overhead and latency at the physical layer. Meanwhile, emerging 6G demands such as immersive communications and environmental sensing pose challenges to traditional signal processing. To address these issues, we propose the ``semantic-aware MIMO'' paradigm, which leverages specialist models and large models to perceive, utilize, and fuse the inherent semantics of channels and sources for improved performance. Moreover, for representative MIMO physical-layer tasks, e.g., random access activity detection, channel feedback, and precoding, we design specialist models that exploit channel and source semantics for better performance. Additionally, in view of the more diversified functions of 6G MIMO, we further explore large models as a scalable solution for multi-task semantic-aware MIMO and review recent advances along with their advantages and limitations. Finally, we discuss the challenges, insights, and prospects of the evolution of specialist models and large models empowered semantic-aware MIMO paradigms.

Millimeter-wave massive multiple-input multiple-output systems employ highly directional beamforming to overcome severe path loss, and their performance critically depends on accurate beam alignment. Conventional codebook-based methods offer low training overhead but suffer from limited angular resolution and sensitivity to hardware impairments. To address these challenges, we propose a deep learning-enhanced super-resolution beam alignment framework with three key components. First, we design the Quaternary Search-based Super-Resolution (QSSR) algorithm, which leverages the monotonic power ratio property between two discrete Fourier transform (DFT) codebook beams to achieve super-resolution angle estimation without increasing measurement complexity relative to binary search. Second, we develop QSSR-Net, a gated recurrent unit-based neural network that exploits sequential multi-layer beam measurements to capture angular dependencies, thereby improving estimation accuracy, robustness to noise, and generalization across diverse propagation environments. Third, to mitigate the adverse effects of hardware impairments such as antenna position and phase errors, we propose a parametric self-calibration method that requires no additional hardware overhead and adapts compensation parameters in real time. Simulation results show that the proposed framework consistently outperforms binary search and even exhaustive search at high signal-to-noise ratios, achieving substantial performance gains while maintaining low overhead.

Accurate channel state information (CSI) is vital for multiple-input multiple-output (MIMO) systems. However, superimposed pilots (SIP), which reduce overhead, introduce severe pilot contamination and data interference, complicating joint channel estimation and data detection. This paper proposes a conditional flow matching receiver (CFM-Rx), an unsupervised generative framework that learns directly from received signals, eliminating the need for labeled data and improving adaptability across diverse system settings. By leveraging flow-based generative modeling, CFM-Rx enables deterministic, low-latency inference and exploits model invertibility to capture the bidirectional nature of signal propagation. This framework unifies flow matching with score-based diffusion modeling via a moment-consistent ordinary differential equation (ODE), replacing stochastic differential equation (SDE) sampling with a deterministic and efficient process. Furthermore, it integrates receiver-side priors to ensure stable, data-consistent inference. Extensive simulation results across various MIMO configurations demonstrate that CFM-Rx consistently outperforms conventional estimators and state-of-the-art data-driven receivers, achieving notable gains in channel estimation accuracy and symbol detection robustness, particularly under severe pilot contamination.

This paper aims to design multiple access (MA) schemes to improve the max-min fairness (MMF) for pinching antennas (PAs)-based multigroup multicast communications, where PA placement and resource allocation are jointly optimized. Specifically, three MA schemes are considered to facilitate the multicast transmission: i) treating interference as noise (TIN), ii) non-orthogonal multiple access (NOMA), and iii) time-division multiple access (TDMA) with two PA reconfiguration protocols, namely pinching switching (PS) and pinching multiplexing (PM). i) For TIN, a closed-form solution is derived for optimal power allocation, while a sequential element-wise optimization (SEO) is developed for the PA placement. ii) For NOMA, a recursive power allocation framework incorporating a bisection search is developed, and a hierarchical objective evaluation (HOE) mechanism is incorporated to simplify the SEO process for PA location update. iii) For TDMA, the PS protocol allows the PA locations to be optimized separately using the SEO method, after which the time-power allocation is solved as a convex problem with a global optimum. Under the PM protocol, the PA locations are jointly optimized with the time-power resources through a Karush-Kuhn-Tucker (KKT)-based analytical solution. Numerical results demonstrate that: i) the pinching-antenna system (PASS) architecture significantly outperforms traditional fixed-antenna systems. ii) TDMA-PS achieves superior performance by fully leveraging the flexible PA reconfiguration and benefiting from interference-free transmission, whereas TIN serves as a practical lower-bound solution due to its simplicity despite its limited performance. iii) NOMA consistently outperforms TDMA-PM and, in high transmit power regimes with heterogeneous multicast group distributions, can even surpass the performance achieved by TDMA-PS.

It has been shown that WiFi devices enable sensing of environments and targets through their channel state information. However, the same devices have not been used for delay-synchronous channel sounding due to challenges related to the stability of synchronization and lack of reference power levels. Due to factors such as uncertainty in symbol reception timing, impulse responses are discontinuous across acquisitions. The present paper addresses the challenges to perform delay-synchronous channel sounding using off-the-shelf multiple-antenna IEEE 802.11ax WiFi devices, referred to as SoundiFi. Stable delay synchronization and power level reference are realized by remoting the antennas with coaxial cables and devoting one of the antennas as a reference channel, with which the gain and delay of other simultaneous channels are defined. Indoor experiments confirmed that the impulse response becomes continuous across successive acquisitions and provide the absolute delay. The impulse response has a noise level at -115 dB, indicating the maximum path gain value that can be measured with the devices. The impulse response also revealed the existence of long-delayed multipaths up to 132 m propagation distance in a reverberant 30-m-long corridor.

Extremely large antenna arrays and high-frequency operation are two key technologies that advance performance metrics such as higher data rates, lower latency, and wider coverage in sixth-generation communications. However, the adoption of these technologies fundamentally changes the characteristics of wavefronts, forcing communication systems to operate in the near-field region. The transition from planar far-field communications to spherical near-field propagation necessitates novel channel estimation algorithms to fully exploit the unique features of spherical wavefronts for advanced transceiver design. To this end, we propose a novel semi-gridless channel estimation approach based on a variational Bayesian (VB) inference framework. Specifically, we reformulate the near-field channel model for both uniform linear arrays and uniform planar arrays into separate direction-of-arrival (DoAs) and distance components. Building on these new representations, we employ a gridless approach for DoAs estimation using a von Mises distribution, and a coarse-to-fine grid search for distance estimation. We then develop a semi-gridless variational Bayesian (SG-VB) algorithm with efficient update rules that enables accurate channel reconstruction. Simulation results validate the effectiveness of the proposed SG-VB algorithm, demonstrating enhanced near-field channel reconstruction accuracy and superior estimation performance for both DoAs and distance components embedded in near-field channels.

An index of an effective number of variables (ENV) is introduced for model selection in nested models. This is the case, for instance, when we have to decide the order of a polynomial function or the number of bases in a nonlinear regression, choose the number of clusters in a clustering problem, or the number of features in a variable selection application (to name few examples). It is inspired by the idea of the maximum area under the curve (AUC). The interpretation of the ENV index is identical to the effective sample size (ESS) indices concerning a set of samples. The ENV index improves {drawbacks of} the elbow detectors described in the literature and introduces different confidence measures of the proposed solution. These novel measures can be also employed jointly with the use of different information criteria, such as the well-known AIC and BIC, or any other model selection procedures. Comparisons with classical and recent schemes are provided in different experiments involving real datasets. Related Matlab code is given.

Traditional protection systems for microgrids, which rely on high fault currents and continuous communication, struggle to keep up with the changing dynamics and cybersecurity concerns of decentralized networks. In this study, we introduce a novel biologically inspired protection system based on neuromorphic principles, where each distributed energy resource (DER) functions as a simple neuron. These neurons process local changes in voltage, current signals, and converting them into spike patterns that represent the severity of disturbances. Just as neurons communicate via synapses in biological systems, we exploit transmission cables to coordinate between DERs, enabling them to share information and respond to faults collectively. Fault detection and circuit breaker activation are driven by a First-To-Spike (FTTS) mechanism, similar to the concept of traveling wave protection, but without needing GPS synchronization or communication links. A key innovation is the ability to use the timing of spikes to locally determine the nature of a fault, offering an intelligent, adaptive response to disturbances. Performance shows tripping latency of 10-58 ms, surpassing conventional relays and even traveling-wave methods (60 ms), while maintaining detection accuracy above 98% and spatial selectivity over 97%, enabling real-time, communication-free, scalable protection for plug-and-play microgrids.

Continuous signal representations are naturally suited for inverse problems, such as magnetic resonance imaging (MRI) and computed tomography, because the measurements depend on an underlying physically continuous signal. While classical methods rely on predefined analytical bases like B-splines, implicit neural representations (INRs) have emerged as a powerful alternative that use coordinate-based networks to parameterize continuous functions with implicitly defined bases. Despite their empirical success, direct comparisons of their intrinsic representation capabilities with conventional models remain limited. This preliminary empirical study compares a positional-encoded INR with a cubic B-spline model for continuous function fitting from sparse random samples, isolating the representation capacity difference by only using coefficient-domain Tikhonov regularization. Results demonstrate that, under oracle hyperparameter selection, the INR achieves a lower normalized root-mean-squared error, yielding sharper edge transitions and fewer oscillatory artifacts than the oracle-tuned B-spline model. Additionally, we show that a practical bilevel optimization framework for INR hyperparameter selection based on measurement data split effectively approximates oracle performance. These findings empirically support the superior representation capacity of INRs for sparse data fitting.

High-rate quantum error correcting (QEC) codes encode many logical qubits in a given number of physical qubits, making them promising candidates for quantum computation. Implementing high-rate codes at a scale that both frustrates classical computing and improves performance by encoding requires both high fidelity gates and long-range qubit connectivity -- both of which are offered by trapped-ion quantum computers. Here, we demonstrate computations that outperform their unencoded counterparts in the high-rate $[[ k+2,\, k,\, 2 ]]$ iceberg quantum error detecting (QED) and $[[ (k_2 + 2)(k_1 + 2),\, k_2k_1,\, 4 ]]$ two-level concatenated iceberg QEC codes, using the 98-qubit Quantinuum Helios trapped-ion quantum processor. Utilizing new gadgets for encoded operations, we realize this "beyond break-even" performance with reasonable postselection rates across a range of fault-tolerant (FT) and partially-fault-tolerant (pFT) component and application benchmarks with between $48$ and $94$ logical qubits. These benchmarks include FT state preparation and measurement, QEC cycle benchmarking, logical gate benchmarking, GHZ state preparation, and a pFT quantum simulation of the three-dimensional $XY$ model of quantum magnetism. Additionally, we illustrate that postselection rates can be suppressed by increasing the code distance via concatenation. Our results represent state-of-the-art logical component and state fidelities and provide evidence that high-rate QED/QEC codes are viable on contemporary quantum computers for near-term beyond-classical-scale computation.

Exceptional points, where two or more eigenstates of a non-Hermitian system coalesce, are now of interest across many fields of physics, from the perspective of open-system dynamics, sensing, nonreciprocal transport, and topological phase transitions. In this work, we investigate exceptional points in cavity optomechanics, a platform of interest to diverse communities working on gravitational-wave detection, macroscopic quantum mechanics, quantum transduction, etc. Specifically, we clarify the role of quantum jumps in making a clear distinction between Liouvillian and Hamiltonian exceptional points in optomechanical systems. While the Liouvillian exceptional point arises from the unconditional Lindblad dynamics and is independent of the phonon-bath temperature, the Hamiltonian exceptional point emerges from the conditional no-jump evolution and acquires a thermal shift due to an enhanced conditional damping. Employing the thermofield formalism, we derive a unified spectral framework that interpolates between these regimes via an analytical hybrid-Liouvillian description. Remarkably, in the weak-quantum-jump regime, the exceptional point is perturbed only at the second order, highlighting the robustness of the Hamiltonian exceptional point under small hybrid perturbations. Our work reveals a continuous family of hybrid exceptional points, clarifies the operational and physical differences between the conditional and unconditional dissipative dynamics in optomechanical systems, and provides a probe for thermal baths.

We develop a simple and systematic route to higher levels of the qubit Clifford hierarchy by coherently controlling Clifford operations. Our approach is based on Pauli periodicity, defined for a Clifford unitary $U$ as the smallest integer $m\ge 1$ such that $U^{2^{m}}$ is a Pauli operator up to phase. We prove a sharp controlled-jump rule showing that the controlled gate $CU$ lies strictly in level $m+2$ of the hierarchy, and equivalently that $CU$ lies in level $k$ if $U^{2^{k-2}}$ is Pauli while no smaller positive power of $U$ is Pauli. We further quantify the resources required to realize large level jumps in the Clifford hierarchy by proving an essentially tight upper bound on Pauli periodicity as a function of the number of qubits, which implies that accessing high hierarchy levels through controlled Cliffords requires a number of target qubits that grows exponentially with the desired level. We complement this limitation with explicit infinite families of Pauli-periodic Cliffords whose controlled versions achieve asymptotically optimal jumps. As an application, we propose a protocol for preparing logical catalyst states that enable logical $Z^{1/2^k}$ phase gates via phase kickback from a single jumped Clifford.

Sybil resistance is a key requirement of decentralized consensus protocols. It is achieved by introducing a scarce resource (such as computational power, monetary stake, disk space, etc.), which prevents participants from costlessly creating multiple fake identities and hijacking the protocol. Quantum states are generically uncloneable, which suggests that they may serve naturally as an unconditionally scarce resource. In particular, uncloneability underlies quantum position based-cryptography, which is unachievable classically. We design a consensus protocol that combines classical hybrid consensus protocols with quantum position verification as the Sybil resistance mechanism, providing security in the standard model, and achieving improved energy efficiency compared to hybrid protocols based on Proof-of-Work. Our protocol inherits the benefits of other hybrid protocols, namely the faster confirmation times compared to pure Proof-of-Work protocols, and resilience against the compounding wealth issue that plagues protocols based on Proof-of-Stake Sybil resistance. We additionally propose a spam prevention mechanism for our protocol in the Random Oracle model.

Single-shot fidelity is the standard benchmark for superconducting qubit readout, but it does not directly minimize the total wall-clock time required to certify a quantum state. We formulate an information-theoretic description of dispersive readout that treats the measurement record as a stochastic communication channel and compute the classical Chernoff information governing the multi-shot error exponent using a trajectory model that incorporates T1 relaxation with full cavity memory. We find a consistent separation between the integration time that maximizes single-shot fidelity and the time that minimizes total certification time. For representative transmon parameters and hardware overheads, the throughput-optimal integration window is longer than the fidelity-optimal one, yielding certification speedups of approximately 9-11%, with the gain saturating near 1.13x in the high-readout-power and high-overhead regime. Comparing the extracted classical information to the Gaussian Chernoff limit defines an information-extraction efficiency metric and shows that typical dispersive schemes are limited to about 45% capture at short integration times by detection efficiency, decreasing to approximately 12% at the throughput-optimal integration time of approximately 1.22 us due to T1-induced trajectory smearing. This formulation connects readout calibration directly to the operational objective of minimizing certification time in high-throughput superconducting processors.

Assuming that dark matter is an ultralight pseudoscalar particle which couples to electromagnetism like an axion (an ALP), we demonstrate that the coupling of the cosmological magnetic field produced by the ALP field oscillations to the primordial dark matter fluctuations yields a spectrum of gauge field fluctuations which can produce a sufficient flux of Lyman-Werner photons to enable the Direct Collapse Black Hole formation scenario. The induced flux is consistent with the bounds on the excess flux of radio photons from ARCADE2 and EDGES measurements.

Power consumption is a critical constraint for CubeSat based quantum communication, where tracking systems often dominate the onboard power budget. We demonstrate an energy-efficient approach that enables reliable satellite tracking at substantially reduced beacon power by treating tracking as a weak-signal estimation task. Using a closed-loop system with fine steering mirrors and higher-order Kalman filters on ground, we can maintain stable tracking at a transmitted power equivalent to 34 mW over a -60 dB satellite to ground optical channel. Our results show that the resulting penalties on QKD bit error rates and signal-to-noise ratios are negligible, allowing for more efficient power allocation to quantum payloads in CubeSat missions.

What are gauge-invariant local observables in a subregion in quantum gravity? How does one even define such a subregion non-perturbatively? We study these questions in JT gravity. One can define a subregion by specifying the value of the dilaton at the boundary of the region. We study conformal matter correlators in such a subregion. There is a gravitational constraint associated with York time evolution within the causal diamond of the subregion. This constraint can be leveraged to construct gauge-invariant observables in quantum gravity, using a crossed product construction. The extrinsic curvature of Cauchy slices acts as the physical clock. This is a simple example of how gauge-invariant observables can be obtained by dressing to features of a spacetime (or other fields), without the need for introducing an external observer. The entropy associated with this algebra of observables is not an area, or any boundary term. We show that gravitational constraints only give boundary formulas for entropy when gauging isometric diffeomorphisms. York time flow is merely a conformal isometry, not an actual isometry, and thus leads to bulk contributions to entropy. We repeat our construction for Milne-type closed Big-Bang universes, which may be of independent interest.

We study the existence and radial stability of static, spherically symmetric thin shells separating two Schwarzschild--de Sitter spacetimes with parameters $(m_\pm,Λ_\pm)$. Using the Israel junction formalism and a linear barotropic equation of state $p = λ(σ- σ_1)c^2$, we decouple the sound speed $c_s^2 = λc^2$ from the equilibrium equation-of-state parameter $w_0 \equiv p_0 / (σ_0 c^2)$ and derive the effective potential governing radial dynamics. For observationally motivated parameters, stable configurations with $σ_0>0$ and $0<λ\leq1$ exist only when $m_+/m_- > 1$. Three distinct stability windows emerge, when $λ=1$: $-3/7 \lesssim w_0 \lesssim 1/2$ for $Λ_+ = Λ_-$, $-2/3 \lesssim w_0 \lesssim 1/2$ for $Λ_+ > Λ_-$ and $0 \lesssim w_0 \lesssim 1$ for $Λ_+ < Λ_-$. Positive-pressure shells ($w_0>0$) reside near the photon sphere, whereas negative-pressure shells ($w_0<0$) extend outward, reaching either the cosmological horizon or the static radius. Stability relies on the variation of $w(σ)$ with the surface energy density. Negative pressure (tension) stabilizes the system because the tension increases during expansion. Conversely, positive pressure stabilizes the system because the pressure increases during contraction. Finally, a static, stable, dark, fluid thin shell acts as a gravitational refractive layer that enlarges the black hole's shadow for a distant, static observer outside the shell. The effect depends on the shell radius $R_0$, the background parameters $(m_{\pm}, Λ_{\pm})$, and the equation-of-state. Dark fluid shells can be considered as theoretical toy models that illustrate qualitative effects. Future high-resolution black hole shadow observations could, in principle, use such models to explore how different equations of state might influence observable signatures.

Learning properties of quantum states and channels is known to benefit from resources such as entangled operations, auxiliary qubits, and adaptivity, whereas the resource structure of measurement learning, namely, learning properties of quantum measurement operators, remains poorly understood. In this work, we identify a measurement learning task for which access limited to classical measurement outcomes leads to an exponential lower bound on the query complexity, established via a distinguishing task between a genuine quantum projective measurement and a purely classical random number generator. Remarkably, this hardness persists even when arbitrary entangled operations, auxiliary systems, and fully adaptive strategies are allowed, indicating that conventional resources for state and channel learning are ineffective in this task. In contrast, when access to the post-measurement quantum state is available, the same task can be solved with constant query complexity using a simple measuring-twice protocol, without requiring resources that are useful for state and channel learning. Our results reveal post-measurement states as a qualitatively new and decisive resource for measurement learning, suggesting potential implications for the design of practical quantum certification protocols.

Gauss's law-based quantum error correction (GLQEC) offers a promising approach to reducing qubit overhead in lattice gauge theory simulations by leveraging built-in symmetries. For applications of GLQEC to 1+1D lattice quantum electrodynamics (QED), we identify two significant trade-offs. First, we prove via dimension-counting arguments that GLQEC requires periodic electric fields, thereby constraining the design space for lattice QED simulations. Second, we numerically compare GLQEC with a universal quantum error correction (UQEC) code, specifically the $d=3$ bitflip repetition code, and find that while GLQEC can achieve lower logical error rates in single-round error correction, it exhibits faster decoherence to the steady-state mixed ensemble under multiple rounds. The mixing speed penalty is manifest in observables of interest for both memory experiments and Hamiltonian evolution. We identify a mixing speed threshold, $p_{th}=0.277(2)$, above which using GLQEC exhibits even faster decoherence than without error correction. Our results highlight fundamental limitations of symmetry-based error correction schemes and inform corresponding constraints on formulations of lattice gauge theories compatible with error-robust quantum simulation techniques.

