arXiv Daily Digest - 2026-02-23

Clustering is widely used for unsupervised structure discovery, yet it offers limited insight into how reliable each individual assignment is. Diagnostics, such as convergence behavior or objective values, may reflect global quality, but they do not indicate whether particular instances are assigned confidently, especially for initialization-sensitive algorithms like k-means. This assignment-level instability can undermine both accuracy and robustness. Ensemble approaches improve global consistency by aggregating multiple runs, but they typically lack tools for quantifying pointwise confidence in a way that combines cross-run agreement with geometric support from the learned cluster structure. We introduce CAKE (Confidence in Assignments via K-partition Ensembles), a framework that evaluates each point using two complementary statistics computed over a clustering ensemble: assignment stability and consistency of local geometric fit. These are combined into a single, interpretable score in [0,1]. Our theoretical analysis shows that CAKE remains effective under noise and separates stable from unstable points. Experiments on synthetic and real-world datasets indicate that CAKE effectively highlights ambiguous points and stable core members, providing a confidence ranking that can guide filtering or prioritization to improve clustering quality.

Large Language Models (LLMs) have made significant progress in reasoning tasks across various domains such as mathematics and coding. However, their performance deteriorates in tasks requiring rich socio-cultural knowledge and diverse local contexts, particularly those involving Indian Culture. Existing Cultural benchmarks are (i) Manually crafted, (ii) contain single-hop questions testing factual recall, and (iii) prohibitively costly to scale, leaving this deficiency largely unmeasured. To address this, we introduce VIRAASAT, a novel, semi-automated multi-hop approach for generating cultural specific multi-hop Question-Answering dataset for Indian culture. VIRAASAT leverages a Knowledge Graph comprising more than 700 expert-curated cultural artifacts, covering 13 key attributes of Indian culture (history, festivals, etc). VIRAASAT spans all 28 states and 8 Union Territories, yielding more than 3,200 multi-hop questions that necessitate chained cultural reasoning. We evaluate current State-of-the-Art (SOTA) LLMs on VIRAASAT and identify key limitations in reasoning wherein fine-tuning on Chain-of-Thought(CoT) traces fails to ground and synthesize low-probability facts. To bridge this gap, we propose a novel framework named Symbolic Chain-of-Manipulation (SCoM). Adapting the Chain-of-Manipulation paradigm, we train the model to simulate atomic Knowledge Graph manipulations internally. SCoM teaches the model to reliably traverse the topological structure of the graph. Experiments on Supervised Fine-Tuning (SFT) demonstrate that SCoM outperforms standard CoT baselines by up to 20%. We release the VIRAASAT dataset along with our findings, laying a strong foundation towards building Culturally Aware Reasoning Models.

Autonomous (noise-agnostic) generative models, such as Equilibrium Matching and blind diffusion, challenge the standard paradigm by learning a single, time-invariant vector field that operates without explicit noise-level conditioning. While recent work suggests that high-dimensional concentration allows these models to implicitly estimate noise levels from corrupted observations, a fundamental paradox remains: what is the underlying landscape being optimized when the noise level is treated as a random variable, and how can a bounded, noise-agnostic network remain stable near the data manifold where gradients typically diverge? We resolve this paradox by formalizing Marginal Energy, $E_{\text{marg}}(\mathbf{u}) = -\log p(\mathbf{u})$, where $p(\mathbf{u}) = \int p(\mathbf{u}|t)p(t)dt$ is the marginal density of the noisy data integrated over a prior distribution of unknown noise levels. We prove that generation using autonomous models is not merely blind denoising, but a specific form of Riemannian gradient flow on this Marginal Energy. Through a novel relative energy decomposition, we demonstrate that while the raw Marginal Energy landscape possesses a $1/t^p$ singularity normal to the data manifold, the learned time-invariant field implicitly incorporates a local conformal metric that perfectly counteracts the geometric singularity, converting an infinitely deep potential well into a stable attractor. We also establish the structural stability conditions for sampling with autonomous models. We identify a ``Jensen Gap'' in noise-prediction parameterizations that acts as a high-gain amplifier for estimation errors, explaining the catastrophic failure observed in deterministic blind models. Conversely, we prove that velocity-based parameterizations are inherently stable because they satisfy a bounded-gain condition that absorbs posterior uncertainty into a smooth geometric drift.

Comprehensively retrieving diverse documents is crucial to address queries that admit a wide range of valid answers. We introduce retrieve-verify-retrieve (RVR), a multi-round retrieval framework designed to maximize answer coverage. Initially, a retriever takes the original query and returns a candidate document set, followed by a verifier that identifies a high-quality subset. For subsequent rounds, the query is augmented with previously verified documents to uncover answers that are not yet covered in previous rounds. RVR is effective even with off-the-shelf retrievers, and fine-tuning retrievers for our inference procedure brings further gains. Our method outperforms baselines, including agentic search approaches, achieving at least 10% relative and 3% absolute gain in complete recall percentage on a multi-answer retrieval dataset (QAMPARI). We also see consistent gains on two out-of-domain datasets (QUEST and WebQuestionsSP) across different base retrievers. Our work presents a promising iterative approach for comprehensive answer recall leveraging a verifier and adapting retrievers to a new inference scenario.

This study presents an ensemble technique, SPQ (SVD-Pruning-Quantization), for large language model (LLM) compression that combines variance-retained singular value decomposition (SVD), activation-based pruning, and post-training linear quantization. Each component targets a different source of inefficiency: i) pruning removes redundant neurons in MLP layers, ii) SVD reduces attention projections into compact low-rank factors, iii) and 8-bit quantization uniformly compresses all linear layers. At matched compression ratios, SPQ outperforms individual methods (SVD-only, pruning-only, or quantization-only) in perplexity, demonstrating the benefit of combining complementary techniques. Applied to LLaMA-2-7B, SPQ achieves up to 75% memory reduction while maintaining or improving perplexity (e.g., WikiText-2 5.47 to 4.91) and preserving accuracy on downstream benchmarks such as C4, TruthfulQA, and GSM8K. Compared to strong baselines like GPTQ and SparseGPT, SPQ offers competitive perplexity and accuracy while using less memory (6.86 GB vs. 7.16 GB for GPTQ). Moreover, SPQ improves inference throughput over GPTQ, achieving up to a 1.9x speedup, which further enhances its practicality for real-world deployment. The effectiveness of SPQ's robust compression through layer-aware and complementary compression techniques may provide practical deployment of LLMs in memory-constrained environments. Code is available at: https://github.com/JiaminYao/SPQ_LLM_Compression/

Graph neural networks (GNNs) are increasingly applied to hard optimization problems, often claiming superiority over classical heuristics. However, such claims risk being unsolid due to a lack of standard benchmarks on truly hard instances. From a statistical physics perspective, we propose new hard benchmarks based on random problems. We provide these benchmarks, along with performance results from both classical heuristics and GNNs. Our fair comparison shows that classical algorithms still outperform GNNs. We discuss the challenges for neural networks in this domain. Future claims of superiority can be made more robust using our benchmarks, available at https://github.com/ArtLabBocconi/RandCSPBench.

This paper presents a direct framework for sequence models with hidden states on closed subgroups of U(d). We use a minimal axiomatic setup and derive recurrent and transformer templates from a shared skeleton in which subgroup choice acts as a drop-in replacement for state space, tangent projection, and update map. We then specialize to O(d) and evaluate orthogonal-state RNN and transformer models on Tiny Shakespeare and Penn Treebank under parameter-matched settings. We also report a general linear-mixing extension in tangent space, which applies across subgroup choices and improves finite-budget performance in the current O(d) experiments.

The expressive power of Graph Neural Networks (GNNs) is often analysed via correspondence to the Weisfeiler-Leman (WL) algorithm and fragments of first-order logic. Standard GNNs are limited to performing aggregation over immediate neighbourhoods or over global read-outs. To increase their expressivity, recent attempts have been made to incorporate substructural information (e.g. cycle counts and subgraph properties). In this paper, we formalize this architectural trend by introducing Template GNNs (T-GNNs), a generalized framework where node features are updated by aggregating over valid template embeddings from a specified set of graph templates. We propose a corresponding logic, Graded template modal logic (GML(T)), and generalized notions of template-based bisimulation and WL algorithm. We establish an equivalence between the expressive power of T-GNNs and GML(T), and provide a unifying approach for analysing GNN expressivity: we show how standard AC-GNNs and its recent variants can be interpreted as instantiations of T-GNNs.

Despite the successes of deep learning in computer vision, difficulties persist in recognizing objects that have undergone group-symmetric transformations rarely seen during training-for example objects seen in unusual poses, scales, positions, or combinations thereof. Equivariant neural networks are a solution to the problem of generalizing across symmetric transformations, but require knowledge of transformations a priori. An alternative family of architectures proposes to earn equivariant operators in a latent space from examples of symmetric transformations. Here, using simple datasets of rotated and translated noisy MNIST, we illustrate how such architectures can successfully be harnessed for out-of-distribution classification, thus overcoming the limitations of both traditional and equivariant networks. While conceptually enticing, we discuss challenges ahead on the path of scaling these architectures to more complex datasets.

Accurate prediction of main engine power is essential for vessel performance optimization, fuel efficiency, and compliance with emission regulations. Conventional machine learning approaches, such as Support Vector Machines, variants of Artificial Neural Networks (ANNs), and tree-based methods like Random Forests, Extra Tree Regressors, and XGBoost, can capture nonlinearities but often struggle to respect the fundamental propeller law relationship between power and speed, resulting in poor extrapolation outside the training envelope. This study introduces a hybrid modeling framework that integrates physics-based knowledge from sea trials with data-driven residual learning. The baseline component, derived from calm-water power curves of the form $P = cV^n$, captures the dominant power-speed dependence, while another, nonlinear, regressor is then trained to predict the residual power, representing deviations caused by environmental and operational conditions. By constraining the machine learning task to residual corrections, the hybrid model simplifies learning, improves generalization, and ensures consistency with the underlying physics. In this study, an XGBoost, a simple Neural Network, and a Physics-Informed Neural Network (PINN) coupled with the baseline component were compared to identical models without the baseline component. Validation on in-service data demonstrates that the hybrid model consistently outperformed a pure data-driven baseline in sparse data regions while maintaining similar performance in populated ones. The proposed framework provides a practical and computationally efficient tool for vessel performance monitoring, with applications in weather routing, trim optimization, and energy efficiency planning.

Biological neural networks (like the hippocampus) can internally generate "replay" resembling stimulus-driven activity. Recent computational models of replay use noisy recurrent neural networks (RNNs) trained to path-integrate. Replay in these networks has been described as Langevin sampling, but new modifiers of noisy RNN replay have surpassed this description. We re-examine noisy RNN replay as sampling to understand or improve it in three ways: (1) Under simple assumptions, we prove that the gradients replay activity should follow are time-varying and difficult to estimate, but readily motivate the use of hidden state leakage in RNNs for replay. (2) We confirm that hidden state adaptation (negative feedback) encourages exploration in replay, but show that it incurs non-Markov sampling that also slows replay. (3) We propose the first model of temporally compressed replay in noisy path-integrating RNNs through hidden state momentum, connect it to underdamped Langevin sampling, and show that, together with adaptation, it counters slowness while maintaining exploration. We verify our findings via path-integration of 2D triangular and T-maze paths and of high-dimensional paths of synthetic rat place cell activity.