Circuit quantum acoustodynamics (cQAD) devices have a wide range of applications in quantum science, all of which depend crucially on the quantum coherence of the mechanical subsystem. In this context, high-overtone bulk acoustic-wave resonators (HBARs) are particularly promising, since they have shown very high quality factors with negligible dephasing. However, the introduction of piezoelectric films, which are necessary for coupling to a superconducting circuit, can lead to additional loss channels, such as surface scattering and two-level systems (TLS). Here, we study the acoustic dissipation of HBAR resonators in cQAD systems and find that the defect density of the piezoelectric material and its interface with the bulk are limiting factors for the coherence. We measure acoustic modes with phonon lifetimes up to 400 $μ$s and lifetime-limited coherence times approaching one millisecond in the quantum regime. When coupled to a superconducting qubit, this leads to a hybrid system with a large quantum coherence cooperativity of $C_{T_2}=1.1\times10^5$. These results represent a new milestone for the performance of cQAD devices and offer concrete paths forward for further improvements.

We investigate the modifications to inflationary observables that arise when adopting an $α$-vacuum instead of the standard Bunch--Davies vacuum for quantum fluctuations during inflation. Within the Starobinsky inflationary model, we compute and compare the scalar spectral index, its running, and the running of the running arising from different choices of the initial vacuum state. We further examine the energy scales associated with $α$-vacua and argue that, for any number of extra spatial dimensions, the relevant scale can be truncated at the Hubble scale, $\sim$$\mathcal{O}(10^{13})\,\mathrm{GeV}$, without conflict with current Cavendish-type experimental bounds on sub-millimeter gravity ($\sim$$250\,μ\mathrm{m}$). Our analysis demonstrates that the $α$-vacuum is subject to stringent constraints as a viable de~Sitter-invariant alternative to the Euclidean (Bunch--Davies) vacuum, with the corrections that it induces in the inflationary observables being strongly limited by the latest Planck data.

This work presents a comprehensive investigation of the gravitational phenomena that correspond to a non-commutative (NC) charged black hole, by incorporating NC geometry through a Moyal twist. We derive the deformed metric up to the second order of the NC parameter, utilizing the Seiberg-Witten map for the Reissner-Nordstrom black hole. We explore how non-commutativity modifies key thermodynamic properties, such as the Hawking temperature and heat capacity, and the existence of a remnant mass at the final stage of evaporation. Additionally, the study of Hawking radiation for bosonic and fermionic particles is discussed. Applying a perturbative method, scalar quasinormal modes are analyzed numerically. Furthermore, null geodesics and photon sphere stability are explored via curvature and topological methods. The shadow radius and deflection angle are computed to understand observational signatures. Lensing observables are compared to Event Horizon Telescope observations to provide probable constraints on the non-commutativity parameter. This study bridges theoretical predictions with astrophysical observations, offering insights into quantum gravity effects on black hole physics.

Representing the time-evolution operator as a tensor network constitutes a key ingredient in several algorithms for studying quantum lattice systems at finite temperature or in a non-equilibrium setting. For a Hamiltonian composed of strictly short-ranged interactions, the Suzuki-Trotter decomposition is the main technique for obtaining such a representation. In [B.~Vanhecke, L.~Vanderstraeten and F.~Verstraete, Physical Review A, L020402 (2021)], an alternative strategy, the cluster expansion, was introduced. This approach naturally preserves internal and lattice symmetries and can more easily be extended to higher-order representations or longer-ranged interactions. We extend the cluster expansion to two-dimensional fermionic systems, and employ it to construct projected entangled-pair operator (PEPO) approximations of Gibbs states. We also discuss and benchmark different truncation schemes for multiplying layers of PEPOs together. Applying the resulting framework to a two-dimensional spinless fermion model with attractive interactions, we resolve a clear phase boundary at finite temperature.

Quantum communication technologies capable of operating reliably across heterogeneous optical channels are essential for scalable metropolitan quantum networks. Here we demonstrate high-dimensional time-bin-encoded quantum key distribution over a hybrid metropolitan link comprising 1.7 km free-space transmission and 685 m of optical fiber. Operating at a clock rate of 500 MHz in the C-band, we implement both 2- and 4-dimensional protocols, and obtain estimated secure finite-key rates of (95 +- 28) kbit/s for 4D at (25.0 +- 2.0) dB loss and (59 +- 27) kbit/s for 2D at (23.5 +- 2.3) dB loss. Crucially, we achieve continuous operation over 48 h in a fully self-referenced architecture: initial synchronization, interferometric phase stabilization, and long-term drift compensation are performed exclusively using the detected quantum signals, without auxiliary optical reference channels. Our results thus establish a practical and versatile platform for hybrid free-space-to-fiber quantum communication and show that the encoding dimensionality can be adapted to the optimal operating regime of realistic metropolitan channels, providing a pathway toward efficient, autonomous and deployable quantum network nodes.

The stochastic--quantum correspondence reinterprets quantum dynamics as arising from an underlying stochastic process on a configuration space. We generalize the correspondence by lifting an arbitrary stochastic kernel $Γ$ in finite dimension to a map $φ$ on $B(\mathcal H)$, formulating the associated lift-compatibility relation, and giving an explicit dictionary between $Γ$ and CPTP (Kraus) maps. We isolate Chapman--Kolmogorov divisibility of the lifted family as the decisive additional constraint: when a CK-consistent CPTP family exists, the lift admits a Lindblad master equation form. In this picture, off-diagonal (phase) degrees of freedom act as a compressed carrier of history dependence not fixed by transition kernels alone; conversely, the apparent emergence of quantum phase information from a phase-blind stochastic description is explained as a memory effect. Finally, we state and prove a divisibility criterion for the underlying stochastic kernels, expressed as a condition involving divisibility of the lifted map together with a diagonality requirement on the density operator.

Many traditional algorithms applied in gravitational-wave astronomy rely on the assumption of Gaussian noise, a condition not always met. To meet this need, this study extends a robust statistical framework, advancing previous work on heavy-tailed likelihoods, that adapts the hyperbolic likelihood method for full frequency domain applications. The framework is designed to maintain high performance under ideal conditions while improving robustness against non-Gaussian noise and outliers in real-world data. We demonstrate the efficacy of this approach through two key case studies. The first case study analyzes a massive black hole binary merger in simulated Laser Interferometer Space Antenna (LISA) data with Gaussian noise, showing that the extended hyperbolic likelihood method performs comparably to the more commonly used Whittle likelihood. The second case study examines a stellar-mass black hole binary merger using real ground-based gravitational-wave data containing non-Gaussian noise or overlapping signals, where our framework exhibits increased robustness and yields more accurate parameter estimations. Our results show that the hyperbolic likelihood better captures the true noise distribution, providing a flexible and physically motivated alternative for GW data analysis across current and future detectors.

Generative models for quantum data pose significant challenges but hold immense potential in fields such as chemoinformatics and quantum physics. Quantum denoising diffusion probabilistic models (QuDDPMs) enable efficient learning of quantum data distributions by progressively scrambling and denoising quantum states; however, existing implementations typically rely on circuit-based random unitary dynamics that can be costly to realize and sensitive to control imperfections, particularly on analog quantum hardware. We propose the chaotic quantum diffusion model, a framework that generates projected ensembles via chaotic Hamiltonian time evolution, providing a flexible and hardware-compatible diffusion mechanism. Requiring only global, time-independent control, our approach substantially reduces implementation overhead across diverse analog quantum platforms while achieving accuracy comparable to QuDDPMs. This method improves trainability and robustness, broadening the applicability of quantum generative modeling.

Quantum state and process tomography are typically analyzed under the assumption that devices emit independent and identically distributed (i.i.d.) states or channels. In realistic experiments, however, noise, drift, feedback, or adversarial behavior violate this assumption. We show that projected least-squares tomography remains statistically optimal even under fully adaptive state and channel preparation. Specifically, we prove that the sample complexity for reconstructing the time-averaged state or channel matches the optimal i.i.d. scaling for non-adaptive, single-copy measurements. For rank-$r$ states, the sample complexity is $\mathcal{O}(d r^2/ε^2)$ to achieve accuracy $ε$ in trace distance, while for process tomography it is $\mathcal{O}(d^6/ε^2)$ to achieve accuracy $ε$ in diamond distance. Thus, dropping the i.i.d. assumption does not increase the fundamental sample complexity of quantum tomography, but only changes the interpretation of the reconstructed object.

We study the propagation of a constant speed gaussian scalar field wave-packet (GWP) in Minkowski space, showing that the energy conditions are violated for superluminal speed. We then apply the effective speed approach to the GWP propagation, deriving the corresponding effective metric, effective Lagrangian and effective stress-energy tensor, showing that the null, weak and strong energy conditions are satisfied.

Open quantum systems host mixed-state phases that go beyond the symmetry-protected topological and spontaneous symmetry-breaking paradigms established for closed, pure-state systems. Developing a unified and physically transparent classification of such phases remains a central challenge. In this work, we introduce a purification-based framework that systematically characterizes all mixed-state phases in one-dimensional systems with $\mathbb{Z}_2^σ \times \mathbb{Z}_2^τ$ symmetry. By introducing an ancillary $κ$ chain and employing decorated domain-wall constructions, we derive eight purified fixed-point Hamiltonians labeled by topological indices $(μ_{στ},μ_{τκ},μ_{κσ}) \in \{\pm1\}^3$. Tracing out the ancilla recovers the full structure of mixed-state phases, including symmetric, strong-to-weak spontaneous symmetry breaking, average symmetry-protected topological phases, and their nontrivial combinations. Interpolations between the eight fixed points naturally define a three-dimensional phase diagram with a cube geometry. The edges correspond to elementary transitions associated with single topological indices, while the faces host intermediate phases arising from competing domain-wall decorations. Along the edges, we identify a class of critical behavior that connects distinct strong-to-weak symmetry-breaking patterns associated with distinct strong subgroups, highlighting a mechanism unique to mixed-state settings. Large-scale tensor-network simulations reveal a rich phase structure, including pyramid-shaped symmetry-breaking regions and a fully symmetry-broken phase at the cube center. Overall, our purification approach provides a geometrically transparent and physically complete classification of mixed-state phases, unified with a single $\mathbb{Z}_2^σ \times \mathbb{Z}_2^τ \times \mathbb{Z}_2^κ$ model.

We present a comprehensive study of the thermodynamic phase structure for Anti-de Sitter black holes in Einstein-Maxwell-power-Yang-Mills gravity, reformulated through holographic duality as an ensemble problem in the dual conformal field theory (CFT). By deriving an extended first law where the central charge \(C\) is a thermodynamic variable, we systematically explore both canonical and mixed ensembles. In the canonical ensemble with fixed charges, we identify a van der Waals-like phase transition between small and large black holes, marked by a characteristic swallowtail structure and coexistence curves with a negative slope. In contrast, within the mixed ensemble of fixed electric potential, the system exhibits a Hawking-Page transition between confined and deconfined phases of the boundary CFT. Our key finding is the suppressive role of the non-Abelian Yang-Mills charge \(\tilde{q}\): increasing \(\tilde{q}\) lowers both the minimum and the Hawking-Page transition temperatures, significantly narrowing the stability window of the confined phase. These results, supported by detailed numerical analysis, reveal a rich, ensemble-dependent phase landscape and establish the non-linear Yang-Mills sector as a critical controller of confinement physics in strongly coupled holographic systems.

The search for dark matter candidates includes primordial black holes (PBHs) as possible constituents. Recent studies show that some PBHs can survive to the present epoch by gaining angular momentum through Hawking radiating photons and becoming extremal before complete evaporation. While this provides a plausible model in a two-derivative theory of gravity, additional issues arise in EFT-corrected theories of gravity. In such theories, a rapidly spinning black hole can lead to extremely high tidal forces on a near-horizon observer, with possible observational consequences. In this work, by modeling Hawking radiation as a biased random walk within an EFT of gravity, we show that nearly the same fraction of PBHs survives as in GR. We argue that the resultant near horizon tidal effects should be detectable in future gravitational-wave observables.

We construct an exact spinning generalisation of the Morris-Thorne traversable wormhole supported by an anisotropic fluid. Within the Teo wormhole ansatz with unit lapse and Morris-Thorne shape function, we solve analytically for the frame-dragging function and obtain a two-parameter family of asymptotically flat solutions labelled by the throat radius $r_0$ and total angular momentum $J$. Curvature scalars and stress-energy components are given in closed form, showing a regular throat, equatorial reflection symmetry, and violations of all standard energy conditions, as required for traversable wormholes. We analyse the causal structure and show that, despite the presence of an ergoregion for sufficiently large $|J|$, the coordinate time defines a global temporal function, so the spacetime is stably causal and free of closed timelike curves. The optical appearance is studied via photon trajectories. The resulting shadows are smaller than Kerr's and depend on the wormhole shape. Finally, we compute the Geroch-Hansen multipole moments and find a massless but spinning configuration with distinctive higher multipoles that encode the throat scale.

We introduce the prodiabatic elimination, a powerful approximation technique that systematically extends the adiabatic elimination of fast degrees of freedom in light-matter coupled systems. Through a controlled expansion of operators, the prodiabatic elimination incorporates higher-order corrections and consistently includes noise contributions, leading to a significantly improved performance compared to standard adiabatic elimination. Importantly, it retains the simplicity and computational efficiency of the adiabatic elimination, making it convenient for practical applications. We demonstrate the approach on two setups: a driven dissipative Jaynes-Cummings model and a three-level system in a two-mode cavity that performs stimulated Raman adiabatic passage (STIRAP). These examples establish the prodiabatic elimination as a robust and broadly applicable tool for analyzing open quantum systems.

Qudits, or multi-level quantum information carriers, present a promising path for scaling quantum computers. However, their use introduces increased complexity in quantum logic, necessitating careful control of relative phases between different qudit levels. In trapped-ion systems, entangling operations accumulate phases on specific levels that are no longer global, unlike in qubit architectures. Furthermore, the structure of multi-level gates becomes increasingly intricate with higher-dimensional Hilbert spaces. This work explores the theory of these additional entangling and non-entangling phases, accumulated in Mølmer--Sørensen and Light-shift gates. We propose methods to actively compensate for these phases, enhance gate robustness against parameter fluctuations, and simplify native gates for more efficient circuit decomposition. Our results pave the way toward the practical and scalable implementation of qudit-based quantum processors.

We present a new analytic model of force-free electromagnetic jet launched from a disk-fed rotating black hole. The jet solution is obtained through a systematic construction from previously developed methods. The resulting physical jet solution exhibits an asymptotically parabolic structure and is parametrized by the location of localized current concentration and sign reversal in the disk. We find, however, that the jet properties show negligible dependence on the disk parameter. The black hole jet captures the basic feature of the Blandford-Znajek mechanism for energy extraction and jet formation.

Greenberger-Horne-Zeilinger (GHZ) state is a typical maximally entangled state which is pursued in both fundamental research and emerging quantum technologies. Preparing large-scale GHZ states in lattice spin systems is particularly appealing for quantum advantages, but conventional schemes face great challenges in scalability. Here we propose a universal and scalable scheme to generate large-scale GHZ-like states, which share similar entanglement and metrological properties with standard GHZ states, in lattice spin systems through global Floquet engineering. Our scheme requires only global operations and shows great advantage for large particle number. It is applicable to systems with arbitrary interaction ranges, offering a practical pathway for large-scale implementation of many-body entangled states in various systems.

Using the hotspot imaging method, this paper investigates the motion trajectory of plasma in the plunging region before and after the Comisso-Asenjo mechanism. Following a brief review of the magnetic reconnection process in the plunging region of a Kerr black hole, we introduce the hotspot model and imaging method. Based on numerical simulations, we separately study the hotspot images in the plunging region without magnetic reconnection, with magnetic reconnection, and when the escape condition is not met. We also compare these with hotspot images in the circular orbit region. The results show that for hotspot images without magnetic reconnection, if the plasma follow plunging orbits, the flare intensity gradually decreases, whereas if they follow circular orbits, the flare intensity remains nearly constant. Additionally, we find that in the plunging region, the signal for energy extraction is weaker compared to that in the circular orbit region.

The Glavan-Lin proposal for 4D Einstein-Gauss-Bonnet (EGB) gravity introduces a singular dimensional scaling to bypass Lovelock's theorem, though its fundamental origin remains debated. In this work, we demonstrate that this specific dimension-dependent scaling naturally emerges from the one-loop self-energy corrections of gravitons. By employing real-space techniques to evaluate graviton interactions with minimally coupled scalar and electromagnetic fields in a de Sitter background, we show that the $1/(D-4)$ pole universally generates a dynamical Gauss-Bonnet term. This confirms that the scaling is not an ad-hoc classical limit but a necessary consequence of quantum field-theoretic renormalization. Furthermore, canceling the remaining divergences strictly requires the inclusion of quadratic curvature counterterms, specifically Weyl-squared and $R^2$ invariants. We discuss the implications of this in the early-Universe and consequences in strong gravity regime.

We develop a dark-sector selective trace-coupled extension of gravity in which the matter--curvature coupling depends exclusively on the trace of the dark-matter energy--momentum tensor, $T_χ$, defined from a canonical dark-matter field $χ$. This construction provides a microphysically specified trace sector, removes the usual matter-Lagrangian ambiguity of $f(R,T)$-type models, and preserves minimal coupling of visible matter by design. We derive the full field equations, the exact dark-sector exchange structure, and the linear scalar-perturbation system in gauge-ready form. In the sub-horizon regime, we derive effective modified-gravity functions governing structure growth and light deflection, and show that the model generically produces correlated, scale- and time-dependent departures from General Relativity in growth and lensing observables. Building on this structure, we formulate a perturbation-focused multi-probe framework based on redshift-space distortions, weak lensing, and CMB lensing, explicitly targeting degeneracy breaking beyond background-expansion tests. The analysis establishes the action-level and perturbation-level foundations of the model and provides a conservative, reproducible framework for translated linear-regime constraints within a dark-sector selective modified-gravity setting.

Across diverse synthetic and real-world interaction graphs, the variational landscapes of reduced Quantum Approximate Optimization Algorithm (QAOA) instances obtained via variable freezing exhibit a robust universality. Leveraging this structure, we introduce Doubly Optimized QAOA (DO-QAOA), which lowers runtime and quantum measurement overhead while maintaining a competitive approximation ratio gap (ARG). Adapting the replica-overlap framework of spin-glass physics, we define a landscape-overlap order parameter $q$ to quantify geometric correlations between energy landscapes, revealing a sharp landscape-similarity transition as graph connectivity is tuned. Notwithstanding this transition, the dominant convex features of nearly all conditioned sub-instances remain aligned across both phases. Exploiting this persistence, DO-QAOA collapses the nominal $2^m$ reduced instances generated by freezing $m$ qubits into $K = O(1)$ effective landscape classes, eliminating the exponential proliferation in $m$. By leveraging landscape structure, DO-QAOA provides a scalable route to hybrid quantum-classical optimization under realistic hardware constraints, with potential applicability across variational quantum algorithms.

We formulate an infinite hierarchy of continuous-variable separability criteria in terms of quasiprobability distributions and their derivatives evaluated at individual points in phase space. Our approach is equivalent to the Peres--Horodecki criterion and sheds light on how distillable entanglement manifests in the phase-space picture. We demonstrate that already the lowest-order variant constitutes a powerful method for detecting the elusive non-Gaussian entanglement of relevant state families. Further, we devise a simple measurement scheme that relies solely on passive linear transformations and coherent ancillas. By strategically probing specific phase-space regions, our method offers clear advantages over existing techniques that rely on access to the full phase-space distributions.

We consider a massive Kalb-Ramond field with a general non-minimal coupling to gravity. We first study the theory in flat space-time, taking into account the non-linearities. We show that the coupling with the Ricci scalar gives rise to the strong coupling of the two transverse pseudo-vector degrees of freedom, which are absent in the massless theory. We then show that if the theory is instead coupled to the Ricci tensor or the Riemann tensor, the two tensor modes become strongly coupled in addition to the transverse pseudo-vector modes. We then extend our analysis to homogeneous and isotropic space-time, with vanishing background value of the Kalb-Ramond field. We show that in this case, the couplings with the Ricci and Riemann tensor give rise to the runaway instability. Finally, we discuss the inclusion of the disformal coupling as a possible resolution to this unnatural behavior.

Classical-quantum coexistence enables cost-effective transmission of data and quantum signals over the same fiber-optic channel. Nevertheless, weak quantum-key distribution (QKD) signals are susceptible to non-linear interference generated from the classical traffic, primarily spontaneous Raman scattering (SpRS) and four-wave-mixing (FWM), as well as to unfiltered noise. In QKD protocols, increased channel loss and excess noise both reduce the secret key rates (SKRs), as illustrated in this work for the two-state BB84 and Gaussian-modulated coherent-states (GMCS) protocols. In this study, we derive closed-form expressions for evaluating the accumulated interference power from coexisting classical signals in a quantum frequency channel. Our model enables effective design of classical-quantum systems in single-mode fibers (SMFs), capturing the evolution of interference arising from the relevant physical phenomena. We utilize the model to examine frequency allocation in multiband transmission systems, demonstrating that, contrary to common practice of allocating QKD channels in the O-band, increased SKR is achieved by placing quantum channels in the upper E-/lower S-band across the relevant scenarios.

Past-geodesically-complete cosmologies are thought to require either contraction, or an asymptotically static past. We introduce a third possibility: Einstein-frame time can dynamically attain a local minimum. This time-reversal is caused by phantom Chaplygin gas, whose acoustic cone defines a `causal-frame' geometry that is geodesically-complete. While gravity experiences time-reversal, the Chaplygin gas always evolves forward in time, realizing a transient mismatch in thermodynamic arrows of time. Time-reversal affects the scale-factor, enforcing a non-singular bounce in causal frame that is robust against any additional matter.