We propose PRISM-FCP (Partial shaRing and robust calIbration with Statistical Margins for Federated Conformal Prediction), a Byzantine-resilient federated conformal prediction framework that utilizes partial model sharing to improve robustness against Byzantine attacks during both model training and conformal calibration. Existing approaches address adversarial behavior only in the calibration stage, leaving the learned model susceptible to poisoned updates. In contrast, PRISM-FCP mitigates attacks end-to-end. During training, clients partially share updates by transmitting only $M$ of $D$ parameters per round. This attenuates the expected energy of an adversary's perturbation in the aggregated update by a factor of $M/D$, yielding lower mean-square error (MSE) and tighter prediction intervals. During calibration, clients convert nonconformity scores into characterization vectors, compute distance-based maliciousness scores, and downweight or filter suspected Byzantine contributions before estimating the conformal quantile. Extensive experiments on both synthetic data and the UCI Superconductivity dataset demonstrate that PRISM-FCP maintains nominal coverage guarantees under Byzantine attacks while avoiding the interval inflation observed in standard FCP with reduced communication, providing a robust and communication-efficient approach to federated uncertainty quantification.

Pure Pursuit (PP) is widely used in autonomous racing for real-time path tracking due to its efficiency and geometric clarity, yet performance is highly sensitive to how key parameters-lookahead distance and steering gain-are chosen. Standard velocity-based schedules adjust these only approximately and often fail to transfer across tracks and speed profiles. We propose a reinforcement-learning (RL) approach that jointly chooses the lookahead Ld and a steering gain g online using Proximal Policy Optimization (PPO). The policy observes compact state features (speed and curvature taps) and outputs (Ld, g) at each control step. Trained in F1TENTH Gym and deployed in a ROS 2 stack, the policy drives PP directly (with light smoothing) and requires no per-map retuning. Across simulation and real-car tests, the proposed RL-PP controller that jointly selects (Ld, g) consistently outperforms fixed-lookahead PP, velocity-scheduled adaptive PP, and an RL lookahead-only variant, and it also exceeds a kinematic MPC raceline tracker under our evaluated settings in lap time, path-tracking accuracy, and steering smoothness, demonstrating that policy-guided parameter tuning can reliably improve classical geometry-based control.

Federated Learning (FL) enables distributed model training on edge devices while preserving data privacy. However, clients tend to have non-Independent and Identically Distributed (non-IID) data, which often leads to client-drift, and therefore diminishing convergence speed and model performance. While adaptive optimizers have been proposed to mitigate these effects, they frequently introduce computational complexity or communication overhead unsuitable for resource-constrained IoT environments. This paper introduces Federated Zero Mean Gradients (FedZMG), a novel, parameter-free, client-side optimization algorithm designed to tackle client-drift by structurally regularizing the optimization space. Advancing the idea of Gradient Centralization, FedZMG projects local gradients onto a zero-mean hyperplane, effectively neutralizing the "intensity" or "bias" shifts inherent in heterogeneous data distributions without requiring additional communication or hyperparameter tuning. A theoretical analysis is provided, proving that FedZMG reduces the effective gradient variance and guarantees tighter convergence bounds compared to standard FedAvg. Extensive empirical evaluations on EMNIST, CIFAR100, and Shakespeare datasets demonstrate that FedZMG achieves better convergence speed and final validation accuracy compared to the baseline FedAvg and the adaptive optimizer FedAdam, particularly in highly non-IID settings.

Quantum reservoir computers (QRCs) have emerged as a promising approach to quantum machine learning, since they utilize the natural dynamics of quantum systems for data processing and are simple to train. Here, we consider n-qubit quantum extreme learning machines (QELMs) with continuous-time reservoir dynamics. QELMs are memoryless QRCs capable of various ML tasks, including image classification and time series forecasting. We apply the Pauli transfer matrix (PTM) formalism to theoretically analyze the influence of encoding, reservoir dynamics, and measurement operations, including temporal multiplexing, on the QELM performance. This formalism makes explicit that the encoding determines the complete set of (nonlinear) features available to the QELM, while the quantum channels linearly transform these features before they are probed by the chosen measurement operators. Optimizing a QELM can therefore be cast as a decoding problem in which one shapes the channel-induced transformations such that task-relevant features become available to the regressor. The PTM formalism allows one to identify the classical representation of a QELM and thereby guide its design towards a given training objective. As a specific application, we focus on learning nonlinear dynamical systems and show that a QELM trained on such trajectories learns a surrogate-approximation to the underlying flow map.

Interactive perception (IP) enables robots to extract hidden information in their workspace and execute manipulation plans by physically interacting with objects and altering the state of the environment -- crucial for resolving occlusions and ambiguity in complex, partially observable scenarios. We present Zero-Shot IP (ZS-IP), a novel framework that couples multi-strategy manipulation (pushing and grasping) with a memory-driven Vision Language Model (VLM) to guide robotic interactions and resolve semantic queries. ZS-IP integrates three key components: (1) an Enhanced Observation (EO) module that augments the VLM's visual perception with both conventional keypoints and our proposed pushlines -- a novel 2D visual augmentation tailored to pushing actions, (2) a memory-guided action module that reinforces semantic reasoning through context lookup, and (3) a robotic controller that executes pushing, pulling, or grasping based on VLM output. Unlike grid-based augmentations optimized for pick-and-place, pushlines capture affordances for contact-rich actions, substantially improving pushing performance. We evaluate ZS-IP on a 7-DOF Franka Panda arm across diverse scenes with varying occlusions and task complexities. Our experiments demonstrate that ZS-IP outperforms passive and viewpoint-based perception techniques such as Mark-Based Visual Prompting (MOKA), particularly in pushing tasks, while preserving the integrity of non-target elements.

Providing scaffolding through educational chatbots built on Large Language Models (LLM) has potential risks and benefits that remain an open area of research. When students navigate impasses, they ask for help by formulating impasse-driven questions. Within interactions with LLM chatbots, such questions shape the user prompts and drive the pedagogical effectiveness of the chatbot's response. This paper focuses on such student questions from two datasets of distinct learning contexts: formative self-study, and summative assessed coursework. We analysed 6,113 messages from both learning contexts, using 11 different LLMs and three human raters to classify student questions using four existing schemas. On the feasibility of using LLMs as raters, results showed moderate-to-good inter-rater reliability, with higher consistency than human raters. The data showed that 'procedural' questions predominated in both learning contexts, but more so when students prepare for summative assessment. These results provide a basis on which to use LLMs for classification of student questions. However, we identify clear limitations in both the ability to classify with schemas and the value of doing so: schemas are limited and thus struggle to accommodate the semantic richness of composite prompts, offering only partial understanding the wider risks and benefits of chatbot integration. In the future, we recommend an analysis approach that captures the nuanced, multi-turn nature of conversation, for example, by applying methods from conversation analysis in discursive psychology.

Recent works have proposed various explanations for the ability of modern large language models (LLMs) to perform in-context prediction. We propose an alternative conceptual viewpoint from an information-geometric and statistical perspective. Motivated by Bach[2023], we model training as learning an embedding of probability distributions into the space of quantum density operators, and in-context learning as maximum-likelihood prediction over a specified class of quantum models. We provide an interpretation of this predictor in terms of quantum reverse information projection and quantum Pythagorean theorem when the class of quantum models is sufficiently expressive. We further derive non-asymptotic performance guarantees in terms of convergence rates and concentration inequalities, both in trace norm and quantum relative entropy. Our approach provides a unified framework to handle both classical and quantum LLMs.

The quality assessment of Artificial Intelligence (AI) systems is a fundamental challenge due to their inherently probabilistic nature. Standards such as ISO/IEC 25059 provide a quality model, but they lack practical and statistically robust methods for assessing functional correctness. This paper proposes and evaluates the Statistical Confidence in Functional Correctness (SCFC) approach, which seeks to fill this gap by connecting business requirements to a measure of statistical confidence that considers both the model's average performance and its variability. The approach consists of four steps: defining quantitative specification limits, performing stratified and probabilistic sampling, applying bootstrapping to estimate a confidence interval for the performance metric, and calculating a capability index as a final indicator. The approach was evaluated through a case study on two real-world AI systems in industry involving interviews with AI experts. Valuable insights were collected from the experts regarding the utility, ease of use, and intention to adopt the methodology in practical scenarios. We conclude that the proposed approach is a feasible and valuable way to operationalize the assessment of functional correctness, moving the evaluation from a point estimate to a statement of statistical confidence.

Qualitative data analysis is labor-intensive, yet the privacy risks associated with commercial Large Language Models (LLMs) often preclude their use in sensitive research. To address this, we introduce ChatQDA, an on-device framework powered by open-source LLMs designed for privacy-preserving open coding. Our mixed-methods user study reveals that while participants rated the system highly for usability and perceived efficiency, they exhibited "conditional trust", valuing the tool for surface-level extraction while questioning its interpretive nuance and consistency. Furthermore, despite the technical security of local deployment, participants reported epistemic uncertainty regarding data protection, suggesting that invisible security measures are insufficient to foster trust. We conclude with design recommendations for local-first analysis tools that prioritize verifiable privacy and methodological rigor.

Real-world knowledge representation often requires capturing subjective, continuous attributes -- such as political positions -- that conflict with pairwise validation, the widely accepted gold standard for human evaluation. We address this challenge through a dual-scale validation framework applied to political stance prediction in argumentative discourse, combining pointwise and pairwise human annotation. Using 22 language models, we construct a large-scale knowledge base of political position predictions for 23,228 arguments drawn from 30 debates that appeared on the UK politicial television programme \textit{Question Time}. Pointwise evaluation shows moderate human-model agreement (Krippendorff's $α=0.578$), reflecting intrinsic subjectivity, while pairwise validation reveals substantially stronger alignment between human- and model-derived rankings ($α=0.86$ for the best model). This work contributes: (i) a practical validation methodology for subjective continuous knowledge that balances scalability with reliability; (ii) a validated structured argumentation knowledge base enabling graph-based reasoning and retrieval-augmented generation in political domains; and (iii) evidence that ordinal structure can be extracted from pointwise language models predictions from inherently subjective real-world discourse, advancing knowledge representation capabilities for domains where traditional symbolic or categorical approaches are insufficient.

We demonstrate the application of a quantum feature extraction method to enhance multi-class image classification for space applications. By harnessing the dynamics of many-body spin Hamiltonians, the method generates expressive quantum features that, when combined with classical processing, lead to quantum-enhanced classification accuracy. Using a strong and well-established ResNet50 baseline, we achieved a maximum classical accuracy of 83%, which can be improved to 84% with a transfer learning approach. In contrast, applying our quantum-classical method the performance is increased to 87% accuracy, demonstrating a clear and reproducible improvement over robust classical approaches. Implemented on several of IBM's quantum processors, our hybrid quantum-classical approach delivers consistent gains of 2-3% in absolute accuracy. These results highlight the practical potential of current and near-term quantum processors in high-stakes, data-driven domains such as satellite imaging and remote sensing, while suggesting broader applicability in real-world machine learning tasks.

AutoClustering methods aim to automate unsupervised learning tasks, including algorithm selection (AS), hyperparameter optimization (HPO), and pipeline synthesis (PS), by often leveraging meta-learning over dataset meta-features. While these systems often achieve strong performance, their recommendations are often difficult to justify: the influence of dataset meta-features on algorithm and hyperparameter choices is typically not exposed, limiting reliability, bias diagnostics, and efficient meta-feature engineering. This limits reliability and diagnostic insight for further improvements. In this work, we investigate the explainability of the meta-models in AutoClustering. We first review 22 existing methods and organize their meta-features into a structured taxonomy. We then apply a global explainability technique (i.e., Decision Predicate Graphs) to assess feature importance within meta-models from selected frameworks. Finally, we use local explainability tools such as SHAP (SHapley Additive exPlanations) to analyse specific clustering decisions. Our findings highlight consistent patterns in meta-feature relevance, identify structural weaknesses in current meta-learning strategies that can distort recommendations, and provide actionable guidance for more interpretable Automated Machine Learning (AutoML) design. This study therefore offers a practical foundation for increasing decision transparency in unsupervised learning automation.

In jurisdictions like India, where courts face an extensive backlog of cases, artificial intelligence offers transformative potential for legal judgment prediction. A critical subset of this backlog comprises appellate cases, which are formal decisions issued by higher courts reviewing the rulings of lower courts. To this end, we present Vichara, a novel framework tailored to the Indian judicial system that predicts and explains appellate judgments. Vichara processes English-language appellate case proceeding documents and decomposes them into decision points. Decision points are discrete legal determinations that encapsulate the legal issue, deciding authority, outcome, reasoning, and temporal context. The structured representation isolates the core determinations and their context, enabling accurate predictions and interpretable explanations. Vichara's explanations follow a structured format inspired by the IRAC (Issue-Rule-Application-Conclusion) framework and adapted for Indian legal reasoning. This enhances interpretability, allowing legal professionals to assess the soundness of predictions efficiently. We evaluate Vichara on two datasets, PredEx and the expert-annotated subset of the Indian Legal Documents Corpus (ILDC_expert), using four large language models: GPT-4o mini, Llama-3.1-8B, Mistral-7B, and Qwen2.5-7B. Vichara surpasses existing judgment prediction benchmarks on both datasets, with GPT-4o mini achieving the highest performance (F1: 81.5 on PredEx, 80.3 on ILDC_expert), followed by Llama-3.1-8B. Human evaluation of the generated explanations across Clarity, Linking, and Usefulness metrics highlights GPT-4o mini's superior interpretability.

Despite the remarkable practical success of transformer-based language models, recent work has raised concerns about their ability to perform state tracking. In particular, a growing body of literature has shown this limitation primarily through failures in out-of-distribution (OOD) generalization, such as length extrapolation. In this work, we shift attention to the in-distribution implications of these limitations. We conduct a large-scale experimental study of the data efficiency of transformers and recurrent neural networks (RNNs) across multiple supervision regimes. We find that the amount of training data required by transformers grows much more rapidly with state-space size and sequence length than for RNNs. Furthermore, we analyze the extent to which learned state-tracking mechanisms are shared across different sequence lengths. We show that transformers exhibit negligible or even detrimental weight sharing across lengths, indicating that they learn length-specific solutions in isolation. In contrast, recurrent models exhibit effective amortized learning by sharing weights across lengths, allowing data from one sequence length to improve performance on others. Together, these results demonstrate that state tracking remains a fundamental challenge for transformers, even when training and evaluation distributions match.

We describe a modern deep learning system that automatically identifies informative contextual examples (\qu{contexts}) for first language vocabulary instruction for high school student. Our paper compares three modeling approaches: (i) an unsupervised similarity-based strategy using MPNet's uniformly contextualized embeddings, (ii) a supervised framework built on instruction-aware, fine-tuned Qwen3 embeddings with a nonlinear regression head and (iii) model (ii) plus handcrafted context features. We introduce a novel metric called the Retention Competency Curve to visualize trade-offs between the discarded proportion of good contexts and the \qu{good-to-bad} contexts ratio providing a compact, unified lens on model performance. Model (iii) delivers the most dramatic gains with performance of a good-to-bad ratio of 440 all while only throwing out 70\% of the good contexts. In summary, we demonstrate that a modern embedding model on neural network architecture, when guided by human supervision, results in a low-cost large supply of near-perfect contexts for teaching vocabulary for a variety of target words.

Psychological corpora in NLP are collections of texts used to analyze human psychology, emotions, and mental health. These texts allow researchers to study psychological constructs, detect mental health issues and analyze emotional language. However, mental health data can be difficult to collect correctly from social media, due to suppositions made by the collectors. A more pragmatic strategy involves gathering data through open-ended questions and then assessing this information with self-report screening surveys. This method was employed successfully for English, a language with a lot of psychological NLP resources. However, this cannot be stated for Romanian, which currently has no open-source mental health corpus. To address this gap, we have created the first corpus for depression and anxiety in Romanian, by utilizing a form with 6 open-ended questions along with the standardized PHQ-9 and GAD-7 screening questionnaires. Consisting of the texts of 205 respondents and although it may seem small, PsihoRo is a first step towards understanding and analyzing texts regarding the mental health of the Romanian population. We employ statistical analysis, text analysis using Romanian LIWC, emotion detection and topic modeling to show what are the most important features of this newly introduced resource to the NLP community.

We present the first motion generation system for playtesting virtual reality (VR) games. Our player model generates VR headset and handheld controller movements from in-game object arrangements, guided by style exemplars and aligned to maximize simulated gameplay score. We train on the large BOXRR-23 dataset and apply our framework on the popular VR game Beat Saber. The resulting model Robo-Saber produces skilled gameplay and captures diverse player behaviors, mirroring the skill levels and movement patterns specified by input style exemplars. Robo-Saber demonstrates promise in synthesizing rich gameplay data for predictive applications and enabling a physics-based whole-body VR playtesting agent.

Machine learning (ML) approaches have shown promising results for predicting molecular properties relevant for chemical process design. However, they are often limited by scarce experimental property data and lack thermodynamic consistency. As such, thermodynamics-informed ML, i.e., incorporating thermodynamic relations into the loss function as regularization term for training, has been proposed. We herein transfer the concept of thermodynamics-informed graph neural networks (GNNs) from the Gibbs-Duhem to the Clapeyron equation, predicting several pure component properties in a multi-task manner, namely: vapor pressure, liquid molar volume, vapor molar volume and enthalpy of vaporization. We find improved prediction accuracy of the Clapeyron-GNN compared to the single-task learning setting, and improved approximation of the Clapeyron equation compared to the purely data-driven multi-task learning setting. In fact, we observe the largest improvement in prediction accuracy for the properties with the lowest availability of data, making our model promising for practical application in data scarce scenarios of chemical engineering practice.

Recent advances in deep learning, exemplified by Hyper-Connections (HC), have expanded the residual connection paradigm by introducing wider residual streams and diverse connectivity patterns. While these innovations yield significant performance gains, they compromise the identity mapping property of residual connections, leading to training instability, limited scalability, and increased memory overhead. To address these challenges, we propose JPmHC (Jacobian-spectrum Preserving manifold-constrained Hyper-Connections), a framework that replaces identity skips with a trainable linear mixer acting on n parallel streams while explicitly controlling gradient conditioning. By constraining the mixer M on operator-norm-bounded manifolds (e.g., bistochastic, Stiefel, Grassmann), JPmHC prevents gradient pathologies and enhances stability. JPmHC introduces three key contributions: (i) a free-probability analysis that predicts Jacobian spectra for structured skips, providing actionable design rules for mixer selection; (ii) memory-efficient implicit differentiation for fixed-point projections, reducing activation memory and synchronization overhead; and (iii) a Stiefel-constrained mixer via Cayley transforms, ensuring orthogonality without post-hoc normalization. Empirical evaluations on ARC-AGI demonstrate that JPmHC achieves faster convergence, higher accuracy, and lower computational cost compared to bistochastic baselines. As a flexible and scalable extension of HC, JPmHC advances spectrum-aware, stable, and efficient deep learning, offering insights into topological architecture design and foundational model evolution.

Large language models have achieved striking results in interactive theorem proving, particularly in Lean. However, most benchmarks for LLM-based proof automation are drawn from mathematics in the Mathlib ecosystem, whereas proofs in software verification are developed inside definition-rich codebases with substantial project-specific libraries. We introduce VeriSoftBench, a benchmark of 500 Lean 4 proof obligations drawn from open-source formal-methods developments and packaged to preserve realistic repository context and cross-file dependencies. Our evaluation of frontier LLMs and specialized provers yields three observations. First, provers tuned for Mathlib-style mathematics transfer poorly to this repository-centric setting. Second, success is strongly correlated with transitive repository dependence: tasks whose proofs draw on large, multi-hop dependency closures are less likely to be solved. Third, providing curated context restricted to a proof's dependency closure improves performance relative to exposing the full repository, but nevertheless leaves substantial room for improvement. Our benchmark and evaluation suite are released at https://github.com/utopia-group/VeriSoftBench.

With the rapid improvement of LLMs' coding capabilities, the bottleneck of LLM-based automated software development is shifting from generating correct code to eliciting users' requirements. Despite growing interest, the interview competence of LLMs in conversational requirements elicitation remains fully underexplored. Existing evaluations often depend on a few scenarios, real user interaction, and subjective human scoring, which hinders systematic and quantitative comparison. To address these challenges, we propose ReqElicitGym, an interactive and automatic evaluation environment for assessing interview competence in conversational requirements elicitation. Specifically, ReqElicitGym introduces a new evaluation dataset and designs both an interactive oracle user and a task evaluator. The dataset contains 101 website requirements elicitation scenarios spanning 10 application types. Both the oracle user and the task evaluator achieve high agreement with real users and expert judgment. Using our ReqElicitGym, any automated conversational requirements elicitation approach (e.g., LLM-based agents) can be evaluated in a reproducible and quantitative manner through interaction with the environment. Based on our ReqElicitGym, we conduct a systematic empirical study on seven representative LLMs, and the results show that current LLMs still exhibit limited interview competence in uncovering implicit requirements. Particularly, they elicit less than half of the users' implicit requirements, and their effective elicitation questions often emerge in later turns of the dialogue. Besides, we found LLMs can elicit interaction and content implicit requirements, but consistently struggle with style-related requirements. We believe ReqElicitGym will facilitate the evaluation and development of automated conversational requirements elicitation.