The authors wished to document the sensitivity improvement that has been contributed to the GW detection rate by detection algorithm research and development efforts, and set about re-analyzing S5 and S6 to determine the sensitive time-volumes of a modern pipeline and compare them to that of analysis algorithms of the day. To our surprise, this effort led to the discovery of GW070605, what at first appeared to be a previously unreported high significance binary neutron star merger at a time when only the Livingston detector (L1) was operating -- data that could not have been analyzed and a signal that could not have been discovered previously because the algorithms of the day required coincidence between two or more detectors. GW070605's end time occurs in LIGO's L1 detector at 2007-06-05 18:37:02 UTC, and is estimated to be a merger with component masses of 1.82$M_\odot$ and 1.24$M_\odot$. The GstLAL detection algorithm estimates that noise processes produce false positives at least as significant as GW070605 at a rate of $8.6\times10^{-10}$ per year. Disappointingly, subsequent investigations revealed the presence of a previously undocumented hardware injection in the L1 detector's Y arm end test mass' excitation channel, whose time and properties match that of GW070605. The injection does not appear in the Gravitational Wave Open Science Center list of hardware injections. We determined that while there is no sensitivity improvement between GstLAL and previous algorithms at the null-result threshold, there is marked improvement at above null-result thresholds; specifically, an approximately 55-times detection rate increase from initial-era algorithms at a FAR threshold of 1 per 7000 years.

Can we estimate the direction of arrival (DOA) of a gravitational wave (GW) signal from pulsar timing array observations? The present paper addresses the inverse problem, for which we consider quadrupole moments of pulsar timings due to GWs from a dominant isolated source such as a binary of supermassive black holes over an isotropic stochastic background. Correlations between the quadrupole moments are discussed, where the correlations between pulsar pairs over the full sky are taken into account. The correlations turn out to be in the form of a three-dimensional traceless matrix with rank 2 that can be closely related with a projection tensor for the GW. Thereby, we demonstrate that the rank-2 matrix allows to estimate the DOA of the GW. In expectation of the forthcoming Square Kilometer Array, angular resolutions as well as DOA estimation errors are also examined.

We propose a method to directly recover the degree of entanglement distributed by entanglement swapping in the presence of noise. Our approach introduces a reversing operation that probabilistically undoes the effect of amplitude damping or photon loss on a single entangled pair, enabling heralded recovery of entanglement. We demonstrate that entanglement can be substantially recovered even under strong noise, including parameter regimes where the distributed entanglement would otherwise vanish due to entanglement sudden death. We analyze the effectiveness of the protocol in two representative repeater models, i.e.,~two-way and one-way architectures and identify the optimal reversing strategy. Due to its heralded and single-copy nature, our protocol is readily compatible with other entanglement recovery techniques such as entanglement purification and distillation. Our work provides a practical and experimentally feasible way toward robust entanglement distribution in current and near-term quantum repeater architectures.

The Quantum Approximate Optimization Algorithm (QAOA) is a quantum algorithm proposed for Noisy Intermediate-Scale Quantum (NISQ) devices and is regarded as a promising approach to combinatorial optimization problems, with potential applications in the financial sector. In this study, we apply QAOA to the portfolio optimization problem, which is one of the central challenges in financial engineering. A portfolio consists of a combination of multiple assets, and the portfolio optimization problem aims to determine the optimal asset allocation by balancing expected return and risk. In the context of quantum optimization, portfolio optimization is often formulated using discrete variables. Unlike conventional binary formulations, we consider a ternary portfolio optimization problem that accounts for three states-holding, not holding, and short selling-and compare its performance using different mixer operators. Specifically, we implement QAOA with the standard mixer and several XY Mixers (XY Ring, XY Parity Ring, XY Full, and QAMPA), and conducted simulations using real data based on the German stock index (DAX 30) for portfolios consisting of 5 and 8 assets. Furthermore, we introduce noise based on a depolarizing channel to investigate the behavior of the algorithm in realistic environments. The results show that while XY Mixers exhibit superiority in noiseless settings, their advantage degrades in noisy environments, and the optimal choice of mixer depends on both the number of QAOA depths and the noise strength.

This work presents a comprehensive study of energy extraction via the Comisso-Asenjo magnetic reconnection mechanism from rotating charged black holes in the context of Kalb-Ramond (KR) gravity. We systematically investigate the influence of various parameters on the energy extraction process, comparing the results in two distinct regions: the circular orbit region and the plunging region. {The results reveal that the Lorentz-violating parameter has a significant impact on energy extraction, affecting not only the parameter space where energy extraction is possible, but also the energy extraction power and efficiency.} It is found that the energy extraction process in the circular orbit region can offer a promising avenue for constraining KR gravity. In contrast, although energy extraction from the plunging region remains feasible even for black holes with relatively low spins and takes place nearer to the event horizon, its sensitivity to the Lorentz-violating parameter is significantly reduced. Overall, the Comisso-Asenjo magnetic reconnection mechanism can serve as a probe of the KR field, particularly through the energy extraction process in the circular orbit region.

As quantum information systems mature, efficient and coherent transfer of quantum information through noisy channels becomes increasingly important. We examine how passive environment-assisted quantum communication enhances direct quantum information transfer efficiency. A bosonic pure-loss channel, modeled as transmission through a beam splitter with a vacuum input state at the dark port, has zero quantum capacity when transmissivity is below 50%. Quantum communication through the channel can be enhanced by passive environment assistance, achieved via the selection of an appropriate input state for the ancilla port. Although ideal Gottesman-Kitaev-Preskill (GKP) states enable perfect quantum information transmission at arbitrarily small transmissivity, they are challenging to realize experimentally. We therefore explore more experimentally accessible non-Gaussian ancilla states, such as Fock, cat, and squeezed cat states, and numerically determine the optimal encoding and decoding strategies. We also construct analytical schemes that yield high-fidelity transmission and good information rates.

We proposed a time-delayed quantum extreme learning machine (TD-QELM) for efficient time-series prediction on noisy intermediate-scale quantum (NISQ) devices. By encoding multiple past inputs simultaneously, TD-QELM achieves shallow circuit depth independent of sequence length, thereby, mitigating noise accumulation and reducing computational complexity. Experiments using the NARMA benchmark on both noiseless simulations and IBM's 127-qubit processor demonstrate that TD-QELM consistently outperforms conventional quantum reservoir computing in prediction accuracy and noise robustness. These results highlight TD-QELM as a practical and scalable framework for time-series learning on current NISQ hardware.

We examine Schwarzschild black holes within the framework of gravitational thermodynamics, introducing an ``energy layer'' picture for black-hole mass-energy and exploring a possible energy-extraction mechanism termed ``quasi-superradiance.'' Building on the standard relations for Hawking temperature and Bekenstein--Hawking entropy, we formalize energy layers via quasi-local radial energy accounting (e.g.\ integrating an effective local energy density over spherical shells) and connect this bookkeeping to the free energy $\FHelm=M-Þ\SBH$. We then extend the entropy correction ansatz with explicit series inversion and derive higher-order expansions for $Þ(M)$ and $\FHelm(M)$, including logarithmic and inverse-mass terms. To enhance mathematical transparency, we add intermediate derivations, lemma/theorem statements, and appendices. The quasi-superradiant mechanism is framed as a Carnot-like thought experiment powered by the Tolman temperature gradient between the near-horizon region and infinity; we show that the generalized second law enforces the Carnot bound and yields integrated maximum-work inequalities. Throughout, we stress that the proposal is heuristic and intended as a consistency-checked framework for discussion rather than a claim of definitive new physics.

In this paper, we will provide a complete derivation of the decoherence rate for a magnetic nanoparticle in quantum superposition in the presence of the fluctuating electromagnetic field in a thermal background by using the fluctuation-dissipation theorem in the long-wavelength limit. The long-wavelength limit assumes that the superposition size is much smaller than the wavelength of the electromagentic filed fluctuations. We will extend this computation to two diamagnetic nanoparticles kept in quantum superposition adjacent to each other. We will also show how the drag force on a single nanoparticle arises from external electromagnetic-field fluctuations, and compare our results with those for the nanoparticle's dielectric properties.

Reaching fast and robust two-qubit gates with low infidelities has been an outstanding challenge for the long-term goal of useful quantum computers. Typically, optimizing the pulse shapes can minimize the gate infidelity and improve its robustness to certain types of errors; yet it remains incapable of speeding up the gate execution time which is fundamentally restricted by the attainable Rabi frequency in a realistic setup. In this work, we develop a fast implementation of two-qubit CZ gates using optimized ancillary drive to enhance the two-photon Rabi frequency between the ground and Rydberg states.This ancillary drive can work in an error-robustness framework without increasing the original gate infidelity in the absence of the drive. Considering the experimentally feasible parameters for $^{87}$Rb atoms, we demonstrate that the execution time required for such CZ gates can be shortened by more than 30$\%$ as compared to standard two-photon protocols arising the gate fidelity above 0.9954 by taking account of all relevant error sources. Our results reduce the high-power laser requirement and unlock the potential toward fast, high-fidelity quantum operations for large-scale quantum computation with neutral atoms.

In this work, we show that the sample complexity (equivalently, the number of measurements) required in quantum learning theory within a general parametric framework, is fundamentally governed by the inverse Fisher information matrix. More specifically, we derive upper and lower bounds on the number of samples required to estimate the parameters of a quantum system within a prescribed small additive error and with high success probability under maximum likelihood estimation. The upper bound is governed by the supremum of the largest diagonal entry of the inverse Fisher information matrix, while the lower bound is characterized by any diagonal element evaluated at arbitrary parameter values. We then apply the general bounds to Pauli channel learning and to the estimation of Pauli expectation values in the asymptotic small-error regime, and recover the previously established sample complexity through considerably streamlined derivations. Furthermore, we identify the structural origin of exponential sample complexity in Pauli channel learning without entanglement and in Pauli expectation value estimation without quantum memory. We then extend the analysis to an error criterion based on the Euclidean distance between the true parameter values and their estimators. We derive the corresponding upper and lower bounds on the sample complexity, which are likewise characterized by the inverse Fisher information matrix. As an application, we consider Pauli expectation estimation with entangled probes. Finally, we highlight two fundamental contributions to quantum learning theory. First, we establish a systematic framework that determines the task-independent sample complexity under maximum-likelihood estimation. Second, we show that, in the small-error regime, learning sample complexity is governed by the inverse Fisher information matrix.

We demonstrate a robust, high-visibility nonlocal Franson interferometry for fiber-based quantum networks by co-propagating a classical Radio-over-Fiber clock signal with energy-time entangled photon pairs in the same fiber. Utilizing cross-band allocation (O-band for classical, L-band for quantum signals), the spontaneous Raman scattering noise photons are effectively suppressed. At the same time, their environmental delay fluctuations remain highly correlated for common-mode noise cancellation, achieving a passive synchronization with picoseconds precision. Over 50 km of single-mode fiber, this co-propagation enables nonlocal quantum interference with a visibility of (88.35\pm3.62)%, without relying on external dedicated timing infrastructure. This work provides a practical, scalable synchronization solution for metropolitan-scale entanglement-based quantum networks.

We investigate the interplay between nonlinearity and inhomogeneities in discrete-time quantum walks on one-dimensional lattices. Nonlinear effects are introduced through a Kerr-like, intensity-dependent local phase, while spatial and temporal inhomogeneities are implemented via random variations of the quantum gate operations. By analyzing typical quantities, such as the return probability and the participation function, we identify distinct quantum walking regimes as the nonlinear parameter $χ$ and the quantum gate parameter $θ$ are varied. Spatial inhomogeneities weaken nonlinear self-trapping and constrict the region of robust localization. In this process, partially localized regimes emerge, characterized by the coexistence of a confined core and dispersive wave-packet components. In contrast, temporal inhomogeneities act as time-dependent perturbations that continuously disrupt the phase coherence required for self-trapped excitation, thereby enhancing dispersive emission and promoting delocalization. By using $χ$ versus $θ$ diagrams, we display a comprehensive characterization of how inhomogeneities modify the stability and extent of prevailing dynamical regimes, elucidating the competition between nonlinearity and inhomogeneities in discrete-time quantum walks.

We study the fully entangled fraction of quantum states based on the Bloch representation of density matrices. Analytical upper bounds on the fully entangled fraction are obtained for arbitrary $d\otimes d$ bipartite systems. The fully entangled fractions for classes of $d\otimes d$ quantum states are analytically derived. Detailed examples are given to illustrate the advantages of our results.

A quantum-mechanical description is presented for the three-body physics of shielded dipolar molecules, including a prediction of observable Efimov physics. Despite the anisotropic and long-range nature of the interaction, shielding enables a regime in which universality emerges already at the two-body level and extends to the three-body sector, where Efimov physics emerges. On the negative side of the scattering-length resonance, computed trimer binding energies display the characteristic scaling expected for Efimov resonances. Finally, the sudden approximation can be used to create trimer bound states, starting from positive energy trap states as a way to create or detect these molecular trimers. Moreover, the three-body parameter expressed in dipolar units is found to be universal.

In this paper, we investigate the dynamic phenomenological signatures of a Schwarzschild-MOG black hole shadow perturbed by passing gravitational waves. By perturbing the Hamilton-Jacobi equation for photon null geodesics, we demonstrate that the unique field content of MOG breaks the observational degeneracy with standard General Relativity. We mathematically prove two distinct, time-dependent signatures. First, the massless MOG scalar field induces a volumetric ``breathing mode'' polarization, causing the total apparent area of the shadow to rhythmically expand and contract. Second, the massive MOG vector field undergoes quantum vacuum dispersion, arriving at the observer with a predictable time delay. This delayed massive wave sources secondary longitudinal metric perturbations that manifest as a sudden, asymmetric translational wobble of the shadow on the celestial screen. These dynamic geometric shifts offer a robust observational template for next-generation interferometry to strictly test the existence of massive force carriers and scalar fields in gravity.

Embedding non-Markovian open quantum dynamics into an enlarged Markovian space offers a powerful route to nonperturbative simulations, where the dynamics of the extended space can be governed by multiple distinct Markovian equations. We show that these distinct embeddings arise from different unravelings of Gaussian bath self-energies, generating a family of deterministic, time-local equations for the extended system. Using the Brownian-oscillator spectral density as an illustrative example, we clarify the relationships among existing approaches, including the Hierarchical Equations of Motion (HEOM) and the Lindblad--pseudomode formalism, and demonstrate how this framework enables numerically stable and efficient simulations. This work provides both a transparent theoretical foundation for embedding techniques and a flexible platform for developing new methods to simulate non-Markovian quantum dynamics.

We introduce Natural Qubit Algebra (NQA), a compact real operator calculus for qubit systems based on a $2\times2$ block alphabet $\{I,X,Z,W\}\subset\mathrm{Mat}(2,\mathbb{R})$ and tensor-word representations. The resulting multiplication law induces a canonical $(\mathbb{Z}_2)^{2m}$-grading with a bicharacter that controls commutation signs, placing the framework naturally within the theory of color-graded and Clifford-type algebras. Within this language, we provide: (i) an explicit real Clifford normal form for two-qubit operators via the identification $\mathrm{Mat}(4,\mathbb{R})\cong\mathrm{Cl}(2,2;\mathbb{R})$; (ii) a purely algebraic reformulation of the Bell--CHSH scenario, where the quantum violation is expressed as a spectral non-embeddability of a noncommutative spinor algebra into any commutative Kolmogorov algebra; and (iii) compact factored representations of the Bernstein--Vazirani and Grover phase oracles, showing that both Clifford and non-Clifford examples can admit similarly structured symbolic descriptions. We clarify that Grover's iterate remains outside the Clifford group due to its continuous spectral rotation, consistent with the Gottesman--Knill theorem, while retaining a compact tensor-block form in NQA. The framework isolates spectral, algebraic, and syntactic aspects of operator structure, providing a graded operator language compatible with standard quantum mechanics.

Demonstrating genuine many-body quantum coherence in large-scale quantum processors remains a central challenge for near-term quantum technologies. Recent experiments on D-Wave quantum annealers have investigated quenches of Ising chains and observed defect densities that show Kibble-Zurek scaling, consistent with coherent quantum dynamics. However, identical scaling can arise from classical or thermal processes. Here we propose the use of many-body coherent oscillations (MBCO) as a diagnostic for the identification of system-wide coherence in analog quantum simulators. Solving the time-dependent Schrodinger equation, we show that quenches of a staggered one-dimensional Ising chain across a quantum critical point produce oscillatory signatures in defect observables. We implement this model on the D-Wave Advantage quantum annealer. Using fast-anneal protocols, we find that, although defect densities follow Kibble-Zurek scaling, the expected oscillatory behavior is absent. We demonstrate that static disorder associated with individual qubits is not likely responsible for the absence of MBCO. Modest modifications to annealing schedules can dramatically enhance oscillation visibility. This work gives a general roadmap for the search for quantum coherence in noisy, large-scale quantum platforms.

In Quantum Machine Learning (QML) and Quantum Finance, amplitude encoding is often motivated by its logarithmic storage capacity arXiv:1307.0411. This paradigm typically relies on the mapping $ψ= \sqrt{P}$, treating the quantum state as a derivative of a classical probability distribution $P$. By restricting the data manifold to the positive real orthant $\mathcal{S}^+$, the accessible Hilbert space is effectively abelianized, rendering the representation ``phase-deaf''. We rigorously establish that while $P$ is a projection of $|ψ|^2$, the simple square-root mapping fails to recover the non-commutative structure necessary for genuine quantum advantage in classification tasks. Furthermore, we clarify why applying basis changes (like Hadamard gates) to these states fails to replicate the computational power of active phase-kickback mechanisms. Finally, we advocate for Dynamical Hamiltonian Encoding (based on QIFT), where data generates non-commutative evolution rather than serving as a static, phase-locked vector.

Precision measurements at electron-positron colliders provide stringent tests of the Standard Model and powerful probes of possible higher-dimensional interactions. We present a hybrid quantum-classical framework for computing leading-order helicity amplitudes for $e^+e^-\to \ell^+\ell^-$ scattering on gate-based quantum hardware and using the resulting cross sections to constrain both Standard Model couplings and effective field theory operators. In our approach, external kinematics are encoded into single-qubit Weyl spinors, and full helicity amplitudes are reconstructed by coherently combining diagrammatic contributions within a single quantum circuit. Classical post-processing yields physical amplitudes and differential cross sections that can be directly compared with collider data. As a proof of concept, we compute unpolarised angular distributions and perform binned likelihood fits to precision electron-positron measurements. The extracted bounds are statistically consistent with Standard Model expectations, demonstrating that quantum-assisted amplitude evaluation can interface directly with phenomenological analyses and experimental data. This work establishes a concrete pathway toward applying quantum computing to precision collider physics and effective field theory studies.

In this work, we investigate exact black string solutions in the context of $f(R,T)$ gravity. Adopting the specific form $f(R,T) = R + 2χT$, we consider an anisotropic Kiselev fluid as the matter content and obtain static cylindrical solutions, which are then extended to the rotating case through a suitable coordinate transformation. The influence of the quintessence state parameter $w_q$ and the matter--geometry coupling constant $χ$ on the geometry is analyzed. We examine the weak, null, and strong energy conditions, identifying the regions in the parameter space where they are satisfied. Furthermore, we apply the Hamilton--Jacobi method to study the tunneling of scalar particles across the event horizon and derive the corresponding Hawking temperature. The thermodynamic stability of the solutions is investigated by computing the heat capacity, and the conditions for phase transitions are discussed. The results provide a characterization of black strings in $f(R,T)$ gravity surrounded by quintessence, highlighting the combined effects of anisotropic matter and modified gravity on their physical properties.

We study a novel repumping transition for $^{88}$Sr atoms trapped in a 'blue' magneto-optical trap. We show that, while the repumping efficiency is about three orders of magnitude smaller than for traditional schemes, it is sufficient for recycling all atoms, provided the repumping laser beams are arranged to form a 'green' magneto- optical trap (MOT) helping to cool and confine the atoms and preventing their loss. Our main findings are: (i) that the green MOT configuration is able to trap 10 times more atoms in the blue MOT than using the green transition merely as a repump, and (ii) that the atom numbers in the two-color MOT can be balanced through experimental control parameters. The interest of this scheme lies in its capability of reaching low temperature and its suitability for continuous atomic beam generation.

In this study, by revisiting the quantum interpretation of the cosmological constant, we introduce its formal representation within standard General Relativity. Examining its behavior in a Friedmann-Robertson-Walker spacetime reveals a mechanism in which the symmetry between energy and momentum is dynamically broken. Applying this concept naturally leads to the derivation of the familiar LCDM model, while simultaneously alleviating both the fine-tuning and coincidence problems. Comparison of the ground-state energy behavior in the Friedmann equations with a dust matter field further indicates that large-scale matter exhibits the same symmetry-breaking behavior. Remarkably, due to this broken symmetry, the interactions between local regions of matter in the large-scale structure generate effective pressure, driving late-time acceleration and reproducing the LCDM expansion history without invoking exotic fields or negative-pressure components. This framework provides a self-consistent realization of LCDM within General Relativity, emerging entirely from the intrinsic dynamics of standard matter without fine-tuning and coincidence problems.