Autoregressive large language models (LLMs) generate text token-by-token, requiring n forward passes to produce a sequence of length n. Recent work, Exploring the Latent Capacity of LLMs for One-Step Text Reconstruction (Mezentsev and Oseledets), shows that frozen LLMs can reconstruct hundreds of tokens from only two learned proto-tokens in a single forward pass, suggesting a path beyond the autoregressive paradigm. In this paper, we study what information these proto-tokens encode and how they behave under reconstruction and controlled constraints. We perform a series of experiments aimed at disentangling semantic and syntactic content in the two proto-tokens, analyzing stability properties of the e-token, and visualizing attention patterns to the e-token during reconstruction. Finally, we test two regularization schemes for "imposing" semantic structure on the e-token using teacher embeddings, including an anchor-based loss and a relational distillation objective. Our results indicate that the m-token tends to capture semantic information more strongly than the e-token under standard optimization; anchor-based constraints trade off sharply with reconstruction accuracy; and relational distillation can transfer batch-level semantic relations into the proto-token space without sacrificing reconstruction quality, supporting the feasibility of future non-autoregressive seq2seq systems that predict proto-tokens as an intermediate representation.

Chain-of-thought (CoT) monitors are LLM-based systems that analyze reasoning traces to detect when outputs may exhibit attributes of interest, such as test-hacking behavior during code generation. In this paper, we use information-theoretic analysis to show that non-zero mutual information between CoT and output is a necessary but not sufficient condition for CoT monitorability. We identify two sources of approximation error that may undermine the performance of CoT monitors in practice: information gap, which measures the extent to which the monitor can extract the information available in CoT, and elicitation error, which measures the extent to which the monitor approximates the optimal monitoring function. We further demonstrate that CoT monitorability can be systematically improved through targeted training objectives. To this end, we propose two complementary approaches: (a) an oracle-based method that directly rewards the monitored model for producing CoTs that maximize monitor accuracy, and (b) a more practical, label-free approach that maximizes conditional mutual information between outputs and CoTs. Across multiple different environments, we show both methods significantly improve monitor accuracy while preventing CoT degeneration even when training against a monitor, thereby mitigating reward hacking when the task reward is imperfectly specified.

Decoding sits between a language model and everything we do with it, yet it is still treated as a heuristic knob-tuning exercise. We argue decoding should be understood as a principled optimisation layer: at each token, we solve a regularised problem over the probability simplex that trades off model score against structural preferences and constraints. This single template recovers greedy decoding, Softmax sampling, Top-K, Top-P, and Sparsemax-style sparsity as special cases, and explains their common structure through optimality conditions. More importantly, the framework makes it easy to invent new decoders without folklore. We demonstrate this by designing Best-of-K (BoK), a KL-anchored coverage objective aimed at multi-sample pipelines (self-consistency, reranking, verifier selection). BoK targets the probability of covering good alternatives within a fixed K-sample budget and improves empirical performance. We show that such samples can improve accuracy by, for example, +18.6% for Qwen2.5-Math-7B on MATH500 at high sampling temperatures.

Online Multi-Agent Reinforcement Learning (MARL) is a prominent framework for efficient agent coordination. Crucially, enhancing policy expressiveness is pivotal for achieving superior performance. Diffusion-based generative models are well-positioned to meet this demand, having demonstrated remarkable expressiveness and multimodal representation in image generation and offline settings. Yet, their potential in online MARL remains largely under-explored. A major obstacle is that the intractable likelihoods of diffusion models impede entropy-based exploration and coordination. To tackle this challenge, we propose among the first \underline{O}nline off-policy \underline{MA}RL framework using \underline{D}iffusion policies (\textbf{OMAD}) to orchestrate coordination. Our key innovation is a relaxed policy objective that maximizes scaled joint entropy, facilitating effective exploration without relying on tractable likelihood. Complementing this, within the centralized training with decentralized execution (CTDE) paradigm, we employ a joint distributional value function to optimize decentralized diffusion policies. It leverages tractable entropy-augmented targets to guide the simultaneous updates of diffusion policies, thereby ensuring stable coordination. Extensive evaluations on MPE and MAMuJoCo establish our method as the new state-of-the-art across $10$ diverse tasks, demonstrating a remarkable $2.5\times$ to $5\times$ improvement in sample efficiency.

The environmental sustainability of Information Technology (IT) has emerged as a critical concern, driven by the need to reduce both energy consumption and greenhouse gas (GHG) emissions. In the context of cloud-native applications deployed across the cloud-edge continuum, this challenge translates into identifying energy-efficient deployment strategies that consider not only the computational demands of application components but also the environmental impact of the nodes on which they are executed. Generating deployment plans that account for these dynamic factors is non-trivial, due to fluctuations in application behaviour and variations in the carbon intensity of infrastructure nodes. In this paper, we present an approach for the automatic generation of deployment plans guided by green constraints. These constraints are derived from a continuous analysis of energy consumption patterns, inter-component communication, and the environmental characteristics of the underlying infrastructure. This paper introduces a methodology and architecture for the generation of a set of green-aware constraints that inform the scheduler to produce environmentally friendly deployment plans. We demonstrate how these constraints can be automatically learned and updated over time using monitoring data, enabling adaptive, energy-aware orchestration. The proposed approach is validated through realistic deployment scenarios of a cloud-native application, showcasing its effectiveness in reducing energy usage and associated emissions.

Modeling long sequences of user behaviors has emerged as a critical frontier in generative recommendation. However, existing solutions face a dilemma: linear attention mechanisms achieve efficiency at the cost of retrieval precision due to limited state capacity, while softmax attention suffers from prohibitive computational overhead. To address this challenge, we propose HyTRec, a model featuring a Hybrid Attention architecture that explicitly decouples long-term stable preferences from short-term intent spikes. By assigning massive historical sequences to a linear attention branch and reserving a specialized softmax attention branch for recent interactions, our approach restores precise retrieval capabilities within industrial-scale contexts involving ten thousand interactions. To mitigate the lag in capturing rapid interest drifts within the linear layers, we furthermore design Temporal-Aware Delta Network (TADN) to dynamically upweight fresh behavioral signals while effectively suppressing historical noise. Empirical results on industrial-scale datasets confirm the superiority that our model maintains linear inference speed and outperforms strong baselines, notably delivering over 8% improvement in Hit Rate for users with ultra-long sequences with great efficiency.

This work studies heterogeneous Multi-Objective Reinforcement Learning (MORL), where objectives can differ sharply in temporal frequency. Such heterogeneity allows dense objectives to dominate learning, while sparse long-horizon rewards receive weak credit assignment, leading to poor sample efficiency. We propose a Parallel Reward Integration with Symmetry (PRISM) algorithm that enforces reflectional symmetry as an inductive bias in aligning reward channels. PRISM introduces ReSymNet, a theory-motivated model that reconciles temporal-frequency mismatches across objectives, using residual blocks to learn a scaled opportunity value that accelerates exploration while preserving the optimal policy. We also propose SymReg, a reflectional equivariance regulariser that enforces agent mirroring and constrains policy search to a reflection-equivariant subspace. This restriction provably reduces hypothesis complexity and improves generalisation. Across MuJoCo benchmarks, PRISM consistently outperforms both a sparse-reward baseline and an oracle trained with full dense rewards, improving Pareto coverage and distributional balance: it achieves hypervolume gains exceeding 100\% over the baseline and up to 32\% over the oracle. The code is at \href{https://github.com/EVIEHub/PRISM}{https://github.com/EVIEHub/PRISM}.

Static security analysis is a widely used technique for detecting software vulnerabilities across a wide range of weaknesses, application domains, and programming languages. While prior work surveyed static analyzes for specific weaknesses or application domains, no overview of the entire security landscape exists. We present a systematic literature review of 246 static security analyzers concerning their targeted vulnerabilities, application domains, analysis techniques, evaluation methods, and limitations. We observe that most analyzers focus on a limited set of weaknesses, that the vulnerabilities they detect are rarely exploitable, and that evaluations use custom benchmarks that are too small to enable robust assessment.

Automated methods for discovering mechanistic simulator models from observational data offer a promising path toward accelerating scientific progress. Such methods often take the form of agentic-style iterative workflows that repeatedly propose and revise candidate models by imitating human discovery processes. However, existing LLM-based approaches typically implement such workflows via hand-crafted heuristic procedures, without an explicit probabilistic formulation. We recast model discovery as probabilistic inference, i.e., as sampling from an unknown distribution over mechanistic models capable of explaining the data. This perspective provides a unified way to reason about model proposal, refinement, and selection within a single inference framework. As a concrete instantiation of this view, we introduce ModelSMC, an algorithm based on Sequential Monte Carlo sampling. ModelSMC represents candidate models as particles which are iteratively proposed and refined by an LLM, and weighted using likelihood-based criteria. Experiments on real-world scientific systems illustrate that this formulation discovers models with interpretable mechanisms and improves posterior predictive checks. More broadly, this perspective provides a probabilistic lens for understanding and developing LLM-based approaches to model discovery.

While mechanistic interpretability has developed powerful tools to analyze the internal workings of Large Language Models (LLMs), their complexity has created an accessibility gap, limiting their use to specialists. We address this challenge by designing, building, and evaluating ELIA (Explainable Language Interpretability Analysis), an interactive web application that simplifies the outcomes of various language model component analyses for a broader audience. The system integrates three key techniques -- Attribution Analysis, Function Vector Analysis, and Circuit Tracing -- and introduces a novel methodology: using a vision-language model to automatically generate natural language explanations (NLEs) for the complex visualizations produced by these methods. The effectiveness of this approach was empirically validated through a mixed-methods user study, which revealed a clear preference for interactive, explorable interfaces over simpler, static visualizations. A key finding was that the AI-powered explanations helped bridge the knowledge gap for non-experts; a statistical analysis showed no significant correlation between a user's prior LLM experience and their comprehension scores, suggesting that the system reduced barriers to comprehension across experience levels. We conclude that an AI system can indeed simplify complex model analyses, but its true power is unlocked when paired with thoughtful, user-centered design that prioritizes interactivity, specificity, and narrative guidance.

Data-efficient neural decoding is a central challenge for speech brain-computer interfaces. We present the first demonstration of transfer learning and cross-task decoding for MEG-based speech models spanning perception and production. We pre-train a Conformer-based model on 50 hours of single-subject listening data and fine-tune on just 5 minutes per subject across 18 participants. Transfer learning yields consistent improvements, with in-task accuracy gains of 1-4% and larger cross-task gains of up to 5-6%. Not only does pre-training improve performance within each task, but it also enables reliable cross-task decoding between perception and production. Critically, models trained on speech production decode passive listening above chance, confirming that learned representations reflect shared neural processes rather than task-specific motor activity.

Discrete image tokenizers encode visual inputs as sequences of tokens from a finite vocabulary and are gaining popularity in multimodal systems, including encoder-only, encoder-decoder, and decoder-only models. However, unlike CLIP encoders, their vulnerability to adversarial attacks has not been explored. Ours being the first work studying this topic, we first formulate attacks that aim to perturb the features extracted by discrete tokenizers, and thus change the extracted tokens. These attacks are computationally efficient, application-agnostic, and effective across classification, multimodal retrieval, and captioning tasks. Second, to defend against this vulnerability, inspired by recent work on robust CLIP encoders, we fine-tune popular tokenizers with unsupervised adversarial training, keeping all other components frozen. While unsupervised and task-agnostic, our approach significantly improves robustness to both unsupervised and end-to-end supervised attacks and generalizes well to unseen tasks and data. Unlike supervised adversarial training, our approach can leverage unlabeled images, making it more versatile. Overall, our work highlights the critical role of tokenizer robustness in downstream tasks and presents an important step in the development of safe multimodal foundation models.

We propose a proof of concept for a variational distributional neuron: a compute unit formulated as a VAE brick, explicitly carrying a prior, an amortized posterior and a local ELBO. The unit is no longer a deterministic scalar but a distribution: computing is no longer about propagating values, but about contracting a continuous space of possibilities under constraints. Each neuron parameterizes a posterior, propagates a reparameterized sample and is regularized by the KL term of a local ELBO - hence, the activation is distributional. This "contraction" becomes testable through local constraints and can be monitored via internal measures. The amount of contextual information carried by the unit, as well as the temporal persistence of this information, are locally tuned by distinct constraints. This proposal addresses a structural tension: in sequential generation, causality is predominantly organized in the symbolic space and, even when latents exist, they often remain auxiliary, while the effective dynamics are carried by a largely deterministic decoder. In parallel, probabilistic latent models capture factors of variation and uncertainty, but that uncertainty typically remains borne by global or parametric mechanisms, while units continue to propagate scalars - hence the pivot question: if uncertainty is intrinsic to computation, why does the compute unit not carry it explicitly? We therefore draw two axes: (i) the composition of probabilistic constraints, which must be made stable, interpretable and controllable; and (ii) granularity: if inference is a negotiation of distributions under constraints, should the primitive unit remain deterministic or become distributional? We analyze "collapse" modes and the conditions for a "living neuron", then extend the contribution over time via autoregressive priors over the latent, per unit.

Deep learning-based methods have shown remarkable effectiveness in solving PDEs, largely due to their ability to enable fast simulations once trained. However, despite the availability of high-performance computing infrastructure, many critical applications remain constrained by the substantial computational costs associated with generating large-scale, high-quality datasets and training models. In this work, inspired by studies on the structure of Green's functions for elliptic PDEs, we introduce Neural-HSS, a parameter-efficient architecture built upon the Hierarchical Semi-Separable (HSS) matrix structure that is provably data-efficient for a broad class of PDEs. We theoretically analyze the proposed architecture, proving that it satisfies exactness properties even in very low-data regimes. We also investigate its connections with other architectural primitives, such as the Fourier neural operator layer and convolutional layers. We experimentally validate the data efficiency of Neural-HSS on the three-dimensional Poisson equation over a grid of two million points, demonstrating its superior ability to learn from data generated by elliptic PDEs in the low-data regime while outperforming baseline methods. Finally, we demonstrate its capability to learn from data arising from a broad class of PDEs in diverse domains, including electromagnetism, fluid dynamics, and biology.

Recent work on test-time scaling for large language model (LLM) reasoning typically assumes that allocating more inference-time computation uniformly improves correctness. However, prior studies show that reasoning uncertainty is highly localized: a small subset of low-confidence tokens disproportionately contributes to reasoning errors and unnecessary output expansion. Motivated by this observation, we propose Thinking by Subtraction, a confidence-driven contrastive decoding approach that improves reasoning reliability through targeted token-level intervention. Our method, Confidence-Driven Contrastive Decoding, detects low-confidence tokens during decoding and intervenes selectively at these positions. It constructs a contrastive reference by replacing high-confidence tokens with minimal placeholders, and refines predictions by subtracting this reference distribution at low-confidence locations. Experiments show that CCD significantly improves accuracy across mathematical reasoning benchmarks while substantially reducing output length, with minimal KV-cache overhead. As a training-free method, CCD enhances reasoning reliability through targeted low-confidence intervention without computational redundancy. Our code will be made available at: https://github.com/bolo-web/CCD.

Large Language Models (LLMs) demonstrate significant potential in multi-agent negotiation tasks, yet evaluation in this domain remains challenging due to a lack of robust and generalizable benchmarks. Abdelnabi et al. (2024) introduce a negotiation benchmark based on Scoreable Games, with the aim of developing a highly complex and realistic evaluation framework for LLMs. Our work investigates the reproducibility of claims in their benchmark, and provides a deeper understanding of its usability and generalizability. We replicate the original experiments on additional models, and introduce additional metrics to verify negotiation quality and evenness of evaluation. Our findings reveal that while the benchmark is indeed complex, model comparison is ambiguous, raising questions about its objectivity. Furthermore, we identify limitations in the experimental setup, particularly in information leakage detection and thoroughness of the ablation study. By examining and analyzing the behavior of a wider range of models on an extended version of the benchmark, we reveal insights that provide additional context to potential users. Our results highlight the importance of context in model-comparative evaluations.

Accurate AC-PF prediction under domain shift is critical when models trained on medium-voltage (MV) grids are deployed on high-voltage (HV) networks. Existing physics-informed graph neural solvers typically rely on full fine-tuning for cross-regime transfer, incurring high retraining cost and offering limited control over the stability-plasticity trade-off between target-domain adaptation and source-domain retention. We study parameter-efficient domain adaptation for physics-informed self-attention based GNN, encouraging Kirchhoff-consistent behavior via a physics-based loss while restricting adaptation to low-rank updates. Specifically, we apply LoRA to attention projections with selective unfreezing of the prediction head to regulate adaptation capacity. This design yields a controllable efficiency-accuracy trade-off for physics-constrained inverse estimation under voltage-regime shift. Across multiple grid topologies, the proposed LoRA+PHead adaptation recovers near-full fine-tuning accuracy with a target-domain RMSE gap of $2.6\times10^{-4}$ while reducing the number of trainable parameters by 85.46%. The physics-based residual remains comparable to full fine-tuning; however, relative to Full FT, LoRA+PHead reduces MV source retention by 4.7 percentage points (17.9% vs. 22.6%) under domain shift, while still enabling parameter-efficient and physically consistent AC-PF estimation.

Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation, leveraging large-scale pre-training to achieve strong performance. The field has rapidly evolved with additional spatial priors and diverse architectural innovations. However, these advancements are often accompanied by varying training recipes and implementation details, which can make it challenging to disentangle the precise source of empirical gains. In this work, we introduce SimVLA, a streamlined baseline designed to establish a transparent reference point for VLA research. By strictly decoupling perception from control, using a standard vision-language backbone and a lightweight action head, and standardizing critical training dynamics, we demonstrate that a minimal design can achieve state-of-the-art performance. Despite having only 0.5B parameters, SimVLA outperforms multi-billion-parameter models on standard simulation benchmarks without robot pretraining. SimVLA also reaches on-par real-robot performance compared to pi0.5. Our results establish SimVLA as a robust, reproducible baseline that enables clear attribution of empirical gains to future architectural innovations. Website: https://frontierrobo.github.io/SimVLA

Real-time sentence comprehension imposes a significant load on working memory, as comprehenders must maintain contextual information to anticipate future input. While measures of such load have played an important role in psycholinguistic theories, they have been formalized, largely, using symbolic grammars, which assign discrete, uniform costs to syntactic predictions. This study proposes a measure of processing storage cost based on an information-theoretic formalization, as the amount of information previous words carry about future context, under uncertainty. Unlike previous discrete, grammar-based metrics, this measure is continuous, theory-neutral, and can be estimated from pre-trained neural language models. The validity of this approach is demonstrated through three analyses in English: our measure (i) recovers well-known processing asymmetries in center embeddings and relative clauses, (ii) correlates with a grammar-based storage cost in a syntactically-annotated corpus, and (iii) predicts reading-time variance in two large-scale naturalistic datasets over and above baseline models with traditional information-based predictors.

Deep Generative models (DGMs) play two key roles in modern machine learning: (i) producing new information (e.g., image synthesis) and (ii) reducing dimensionality. However, traditional architectures often rely on auxiliary networks such as encoders in Variational Autoencoders (VAEs) or discriminators in Generative Adversarial Networks (GANs), which introduce training instability, computational overhead, and risks like mode collapse. We present NeuroSQL, a new generative paradigm that eliminates the need for auxiliary networks by learning low-dimensional latent representations implicitly. NeuroSQL leverages an asymptotic approximation that expresses the latent variables as the solution to an optimal transportation problem. Specifically, NeuroSQL learns the latent variables by solving a linear assignment problem and then passes the latent information to a standalone generator. We benchmark its performance against GANs, VAEs, and a budget-matched diffusion baseline on four datasets: handwritten digits (MNIST), faces (CelebA), animal faces (AFHQ), and brain images (OASIS). Compared to VAEs, GANs, and diffusion models: (1) in terms of image quality, NeuroSQL achieves overall lower mean pixel distance between synthetic and authentic images and stronger perceptual/structural fidelity; (2) computationally, NeuroSQL requires the least training time; and (3) practically, NeuroSQL provides an effective solution for generating synthetic data with limited training samples. By embracing quantile assignment rather than an encoder, NeuroSQL provides a fast, stable, and robust way to generate synthetic data with minimal information loss.

We review recent advances in machine learning (ML) force-field methods for Landau-Lifshitz-Gilbert (LLG) simulations of itinerant electron magnets, focusing on scalability and transferability. Built on the principle of locality, a deep neural network model is developed to efficiently and accurately predict the electron-mediated forces governing spin dynamics. Symmetry-aware descriptors constructed through a group-theoretical approach ensure rigorous incorporation of both lattice and spin-rotation symmetries. The framework is demonstrated using the prototypical s-d exchange model widely employed in spintronics. ML-enabled large-scale simulations reveal novel nonequilibrium phenomena, including anomalous coarsening of tetrahedral spin order on the triangular lattice and the freezing of phase separation dynamics in lightly hole-doped, strong-coupling square-lattice systems. These results establish ML force-field frameworks as scalable, accurate, and versatile tools for modeling nonequilibrium spin dynamics in itinerant magnets.