Next-generation gravitational-wave detectors such as Cosmic Explorer and the Einstein Telescope will operate 10-40 km Fabry-Perot arm cavities inside vacuum beam tubes. FFT-based paraxial tools treat propagation in free space and therefore do not explicitly enforce beam tube boundary conditions. We introduce a waveguide-like mode description of the optical field that incorporates an imposed beam tube boundary condition and enables an independent benchmark of free-space FFT tools We derive the associated modal-mixing matrices for mirrors and baffles, including a closed-form series for axisymmetric circular apertures. We quantify the strain-equivalent couplings from baffle miscentering and from a localized near-wall tube defect, and show that they are suppressed as baffle density increases. In the relevant regime of densely baffled cavities and small perturbations, beam tube boundary effects are subdominant, which supports the continued use of FFT-based codes to guide the design of 3G detectors.

Higher Berry phase has recently been proposed to study the topology of the space of gapped many-body quantum systems. In this work, we develop a boundary-scattering approach to detect higher Berry phases in one-dimensional gapped free-fermion systems. By coupling a gapless lead to the gapped system, we demonstrate that the higher Berry invariant can be obtained by studying the higher winding number of the boundary reflection matrix. The resulting topological invariant is robust against perturbations such as disorder. Our approach establishes a connection between higher Berry invariants and transport properties, thereby providing a potentially experimentally accessible probe of parametrized topological phases.

Baryogenesis remains an unresolved problem in cosmology, with existing mechanisms facing significant caveats. We show that the effects of primordial black holes (PBHs) on neutrinos produce the lepton asymmetry $\sim 10^{-10}$ which subsequently produces the baryon asymmetry. We consider the Dirac Lagrangian in curved spacetime in local coordinates exhibiting Hermitian pseudo-vector and non-Hermitian vector terms. These terms lead to energy splitting between weakly interacting neutrinos and antineutrinos, resulting in their unequal number densities and hence a lepton asymmetry. While the non-Hermitian effect leads to a non-conserved total probability of neutrinos, the leptogenesis due to gravitational effects of a PBH could be significant until the nucleosynthesis era. This in turn produces baryon asymmetry from the symmetry of lepton and baryon numbers via the sphaleron process in the electro-weak era. We show that in the most conservative scenario, the PBHs of mass $\sim 10^{12}$ g and spin $\sim 0.01$ produce the observed baryogenesis at temperature 130 GeV, when such PBHs are available abundantly. However, massive PBHs also could produce the observed asymmetry, assuming the non/anti-Hermitian vector couplings for neutrino and anti-neutrino get canceled from the Lagrangian, leading the system to be Hermitian.

We analyze decoherence in a one-loop Stern--Gerlach--type matter-wave interferometer for a massive nanoparticle embedded with a nitrogen vacancy (NV)-centred nanodiamond evolving under an effective harmonic-oscillator dynamics in a magnetic-field gradient. We assume that the Stern-Gerlach interferometer is subjected to a random acceleration noise external to the system. This could be along the direction of the superposition at an angle which can be varied. We quantify dephasing from two noise channels: fluctuations in the external acceleration $a(t)$ magnitude and direction as specified by the tilt angle $θ_0(t)$ between the superposition axis and the acceleration. At the level of the action, we treat these two external noise as stochastic inputs, and compute the resulting stochastic arm-phase difference, and obtain the dephasing rate $Γ$ using the Wiener--Khinchin theorem. For a white noise and a coherence target $Γτ\leq 1$ and by assuming that we finish the one-loop interferometer within $τ=2π/ω_0\simeq 0.015~\mathrm{s}$, for a reasonable choice of the magnetic field gradient, $η_0=6\times 10^{3}~\mathrm{T\,m^{-1}}$ and mass of the nanodiamond, $m=10^{-15}~\mathrm{kg}$) to create a superposition size of $Δx\sim 1$nm. We find $\sqrt{\mathcal{S}_{aa}}\lesssim \mathcal{O}(10^{-11})~\mathrm{m\,s^{-2}\,Hz^{-1/2}}$ even if we take the external acceleration, $a=0~{\rm ms^{-2}}$ and $θ_0=0^\circ$ (along the dirction of the superposition), and $\sqrt{\mathcal{S}_{θθ}}\lesssim \mathcal{O}(10^{-10})~\mathrm{rad\,Hz^{-1/2}}$ for $a=g= 9.81~\mathrm{m\,s^{-2}}$ and $θ_0=0^\circ$ (superposition direction is perpendicular to the Earth's gravity). We have also found an operating regime where the acceleration noise can be minimized by either varying $θ_0$ or $a$ for a fixed set of other experimental parameters.

We compute leading order quantum gravity contributions to a simple scalar scattering amplitude in Asymptotic Safety. Our model admits an analytic treatment so that several subtleties can be analysed. We find that (i) the existence of an asymptotically safe renormalisation group fixed point alone does not imply the boundedness of scattering amplitudes, (ii) gravitational logarithms can dominate the infrared regime of massless theories, (iii) a derivative expansion of the effective action fails quantitatively to predict the correct Wilson coefficients in massless theories, and (iv) standard renormalisation group improvement techniques fail qualitatively to describe the momentum dependence of correlation functions. Only momentum-dependent computations can resolve these issues. For theories that include massive fields, the derivative expansion can work effectively in most cases, but it can still fail for classically marginal couplings, and purely gravitational couplings. We also speculate about an effective realisation of the no-global-symmetries conjecture in Asymptotic Safety.

We analyze the emergence of classical cosmological spacetimes in quantum cosmology by computing the reduced density matrix for long-wavelength curvature perturbations. Starting from standard Hartle--Hawking and tunneling boundary conditions, we emphasize that semiclassical WKB structure and inflationary squeezing do not by themselves yield classicality. Tracing over unobserved degrees of freedom and using the influence functional formalism, we derive the decoherence functional for superhorizon curvature modes during inflation. For a light massive environmental scalar field in the Bunch--Davies vacuum, we obtain an explicit noise kernel and show how a nonzero mass regulates the infrared behavior. We then evaluate decoherence under horizon-based and EFT-motivated coarse grainings, finding efficient suppression of interference between macroscopically distinct perturbation histories in both cases. The analysis clarifies the distinct roles of boundary conditions (branch amplitudes) and decoherence (classical branch selection) and yields an emergent cosmological arrow of time through environment-induced entanglement.

Black holes are some of the most interesting objects in the universe. While they first arise in the complicated behavior of general relativity, the physical laws ruling their behavior are surprisingly simple. For example, one of the core facts about black holes is that their area never decreases, much alike the entropy in thermodynamics. In this note directed at introductory physics students and their instructors, we use this similarity to understand properties of black hole physics using standard techniques from an undergraduate course in thermal physics. We explore the never-decreasing nature of black hole area to obtain bounds on the energy emitted in a black hole merger (a calculation originally done by Hawking). We show how this allows us to think of black holes in manners very similar to heat engines, and how these ideas have been used in modern gravitational wave observatories to test general relativity. This allows a research-level topic to be discussed in introductory physics lectures.

Resonant inelastic X-ray scattering (RIXS) is the workhorse experimental technique for probing the structural degradation of higher-capacity cathode materials. However, the interpretation of experimental spectra is challenging due to the lack of accurate simulations. In this work, we propose a quantum algorithm for simulating the RIXS spectrum of molecular clusters hypothesized to form in Li-excess cathodes. The algorithm uses quantum phase estimation to sample the spectrum from a state encoding the scattering transition amplitudes of the cluster valence excitations. We prepare this state in the quantum computer using a block-encoding of the dipole operator and quantum signal processing to implement the Green's function propagator over intermediate core-excited states. To showcase the algorithm, we use a model cluster proposed in recent experimental works consisting of an oxygen dimer bonded to a manganese atom. Using the PennyLane software platform, we report resource estimation for simulating RIXS spectra for chemically motivated active spaces of increasing sizes. For a classically challenging active space with 20 orbitals, the algorithm requires $2.0 \times 10^{10}$ Toffoli gates and $414$ logical qubits.

In \cite{salvio} a prescription for calculating the correlation functions in Quadratic Gravity \cite{stelle1}-\cite{stelle2} was presented. This procedure does not enter in conflict with unitarity The Gauss-Ostrogradsky method for higher order theories defines two momentum densities $P_1$ and $P_2$ and two coordinate densities $Q_1$ and $Q_2$, one pair is standard, the other ghost like. The approach in \cite{salvio} involves the continuation $P_2\to i P_2$ and $Q_2\to i Q_2$ of the ghost variables. In the present work, following \cite{yomismo}, the LSZ rules are derived, but with a formalism adapted to full quartic or higher order theories. The hypothesis for quantization are that $[Q_1, Q_2]=\text{"gauge terms"}$, $[P_1, P_2]=\text{"gauge terms"}$ and $[P_1, Q_1]=iI+\text{"gauge terms"}$. This alone leads to the conclusion that $[P_2, Q_2]=-iI+\text{"gauge terms"}$, therefore this last pair of variables is ghost like. The graviton contains a massless mode, which is the standard graviton, plus two massive modes with masses $m_2$ and $m_3$. The third mode is usually interpreted as a ghost in the literature \cite{stelle2}. Here it is shown that, even after making the continuation $P_2\to i P_2$ and $Q_2\to i Q_2$, the creation and annihilation operators for this mode commute, the third mode does not appear as a free wave. This does not invalidate the model. The effective action $Γ$ can be calculated following \cite{stelle1}, and can be constrained by the Slavnov-Taylor identities \cite{Slavnov}-\cite{Taylor}, and the scattering rules may be worked out consequently.

It is well known that the diffusion equation, when treated as a stand-alone partial differential equation, exhibits exponential instabilities in boosted frames, which render the corresponding initial-value problem ill-posed. Recently, however, it was shown that Fick-type diffusion arises as the exact hydrodynamic sector of relativistic Fokker-Planck kinetic theory. In this work, we exploit this kinetic embedding to formulate a modified initial-value problem for one-dimensional Lorentz-boosted diffusion. We show that the resulting dynamics are well posed both forward and backward in time, provided the boosted density profiles admit a kinetic-theory realization. Such profiles form a space of band-limited functions, within which the evolution can be expressed as a discrete superposition of spatially sampled initial data, weighted by a Shannon-Whittaker-type Green function defined on the full Minkowski plane. The Green function is obtained in closed analytic form.

We conjecture a family of matrix quantum mechanical models that are holographically dual to discrete light-cone quantization of M-theory in pp-wave-like backgrounds. These backgrounds can be obtained from a Penrose limit of AdS$_4\times X_7$, where $X_7$ is Einstein. The matrix models arise from a classically consistent dimensional reduction of the UV Lagrangians of $\mathcal{N}=1$ superconformal field theories, in close analogy with how the BMN matrix model is obtained by dimensional reduction from $\mathcal{N}=4$ super Yang-Mills theory. We also discuss about supersymmetric black objects in pp-wave background by studying the Witten index and speculate that the area of the horizon is bounded from above for a fixed $N$.

Direct searches for dark matter in the form of WIMPs with argon-based detectors require precise measurements of the ionization yield \Qy\ for nuclear recoils at low energies. Prior to this work, direct experimental data were available only above 6.7 keV, leaving a critical gap in the energy region most relevant for low-mass WIMP searches. The Recoil Directionality (ReD) experiment addressed this limitation by measuring the argon \Qy\ for nuclear recoils between 2 and 10 keV using a dual-phase TPC irradiated with neutrons from a \Cf\ fission source. The results extend existing direct measurements to lower energies, show consistency with previous data above 7 keV, and indicate an enhanced ionization yield at low recoil energies. These measurements provide essential input for next-generation argon-based dark matter searches and directly motivate the upgraded ReD+ phase, designed to further extend sensitivity into the sub-keV recoil-energy regime.

We present measurements of the total and differential cross sections for near-threshold J/$ψ$ photoproduction obtained with the CLAS12 detector at the Thomas Jefferson National Accelerator Facility. The results are based on data collected during the Fall 2018 and Spring 2019 running periods, using electron beams with energies of 10.6 and 10.2 GeV, respectively, scattered off a liquid-hydrogen target. Near-threshold J$/ψ$ photoproduction offers a unique sensitivity to the strong interaction in the non-perturbative regime of Quantum Chromodynamics (QCD). The energy dependence of the cross section constrains the underlying J$/ψ$ production mechanisms, including multi-gluon exchange and potential baryonic excitations. Additionally, the $t$-dependence of the differential cross section can be related to the transverse spatial distribution of gluons in the proton, providing critical input for theoretical descriptions of the gluonic structure of the proton. An interpretation of the results in terms of the gluon content of the proton is presented, providing new experimental constraints on QCD-inspired models of the proton structure and the role of gluonic degrees of freedom in hadronic mass generation.

Hyperon physics offers a distinctive laboratory for probing the intensity frontier and searching for physics beyond the Standard Model. This review summarizes recent results from the BESIII experiment, including pioneering studies of dark baryons, massless BSM particles, and invisible decay modes, together with investigations of baryon- and lepton-number violation. A central highlight is the determination of the $Λ$ electric dipole moment using quantum-entangled hyperon-antihyperon pairs, achieving a sensitivity three orders of magnitude beyond previous limits. These measurements provide world-leading constraints on new physics scenarios and establish a robust foundation for next-generation precision studies. By integrating experimental progress with theoretical developments and future facility prospects, this review emphasizes the critical role of hyperon probes in testing the fundamental laws of nature.

We show how to construct a complete set of lowering operators, whose successive application reduces an arbitrary Fenyman integral to a combination of master integrals. The construction builds systems of equations for generic integral indices using IBP-generating vectors. The solution to each system is a lowering operator.

Primary proton-air interactions at ultra-high energies leave a physically interpretable imprint on the correlated fluctuations of the depth of shower maximum and the muon content in extensive air showers. This imprint reflects the stochasticity in the partition of the primary energy among secondary particles in the first interaction. We show that these fluctuations can be accessed through a probabilistic description that isolates sensitivity to hadronic physics in the initial collision, while treating the subsequent shower development as effectively universal. The uncertainties resulting from this universality are smaller than the spread among current hadronic interaction models and comparable to current experimental uncertainties. Consequently, the joint observable space defined by these two quantities provides a new probe of hadron production in kinematic regimes far beyond the reach of human-made accelerators.

The spectral problem of four-dimensional superconformal quiver gauge theories can be mapped to one-dimensional spin chains with restricted Hilbert spaces, where the composition of neighbouring spins follows the path algebra of the quiver. To better understand such spin chains, we compute the one-loop planar dilatation operator for the 4d N=2 ADE quiver gauge theories obtained by orbifolding the N=4 Super-Yang-Mills theory and marginally deforming by independently varying the gauge couplings. This extends previous work which was mainly focused on the Z2 quiver. We characterise the general features of the resulting ADE spin-chain models and construct the 2-magnon Bethe ansatz for holomorphic states. We also evaluate, at large N, the N=2 superconformal index of these gauge theories and use it to study their protected spectrum in specific sectors.

We propose an approach to construct three-dimensional lattice models using line defects in state integral models on shaped triangulations of 3-manifolds. The Boltzmann weights for these models satisfy a variant of the tetrahedron equation, which implies integrability under suitable assumptions on R-matrices and transfer matrices. As an explicit example, we present a solution produced by Teichmüller TQFT.

We present full off-shell NLO corrections in QCD obtained with the MoCaNLO code matched to parton shower. A resonance-aware matching procedure has been devised for the MC@NLO method tuned to the Catani-Seymour dipole subtraction. Specifically, we consider the off-shell production of a top-antitop pair in the semi-leptonic decay channel in electron-positron collisions and match it to the final-state QCD parton shower of PYTHIA8. Distortions of resonances' line shapes are avoided by providing the details of the resonance-cascade chain on an event-by-event basis to the parton shower and by adapting the matching accordingly through the introduction of dedicated counterterms.

Three-dimensional Weyl materials provide a controlled setting for exploring Floquet dynamics in open quantum systems, including nonequilibrium steady states (NESS). Motivated by the desire for a strongly-coupled description, we employ holography to analyze the formation and stability of a NESS in a Weyl semi-metal induced by an external circularly polarized electric field. A time-periodic steady-state solution is constructed and its stability is determined from the spectrum of out-of-equilibrium quasinormal modes (Floquet exponents). A stable region in the drive parameter space is identified; beyond a critical curve, the Floquet exponents enter the upper half of the complex plane, leading to a superharmonic response. At sufficiently strong driving, chaotic time evolution emerges in the fully nonlinear initial-boundary value problem. The anomaly-induced response of the NESS to an external magnetic field is also computed, and the resulting behavior is related to the previously proposed chiral pumping effect.

Stiff differential equations pose a major challenge for Physics-Informed Neural Networks (PINNs), often causing poor convergence. We propose a simple, hyperparameter-free method to address stiffness by normalizing loss residuals with the Jacobian. We provide theoretical indications that Jacobian-based normalization can improve gradient descent and validate it on benchmark stiff ordinary differential equations. We then apply it to a realistic system: the stiff Boltzmann equations (BEs) governing weakly interacting massive particle (WIMP) dark matter (DM). Our approach achieves higher accuracy than attention mechanisms previously proposed for handling stiffness, recovering the full solution where prior methods fail. This is further demonstrated in an inverse problem with a single experimental data point - the observed DM relic density - where our inverse PINNs correctly infer the cross section that solves the BEs in both Standard and alternative cosmologies.

The SND@LHC experiment is currently taking data at the Large Hadron Collider (LHC), exploring the unique forward region at pseudorapidities from 7.2 to 8.4. Its physics programme covers neutrinos originating from heavy-flavour decays and feebly interacting particles produced in proton proton collisions. Building upon the successful operation of the present detector, this paper presents the physics reach of the approved SND@LHC upgrade for Run4 of the LHC, and compares it with an alternative installation scenario. Lowering the detector by approximately 40 cm and shifting it horizontally by about 30 cm, while keeping it off-axis, increases the total neutrino interaction rate by a factor of five. The paper describes the design of the upgraded detector and compare the physics performance in both installation scenarios.

The recent observation of a family of fully-charm tetraquark states by the LHCb, ATLAS and CMS Collaborations suggests the possible existence of fully-heavy dibaryons. In this work, we investigate the $Ω_{ccc}Ω_{ccc}$ and $Ω_{bbb}Ω_{bbb}$ dibaryons in both the $^1S_0$ and $^5S_2$ channels using the method of QCD sum rules. We employ the iterative dispersion relation (IDR) method to efficiently compute the massive five-loop banana diagrams that appear in these systems, and properly address the tricky small-circle divergence problem in the nonperturbative terms. Our analyses reveal that for both charm and bottom systems, the scalar dibaryon lies lower than its tensor counterpart. The mass of the scalar $Ω_{ccc}Ω_{ccc}$ dibaryon is found to be slightly above the $2Ω_{ccc}$ mass threshold, while the $Ω_{bbb}Ω_{bbb}$ systems may form bound states.

Assuming that the $f_0(1370)$ is a $κ\barκ$ molecular state, the partial widths of its various decay channels are calculated, including the two-body decay $K \bar{K}$, $ππ$, $ηη$ and the four-body decay $ρρ/ σσ\to 4 π$ and $K \bar{K} ππ$. The coupling of $g_{f_0(1370) κ\barκ}\approx 13$ GeV estimated from the Weinberg criterion appears to be significantly underestimated. If this coupling is adjusted to $25 \sim 40$ GeV, the total width of $f_0(1370)$ can be fitted to the measured value $200\sim 500$ MeV. At the center-of-mass energy $\sqrt{s}=1.37$ GeV, the channels that mainly contribute to the total width are $K \bar{K}$, $ππ$ and $4 π$ ranked as $Γ(K \bar{K }) > Γ(4 π) \approx Γ(ππ) $ with $g_{f_0(1370) κ\barκ}= 35$ GeV. Around $1.37$ GeV, the decay widths of the two-body channels $K \bar{K}$, $ππ$ and $ηη$ remain stable with variation in $\sqrt{s}$, whereas the decay widths of the four-body channels $4 π$ and $K \bar{K }ππ$ increase continuously with $\sqrt{s}$. Most current data are model-dependent and conflicting, such as the $4 π$ dominant conclusion and the $K \bar{K}$ to $ππ$ ratios. The current data can not rule out the $κ\barκ$ assignment for $f_0(1370)$. Further reliable theoretical and experimental analyses of $f_0(1370)$ are required to reveal its nature.

We give an introduction to equivariant localization in supergravity, focusing on the application to four-dimensional theories and supersymmetric black holes.

We propose a scheme to correctly incorporate the contribution of the chiral anomaly in the D3/D7 model to calculate chiral transport phenomena. To ensure the D7-brane wraps S^5 appropriately and the Wess-Zumino term is switched on, we allow the D7-brane to rotate in the compactified extra directions and perform the analysis accordingly. To demonstrate that this calculation procedure works well, we specifically compute the magnetoresistance in the D3/D7 model. We find that a finite axial chemical potential is realized and the negative magnetoresistance is enhanced by the anomaly contribution.