Unsupervised representations are widely assumed to be neutral with respect to sensitive attributes when those attributes are withheld from training. We show that this assumption is false. Using SOMtime, a topology-preserving representation method based on high-capacity Self-Organizing Maps, we demonstrate that sensitive attributes such as age and income emerge as dominant latent axes in purely unsupervised embeddings, even when explicitly excluded from the input. On two large-scale real-world datasets (the World Values Survey across five countries and the Census-Income dataset), SOMtime recovers monotonic orderings aligned with withheld sensitive attributes, achieving Spearman correlations of up to 0.85, whereas PCA and UMAP typically remain below 0.23 (with a single exception reaching 0.31), and against t-SNE and autoencoders which achieve at most 0.34. Furthermore, unsupervised segmentation of SOMtime embeddings produces demographically skewed clusters, demonstrating downstream fairness risks without any supervised task. These findings establish that \textit{fairness through unawareness} fails at the representation level for ordinal sensitive attributes and that fairness auditing must extend to unsupervised components of machine learning pipelines. We have made the code available at~ https://github.com/JosephBingham/SOMtime

Structured dilated attention has an appealing inference-time efficiency knob: it reduces the FLOPs of the attention and the KV cache size by a factor of the dilation size D, while preserving long-range connectivity. However, we find a persistent failure mode of them -- sparsifying a pretrained attention model to a dilated pattern leads to severe accuracy degradation. We introduce RAT+, a dense-pretraining architecture that augments attention with full-sequence recurrence and active recurrence learning. A single RAT+ model is pretrained densely once, then flexibly switched at inference time to dilated attention (optionally with local windows) or hybrid layer/head compositions, requiring only a short 1B-token resolution adaptation rather than retraining separate sparse models. At 1.5B parameters trained on 100B tokens, RAT+ closely matches dense accuracy at 16 and drops by about 2-3 points at 64 on commonsense reasoning and LongBench tasks, respectively. Moreover, RAT+ outperforms attention when sparsifying to the top-k block attention. We further scale to 2.6B parameters and 200B tokens and observe the same trend.

Alzheimer's disease (AD) alters brain electrophysiology and disrupts multichannel EEG dynamics, making accurate and clinically useful EEG-based diagnosis increasingly important for screening and disease monitoring. However, many existing approaches rely on black-box classifiers and do not explicitly model the underlying dynamics that generate observed signals. To address these limitations, we propose LERD, an end-to-end Bayesian electrophysiological neural dynamical system that infers latent neural events and their relational structure directly from multichannel EEG without event or interaction annotations. LERD combines a continuous-time event inference module with a stochastic event-generation process to capture flexible temporal patterns, while incorporating an electrophysiology-inspired dynamical prior to guide learning in a principled way. We further provide theoretical analysis that yields a tractable bound for training and stability guarantees for the inferred relational dynamics. Extensive experiments on synthetic benchmarks and two real-world AD EEG cohorts demonstrate that LERD consistently outperforms strong baselines and yields physiology-aligned latent summaries that help characterize group-level dynamical differences.

The adoption of Generative AI (GenAI) suggests major changes for software engineering, including technical aspects but also human aspects of the professionals involved. One of these aspects is how individuals perceive themselves regarding their work, i.e., their work identity, and the processes they perform to form, adapt and reject these identities, i.e., identity work. Existent studies provide evidence of such identity work of software professionals triggered by the adoption of GenAI, however they do not consider differences among diverse roles, such as developers and testers. In this paper, we argue the need for considering the role as a factor defining the identity work of software professionals. To support our claim, we review some studies regarding different roles and also recent studies on how to adopt GenAI in software engineering. Then, we propose a research agenda to better understand how the role influences identity work of software professionals triggered by the adoption of GenAI, and, based on that, to propose new artifacts to support this adoption. We also discuss the potential implications for practice of the results to be obtained.

Locally checkable labeling problems (LCLs) form the foundation of the modern theory of distributed graph algorithms. First introduced in the seminal paper by Naor and Stockmeyer [STOC 1993], these are graph problems that can be described by listing a finite set of valid local neighborhoods. This seemingly simple definition strikes a careful balance between two objectives: they are a family of problems that is broad enough so that it captures numerous problems that are of interest to researchers working in this field, yet restrictive enough so that it is possible to prove strong theorems that hold for all LCL problems. In particular, the distributed complexity landscape of LCL problems is now very well understood. In this work we show that the family of LCL problems is extremely robust to variations. We present a very restricted family of locally checkable problems (essentially, the "node-edge checkable" formalism familiar from round elimination, restricted to regular unlabeled graphs); most importantly, such problems cannot directly refer to e.g. the existence of short cycles. We show that one can translate between the two formalisms (there are local reductions in both directions that only need access to a symmetry-breaking oracle, and hence the overhead is at most an additive $O(\log^* n)$ rounds in the LOCAL model).

We introduce Box Thirding (B3), a flexible and efficient algorithm for Best Arm Identification (BAI) under fixed-budget constraints. It is designed for both anytime BAI and scenarios with large N, where the number of arms is too large for exhaustive evaluation within a limited budget T. The algorithm employs an iterative ternary comparison: in each iteration, three arms are compared--the best-performing arm is explored further, the median is deferred for future comparisons, and the weakest is discarded. Even without prior knowledge of T, B3 achieves an epsilon-best arm misidentification probability comparable to Successive Halving (SH), which requires T as a predefined parameter, applied to a randomly selected subset of c0 arms that fit within the budget. Empirical results show that B3 outperforms existing methods under limited-budget constraints in terms of simple regret, as demonstrated on the New Yorker Cartoon Caption Contest dataset.

AI evaluation has primarily focused on measuring capabilities, with formal approaches inspired from Item Response Theory (IRT) being increasingly applied. Yet propensities - the tendencies of models to exhibit particular behaviours - play a central role in determining both performance and safety outcomes. However, traditional IRT describes a model's success on a task as a monotonic function of model capabilities and task demands, an approach unsuited to propensities, where both excess and deficiency can be problematic. Here, we introduce the first formal framework for measuring AI propensities by using a bilogistic formulation for model success, which attributes high success probability when the model's propensity is within an "ideal band". Further, we estimate the limits of the ideal band using LLMs equipped with newly developed task-agnostic rubrics. Applying our framework to six families of LLM models whose propensities are incited in either direction, we find that we can measure how much the propensity is shifted and what effect this has on the tasks. Critically, propensities estimated using one benchmark successfully predict behaviour on held-out tasks. Moreover, we obtain stronger predictive power when combining propensities and capabilities than either separately. More broadly, our framework showcases how rigorous propensity measurements can be conducted and how it yields gains over solely using capability evaluations to predict AI behaviour.

This work presents a new approach to decentralized training-SeedFlood-designed to scale for large models across complex network topologies and achieve global consensus with minimal communication overhead. Traditional gossip-based methods suffer from message communication costs that grow with model size, while information decay over network hops renders global consensus inefficient. SeedFlood departs from these practices by exploiting the seed-reconstructible structure of zeroth-order updates and effectively making the messages near-zero in size, allowing them to be flooded to every client in the network. This mechanism makes communication overhead negligible and independent of model size, removing the primary scalability bottleneck in decentralized training. Consequently, SeedFlood enables training in regimes previously considered impractical, such as billion-parameter models distributed across hundreds of clients. Our experiments on decentralized LLM fine-tuning demonstrate thatSeedFlood consistently outperforms gossip-based baselines in both generalization performance and communication efficiency, and even achieves results comparable to first-order methods in large scale settings.

Masked Diffusion Models (MDMs) offer greater flexibility in decoding order than autoregressive models but require careful planning to achieve high-quality generation. Existing samplers typically adopt greedy heuristics, prioritizing positions with the highest local certainty to decode at each step. Through failure case analysis, we identify a fundamental limitation of this approach: it neglects the downstream impact of current decoding choices on subsequent steps and fails to minimize cumulative uncertainty. In particular, these methods do not fully exploit the non-causal nature of MDMs, which enables evaluating how a decoding decision reshapes token probabilities/uncertainty across all remaining masked positions. To bridge this gap, we propose the Info-Gain Sampler, a principled decoding framework that balances immediate uncertainty with information gain over future masked tokens. Extensive evaluations across diverse architectures and tasks (reasoning, coding, creative writing, and image generation) demonstrate that Info-Gain Sampler consistently outperforms existing samplers for MDMs. For instance, it achieves a 3.6% improvement in average accuracy on reasoning tasks and a 63.1% win-rate in creative writing. Notably, on reasoning tasks it reduces cumulative uncertainty from 78.4 to 48.6, outperforming the best baseline by a large margin. The code will be available at https://github.com/yks23/Information-Gain-Sampler.

Capture-the-Flag (CTF) competitions serve as gateways into offensive cybersecurity, yet they often present steep barriers for novices due to complex toolchains and opaque workflows. Recently, agentic AI frameworks for cybersecurity promise to lower these barriers by automating and coordinating penetration testing tasks. However, their role in shaping novice learning remains underexplored. We present a human-centered, mixed-methods case study examining how agentic AI frameworks -- here Cybersecurity AI (CAI) -- mediates novice entry into CTF-based penetration testing. An undergraduate student without prior hacking experience attempted to approach performance benchmarks from a national cybersecurity challenge using CAI. Quantitative performance metrics were complemented by structured reflective analysis of learning progression and AI interaction patterns. Our thematic analysis suggest that agentic AI reduces initial entry barriers by providing overview, structure and guidance, thereby lowering the cognitive workload during early engagement. Quantitatively, the observed extensive exploration of strategies and low per-strategy execution time potetially facilitatates cybersecurity training on meta, i.e. strategic levels. At the same time, AI-assisted cybersecurity education introduces new challenges related to trust, dependency, and responsible use. We discuss implications for human-centered AI-supported cybersecurity education and outline open questions for future research.

Clickbait headlines degrade the quality of online information and undermine user trust. We present a hybrid approach to clickbait detection that combines transformer-based text embeddings with linguistically motivated informativeness features. Using natural language processing techniques, we evaluate classical vectorizers, word embedding baselines, and large language model embeddings paired with tree-based classifiers. Our best-performing model, XGBoost over embeddings augmented with 15 explicit features, achieves an F1-score of 91\%, outperforming TF-IDF, Word2Vec, GloVe, LLM prompt based classification, and feature-only baselines. The proposed feature set enhances interpretability by highlighting salient linguistic cues such as second-person pronouns, superlatives, numerals, and attention-oriented punctuation, enabling transparent and well-calibrated clickbait predictions. We release code and trained models to support reproducible research.

Accurately modeling the spatio-temporal dynamics of blast wave propagation remains a longstanding challenge due to its highly nonlinear behavior, sharp gradients, and burdensome computational cost. While machine learning-based surrogate models offer fast inference as a promising alternative, they suffer from degraded accuracy, particularly evaluated on complex urban layouts or out-of-distribution scenarios. Moreover, autoregressive prediction strategies in such models are prone to error accumulation over long forecasting horizons, limiting their robustness for extended-time simulations. To address these limitations, we propose RGD-Blast, a robust and generalizable deep surrogate model for high-fidelity, long-term blast wave forecasting. RGD-Blast incorporates a multi-scale module to capture both global flow patterns and local boundary interactions, effectively mitigating error accumulation during autoregressive prediction. We introduce a dynamic-static feature coupling mechanism that fuses time-varying pressure fields with static source and layout features, thereby enhancing out-of-distribution generalization. Experiments demonstrate that RGD-Blast achieves a two-order-of-magnitude speedup over traditional numerical methods while maintaining comparable accuracy. In generalization tests on unseen building layouts, the model achieves an average RMSE below 0.01 and an R2 exceeding 0.89 over 280 consecutive time steps. Additional evaluations under varying blast source locations and explosive charge weights further validate its generalization, substantially advancing the state of the art in long-term blast wave modeling.

Previous work has explored the computational complexity of deriving two fundamental types of explanations for ML model predictions: (1) *sufficient reasons*, which are subsets of input features that, when fixed, determine a prediction, and (2) *contrastive reasons*, which are subsets of input features that, when modified, alter a prediction. Prior studies have examined these explanations in different contexts, such as non-probabilistic versus probabilistic frameworks and local versus global settings. In this study, we introduce a unified framework for analyzing these explanations, demonstrating that they can all be characterized through the minimization of a unified probabilistic value function. We then prove that the complexity of these computations is influenced by three key properties of the value function: (1) *monotonicity*, (2) *submodularity*, and (3) *supermodularity* - which are three fundamental properties in *combinatorial optimization*. Our findings uncover some counterintuitive results regarding the nature of these properties within the explanation settings examined. For instance, although the *local* value functions do not exhibit monotonicity or submodularity/supermodularity whatsoever, we demonstrate that the *global* value functions do possess these properties. This distinction enables us to prove a series of novel polynomial-time results for computing various explanations with provable guarantees in the global explainability setting, across a range of ML models that span the interpretability spectrum, such as neural networks, decision trees, and tree ensembles. In contrast, we show that even highly simplified versions of these explanations become NP-hard to compute in the corresponding local explainability setting.

A growing number of critical workflow applications leverage a streamlined edge-hub-cloud architecture, which diverges from the conventional edge computing paradigm. An edge device, in collaboration with a hub device and a cloud server, often suffices for their reliable and efficient execution. However, task allocation in this streamlined architecture is challenging due to device limitations and diverse operating conditions. Given the inherent criticality of such workflow applications, where reliability and latency are vital yet conflicting objectives, an exact task allocation approach is typically required to ensure optimal solutions. As no existing method holistically addresses these issues, we propose an exact multi-objective task allocation framework to jointly optimize the overall reliability and latency of a workflow application in the specific edge-hub-cloud architecture. We present a comprehensive binary integer linear programming formulation that considers the relative importance of each objective. It incorporates time redundancy techniques, while accounting for crucial constraints often overlooked in related studies. We evaluate our approach using a relevant real-world workflow application, as well as synthetic workflows varying in structure, size, and criticality. In the real-world application, our method achieved average improvements of 84.19% in reliability and 49.81% in latency over baseline strategies, across relevant objective trade-offs. Overall, the experimental results demonstrate the effectiveness and scalability of our approach across diverse workflow applications for the considered system architecture, highlighting its practicality with runtimes averaging between 0.03 and 50.94 seconds across all examined workflows.

Jailbreaking poses a significant risk to the deployment of Large Language Models (LLMs) and Vision Language Models (VLMs). VLMs are particularly vulnerable because they process both text and images, creating broader attack surfaces. However, available resources for jailbreak detection are scarce, particularly in finance. To address this gap, we present FENCE, a bilingual (Korean-English) multimodal dataset for training and evaluating jailbreak detectors in financial applications. FENCE emphasizes domain realism through finance-relevant queries paired with image-grounded threats. Experiments with commercial and open-source VLMs reveal consistent vulnerabilities, with GPT-4o showing measurable attack success rates and open-source models displaying greater exposure. A baseline detector trained on FENCE achieves 99 percent in-distribution accuracy and maintains strong performance on external benchmarks, underscoring the dataset's robustness for training reliable detection models. FENCE provides a focused resource for advancing multimodal jailbreak detection in finance and for supporting safer, more reliable AI systems in sensitive domains. Warning: This paper includes example data that may be offensive.

Large language models generate text through probabilistic sampling from high-dimensional distributions, yet how this process reshapes the structural statistical organization of language remains incompletely characterized. Here we show that lossless compression provides a simple, model-agnostic measure of statistical regularity that differentiates generative regimes directly from surface text. We analyze compression behavior across three progressively more complex information ecosystems: controlled human-LLM continuations, generative mediation of a knowledge infrastructure (Wikipedia vs. Grokipedia), and fully synthetic social interaction environments (Moltbook vs. Reddit). Across settings, compression reveals a persistent structural signature of probabilistic generation. In controlled and mediated contexts, LLM-produced language exhibits higher structural regularity and compressibility than human-written text, consistent with a concentration of output within highly recurrent statistical patterns. However, this signature shows scale dependence: in fragmented interaction environments the separation attenuates, suggesting a fundamental limit to surface-level distinguishability at small scales. This compressibility-based separation emerges consistently across models, tasks, and domains and can be observed directly from surface text without relying on model internals or semantic evaluation. Overall, our findings introduce a simple and robust framework for quantifying how generative systems reshape textual production, offering a structural perspective on the evolving complexity of communication.

Hardware heterogeneity across diverse user devices poses new challenges for beam-based communication in 5G and beyond. This heterogeneity limits the applicability of machine learning (ML)-based algorithms. This article highlights the critical need to treat hardware heterogeneity as a first-class design concern in ML-aided beam management. We analyze key failure modes in the presence of heterogeneity and present case studies demonstrating their performance impact. Finally, we discuss potential strategies to improve generalization in beam management.

Inverse design, particularly geometric shape optimization, provides a systematic approach for developing high-performance nanophotonic devices. While numerous optimization algorithms exist, previous global approaches exhibit slow convergence and conversely local search strategies frequently become trapped in local optima. To address the limitations inherent to both local and global approaches, we introduce BONNI: Bayesian optimization through neural network ensemble surrogates with interior point optimization. It augments global optimization with an efficient incorporation of gradient information to determine optimal sampling points. This capability allows BONNI to circumvent the local optima found in many nanophotonic applications, while capitalizing on the efficiency of gradient-based optimization. We demonstrate BONNI's capabilities in the design of a distributed Bragg reflector as well as a dual-layer grating coupler through an exhaustive comparison against other optimization algorithms commonly used in literature. Using BONNI, we were able to design a 10-layer distributed Bragg reflector with only 4.5% mean spectral error, compared to the previously reported results of 7.8% error with 16 layers. Further designs of a broadband waveguide taper and photonic crystal waveguide transition validate the capabilities of BONNI.

Accurate long-horizon prediction of spatiotemporal fields on complex geometries is a fundamental challenge in scientific machine learning, with applications such as additive manufacturing where temperature histories govern defect formation and mechanical properties. High-fidelity simulations are accurate but computationally costly, and despite recent advances, machine learning methods remain challenged by long-horizon temperature and gradient prediction. We propose a deep learning framework for predicting full temperature histories directly on meshes, conditioned on geometry and process parameters, while maintaining stability over thousands of time steps and generalizing across heterogeneous geometries. The framework adopts a temporal multiscale architecture composed of two coupled models operating at complementary time scales. Both models rely on a latent recurrent graph neural network to capture spatiotemporal dynamics on meshes, while a variational graph autoencoder provides a compact latent representation that reduces memory usage and improves training stability. Experiments on simulated powder bed fusion data demonstrate accurate and temporally stable long-horizon predictions across diverse geometries, outperforming existing baseline. Although evaluated in two dimensions, the framework is general and extensible to physics-driven systems with multiscale dynamics and to three-dimensional geometries.

Hallucination detection is critical for ensuring the reliability of large language models (LLMs) in context-based generation. Prior work has explored intrinsic signals available during generation, among which attention offers a direct view of grounding behavior. However, existing approaches typically rely on coarse summaries that fail to capture fine-grained instabilities in attention. Inspired by signal processing, we introduce a frequency-aware perspective on attention by analyzing its variation during generation. We model attention distributions as discrete signals and extract high-frequency components that reflect rapid local changes in attention. Our analysis reveals that hallucinated tokens are associated with high-frequency attention energy, reflecting fragmented and unstable grounding behavior. Based on this insight, we develop a lightweight hallucination detector using high-frequency attention features. Experiments on the RAGTruth and HalluRAG benchmarks show that our approach achieves performance gains over verification-based, internal-representation-based, and attention-based methods across models and tasks.

Automotive software increasingly outpaces hardware availability, forcing late integration and expensive hardware-in-the-loop (HiL) bottlenecks. The InnoRegioChallenge project investigated whether a virtual test and integration environment can reproduce electronic control unit (ECU) behavior early enough to run real software binaries before physical hardware exists. We report a prototype that generates instruction-accurate processor models in SystemC/TLM~2.0 using an agentic, feedback-driven workflow coupled to a reference simulator via the GNU Debugger (GDB). The results indicate that the most critical technical risk -- CPU behavioral fidelity -- can be reduced through automated differential testing and iterative model correction. We summarize the architecture, the agentic modeling loop, and project outcomes, and we extrapolate plausible technical details consistent with the reported qualitative findings. While cloud-scale deployment and full toolchain integration remain future work, the prototype demonstrates a viable shift-left path for virtual ECU twins, enabling reproducible tests, non-intrusive tracing, and fault-injection campaigns aligned with safety standards.

Graph Neural Networks (GNNs) often struggle to propagate information across long distances due to oversmoothing and oversquashing. Existing remedies such as graph transformers or rewiring typically incur high computational cost or require altering the graph structure. We introduce a Bakry-Emery graph Laplacian that integrates diffusion and advection through a learnable node-wise potential, inducing task-dependent propagation dynamics without modifying topology. This operator has a well-behaved spectral decomposition and acts as a drop-in replacement for standard Laplacians in spectral GNNs. Building on this insight, we develop mu-ChebNet, a spectral architecture that jointly learns the potential and Chebyshev filters, effectively bridging message-passing adaptivity and spectral efficiency. Our theoretical analysis shows how the potential modulates the spectrum, enabling control of key graph properties. Empirically, mu-ChebNet delivers consistent gains on synthetic long-range reasoning tasks, as well as real-world benchmarks, while offering an interpretable routing field that reveals how information flows through the graph. This establishes the Bakry-Emery Laplacian as a principled and efficient foundation for adaptive spectral graph learning.