Motivated by applications of self-dual theories to the AdS/CFT correspondence, we study self-dual Yang-Mills theory (SDYM) and its relation to Yang-Mills theory and to Chalmers-Siegel theory with Dirichlet, Neumann, and mixed boundary conditions. A Fefferman-Graham analysis of SDYM is performed to identify its boundary CFT data. We make a proposal for self-dual holography that defines $3d$ ``self-dual CFTs''. The bulk-to-bulk and boundary-to-bulk propagators for SDYM and for Yang-Mills/Chalmers-Siegel theory with mixed boundary conditions are derived in Feynman and axial gauges. Three- and four-point functions are computed in the spinor-helicity formalism, and the relations among the results in the various theories are clarified. The flat limit and the gauge-(in)dependence of the results are analyzed.

In this note, we study the Hopf-algebra-based (HAB) formula of Yang-Mills-Scalar (YMS) amplitudes, which expands a YMS amplitude with massive scalars as a combination of propagator matrices that mix massless scalars corresponding to gluons with the original massive scalars. We propose a recursive formula which conveniently expresses the HAB formula. In this formula, gluons are converted into massless scalars. Thus it expresses a YMS amplitude with massive scalars by amplitudes with fewer gluons, massive scalars and massless scalars. We verify this formula by using soft behavior approach. In the massless limit, the HAB formula turns into expressions for YMS amplitudes with massless scalars which was earlier shown to satisfy an alternative recursive expansion formula. In this note, we show the equivalence of these two distinct approaches through explicit calculation on amplitudes with one and two gluons.

We study the equation of state (EoS) of QCD matter in a bottom-up holographic setup that combines an Einstein-Maxwell-dilaton (EMD) sector with an improved Karch-Katz-Son-Stephanov (KKSS) flavor action. In the probe approximation, we perform an inverse reconstruction of the model functions by parameterizing them with neural networks and solving the EMD equations via a differentiable ODE solver (a neural ODE framework), calibrating the model to a $(2+1)$-flavor lattice-QCD EoS at finite temperature and finite baryon chemical potential. The reconstructed model functions are then parametrized and kept fixed across thermodynamic states. Next, viewing the EMD sector as an effective description of pure Yang--Mills theory, we fix its parameters by fitting the $μ_B=0$ lattice pure-glue EoS using a hybrid optimization strategy. Finally, we go beyond the probe limit and solve the coupled EMD$+$KKSS equations with back-reaction, using the pure-glue-calibrated EMD sector as a fixed input and varying the KKSS couplings to compare with the $μ_B=0$ two-flavor lattice EoS. We find a visible mismatch and a high-temperature behavior in which the back-reacted dimensionless ratios approach a nearly $β_1$-insensitive plateau close to the pure-glue baseline, providing a simple structural diagnostic for the present flavor-sector truncation.

We numerically investigate the time evolution of several physical observables for the so-called 2 R-Charge Black Hole (2RCBH) model undergoing Bjorken flow. The 2RCBH model corresponds to a top-down holographic construction describing a strongly interacting conformal fluid defined at finite temperature and R-charge density. Taken together with previous findings for the purely thermal $\mathcal{N}=4$ Supersymmetric Yang-Mills (SYM) plasma, and the 1 R-Charge Black Hole (1RCBH) model, our results for the 2RCBH model provide strong numerical evidence for the existence of far-from-equilibrium correlations between the non-equilibrium holographic entropy defined through the area of the apparent horizon of dynamical bulk black holes, and the expectation value of the energy-momentum tensor of the dual boundary quantum field theory. Such correlations are relevant in the pre-hydrodynamic stages of some initial data evolved in time, and seem to hold at least for strongly interacting conformal fluids, be they charged or neutral.

The determination of the weak mixing angle, $\sin^2θ_W$, at low momentum transfers remains a powerful test of the Standard Model and its potential new physics extensions. In this paper, we explore some physics opportunities at present and future reactor neutrino experiments through elastic neutrino-electron scattering (E$ν$ES). We assess the expected sensitivity to the weak mixing angle considering the CLOUD, TAO, and DANSS experimental configurations. We find that both CLOUD and TAO may achieve a precision that surpasses the current global fit from reactor experiments, while DANSS alone is expected to surpass the benchmark precision set by TEXONO measurement of the weak mixing angle. Additionally, we derive projected upper limits for the non-standard neutrino interactions (NSI), effective neutrino magnetic moment ($μ_ν$) and translate these into constraints on the neutrino transition magnetic moments ($Λ_i$). Our results demonstrate the physics potential of the E$ν$ES channel at current and upcoming reactor-based neutrino experiments.

Crystal electromagnetic calorimeters (ECALs) are essential for high-precision measurements of electrons and photons in particle physics experiments. However, the conventional design, in which long crystal bars point radially toward the interaction region and lack longitudinal segmentation, is incompatible with the three-dimensional shower imaging required by Particle Flow Approach (PFA). We propose a novel perspective on crystal ECAL design to address this limitation. The key innovation is a geometric reconfiguration in which crystal bars are oriented to face the interaction region and arranged orthogonally in adjacent longitudinal layers. This layout achieves fine spatial segmentation of energy deposits by correlating measurements of orthogonal crystal bars. An interleaved structure of regular and inverted trapezoidal modules is incorporated to maximize structural uniformity and detector hermeticity. This design is engineered to preserve the excellent intrinsic energy resolution of crystal ECALs while simultaneously providing the detailed three-dimensional shower imaging essential for PFA. Simulation results confirm the feasibility of achieving excellent energy resolution of $1.14\%/\sqrt{E} \oplus 0.44\%$. Consequently, the proposed design repositions crystal ECAL as a foundational component for PFA-oriented detector systems at facilities such as the Circular Electron Positron Collider (CEPC), offering a new technical pathway to advance the physics goals of future colliders.

Over the past few decades, data from leading neutrino experiments have firmly established neutrino oscillation, implying non-zero neutrino masses and leptonic mixing and thereby providing confirmed evidence of physics beyond the Standard Model. On the backdrop of the precision era of neutrino oscillation, this thesis underscores its relevance by demonstrating the physics reach of the forthcoming long-baseline experiments -- Deep Underground Neutrino Experiment (DUNE) and Hyper-Kamiokande (Hyper-K) -- to establish non-maximal $θ_{23}$, resolve the correct $θ_{23}$ octant, and improve the precision on $θ_{23}$ and $Δm^2_{31}$ by efficiently breaking parameter degeneracies. This is enabled by DUNE's high-resolution LArTPC detector and its wide-band beam, achieving sensitivity at a high confidence level compared to the global fits of world neutrino data. The combined analysis of DUNE and Hyper-K not only significantly enhances sensitivity to these phenomenological studies but also demonstrates their capabilities at lower exposures when operated together, relative to their nominal individual exposures. In addition, we investigate the impact of flavor-dependent long-range interactions arising from anomaly-free U(1)' extensions of the Standard Model, showing that although subdominant long-range interactions can substantially influence the sensitivity and precision of oscillation parameter measurements, the complementary strengths of DUNE and Hyper-K mitigate these challenges to a large extent.

We investigate the thermodynamic evolution and endpoint of black hole evaporation in the framework of metastring theory and its particle excitations, the metaparticles. Metaparticles arise as zero modes of metastrings propagating on modular (doubled) spacetime and obey a modified dispersion relation exhibiting intrinsic UV/IR mixing controlled by a duality scale mu. Using a generalized Bekenstein argument adapted to metaparticles, we derive quantum-corrected entropy contributions associated with geometric and dual (winding-like) sectors of the underlying phase space. When treated independently, these two entropy branches lead to an incomplete thermodynamic description, exhibiting unphysical behavior at small horizon area. We show that consistently treating the metaparticle as a single entangled quantum object -- rather than as two independent sectors -- naturally resolves these pathologies. We propose a pseudo-entangled total entropy that incorporates correlations between the geometric and dual sectors. The reality requirement of the entropy dynamically enforces a minimal horizon area and, equivalently, a minimal effective length scale associated with modular spacetime. The resulting black hole thermodynamics exhibits a finite maximal temperature, a divergence of the heat capacity signaling a continuous phase transition, and a shutdown of Hawking radiation through geometric channels, leaving behind a cold, stable remnant. Unlike matter-supported or curvature-bounded regular black holes, the remnant obtained here is non-material and non-geometric in nature, corresponding to a finite modular core of spacetime rather than a dust-filled interior. We compare this scenario with mimetic gravity and other non-singular black hole models, emphasizing the distinct role played by first-class constraints, entropy, and modular geometry in the present framework.

High-energy partons lose energy while propagating through the hot, strongly interacting medium produced in ultrarelativistic nucleus-nucleus collisions, leading to a suppression of particle production at high transverse momentum ($p_\mathrm{T}$). The dependence of this energy loss on the size of the colliding nuclear system has yet to be firmly established experimentally. This Letter presents a systematic study of charged-particle suppression across four different nucleus-nucleus collision systems using nuclear modification factors ($R_\mathrm{AA}$) measured by the CMS Collaboration at the CERN LHC. Previous CMS measurements of $R_\mathrm{AA}$ in oxygen-oxygen, xenon-xenon, and lead-lead collisions are recast with identical $R_\mathrm{AA}$ intervals and are complemented by the first measurement of the charged-particle $R_\mathrm{AA}$ in neon-neon collisions at $\sqrt{s_\mathrm{NN}}$ = 5.36 TeV. The neon-neon data correspond to an integrated luminosity of 0.76 nb$^{-1}$. The $R_\mathrm{AA}$ in all collision systems examined show similar qualitative trends, but have a magnitude which is ordered with the nucleon number A. The $R_\mathrm{AA}$ feature a downward slope at low $R_\mathrm{AA}$, a local minimum at around 5$-$7 GeV, and an upward slope with increasing $R_\mathrm{AA}$. The $R_\mathrm{AA}$ are also compared in terms of \cubicA, which is proportional to the nuclear radius. Models including only initial-state nuclear effects fail to reproduce the observed trends, whereas energy loss models reproduce the trends in the region $R_\mathrm{AA}$ $\gt$ 9.6 GeV.

Weak-interaction-mediated chiral imbalance generation in idealized massless electrons during core-collapse supernovae was once proposed to be the source of strong magnetic fields found in neutron stars. The effect goes by the name of chiral plasma instability (CPI). However, it was found that a finite electron mass damps out this process, inactivating the instability and preventing magnetic field growth. In this work we show that the instability can survive in the presence of abrupt density and temperature perturbation that drives the system sufficiently far out of weak equilibrium. As an example, we work with such perturbations generated by shockwaves which are common during both core collapse as well as neutron star mergers. We find that the chiral imbalance resulting from shock waves, under the right conditions of density and temperature, can sustain the chiral plasma instability despite the damping from the electron mass. Additionally, in an already magnetized medium, the chiral magnetic effect resulting from shock wave density and temperature perturbation can generate substantial ohmic heating. Our results imply that shockwaves during core-collapse supernovae and merging neutron stars can act as a source of strong heating in a magnetized medium as well as CPI.

It has been proposed by Sen that the D-instanton effects in string theory can be systematically determined through the framework of open-closed string field theory. We apply the latter formalism to analyze the D-instanton corrections to the quantum effective action of a D3-brane in type IIB superstring theory, and determine the leading single and multi-instanton contributions to the $D^4 F^4$ effective coupling which is unprotected by supersymmetry. Notably, while we find that the one-instanton contribution agrees with a conjecture of Green and Gutperle, the multi-instanton contribution disagrees with the conjecture.

The topological properties of field configurations in gauge theory contain important data about the (generalized) global symmetries of the theory as well as potential inconsistencies in the form of gauge anomalies. In this work we modify the topological classes of Abelian $p$-form fields, generating new global variants of gauge theories. These modifications implement constraints directly on the classifying space of the gauge field and its cohomology classes via homotopy fiber construction. This general approach allows us to investigate the universal effects of the constraints on the conserved global charges encoded in gauge characteristic classes. We further demonstrate that this procedure generically leads to new topological sectors introducing additional global charges and anomalies in the modified gauge theories.

We reformulate the treatment of density-dependent chemical potential shifts appearing in excluded-volume implementations of the hadron resonance gas model. An auxiliary classical representation is constructed in which a common energy shift is determined by preserving the scalar number density, ensuring thermodynamic consistency. Hadron radii are parametrized through a liquid-drop inspired mass-radius relation with two parameters: the pion radius and a scaling exponent. The resulting framework reproduces lattice QCD results for lower-order conserved-charge susceptibilities at zero chemical potentials with only two adjustable parameters.

Low-mass dwarf galaxies ($M_{\rm vir} \lesssim 10^9\rm\ M_\odot$) are fundamental cosmological building blocks, yet the physical processes driving their structural diversity remain poorly understood. Recent numerical simulations have suggested a diversity in the stellar-to-halo mass ratio in this halo mass range, but either the number of samples obtained from the same simulation setup or the numerical resolution was limited. We performed high-resolution cosmological zoom-in simulations for eight galaxies with a dark matter halo mass of $\sim 10^9\rm\ M_{\odot}$ up to $t=1.2$ Gyr at which most gas in the galaxies has been expelled. Our samples have a scatter of an order of magnitude in the halo mass at the reionization epoch. The stellar-to-halo mass ratio expected at $z=0$ scatters nearly two orders of magnitude with $5\times10^{-5}$ to $2\times10^{-3}$. We also observed variation in the compactness of their stellar distributions. Some of our simulated galaxies exhibit a stellar half-mass radius of $\sim30$ pc, which is as small as that of ultra-compact dwarfs. The formation condition for such a compact stellar distribution is understood as an analog of the condition for the formation of dense, massive star clusters. We found that when the central gas surface density exceeds a critical threshold ($Σ_{\rm gas} \gtrsim 30\rm\ M_\odot \rm\ {pc}^{-2}$), the star formation becomes highly efficient and results in dense stellar systems. These results suggest that UCDs can form in situ even in isolated dark matter halos.

Statistics of polarized synchrotron radiation carry information about the properties of the underlying turbulence. Different statistical measures constructed from observables probe turbulence properties in different ways. We consider a setup in which synchrotron radiation is emitted in a distant volume and then passes through a turbulent screen that induces Faraday rotation. Using both MHD simulations and synthetic turbulence spectra, we explore the spectrum of observed polarization directions measured at a single frequency as a diagnostic for recovering the statistics of turbulence in both the emitting region and the Faraday-rotation screen. We compare these results with our analytical expectations. We also compare the spectrum of polarization direction (SPD) with the wavelength-derivative diagnostic introduced and analytically explored by Lazarian \& Pogosyan. We demonstrate that the SPD exhibits greater sensitivity to turbulence in the Faraday screen. We provide an observer-friendly criterion to determine whether the SPD samples turbulence in the synchrotron-emitting region or in the Faraday screen. These results open a practical pathway for extracting turbulence statistics from existing and forthcoming single-band radio polarimetry.

In low-luminosity active galactic nuclei like M87* and Sgr A*, the accretion disk around the central supermassive black hole is tenuous and collisionless. As a result, the usual ideal magnetohydrodynamics (MHD) approximation may not be applicable. In this Letter, we report on the first fully kinetic simulations of the accretion process where the plasma initially has finite angular momentum. The simulated accretion flow behaves remarkably similarly to the magnetically arrested disk (MAD) regime of ideal MHD, reproducing episodes of magnetic flux saturation and eruption typical of MADs. The resemblance to fluid models owes largely to kinetic instabilities, which regulate pressure anisotropy in the disk, allowing fluid terms to dominate the angular momentum transfer. In addition, by handling vacuum regions effectively, our kinetic approach probes the matter supply to the jet funnel. We observe no efficient penetration of the accreting material into this region, which suggests that a pair discharge may be required to sustain the Blandford-Znajek process.

The large-scale dynamics of most conventional space and astrophysical plasmas are predominantly governed by Alfvén modes, which are low-frequency magnetohydrodynamic modes existing in magnetized media. At scales smaller than the ion gyroscale or frequencies exceeding the ion cyclotron frequency, the Alfvén modes transform into kinetic-Alfvén or whistler modes that significantly contribute to plasma dynamics. However, this scenario reverses in non-neutral pair plasmas, such as those found in the magnetospheres of pulsars and magnetars, around rotating black holes and in their relativistic jets, as well as in certain laboratory plasmas. In these systems, the large-scale dynamics is governed by hybrid whistler-Alfvén modes, which transform into pure Alfvén modes at smaller scales. We derive the nonlinear equations that describe the dynamics of whistler-Alfvén modes in ultrarelativistic non-neutral magnetically dominated pair plasma and discuss the spectrum of turbulence governed by these equations.

We report the discovery of strongly absorbed [OI]63um in a sample of 12 DSFGs at 4.2<z<5.8 selected from the SPT survey. This is the first systematic survey of the [OI]63um fine-structure line at z>4. Using ALMA Bands 9 and 10, we obtain spatially and spectrally resolved observations that probe the interstellar medium on sub-kpc scales. Despite reaching sensitivities 10-100x deeper than most previous studies, we detect [OI]63um in emission in only 2 sources at low significance, with the remaining galaxies yielding stringent non-detections over the full velocity range covered by robust detections of other far-infrared lines, including [CII] and [NII]205um. We identify several compact (0.05-0.2") regions having [OI]63um absorption against the far-infrared dust continuum, some of which are possibly reaching below rest-frame CMB radiation level. We also detect narrow, spatially localised [OI]63um emission "escape channels" preferentially detected in regions with weak or absent dust continuum emission. We predict that similar absorption effects may appear in the [CII] line, particularly when concentrating on the regions with the densest foreground material along the line of sight. The [OI]63um line appears to be originate from a mix of compact, high optical depth [OI]63um emitting regions and sub-thermally excited, oxygen-rich molecular clouds dispersed throughout high-redshift starbursts that are capable of absorbing the ground-state line emission. Combined with a comparison to cosmological radiation hydrodynamical simulations, this supports the interpretation that regions with higher gas and dust column densities may lead to weakening an intrinsically strong [OI]63um line emission. We argue that the high [OI]63um optical depth is the dominant effect causing the strong absorption, limiting the diagnostic power of this line to trace regions of massive star formation in high-redshift DSFGs.

Newly formed stars have a profound impact on their environment by depositing energy and momentum into the surrounding gas. However, only a fraction of the stellar feedback is retained in the cloud and observational constraints are needed to further our understanding of this process. In a sample of 19 nearby galaxies, we match HII regions from PHANGS$\unicode{x2013}$MUSE to their ionizing stellar source from PHANGS$\unicode{x2013}$HST and measure the percentage of ionizing radiation that is leaking into the surrounding diffuse ionized gas (DIG). Based on a catalogue, where each HII region is powered by a single young and massive stellar association, we measure a photon escape fraction of $f_\mathrm{esc}=82^{+12}_{-24}$ per cent. Comparable results are obtained when different procedures are used to match the ionized gas to its source. All samples we study contain a substantial fraction of objects (up to 20 per cent), where the stellar source is not sufficient to produce the H$α$ flux observed from the nebula. Many of them are probably related to uncertain age estimates, but we also find numerous regions, where a significant fraction of the ionizing photon budget is contributed by stars that reside outside the boundaries of the HII region. This motivates the use of an alternative galaxy-wide approach, in which we include all HII regions and stellar sources, not just the ones that show a clear overlap. When summing up the ionization budget over entire galaxies, we measure slightly lower, but consistent values.

The current Fermi-LAT source catalog (4FGL-DR4: 7194 sources over 14 years) was built incrementally from the 8-year catalog. In a survey mission like Fermi, data accumulate on each source over time, so after 16 years (reached in August 2024) and twice the data for the original 4FGL sources we have more precise localization (by 24% on average). It is thus time to generate a new original catalog, which implies, beyond adding the sources newly detectable after two more years, changing the existing source names (derived from their coordinates) and reviewing the associations. We present an early 16-year list (FL16Y) of 7220 sources, which relocalizes all sources and improves a few aspects of the catalog analysis, but still uses the same model of interstellar diffuse emission as 4FGL-DR4.

Overlapping galaxies, in which a foreground galaxy partially overlaps a background galaxy, offer a unique opportunity to measure dust attenuation, a key nuisance parameter in galaxy studies, empirically and in great detail by modelling the light of both the foreground and background galaxy and inferring the missing light in the overlapping region. However, the current catalogue of overlapping pairs is relatively limited in number compared to catalogues dedicated to individual galaxies. Expanding this catalogue is not only a necessity to facilitate further detailed dust studies beyond the few limited studies conducted thus far, but also to improve pair-to-pair variance and support automated identification through machine learning techniques. To achieve this, we utilise galaxies classified as "overlapping" from Galaxy Zoo DECaLS (GZD-1, -2, and -5), along with images from Data Release 10 (DR10) of the DESI Legacy Survey, in our individual citizen science project to classify these pairs directly using volunteers. This new catalogue will not only provide a wealth of targets for future dust studies but will also contribute to a deeper understanding of these pairs and dust as a whole.

We present results from a multi-epoch Very Long Baseline Array (VLBA) survey of compact radio sources in the Orion complex, conducted within both the DYNAMO-VLBA and the GOBELINS projects. Our observations detected 216 compact radio sources, of which 58 yielded reliable multi-epoch astrometric solutions. For these sources, we derived trigonometric parallaxes and proper motions with typical precisions of about 0.05 mas and 0.10 mas yr$^{-1}$, respectively. The measured parallaxes range between 2.26 and 2.65 mas, corresponding to distances of 380 - 440 pc, and delineate the depth of the Orion star-forming complex. We determine mean distances of $405\pm16$ pc for NGC 2068, $403\pm5$ pc for NGC 2024, $407\pm12$ pc for the $σ$ Orionis region, $388.5\pm1.7$ pc for the Orion Nebula Cluster (ONC), and $438\pm12$ pc for L1641. A comparison with Gaia DR3 astrometry for 28 common sources reveals negligible mean parallax offsets ($Δ\varpi=-0.02\pm0.01$ mas) and small systematic differences in proper motions ($\sim$0.07 mas yr$^{-1}$), likely due to residual rotation of the Gaia reference frame. Our results demonstrate the capability of high-precision radio astrometry to map embedded stellar populations and to provide an independent calibration of the Gaia reference system in obscured regions.

Repeating partial tidal disruption events (pTDEs) provide a direct probe of stellar orbits and episodic mass loss around supermassive black holes, but robust identification requires multi-band and multi-epoch evidence. We investigate whether the late-time radio rebrightening of the nuclear transient IGR J12580+0134 in NGC 4845 can be explained as a repeating pTDE, using multi-epoch Karl G. Jansky VLA observations together with X-ray constraints from Swift/XRT and NICER. The radio light curves show two distinct episodes, with the L-band peaks separated by $\approx1513$ days. Modeling the second episode with a synchrotron afterglow framework using Markov Chain Monte Carlo fitting favors a non-relativistic outflow $v\simeq0.03c$, with an isotropic-equivalent kinetic energy of order $10^{50}$ erg propagating in an approximately constant-density circumnuclear medium. No significant contemporaneous brightening is detected by Swift/XRT during the 2016 radio flare, while faint NICER flares in 2023 suggest intermittent low-level accretion. The recurrence timescale and radio energetics therefore make IGR J12580+0134 a possible candidate for a repeating pTDE system, motivating continued sensitive radio and X-ray monitoring to test future reactivations.

We present a spectroscopic study of low-mass galaxies (LMGs;$10^8\leq\rm M_*/M_\odot\leq10^9$) at $z\sim0.15$ in COSMOS field, and compare it to a control sample of intermediate-mass galaxies (IMGs;$10^9\leq\rm M_*/M_\odot\leq10^{10}$) at $z\sim0.35$. We examine their star formation rates (SFRs), dust attenuation properties, and the relationship between nebular and stellar reddening. For both samples, SFRs derived from H$α$ are strongly correlated with SFRs from fitting simple star formation histories (SFHs) to the galaxies' spectral energy distributions. In fitting a joint SFR-$\rm M_*$ relation, we obtain a slope of $\rm {Δlog(SFR_{Hα})}/{Δlog(M_*/M_\odot)}=1.01\pm0.03$, indicating that fair ensembles of SFHs for galaxies at these stellar masses are well-described by scale-free, self-similar forms. We also examine their dust attenuation properties and the relationship between nebular and stellar reddening, exploring how these quantities vary with stellar mass and specific SFR (sSFR). Nebular attenuation increases with stellar mass for IMGs but is lower and less mass-dependent in LMGs, consistent with their reduced dust content. In all cases, stellar continuum attenuation is lower than nebular attenuation, as expected from the two-component dust model. The nebular-to-stellar color excess ratio in both samples is consistent with the canonical factor of 2.27. The ratio is mass-independent, but rises with sSFR in IMGs and remains constant in LMGs. These results suggest that in LMGs, efficient dispersal of birth clouds keeps the differential attenuation approximately constant across sSFR. Thus, although LMGs follow the same global SFR-$\rm M_*$ scaling as massive galaxies, their lower dust content and feedback-maintained ISM produce distinct attenuation behavior relative to IMGs.

We update and expand the MESA Isochrones and Stellar Tracks (MIST) database to include variations in the alpha-capture elements, specifically [alpha/Fe]=-0.2, 0, +0.2, +0.4, and +0.6 for -3 <= [Fe/H] <= +0.5. Variations in [alpha/Fe] are included in a self-consistent manner from the stellar interior models to the synthetic spectra used to translate these models in the observational plane. We describe a number of updates to the physics utilized in these models as well as new information provided by the models. We validate the models with comparisons to other stellar evolution models including the previous generation of MIST and other models from the literature. MIST data products including stellar evolutiont tracks, isochrones, and bolometric correction tables can be obtained from the MIST project website, https://mist.science. All necessary files to reproduce MIST models are available from Zenodo.

We present scientific and technical justifications of a potential coordinated Euclid/CSST/JPCam/JUST survey of the Euclid Deep Field North (EDF-N), aimed at probing the multi-phase circumgalactic and intergalactic medium (CGM/IGM) at the cosmic noon over ~20 deg$^2$. The survey is structured around three connected goals: (1) improving photometric redshift (photo-z) accuracy through the combination of broad- and narrow-band photometry, enabling reliable identification of large-scale structures; (2) probing extended CGM emission with dedicated narrow-band imaging; and (3) mapping foreground IGM via absorption-line spectroscopy of background galaxies. Together, these components establish an integrated observational framework to investigate galactic ecosystems -- linking galaxies to their circumgalactic and intergalactic environments -- at cosmic noon. We show that the J-PAS-like narrow-band system used in JPCam substantially improves photo-z accuracies from only the Euclid/CSST broad-band data, especially for star-forming galaxies at z~1.0-1.4. This enables the identification of galaxy groups and (proto-)clusters directly from photo-z measurements. Stacked JPCam narrow-band imaging should also detect extended [O II]-emitting CGM halos. We then construct mock 3D gas distribution model and realistic galaxy catalog, and further construct mock CSST and JUST background galaxy spectra adding Lyalpha and Mg II absorptions. The reconstructed 3D H I field from CSST Lyalpha forest reliably recovers large-scale structures; however, our simulations indicate that detecting diffuse IGM Mg II absorption with JUST is infeasible, either through spectral stacking or via the two-point correlation function method. We conclude that constraining the metallicity of the diffuse IGM will require significantly deeper and higher-resolution spectroscopy expected from future facilities such as the 39 m E-ELT.

We present the largest collisionless $N$-body cosmological simulations in a MOdified Newtonian Dynamics (MOND) cosmology to date. Our 4 simulations cover $Λ$CDM as a baseline, a MOND with hot dark matter model known as $ν$HDM, and 2 unphysical models we call $Λ$HDM and $ν$CDM to test the individual contributions of hot dark matter and MOND gravity, respectively. $ν$HDM reproduces the CMB power spectrum whilst also theoretically matching cluster dynamics and preserving MOND predictions for galactic rotation curves. We test its viability on cosmological scales using simulations with $256^{3}$ particles in a box of size $800/h$ comoving Mpc. We find generically that the MOND models massively overproduce large-scale structures by $z=0$, with a most massive cluster in $ν$HDM of $\approx 5 \times 10^{17} M_{\odot}/h$ and typical peculiar velocities of several thousand km/s. We also explore a local void solution to the Hubble tension in these models. Analogues to the observed "Local Hole'' do form in the MOND models, but values for the deceleration parameter $<-1.5$ in these regions prevent a satisfactory resolution to the Hubble tension. Whilst $Λ$CDM significantly underpredicts the observed bulk flow in Cosmicflows-4, the high peculiar velocities that arise in the MOND models create the opposite problem, ruling out $ν$HDM at $>5σ$ confidence. Observations clearly require a much milder enhancement to the rate of structure growth in $Λ$CDM than is provided by the $ν$HDM paradigm. Our results also suggest that replacing cold dark matter with hot dark matter is unlikely to provide a viable cosmological model, regardless of the gravity law.

The dark photon is a new hypothetical gauge boson arising in extensions of the Standard Model, and constitutes a compelling dark matter candidate. As dark photon dark matter (DPDM), it can interact with electromagnetic fields via kinetic mixing, and the inelastic scattering process $γγ' \to e^+ e^-$ becomes kinematically allowed for gamma rays above a characteristic energy threshold. This interaction imprints unique spectral attenuation features at very-high-energies (VHE), offering an observational probe of DPDM models. Using the Cherenkov Telescope Array Observatory (CTAO) Instrument Response Functions (IRFs), we simulate observations of VHE sources and forecast novel sensitivities to the kinetic mixing parameter for the photon-dark photon scattering process. Our study focuses on three key astrophysical targets: the Crab Nebula and the blazars Markarian 421 and Markarian 501. Additionally, we investigate the impact of dark matter spikes around black holes on the upper limits. Our results demonstrate that CTAO can probe the DPDM parameter space down to a mixing parameter of $\varepsilon \sim 10^{-8}$ for masses around $m_{A^{\prime}} \sim 10^{-1}~\textrm{eV}$ through high-energy spectral attenuation, at a $95\%$ confidence level.

Cosmological simulations suggest that various galaxy properties depend on their location within the cosmic web. Yet direct observational evidence of the dependence of star formation activity on distance to filaments remains scarce and is missing at z>1. We investigate how starburst, main-sequence (MS), and quenched galaxies are distributed with respect to cosmic web filaments, and how this distribution evolves with redshift. We first use the SIMBA cosmological simulation to predict the redshift evolution of the mean distance to the closest filament from z=3 to z=0 for different galaxy populations after removing stellar-mass dependencies. We then measure the corresponding signal in the COSMOS field, using COSMOS2020 and COSMOS-Web data, where accurate photometric redshifts enable reconstruction of the projected cosmic web from z=2 to z=0.5, and starbursts are identified through far-infrared spectral energy distribution fitting. In agreement with the results from SIMBA, starburst galaxies are found closer to filaments at z>1 and at larger distances at z<1, MS galaxies occupy intermediate environments with little evolution, and quenched galaxies show progressively shorter distances to filaments toward low redshift, with a crossing between starburst and MS populations around z~1. In COSMOS-Web, the relative evolution in the average distance to filaments between starburst and MS galaxies is detected at a significance level of at least 5σ. We show that a minimal toy model in which the only environmental ingredient is the sSFR-filament distance modulation measured in simulations is sufficient to reproduce the observed differential evolution of the average filament distance between starburst and MS galaxies. These results show that the imprint of large-scale environmental effects on the star formation activity of galaxies, predicted by simulations, is detectable from z=2 down to z=0.5.

A significant fraction of stars in both the Galactic field and stellar clusters are members of binary systems. Understanding their properties is therefore essential for a comprehensive view of stellar structure, evolution, and cluster dynamics. Despite extensive studies of cluster binaries, key issues remain unresolved, particularly for photometric binaries among low-mass stars. While the binary fraction in the field strongly depends on stellar mass, cluster studies have generally suggested an approximately constant fraction over the limited mass ranges explored. In addition, the mass function (MF) of very low-mass stars is still poorly constrained in clusters older than a few hundred Myr. We use deep Hubble Space Telescope imaging of the intermediate-age open cluster NGC 2158 to investigate its binary population and derive the luminosity and MFs down to ~0.14 solar masses, enabling the first detailed analysis of binaries in this cluster. We measure a global binary fraction of 38%, consistent with other open clusters, and find a clear mass dependence: it decreases from ~52% at 1.0 solar masses to ~11% at 0.2 solar masses. This trend mirrors that of Galactic field stars, suggesting similar binary properties. The MF is characterized by three regimes: high-mass stars (alpha= -2.49 +- 0.19), low-mass stars (alpha= -1.11 +- 0.09), and very low-mass stars (alpha= -0.08 +- 0.07). The slope change near 1.0 solar mass agrees with recent surveys, though we find a deficit below ~0.3 solar masses. We also detect a main-sequence discontinuity around ~0.3 solar masses, possibly linked to the 3He-driven instability predicted by stellar models and analogous to the Jao Gap seen in nearby field stars.

It has been 10 years since the initial discovery of Ultra-Diffuse Galaxies (UDGs) in the Coma cluster and the revelation that large, low surface brightness galaxies may constitute a greater fraction of galaxies than first thought. This left an open question: Are UDGs something special, or just an extension of the previously known dwarf galaxy population? Seeking to answer this question, in the decade following, dedicated simulations have studied and proposed a myriad of formation pathways to create UDGs. Observations have then pushed the limits of world-class observatories to perform detailed studies of these galaxies in large numbers across the full range of environments in the local Universe. These observations stress test simulations and challenge previous galaxy formation wisdom, with UDGs posing many open puzzles beyond just their unknown formation mechanism. To provide a few pertinent examples: there is observational evidence that not all UDGs follow the standard stellar mass -- halo mass relationship; there is evidence for UDGs with extraordinarily high levels of alpha enhancement; and there is evidence that some UDGs are much more globular cluster rich than other dwarfs of similar stellar mass. In this Dawes review, we undertake the task of summarising the decade of science since the discovery of UDGs. We focus on the quiescent population of UDGs and review their general properties, their proposed formation scenarios, their internal properties and their globular cluster systems. We also provide a brief conjecture on some future directions for the next decade of UDG research.

We forecast cosmological parameter constraints for a cosmic shear analysis of the Rubin Observatory Legacy Survey of Space and Time (LSST), defining an analysis framework that can accurately recover the $Λ$CDM model in the presence of astrophysical and data-related systematics. When accounting for our present uncertainty on the suppression of the non-linear matter power spectrum through baryon feedback, we find that the error on the composite parameter $S_8=σ_8\sqrt{Ω_{\rm m}/0.3}$ almost doubles compared to an LSST analysis which neglects this astrophysical phenomenon. After the first year of observations, LSST will extend beyond the magnitude limit of existing representative spectroscopic calibration samples, requiring photometric redshifts to be calibrated using an alternative strategy. Adopting literature measurements of the reduced redshift calibration precision found from galaxy cross-correlation techniques, combined with current levels of baryon feedback uncertainty, we forecast final year LSST cosmic shear constraints that barely improve upon the first year analysis. This forecast therefore serves as encouragement to the community to develop methodology and observations to constrain models of baryon feedback and enhance photometric redshift calibration at depths where spectroscopy is unrepresentative. With tight priors on both these systematic terms, we forecast that LSST cosmic shear can deliver constraints on $S_8$ that are more than five times as constraining as existing cosmic shear surveys.

A common task in astronomical research is to estimate the physical parameters (temperature, mass, density etc.) of a gas by using observed line emission. This often requires a calculation of how the radiation propagates via emission and absorption (so-called radiative transfer). In radio and infrared astronomy, the Fortran code RADEX (van der Tak et al., 2007) is a popular tool to solve the non-LTE radiative transfer of a uniform medium in a simplified geometry. I present pythonradex, a Python re-implementation of RADEX. Written in Python, it provides an easy and intuitive user interface, improved performance as well as additional functionality not included in RADEX, such as continuum effects and overlapping lines. In addition, pythonradex provides a self-consistent computation of the total flux for all geometries, including spherical geometries.

All gas-rich galaxies in cluster environments are expected to experience ram-pressure stripping from the intra-cluster medium. However, only a fraction of these develop ongoing star-formation in their stripped tail, becoming the so-called ``jellyfish'' galaxies. In this work we provide observational evidence that magnetic fields can signal differences in the extraplanar star formation and explore what are the physical conditions that lead to the formation of a jellyfish galaxy. We first focus on JO147, a jellyfish galaxy that features weak star formation activity in its tail. Using MeerKAT radio continuum observations, we discover polarized emission only in a small fraction of its tail, with an average fraction of $~10\%$, and a low Mach number $\mathcal{M}=1.3-1.6$, which suggests a possible association between magnetic field draping, shock-compression of the gas, and extraplanar star formation activity. Then, we test this scenario in a sample of 17 jellyfish galaxies from the GASP project. We combine dynamical models for their orbits within the host clusters with realistic cluster temperature profiles to infer their Mach number, and we find a positive correlation between it and the star formation activity in their tail. We conclude that supersonic motion is a necessary condition for triggering star formation in the stripped tails of jellyfish galaxies. Our findings provide empirical evidence that the critical factor preventing the stripped gas evaporation is the shock compression induced by the supersonic motion through the cluster. This process likely enhances the magnetic field surrounding the galaxy and the properties of the stripped material.

Magnetized rotating neutron stars, or pulsars, are a possible end product of massive star evolution. Their relativistic wind successively interacts with the supernova ejecta of their defunct progenitor, then with the circumstellar medium of the progenitor, and eventually with the interstellar medium. If a massive star is static with respect to its ambient medium, then its resulting circumstellar medium is elongated along the direction of the local magnetic field, and its supernova remnant transiently appears as a rectangle. The pulsar wind nebula forming in it is, in its turn, elongated, as long as the pulsar axis of rotation matches the direction of the local magnetization. In this work, we explore how the angle between the direction of the local magnetic field of the interstellar medium and the pulsar axis of rotation influences the shaping of its pulsar wind nebula with 3D MHD simulations are carried out with the PLUTO. We use those models to perform radiative transfer calculations to derive non-thermal radio emission maps of the pulsar wind nebulae. When the polar elongation of the pulsar develop, they bend in opposite directions under the effects of the cavity carved by the stellar wind and already filled by supernova ejecta. This induces a complex distribution of magnetized supernova ejecta and pulsar wind, resulting in various observable structures, appearing as rectangles, circles, or irregular oblong shapes, in the radio waveband. The angle between the direction of the pulsar rotation axis and that of the local ambient magnetization is a governing parameter for the shaping and non-thermal radio properties of the pulsar wind nebulae of static massive stars; however, the mixing of material, once the pulsar wind nebula is old (50 to 80 kyr), is not strongly affected by that factor.

Dust grains and polycyclic aromatic hydrocarbons (PAHs) actively contribute to the thermodynamics, chemistry, and radiative state of the interstellar medium (ISM), yet most ISM models and galaxy simulations either exclude them altogether or adopt simplified treatments. We present CALIMA, a new module for dust and PAH formation and evolution in radiation-hydrodynamics simulations for RAMSES, designed to self-consistently couple dust physics to radiative transfer and non-equilibrium thermochemistry in a multiphase ISM. The model employs a two-size, two-composition dust framework with log-normal grain populations, explicitly evolving stellar dust injection, turbulence-informed gas-phase accretion, shattering, coagulation, thermal and non-thermal sputtering, and shock destruction, while PAHs are separate components with their own evolution. The evolving dust populations and radiation field determine local, wavelength-dependent opacities, photoelectric heating efficiencies, grain-assisted recombination, dust-gas collisional heating/cooling, and H$_2$ formation on both grains and PAHs. Updated treatments of thermal sputtering and collisional cooling that include finite grain sizes and modern ion-solid physics reduce sputtering rates at high temperatures and extend the regime where dust significantly cools hot gas. One-zone ISM tests show that dust and PAH evolution modifies classical thermal phase diagrams and C-bearing chemistry, while isolated disc galaxy simulations reveal environment-dependent variations in dust-to-metal ratio, small-to-large grain ratio, PAH fraction, and interstellar radiation field intensity that drive non-trivial structure in infrared emission, UV transparency, and H$_2$ formation. CALIMA provides a physically motivated framework to interpret dust- and PAH-based observables and to assess dust-mediated feedback in galaxy formation across cosmic time.

We study the formation of stellar bars using 145 simulations of disc galaxies embedded in live and static dark matter haloes. We use the exponential bar growth timescale, $τ_{\rm bar}$, to quantify how disc structure and kinematics regulate the onset and rate of secular bar formation. We extend previous work to thicker and more turbulent discs, motivated by those observed at high redshift ($z>1$). By revisiting several commonly used disc stability criteria - the Efstathiou-Lake-Negroponte parameter ($ε_{\rm ELN}$), the Ostriker-Peebles ratio ($t_{\rm OP}$), and the disc stellar mass fraction within 2.2 disc scale radii ($f_{\rm disc}$) - we find that $τ_{\rm bar}$, when expressed in terms of the disc's orbital period, follows a tight power law with each criteria. In Milky Way-like discs embedded in live haloes, bars form within a Hubble time if $f_{\rm disc} \geq 0.18$, $t_{\rm OP} \geq 0.27$, and $ε_{\rm ELN} \leq 1.44$. We show discs with higher velocity dispersion experience delayed bar growth and introduce an empirical relation that correctly describes the bar formation timescales of all our live halo models. Bars in static haloes grow at roughly half the rate of those in live haloes and require substantially greater disc instability to do so.

QU-fitting is a powerful tool for interpreting spectro-polarimetric radio continuum observations by linking them to physical models, enabling estimates of the magnetic fields in, for example, the Milky Way, galaxy clusters, and radio jets. We present a comprehensive investigation into the effectiveness and limitations of QU-fitting within the ASKAP POSSUM survey frequency ranges (800-1088 MHz and 1296-1440 MHz) with projections to other spectro-polarimetric radio observations. We simulate different physical polarization sources: Faraday simple, Burn slab, internal turbulence, external turbulence, and two-component models in the POSSUM frequencies, and assess their observational degeneracies and fit accuracies. Our results highlight the model-dependent nature of reliable fitting and identify specific regions of parameter space where model selection, and therefore characterization of the physical medium, becomes ambiguous. For QU-fitting we find the Bayes factor, computed using the marginal likelihood, outperforms more traditionally used goodness-of-fit metrics such as Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), and chi-squared for model selection. We provide empirical relationships to delineate the boundaries where model distinguishability is impossible. Finally, we evaluate how accurately QU-fitting recovers model parameters and their associated uncertainties, thereby assessing its ability to correctly characterize the Faraday-rotating medium in both point and extended sources in Faraday depth space.

We present a sound horizon independent measurement of the Hubble parameter using a multiprobe large-scale structure analysis. Removing the dependency on the sound horizon with a rescaling procedure at the matter power spectrum level, we analyse the BOSS full-shape power spectrum and bispectrum (for the first time) using the effective field theory of large-scale structure up to one loop. We combine this analysis with the auto- and cross-angular power spectra from the DESI Legacy Imaging Survey DR9, the $3 \times 2$pt analysis from DES Y3, and the CMB gravitational lensing power spectrum from Planck PR3. Our baseline analysis, that does not rely on supernovae data, yields $h = 0.702^{+0.022}_{-0.024}$, $Ω_m = 0.310 \pm 0.013$, and $σ_8 = 0.799 \pm 0.020$, corresponding to $3-4 \%$ precision measurements. When adding supernovae data from Pantheon+, we obtain a $2.6 \%$ measurement of $h$, with $h = 0.686 \pm 0.018$. We further note that our EFTBOSS analysis indicates a slight deviation of the BAO scale parameter (at $1.8 σ$) from its $Λ$CDM value, caused by the small scales of the bispectrum. We finally use the sound horizon-free EFTBOSS analysis as a diagnosis for the presence of new physics, finding that our results are consistent with the recent hints of evolving dark energy.

On the periphery of galaxy clusters, moderately high galaxy densities and velocity dispersions favour interactions and mergers that influence galaxy evolution prior to cluster infall. Observational studies of this phase in dwarfs remain rare. We present a high-resolution study of the merger remnant VCC 693 in the outskirts of Virgo cluster, using observations from the Atomic gas in Virgo Interacting Dwarf galaxies (AVID) project. We explore the origin of VCC 693 and the consequences of the merger on its star formation and structure through a joint analysis of VLA and FAST HI emission line observations, together with complementary optical imaging and spectroscopy. We employ hydrodynamical simulations to help interpret the observations. Our analysis favours a near-major merger between two dwarfs with a stellar mass ratio of 3:1-4:1, with one likely gas-poor progenitor (i.e., a damp merger). The optical appearance of VCC 693 is dominated by complex tidal structures throughout the system, whereas the HI gas has settled to a regular rotating disk. Compared with similar-mass dwarfs, the central star formation and gas-phase metallicity are moderately enhanced. The global star formation rate, HI gas content, and HI-to-optical size ratio of VCC 693 are broadly consistent with those of typical dwarfs of similar mass, albeit somewhat lower. Decomposition of the HI rotation curve into baryonic and dark matter indicates a high halo concentration, suggesting post-merger relaxation into a more centrally peaked configuration. Together with two recent studies of AVID post-merger systems, these results support the view that even major dwarf mergers can produce remnants with overall stellar structures indistinguishable from ordinary dwarfs, and that the environmental effects in cluster outskirts can promote damp or mixed mergers, constituting an integral part of galactic pre-processing.

We present results from coordinated multiwavelength observations of the SS 433, obtained with XRISM, optical telescopes, and near-infrared camera during 2024 April and 2025 March. The XRISM exposures amounted to ~200 ks in 2024 and ~100 ks in 2025. With XRISM/Resolve's high spectral resolution and large effective area, we clearly resolved numerous emission lines even in short time segments, achieving improved accuracy in Doppler-shift measurements relative to earlier observations. The simultaneously obtained X-ray and optical Doppler shifts suggest a possible tendency for the optical emission to lag slightly behind the X-rays. In the Resolve data, the Doppler shifts of the two jet components exhibited apparent asymmetries, with jet speeds fluctuating around ~0.26$\pm$0.01$c$ in 2024 and ~0.30$\pm$0.01$c$ in 2025. The velocity variations indicated modulations on a timescale of ~6.3 d, with a phase offset of about -90$^{\circ}$ relative to the nutation cycle. The observed line widths and flux of the approaching and receding jets appear consistent with the expected geometrical effects, indicating systematically larger line widths in the inner regions of the jets, as proposed by Shidatsu et al. (2025). Optical light curves show flares of ~400 s in 2024 and ~1600 s in 2025, with amplitudes up to ~15% during out-of-eclipse intervals, while the XRISM/Xtend light curves show no significant variability within the overlapping intervals and given the statistical uncertainties. Near-infrared photometry in 2024, obtained during an out-of-eclipse interval at a different epoch from the optical observations, showed no flare-like variability, and the X-ray band also remained constant within uncertainties. These coordinated observations provide a foundation for future XRISM studies aimed at probing the dynamical properties of the relativistic jets in SS 433.

The GW190814 event, involving a black hole of mass $22.2$--$24.3 M_{\odot}$ and a compact object of mass $2.50$--$2.67 M_{\odot}$, challenges our understanding of the mass gap between the heaviest neutron stars and the lightest black holes. If the secondary is a neutron star exceeding $2.5 M_{\odot}$, hyperons are likely to appear in its core, softening the equation of state. Rapid rotation can offset some of this softening, enabling higher maximum masses, but it may simultaneously excite the Chandrasekhar--Friedman--Schutz $r$-mode instability. Bulk viscosity arising from nonleptonic weak interactions in hyperonic matter provides an efficient damping mechanism that can stabilize such configurations. In this work, we investigate the combined effects of rotation, thermal evolution, and hyperon-induced bulk viscosity on the stability of massive neutron stars. We demonstrate a direct connection between the suppression of $r$-mode instabilities and the long-term dynamical stability of hyperon-rich stars, offering a plausible interpretation of the GW190814 secondary as a rapidly rotating, hyperon-rich neutron star rather than a low-mass black hole. Our unified framework extends beyond previous studies restricted to static equations of state or extreme viscous damping assumptions, providing new insights into the stability of massive, exotic neutron star configurations.

We study AT2024kmq and AT2024lhc, two tidal disruption events (TDEs) with blue featureless spectra associated with high-mass black holes ($M_{\rm BH}\sim 10^8\,M_\odot$). Both events show optical precursors consistent with shock dissipation from stream self-intersection. Their X-ray emission is luminous ($L_{\rm X}\sim 10^{44}\,{\rm erg\,s^{-1}}$), highly variable (with minimum observed variability timescales of 1.3\,hr and 4.8\,hr for factor of $\sim3$ flux changes), long-lasting ($>1\,\rm yr$), emerging no later than the optical peak, and well characterized by power-laws with $1.7<Γ<3$ (where $f_ν\propto ν^{1-Γ}$). The X-ray properties and radio non-detections support a compact corona ($\lesssim 10 r_{\rm g}$) producing Comptonized X-ray emission. Using all published featureless TDEs, we find statistically significant bimodality in the distribution of their peak UV/optical blackbody luminosities and radii. We assemble a comparison TDE sample with early-time X-ray observations with eROSITA, in which we find different $M_{\rm BH}$ distributions in TDEs with different X-ray spectral evolution properties: low-mass black holes ($M_{\rm BH} \sim 10^6 M_\odot$) remain soft ($Γ>4$) within $t\lesssim 2$\,yr, intermediate masses ($\sim 10^7 M_\odot$) transition from soft to hard at $\sim$1 yr, while high masses ($\sim 10^8 M_\odot$) are hard ($1.5<Γ\lesssim 3$) from the outset. We interpret this result as evidence that the soft-to-hard state transition in TDEs occurs at the critical threshold of $\dot{M}_{\rm acc} \sim 0.03 \dot M_{\rm Edd}$ (similar to X-ray binaries), using the fact that the transition timescale predicted by simple disk theory scales with black hole mass as $t_{\rm tr}\propto M_{\rm BH}^{-3/4}$.

HATPI is a recently commissioned time-domain facility at Las Campanas Observatory, Chile, that uses 64 wide-angle, 9.6 cm diameter lenses and back-illuminated CCDs, yielding a mosaic field-of-view of 7,100 square arcdegrees, observing the night sky at a cadence of 45 s and a spatial scale of 19.7 arcsec pixel$^{-1}$. In this paper, we present moving object time-series photometry with this facility, focusing on the interstellar comet 3I/ATLAS. 3I/ATLAS was first robustly recovered by HATPI on the night of 2025 July 2 (one night after its discovery) at a Gaia $G$-band magnitude of $G = 17.796 \pm 0.082$ mag ($\pm 0.030$ mag systematic uncertainty). The comet then increased in brightness to $G = 14.071 \pm 0.073$ mag $\pm 0.030$ mag by 2025 Sep 13, after which it became unobservable by HATPI as it approached perihelion. Before 3I/ATLAS achieved a brightness of $G = 16.396 \pm 0.029$ mag $\pm 0.030$ mag on 2025 Aug 6, it could be detected when stacking all HATPI observations from a single night, while after this date it is sufficiently bright to detect in individual 45 s exposures. We do not detect evidence for significant short-time-scale variations in the brightness of 3I/ATLAS after Aug 6. Compared to other light curves in the literature, the HATPI photometry exhibits a somewhat steeper rise in brightness with decreasing heliocentric distance, $r_{H}$. The HATPI magnitudes are well-fit as a power law function of $r_{H}$, with an exponential index of $n = 5.167 \pm 0.095$, over the range $2.14$ AU $ < r_{H} < 4.44$ AU, compared to $n = 3.94 \pm 0.10$ when fitting together with other literature observations. We find that the phase function is constrained to $β= 0.0552 \pm 0.0032$ mag deg$^{-1}$.

We investigate the gas motions in the core region of the Abell~3395 South subcluster (A3395S) using high-resolution X-ray spectroscopy with XRISM/Resolve. By analyzing the Fe~XXV He$α$ emission line, we directly measure the line-of-sight bulk and turbulent velocities of the intracluster medium. We find that the one-dimensional turbulent velocity is low, at the level of $124\pm21~{\rm km\,s^{-1}}$, while a significant line-of-sight bulk velocity of $263\pm23~{\rm km\,s^{-1}}$ is detected. The coexistence of low turbulence and finite bulk motion suggests that A3395S has not yet reached a dynamically relaxed state. These results are consistent with the non-detection of a radio halo in A3395S, implying that turbulent particle reacceleration is currently inefficient in the cluster core. This study demonstrates that high-resolution X-ray spectroscopy with XRISM provides a powerful means to directly constrain intracluster medium dynamics in merging galaxy clusters, and it provides a reference for future comparative studies of A3395N and A3391 within the same large-scale structure.

High-resolution tri-axial simulations are indispensable for realistically co-modeling the dynamical signatures and the radiative fingerprints of astrophysical jets, which are becoming increasingly important in modern computational studies of jet physics. However, such simulations impose extreme computational requirements that often exceed the capabilities of conventional CPU-based codes. GPU-accelerated simulations offer a transformative solution to mitigate these limitations. In this work, we present a detailed performance benchmarking of the recently developed GPU-enabled PLUTO code (gPLUTO), demonstrating runtime speed-ups ranging from an order of magnitude to (approximately) over 30 relative to CPU-only configurations. A direct comparison between computations of extragalactic jet propagation performed at different grid resolutions confirm the physical fidelity and production readiness of the gPLUTO code, while underscoring the importance of resolving the jet radius adequately to capture the jet dynamics accurately. Leveraging GPU-PLUTO's capabilities, we finally present an application by performing high-resolution simulations of giant radio galaxy jets (GRGs $\gtrsim 1$ Mpc), representing the first such well-resolved 3D study to our knowledge (resolving scales down to 500 pc). These simulations probe a range of environmental effects on GRG jets, clarifying their formation from central galaxies within host cosmic structures, rapid peripheral expansion, and the development of asymmetric cocoon morphologies.

IceCube has detected neutrino emission from the Galactic Plane (GP) at a significance of $4.5σ$, though its origin remains uncertain. Utilizing ten years of IceCube muon-track data, we investigate potential correlations between the GP neutrinos and $γ$-ray sources in the first LHAASO catalog (1LHAASO). To avoid issues caused by spectral extrapolation, this analysis focuses on sources detected by the Water Cherenkov Detector Array (WCDA). We employ an unbinned likelihood analysis to search for neutrino emission and constrain the hadronic $γ$-ray component of these sources. Neither single-source searches nor stacking analyses reveal significant neutrino signals. The stacking analysis indicates that the 1LHAASO WCDA population contributes at most $\sim$20\% to the diffuse GP neutrino flux measured by IceCube. The total hadronic contribution to the cumulative $γ$-ray emission from all WCDA sources is constrained to be at most $\sim$$60\%$, suggesting a predominantly leptonic origin for the $γ$-ray emission from the LHAASO source population. Even accounting for unresolved sources below the detection threshold, we estimate the total neutrino flux from all discrete sources (resolved plus unresolved) reaches at most about 40\% of the observed GP neutrino flux. These results support that the bulk of the GP neutrino emission is mainly from truly diffuse processes, i.e., cosmic-ray interactions with the interstellar medium, rather than from unresolved point sources.

GRB 250221A is a short gamma-ray burst (GRB) at redshift $z=0.768$, with a duration of 1.8 s and no extended emission in either Swift/BAT or Konus-Wind bands. A remarkable re-brightening feature in both optical and X-ray bands was observed at $\sim 0.6$ days after the burst trigger, but no supernova or kilonova signature was detected. The burst properties and empirical correlations or distributions (e.g., duration, spectral hardness, location in the Amati correlation, $\varepsilon-$value, $f_{\rm eff}$ parameter, and physical offset) favor a compact binary merger origin. However, a dense circumburst medium with $n\sim 80\rm~cm^{-3}$, obtained by adopting the energy injection into a jet to interpret the late-time re-brightening is inconsistent with the compact binary merger origin. In this paper, we propose a two-component jet model to explain the multiwavelength afterglow observations of GRB 250221A, in which the relativistic narrow jet ($\rm θ_{c} \sim 3.8^\circ$) produces the prompt and the early decay afterglow emission, while the mildly relativistic wide jet ($\rm θ_{w} \sim 4.4^\circ$) dominates at later times, resulting in the observed re-brightening feature. If this is the case, one can obtain a lower medium density with $n\sim 0.72\rm~cm^{-3}$ which is a little bit higher than that of short GRBs in merger environments, but falls into the reasonable and acceptable range. Finally, a possible kilonova emission is also discussed within the scenario of compact star merger origin of GRB 250221A.

We present independent imaging analyses of Event Horizon Telescope (EHT) observations of the active galactic nuclei in radio galaxy Centaurus~A and quasar 3C~279 using Generative Deep learning Image Reconstruction with Closure Terms (GenDIReCT), a recently developed machine-learning framework built on conditional diffusion models that uses interferometric closure invariants as primary observables. For Centaurus~A, our reconstruction reveals two prominent emission ridges ($\simeq 80\,μ$as each) along the jet sheath with a brightness ratio of $1.4\pm 0.1$ and an opening angle of $12.3\pm 0.3$~deg. For 3C~279, we identify three distinct components in the image, with the southern jet ejecta on sub-parsec scale exhibiting a proper motion of $4.6\pm 1.0\,μ$as over $\approx 5.39$ days away from the northern components, corresponding to an apparent superluminal velocity of $\simeq 10\pm 2$ times light speed. These measurements are consistent with those reported by the EHT Collaboration. The results are significant because we demonstrate that: (1) imaging from interferometric aperture synthesis data, especially in VLBI and most acutely in extremely sparse arrays like the EHT, remains a severely ill-posed and challenging inverse problem, yet closure invariants preserve robust morphological information that can strongly constrain structural features, and (2) more importantly, closure-invariant imaging largely avoids calibration systematics, thus providing a fundamentally independent view of spatial structure with very high angular resolution. The generative nature of GenDIReCT further allows us to sample and characterise clusters of plausible image solutions for each dataset. As a calibration-independent, generative imaging approach, GenDIReCT offers a robust and truly independent blind-imaging tool for current and future VLBI experiments.

It is widely expected that the obscured accretion stage can be the initial turning-on stage of active galactic nuclei from quiescent galaxies. We present mid-infrared light curves of NGC 6447 in 3.5$μ$m and 4.6 $μ$m bands observed by WISE/NEOWISE, which show an almost monotonic increasing trend of 1.2 mag over 14 years. The optical light curve from ASAS-SN during the same period is consistent with a constant showing no variability. The mid-infrared color evolution shows that the galaxy transitioned into an active galactic nucleus (AGN) in 2018. The SPHEREx spectrum reveals an increasing continuum resembling warm to hot dust emission from an AGN. NuSTAR detected an X-ray source with a 2-30 keV luminosity of $8.4\times10^{41}$ ergs/s at the lower boundary of AGN X-ray emission range, and a factor of >7 variability in one year compared to the Swift upper limit. NGC 6447 was classified as a quiescent galaxy in the literature. The multi-wavelength timing and spectral properties of NGC 6447 are consistent with the expected AGN turning on event, where the obscuring material around the AGN central engine is gradually dispersed, revealing the central engine. This example shows that long-term infrared variability can be a powerful tool to find similar sources. Based on the sample selection statistics, we estimate the duration of the episodes of AGN accretion (duty cycle) signified by the turning-on event as $10^4$-$10^6$ yr.

We revisit the extensively debated star-forming main sequence (SFMS)-a tight correlation between the star formation rate and stellar mass in both kiloparsec-resolved and integrated galaxies. We statistically explore the fundamental drivers of star formation at global scales, using a large volume-limited sample of 24,954 local star-forming galaxies to overcome the limitations of previous works. Based on the mid-infrared 12 micron luminosity, stellar mass, and g-r color, we estimate the molecular gas mass for the considered sample. At galaxy-wide scales, we establish global relations between the surface densities of the star formation rate, stellar mass, and molecular gas mass . These global density relations are connected with and follow similar trends as the resolved SFMS, the Kennicutt-Schmidt (KS) relation, and the molecular gas main sequence (MGMS). Taking advantage of this large catalog, we show that the scatters in the global KS and MGMS relations are smaller than that of the global relation between the star formation rate surface density and stellar mass surface density, and their Pearson correlation coefficients are higher. More importantly, multivariate regression and partial correlation analyses demonstrate that the apparent correlation between the star formation rate surface density and stellar mass surface density is entirely mediated by the molecular gas surface density, with its best-fit parameters directly derivable from those of the KS and MGMS relations. Overall, our findings suggest that the correlation between stellar mass and molecular gas, as well as that between molecular gas and star formation, are more direct and fundamental. The star-forming main sequence thus appears to be a natural by-product of these two tighter relations.

With the discovery of gravitational waves and fast radio bursts, wave optics has become increasingly relevant in astrophysics. This paper studies the behaviour of random gravitational and plasma lenses, presenting the refractive and diffractive expansions, with higher-order terms that allow error estimates and embody the counterintuitive resurgence phenomenon. Specifically, we show that the diffractive expansion converges for a broad class of bounded lens models and provides an efficient description of interference patterns across frequency regimes. Next, building on Picard-Lefschetz techniques, we derive the full refractive expansion to arbitrary order, organising it into a transseries. Near caustics, the standard transseries is supplemented with uniform asymptotics. We study this transseries, with both Borel and hyperasymptotic resummation yielding systematic approximations to lensing integrals at all frequencies. Our results give a framework for modelling wave optics lensing near caustics and beyond the geometric optics approximation and thereby illustrate how tools from resurgence and asymptotic analysis can be applied to practical problems in astrophysics. Near caustic singularities, the post-refractive corrections diverge, while the uniform asymptotic expansion becomes accurate. We use the leading uniform approximation to derive the strong wave optics suppression of off-axis caustics, which clarifies their subdominant role.

Galaxy clusters form through hierarchical assembly, where smaller substructures merge to build the largest gravitationally bound objects in the universe. These mergers, combined with feedback from AGN, filamentary accretion, and other energy injection processes, generate turbulence and perturbations within the intra-cluster medium (ICM). X-ray and Sunyaev-Zel'dovich (SZ) observations can be utilized to measure these ICM density and pressure inhomogeneities, in turn providing constraints on the effective Equation of State (EOS) of the perturbations and ICM velocities. In this work, we analyze deep SRG-eROSITA and Planck/SPT observations of Abell 3266 (A3266), a dynamically complex merging cluster with elongated morphology and significant substructure. We measure pressure and density fluctuations, and compute the power spectra and deprojected 3D amplitudes of these perturbations. We estimate the ratio of pressure-to-density fluctuation amplitudes as $1.00 \pm 0.55$ and non-thermal pressure support $0.068 \pm 0.050$. Density fluctuations are found to be stronger in the northern sector of the cluster compared to the south, consistent with ongoing accretion along a filamentary structure revealed by eROSITA. Further, we find the amplitude of density fluctuations increases with radius, qualitatively consistent with the trend found in cosmological simulations. Uncertainties in our results are dominated by the relatively low sensitivity of current Planck/SPT data, suggesting that improvements in SZ data quality could substantially improve our understanding of ICM energy injection, transport, and dissipation from this technique.

Type Ia supernovae are bright thermonuclear explosions that are important to numerous areas of astronomy. However, the origins of these events are poorly understood. One proposed setting is that of a near Chandrasekhar mass white dwarf that undergoes runaway carbon burning in the core. During the thousand years leading up to the explosion, the white dwarf undergoes a simmering phase where slow carbon burning heats the core and drives convection. A poorly understood aspect of this phase is the convective Urca process, which links convection with weak nuclear reactions. We use the low Mach number code MAESTROeX to perform full 3D simulations as is required to accurately capture the turbulent convection. We present simulations with and without the A=23 convective Urca process, which have relaxed to a steady state. We characterize the effects of the convective Urca process on the neutrino losses, the nuclear energy generation, and the convective boundary. We find that the size of the convection zone is substantially reduced by the convective Urca process, though convection still extends past the Urca shell. Our findings on the structure of the convective zone and the compositional changes can be used to inform 1D stellar models that track the longer-timescale evolution.

The detection of ultra-high energy neutrinos (UHE-$ν$), with enegies above 10 PeV, has been a long-time goal in astroparticle physics. Autonomous, radio-frequency (RF) UHE-$ν$ detetectors have been deployed in polar regions that rely on the Askaryan effect in ice for the neutrino signal. The Askaryan effect occurs when the excess negative charge within a UHE-$ν$ cascade radiates in a dense medium. UHE-$ν$ can induce cascades that radiate in the RF bandwidth above thermal backgrounds. To identify UHE-$ν$ signals in data from Askaryan-class detectors, analytic models of the Askaryan electromagnetic field have been created and matched to simulations and laboratory measurements. These models describe the Askaryan electromagnetic field, but leave the effects of signal propagation through polar ice and RF channel response to simulations. In this work, a fully analytic Askaryan model that accounts for these effects is presented. First, formulas for the observed voltage trace and its Hilbert envelope are calculated. Second, the analytic model is compared to UHE-$ν$ signals at 100 PeV from NuRadioMC, a key Monte Carlo toolset in the field. Correlation coefficients between the analytic signal envelope and MC data in excess of $0.94$ are found, and 99.99% of UHE-$ν$ signals pass a correlation threshold of $ρ\geq 0.4$. Analysis of RF thermal noise reveals that just 0.2 background events have $ρ\geq 0.4$ in 5 years at a 1 Hz thermal trigger rate. Finally, we describe future work related to the measurement of the logarithm of the UHE-$ν$ cascade energy.

We present spatially resolved measurements of turbulence driving modes across entire extragalactic discs of NGC7793 and NGC1313, using Atacama Large Millimetre/submillimetre Array (ALMA) CO(J=2-1) observations at 13pc resolution. By applying a kernel-based analysis of density and velocity fluctuations, we map the turbulence driving parameter, b, which characterises the balance between solenoidal ($b\sim0.3$) and compressive ($b\sim1$) turbulent driving regimes. b is quantified as the ratio of the turbulent density fluctuations relative to the turbulent sonic Mach number, M. Both galaxies show predominantly solenoidal driving on average for the regions where we find valid results ($b\geq 0.33(\pm 0.05)^{+0.14}_{-0.10}$ in NGC7793; $b\geq 0.24(\pm 0.03)^{+0.10}_{-0.07}$ in NGC1313), noting that this is without including the influences of magnetic fields, making these measurements lower limits. We find substantial spatial variation of b, including localised regions of strongly compressive driving. NGC1313 exhibits higher turbulent Mach numbers and density dispersions than NGC7793, consistent with the disturbed morphology and recent satellite interaction in NGC1313. The turbulence in both NGC7793 and NGC1313 is supersonic ($3\lesssim M\lesssim 20$), and NGC1313 shows a radially decreasing trend of M with galactocentric radius. Radial trends indicate more solenoidal driving in the galaxy centres, potentially reflecting enhanced shear, and increasingly compressive modes in the outskirts. These results demonstrate that turbulence driving varies systematically with galactic environment and cannot be assumed uniform across discs. Our study applies a previously established method to larger scales and new data, linking local turbulence physics to global star formation regulation in galaxies, providing a new avenue for testing theoretical models with future integral field units and ALMA surveys.

We present a radio-continuum survey of the LMC using the MeerKAT telescope, describe the full-Stokes products included in the first data release, and highlight some initial results. The observations are centred at 1.3 GHz with a bandwidth of 0.8 GHz. The imaging products comprise six fields of view, each encompassing $\sim$5$^\circ$ $\times$ 5$^\circ$ with the resulting images achieving a resolution of 8". The median broad-band Stokes~I image root-mean-square noise value is $\sim$11 $μ$Jy beam$^{-1}$. The survey enables a variety of astrophysical studies, which we showcase with the presentation of a few findings. Within the LMC we identify a new supernova remnant candidate; present planetary nebulae and Wolf-Rayet stars without previous radio detections; and show the MeerKAT view of the well-known star-forming region 30 Doradus. We also present some examples of interesting foreground and background sources in the field, including the AB~Dor multiple-star system, a radio ring galaxy, a possible Odd Radio Circle, and a remarkable bent-tail radio galaxy.

We present Near-Infrared (NIR) and Optical linear polarimetry towards the partially embedded cluster AFGL 6366S. The polarization ranges from 0.44-10.3 per cent in NIR and 0.16-11.22 per cent in Optical bands. The position angle spans $1^\circ - 179^\circ$ in both NIR and Optical bands. About 22 stars exhibit intrinsic polarization signatures. A polarization hole is evident towards the densest ($\sim 3.4 \times 10^{23} \mathrm{cm}^{-2}$) and warmest ($\sim 28.8 \mathrm{K}$) central cluster region. It is attributable to depolarization induced by Radiative Torque Disruption (RAT-D) of large grains and a modest contribution from magnetic-field tangling. The local magnetic field towards the cluster's central region is significantly misaligned with both the large-scale Galactic field and the long axis of the filament present in the region. The field morphology wraps around two dense molecular clumps of radii 0.34 pc and 0.22 pc and $\mathrm N(\mathrm H_2)$ = $(7.9 \pm 1.1) \times 10^{22}$ cm$^{-2}$ and $(4.3 \pm 0.5) \times 10^{22}$ cm$^{-2}$, respectively. The clumps are embedded in the filamentary structure and represent locally accelerated stages of mass accumulation. Gravitationally driven mass flows, largely perpendicular to the local magnetic field, produce a U-shaped field curvature across the filament axis. The plane-of-sky magnetic field strengths towards the two clumps are $ 447.91 \pm 83.81 μ\mathrm{G}$ and $396.66 \pm 73.64 μ\mathrm{G}$. The corresponding mass-to-flux ratios ($λ\sim 1.34$ and $0.82$) indicate that one clump is magnetically supercritical and the other is subcritical. The Alfven Mach numbers ($\mathcal{M}_A$) $\sim$ 0.395 and 0.393 indicate that both the clumps are in sub-Alfvénic state.

We present new constraints on dust attenuation and star-formation rates (SFRs) for 77 galaxies at redshifts $z=1.43-2.65$, using Paschen emission line detections from the JWST FRESCO survey and ground-based Balmer line measurements from the MOSDEF survey. Using nebular and continuum emission maps, we find that Paschen emission covers a smaller area than continuum emission observed in the F210M (2.1 $μ$m; rest-frame optical) and F444W (4.4 $μ$m; rest-frame near-IR) bands, and is preferentially located toward galaxy outskirts. These results suggest that current star formation is concentrated in regions farther from galaxy centers than older stellar populations traced by the continuum, indicative of inside-out star formation. With a careful accounting of slit-loss corrections for ground-based measurements, we calculate nebular reddening and dust-corrected SFRs using the Balmer decrement (H$α$/H$β$) and Paschen-to-Balmer line ratios (Pa$α$/H$α$ and Pa$β$/H$α$), assuming the Milky Way extinction curve. On average, Paschen-derived reddening and SFRs agree with Balmer-derived values; however, two galaxies exhibit significantly higher Paschen reddening and four show significantly higher Paschen SFRs. We find that non-unity dust covering fractions bias the Balmer decrement toward less reddened OB associations, while decrements involving the Paschen lines are less affected by this bias. These results highlight the enhanced sensitivity of the Paschen lines to the most heavily obscured OB associations in $z\sim2$ galaxies, particularly in galaxies with patchy dust geometries. Future studies using Paschen lines exclusively to measure nebular reddening will yield more robust constraints on the dustiest star-forming regions.

Star formation in galaxies is governed by internal and environmental processes, yet their relative roles are not well understood. In particular, uncertainties in measurements of active galactic nuclei (AGN) host galaxies, combined with modeling limitations, obfuscate the impact of supermassive black hole feedback across environments and over time. Here we address this with a comprehensive analysis of ~60,000 nearby AGNs (z < 0.15 and new environment and halo-mass measurements for ~500,000 AGN and non-AGN host galaxies. This benchmark enables unified comparisons with three prominent cosmological simulations--SIMBA, TNG, and EAGLE--and reveals major, contrasting shortcomings. Simulations fail to reproduce observed trends linking star formation, quiescence, AGN luminosity, stellar mass, and halo mass. While simulations qualitatively capture that AGNs are more common in low-mass halos than in rich groups or clusters, detailed host demographics diverge strongly from observations. Partial agreement exists in the stellar mass distribution within large-scale structures, yet all simulations overproduce quenched low-mass satellites in massive halos, while misrepresenting quenched fractions of massive central galaxies and those in low-density environments, which are sensitive to feedback implementation. Improved AGN physics and modeling of multi-phase gas cooling and flows are required to capture the observed interplay between black holes, galaxies, and halos.

The detection of sub-solar mass black holes is a milestone of modern astrophysics as it would open a window either onto new stellar physics or could potentially unveil the nature of Dark Matter as Primordial Black Holes. On November 12, 2025, the LIGO-Virgo-KAGRA (LVK) collaboration reported the compact binary merger candidate S251112cm, a system with no obvious EM counterpart, consistent with binary black hole merger with a chirp mass in the range $0.1-0.87 \, M_\odot$. The probability that at least one component has mass $<$1 $M_{\odot}$ is $>99\%$. Inspired by this trigger, we tested if a population of PBHs formed at Quantum Chromodynamics epoch with a broad mass function could account for a signal of this type. Our results, corresponding to a predicted event rate of $0.8 \,\text{yr}^{-1}$ as seen by LVK O3b, suggest that the observed merger rate of $0.23^{+0.86}_{-0.218}\,\text{yr}^{-1}\;(95\%\;\text{C.L.})$ if the trigger is confirmed as an astrophysical event would be compatible with such a model. Our predicted detection rate is also in agreement with current LVK expectations for stellar-mass binaries, remaining consistent with a scenario in which a non-negligible fraction of the $3-200 \;M_\odot$ mergers observed by LVK originate from Primordial Black Holes. If confirmed, this detection would place a lower limit to the PBH abundance $f_{PBH}>0.04$ for our adopted model.

We investigate the structural, photometric, and dynamical evolution of cosmic voids and their galaxy populations from $z=2.09$ to the present, focusing on void size as a key evolutionary parameter. Using void catalogs from four Millennium Simulation snapshots and SDSS data at $z<0.04$, we perform a unified analysis of void demographics, galaxy properties, and internal kinematics. Our analysis reveals clear evidence that cosmic voids exhibit a significant evolutionary trend of becoming progressively emptier toward low redshift, accompanied by a marked decline in the brightness and clustering of their galaxy populations. The void galaxy luminosity function evolves significantly: $M^{*}$ fades and $α$ flattens with time, with large voids hosting brighter, more rapidly evolving galaxies than small voids. Stacked density profiles exhibit a universal shape when scaled by void radius, deepening and building more pronounced walls toward $z=0$. Galaxy spatial distributions reveal persistent size-dependent segregation, with galaxies in large voids lying farther from the center and more strongly clustered. Dynamical analysis of simulations shows coherent outward flows in all voids, with amplitudes decreasing toward $z=0$, providing a physical basis for observed redshift-space distortions. Comparison with SDSS broadly confirms these evolutionary trends but uncovers a non-zero central galaxy population in observed voids -- absent in $Λ$CDM predictions -- that may challenge current galaxy formation models in extreme underdensities. Future comparisons with additional simulations and deeper high-redshift surveys will provide stronger tests of $Λ$CDM in the most underdense regions.

We present the first end-to-end calculation connecting the accretion-induced collapse (AIC) of a magnetized, rapidly rotating white dwarf to observable kilonova signatures, combining 2D general-relativistic neutrino-magnetohydrodynamic simulations, followed by radiation hydrodynamics with in-situ nuclear network and 2D Monte Carlo radiative transfer with spatially resolved heating rates. Unlike all previous unmagnetized AIC models - which predicted proton-rich, $^{56}$Ni-dominated ejecta - strong magnetic fields eject ${\sim 0.2 M_\odot}$ of neutron-rich material $(\langle Y_e \rangle \sim 0.24)$ on dynamical timescales, before neutrino irradiation can raise the electron fraction, enabling strong $r$-process nucleosynthesis up to and beyond the third peak. The resulting kilonova is lanthanide-rich $(X_{\rm lan} \approx 6\%)$ and dominated by near-infrared emission. We compute synthetic light curves in the LSST and JWST bands and find striking agreement, without parameter tuning, between the observations of AT 2023vfi/GRB 230307A and our broadband light curves for polar viewing angles. These results establish magnetized AIC as a viable channel for heavy $r$-process element production and a compelling progenitor candidate for long-duration gamma-ray bursts with kilonova signatures.

Waves launched from the neutron star surface or inner magnetosphere propagate through the magnetosphere as small perturbations, but can grow relative to the background magnetic field and steepen into ``monster shocks'' -- ultra-relativistic magnetized shocks which can power high-energy emission from magnetars, neutron star mergers and collapse. They occur in magnetically dominated plasma and are described by relativistic magnetohydrodynamics (MHD). We present global relativistic MHD simulations of monster shocks in unperturbed and perturbed (``wrinkled'') backgrounds with a global dipolar geometry. Our simulations confirm analytical predictions for equatorial shocks and provide new insight into the behavior of oblique shocks off the equator. Simulations where the shock is formed through Alfvén mode to fast mode conversion are also presented, demonstrating the generic nature of the monster shock mechanism. We explore how the presence of additional modes in the magnetosphere modifies the shock behavior. Modes of comparable amplitude can fragment the shock front, substantially reduce the magnetization, produce localized enhancements in the Lorentz factor relative to an unperturbed dipole background, and intermittently generate additional shocks along a line of sight.

Historically, the study of the central regions of Type 1 AGN has been limited by the combination of the host galaxy spectrum with strong emission from the accretion disk and NLR/BLR, which prevented us from accurately probing the galactic and AGN properties in the central regions. Integral field spectroscopy allows us to correct for this effect and study both the unobscured cores of AGN host galaxies as well as the uncontaminated spectra of their central engines with unprecedented precision. Using MUSE WFM observations, in this work, we present a combined method for modelling and subtracting QSO light in type-1 AGN alongside results for one such source, PKS 2153-68 (z=0.028), both jetted and gamma-ray emitting. After separating the host galaxy and AGN spectra, we discuss the discovery of an unresolved and yet-to-be-identified high-velocity ($\sim$25000 km/s) short timescale ($\leq$1 yr) variable emission, unlike anything observed in other variable AGN.

A deep, space-based, all-sky near-infrared survey carried out with the Nancy Grace Roman Space Telescope would constitute a foundational astronomical infrastructure for decades to come. In this white paper, we present a concrete and feasible path to imaging the entire sky at $\sim0.1''$ resolution, beginning with high-impact fields in Cycle 1 and scaling to ultra-wide coverage within the nominal mission. This first-epoch survey will reach $\mathrm{H}\sim25.5$ AB mag (5$σ$) and maximize synergies with contemporaneous observatories, while preserving substantial time for other ambitious Roman programs. We outline representative scheduling scenarios and an example Cycle 1 program that triples early Roman-LSST overlap and delivers high-value community data products such as LSST forced photometry, joint \textit{Gaia}-Roman astrometry, and catalogs of Galactic substructure, stong lenses, and other rare systems. The Cycle 1 program will lay the foundation for an eventual all-sky survey, while also delivering high-impact early science. We invite broad community participation in shaping and carrying out both the initial program and the long-term vision of an all-sky Roman survey.

Crater II (CraII), a large and low-density dwarf spheroidal galaxy, has unusual observed properties that are difficult to reproduce in cold dark matter simulations. Ongoing tidal disruption may help explain the discrepancies, as evidenced by the recent discovery of tidal tails. Here we present metallicity-sensitive narrowband photometry of the Ca II H and K lines from the Dark Energy Camera, covering $128$ deg$^2$ across the center and identified tidal tails of CraII as part of the Mapping the Ancient Galaxy in CaHK (MAGIC) survey. Our combined photometric metallicity, color-magnitude, proper motion, and parallax selections identify 162 CraII candidates. Of these, 37 candidates are located in the tidal tails which extend at least $7^\circ$ ($\sim 95$ kpc) from the center of CraII, suggesting it has lost $\gtrsim 25$% of its initial stellar mass. We confirm low contamination rates with dedicated control fields and highlight the extremely low surface brightness stellar features that can be uncovered with CaHK data, as faint as $\sim 36$ mag arcsec$^{-2}$. We also make the first detection of a metallicity gradient ($-0.34\pm0.17~{\rm dex}~{\rm deg}^{-1}$) in the center of the galaxy and infer a stream width of $w\sim 0.8^\circ$, roughly 50% larger than the CraII half-light radius. The detection of candidates in the most distant CraII pointings from its center implies that the tidal tails extend beyond our footprint. We compare the CraII stream to $N$-body models with "cored" and "cuspy" dark matter halo progenitors, determining that CraII's density profile is still ambiguous and warrants further modeling.

The existence of massive, compact galaxies (Little Red Dots, LRDs) at $z \sim 2$ challenges early structure formation models, suggesting rapid stellar and black hole (BH) assembly. While LRDs are efficient environments for BH growth, many show no X-ray evidence of strong AGN emission. We utilize a subsample of X-ray non-detected LRDs to test the compatibility of collision-based BH formation scenarios and constrain physical parameters like metallicity and column density. Our results indicate LRDs are ideal birthplaces for massive BHs, particularly given a mass-radius relation $R_{gal} \propto M_{gal}^{0.6}$. Collision-based models suggest seed masses larger than those in the local Universe, consistent with high-redshift BH mass-radius relations. We modeled BH seed formation and X-ray emission (0.3-7 keV) against observed upper limits. We find that mass-radius exponents $> 0.55$ favor the collision-based scenario; however, consistency with stacked X-ray analysis requires specific accretion and obscuration parameters. Constant or increasing SFR scenarios with high Eddington ratios are feasible but necessitate larger column densities or higher metal enrichment. Alternatively, moderate sub-Eddington accretion reconciles massive seeds with observed masses and X-ray weakness. We conclude that even if LRDs began as starbursts, they should eventually evolve into AGNs.

Fast blue optical transients (FBOTs) represent one of the most exotic astrophysical transients, exhibiting unusually strong emission across X-ray, optical, and radio wavelengths. Their physical origins remain highly debated, with proposed explanations ranging from stellar explosion to tidal disruption event (TDE). Here we report observations of the most luminous FBOT, AT 2024wpp whose post-peak luminosity rebrightens in X ray and becomes flattening in optical in a manner follows the decay rate characteristic of TDEs ($L_{\rm bol} \propto t^{-5/3}$). This invokes energy contribution of accretion by a central compact object, getting further corroborations from hardening of X-ray spectral index and detection of outflow inferred from the emission lines at similar phase. Detailed modeling of luminsoity evolution favors a coalesce explosion of a 34 M$_{\odot}$ Wolf-Rayet star with a 15 M$_{\odot}$ black hole (BH), demonstrating that some FBOTs may be associated with TDE of a stellar blackhole.

Context: We present the first Cosmological Parameter inferences from eROSITA X-ray observations of galaxy clusters using a Machine Learning algorithm. Methods: We train a Random Forest using mock catalogs of clusters from Magneticum multi-cosmology hydrodynamical simulations. We apply the trained ML algorithm to observed X-ray features (gas luminosity, mass, and temperature) at different redshifts from the eROSITA eFEDS and eRASS1 catalogs. Results: We obtain cosmological constraints with precision comparable to those from standard analyses, such as weak lensing and cluster abundances. We infer $Ω_{\rm m}=0.30^{+0.03}_{-0.02}$, $σ_8=0.81\pm0.01$, and $h_0=0.710\pm0.004$. The recovered parameters show no tension in the $Ω_{\rm m}-σ_8$ space, but a significant deviation of $h_0$ from the Planck estimates. These inferences remain rather stable against variations of the input observable set and parameter space coverage. These results indicate that correlations among intracluster properties contain cosmological information beyond that encoded in the cluster abundance alone, which can be captured by machine learning trained on multi-cosmology simulations. Conclusions: ML algorithms trained on multi-cosmology hydrodynamical simulations can effectively infer cosmological parameters directly from galaxy cluster data. This is a change of paradigm in the context of cosmological parameter inferences. This approach complements traditional cluster-count analyses and is particularly suited to large upcoming surveys, where systematic uncertainties in mass calibration may otherwise dominate the error budget. It also highlights the potential of large-scale X-ray surveys to deliver independent tests of the standard cosmological model.