Neuromorphic accelerators promise unparalleled energy efficiency and computational density for spiking neural networks (SNNs), especially in edge intelligence applications. However, most existing platforms exhibit rigid architectures with limited configurability, restricting their adaptability to heterogeneous workloads and diverse design objectives. To address these limitations, we present Flexi-NeurA -- a parameterizable neuromorphic accelerator (core) that unifies configurability, flexibility, and efficiency. Flexi-NeurA allows users to customize neuron models, network structures, and precision settings at design time. By pairing these design-time configurability and flexibility features with a time-multiplexed and event-driven processing approach, Flexi-NeurA substantially reduces the required hardware resources and total power while preserving high efficiency and low inference latency. Complementing this, we introduce Flex-plorer, a heuristic-guided design-space exploration (DSE) tool that determines cost-effective fixed-point precisions for critical parameters -- such as decay factors, synaptic weights, and membrane potentials -- based on user-defined trade-offs between accuracy and resource usage. Based on the configuration selected through the Flex-plorer process, RTL code is configured to match the specified design. Comprehensive evaluations across MNIST, SHD, and DVS benchmarks demonstrate that the Flexi-NeurA and Flex-plorer co-framework achieves substantial improvements in accuracy, latency, and energy efficiency. A three-layer 256--128--10 fully connected network with LIF neurons mapped onto two processing cores achieves 97.23% accuracy on MNIST with 1.1~ms inference latency, utilizing only 1,623 logic cells, 7 BRAMs, and 111~mW of total power -- establishing Flexi-NeurA as a scalable, edge-ready neuromorphic platform.

Large Language Models (LLMs), despite extensive pretraining on broad internet corpora, often struggle to adapt effectively to specialized domains. There is growing interest in fine-tuning these models for such domains; however, progress is constrained by the scarcity and limited coverage of high-quality, task-relevant data. To address this, synthetic data generation methods such as paraphrasing or knowledge extraction are commonly applied. Although these approaches excel at factual recall and conceptual knowledge, they suffer from two critical shortcomings: (i) they provide minimal support for interpretive reasoning capabilities in these specialized domains, and (ii) they often produce synthetic corpora that are excessively large and redundant, resulting in poor sample efficiency. To overcome these gaps, we propose an adversarial question-generation framework that produces a compact set of semantically challenging questions. These questions are constructed by comparing the outputs of the model to be adapted and a robust expert model grounded in reference documents, using an iterative, feedback-driven process designed to reveal and address comprehension gaps. Evaluation on specialized subsets of the LegalBench corpus demonstrates that our method achieves greater accuracy with substantially fewer synthetic samples.

Predictive Coding (PC) is a biologically-inspired learning framework characterised by local, parallelisable operations, properties that enable energy-efficient implementation on neuromorphic hardware. Despite this, extending PC effectively to recurrent neural networks (RNNs) has been challenging, particularly for tasks involving long-range temporal dependencies. Backpropagation Through Time (BPTT) remains the dominant method for training RNNs, but its non-local computation, lack of spatial parallelism, and requirement to store extensive activation histories results in significant energy consumption. This work introduces a novel method combining Temporal Predictive Coding (tPC) with approximate Real-Time Recurrent Learning (RTRL), enabling effective spatio-temporal credit assignment. Results indicate that the proposed method can closely match the performance of BPTT on both synthetic benchmarks and real-world tasks. On a challenging machine translation task, with a 15-million parameter model, the proposed method achieves a test perplexity of 7.62 (vs. 7.49 for BPTT), marking one of the first applications of tPC to tasks of this scale. These findings demonstrate the potential of this method to learn complex temporal dependencies whilst retaining the local, parallelisable, and flexible properties of the original PC framework, paving the way for more energy-efficient learning systems.

Histopathology, the current gold standard for cancer diagnosis, involves the manual examination of tissue samples after chemical staining, a time-consuming process requiring expert analysis. Raman spectroscopy is an alternative, stain-free method of extracting information from samples. Using nnU-Net, we trained a segmentation model on a novel dataset of spatial Raman spectra aligned with tumour annotations, achieving a mean foreground Dice score of 80.9%, surpassing previous work. Furthermore, we propose a novel, interpretable, prototype-based architecture called RamanSeg. RamanSeg classifies pixels based on discovered regions of the training set, generating a segmentation mask. Two variants of RamanSeg allow a trade-off between interpretability and performance: one with prototype projection and another projection-free version. The projection-free RamanSeg outperformed a U-Net baseline with a mean foreground Dice score of 67.3%, offering a meaningful improvement over a black-box training approach.

Generative models have recently demonstrated remarkable success across diverse domains, motivating their adoption as expressive policies in reinforcement learning (RL). While they have shown strong performance in offline RL, particularly where the target distribution is well defined, their extension to online fine-tuning has largely been treated as a direct continuation of offline pre-training, leaving key challenges unaddressed. In this paper, we propose Flow Matching with Injected Noise for Offline-to-Online RL (FINO), a novel method that leverages flow matching-based policies to enhance sample efficiency for offline-to-online RL. FINO facilitates effective exploration by injecting noise into policy training, thereby encouraging a broader range of actions beyond those observed in the offline dataset. In addition to exploration-enhanced flow policy training, we combine an entropy-guided sampling mechanism to balance exploration and exploitation, allowing the policy to adapt its behavior throughout online fine-tuning. Experiments across diverse, challenging tasks demonstrate that FINO consistently achieves superior performance under limited online budgets.

Compressing neural networks without retraining is vital for deployment at scale. We study calibration-free compression through the lens of projection geometry: structured pruning is an axis-aligned projection, whereas model folding performs a low-rank projection via weight clustering. We formalize both as orthogonal operators and show that, within a rank distance of one, folding provably yields smaller parameter reconstruction error, and under mild smoothness assumptions, smaller functional perturbations than pruning. At scale, we evaluate >1000 checkpoints spanning ResNet18, PreActResNet18, ViT-B/32, and CLIP ViT-B/32 on CIFAR-10 and ImageNet-1K, covering diverse training hyperparameters (optimizers, learning rates, augmentations, regularization, sharpness-aware training), as well as multiple LLaMA-family 60M and 130M parameter models trained on C4. We show that folding typically achieves higher post-compression accuracy, with the largest gains at moderate-high compression. The gap narrows and occasionally reverses at specific training setups. Our results position folding as a geometry-aware, calibration-free alternative to pruning that is often superior in practice and principled in theory.

We study online resource allocation under non-stationary demand with a minimum offline data requirement. In this problem, a decision-maker must allocate multiple types of resources to sequentially arriving queries over a finite horizon. Each query belongs to a finite set of types with fixed resource consumption and a stochastic reward drawn from an unknown, type-specific distribution. Critically, the environment exhibits arbitrary non-stationarity -- arrival distributions may shift unpredictably-while the algorithm requires only one historical sample per period to operate effectively. We distinguish two settings based on sample informativeness: (i) reward-observed samples containing both query type and reward realization, and (ii) the more challenging type-only samples revealing only query type information. We propose a novel type-dependent quantile-based meta-policy that decouples the problem into modular components: reward distribution estimation, optimization of target service probabilities via fluid relaxation, and real-time decisions through dynamic acceptance thresholds. For reward-observed samples, our static threshold policy achieves $\tilde{O}(\sqrt{T})$ regret. For type-only samples, we first establish that sublinear regret is impossible without additional structure; under a mild minimum-arrival-probability assumption, we design both a partially adaptive policy attaining the same $\tilde{O}({T})$ bound and, more significantly, a fully adaptive resolving policy with careful rounding that achieves the first poly-logarithmic regret guarantee of $O((\log T)^3)$ for non-stationary multi-resource allocation. Our framework advances prior work by operating with minimal offline data (one sample per period), handling arbitrary non-stationarity without variation-budget assumptions, and supporting multiple resource constraints.

Real-time schedulers must reason about tight deadlines under strict compute budgets. We present TempoNet, a reinforcement learning scheduler that pairs a permutation-invariant Transformer with a deep Q-approximation. An Urgency Tokenizer discretizes temporal slack into learnable embeddings, stabilizing value learning and capturing deadline proximity. A latency-aware sparse attention stack with blockwise top-k selection and locality-sensitive chunking enables global reasoning over unordered task sets with near-linear scaling and sub-millisecond inference. A multicore mapping layer converts contextualized Q-scores into processor assignments through masked-greedy selection or differentiable matching. Extensive evaluations on industrial mixed-criticality traces and large multiprocessor settings show consistent gains in deadline fulfillment over analytic schedulers and neural baselines, together with improved optimization stability. Diagnostics include sensitivity analyses for slack quantization, attention-driven policy interpretation, hardware-in-the-loop and kernel micro-benchmarks, and robustness under stress with simple runtime mitigations; we also report sample-efficiency benefits from behavioral-cloning pretraining and compatibility with an actor-critic variant without altering the inference pipeline. These results establish a practical framework for Transformer-based decision making in high-throughput real-time scheduling.

In voice conversion (VC) applications, diffusion and flow-matching models have exhibited exceptional speech quality and speaker similarity performances. However, they are limited by slow conversion owing to their iterative inference. Consequently, we propose MeanVoiceFlow, a novel one-step nonparallel VC model based on mean flows, which can be trained from scratch without requiring pretraining or distillation. Unlike conventional flow matching that uses instantaneous velocity, mean flows employ average velocity to more accurately compute the time integral along the inference path in a single step. However, training the average velocity requires its derivative to compute the target velocity, which can cause instability. Therefore, we introduce a structural margin reconstruction loss as a zero-input constraint, which moderately regularizes the input-output behavior of the model without harmful statistical averaging. Furthermore, we propose conditional diffused-input training in which a mixture of noise and source data is used as input to the model during both training and inference. This enables the model to effectively leverage source information while maintaining consistency between training and inference. Experimental results validate the effectiveness of these techniques and demonstrate that MeanVoiceFlow achieves performance comparable to that of previous multi-step and distillation-based models, even when trained from scratch. Audio samples are available at https://www.kecl.ntt.co.jp/people/kaneko.takuhiro/projects/meanvoiceflow/.

In this paper, we present a framework for enabling autonomous vehicles to interact with cyclists in a manner that balances safety and optimality. The approach integrates Hamilton-Jacobi reachability analysis with deep Q-learning to jointly address safety guarantees and time-efficient navigation. A value function is computed as the solution to a time-dependent Hamilton-Jacobi-Bellman inequality, providing a quantitative measure of safety for each system state. This safety metric is incorporated as a structured reward signal within a reinforcement learning framework. The method further models the cyclist's latent response to the vehicle, allowing disturbance inputs to reflect human comfort and behavioral adaptation. The proposed framework is evaluated through simulation and comparison with human driving behavior and an existing state-of-the-art method.

Natural language understanding requires interleaving textual and logical reasoning, yet large language models often fail to perform such reasoning reliably. Existing neurosymbolic systems combine LLMs with solvers but remain limited to fully formalizable tasks such as math or program synthesis, leaving natural documents with only partial logical structure unaddressed. We introduce Logitext, a neurosymbolic language that represents documents as natural language text constraints (NLTCs), making partial logical structure explicit. We develop an algorithm that integrates LLM-based constraint evaluation with satisfiability modulo theory (SMT) solving, enabling joint textual-logical reasoning. Experiments on a new content moderation benchmark, together with LegalBench and Super-Natural Instructions, show that Logitext improves both accuracy and coverage. This work is the first that treats LLM-based reasoning as an SMT theory, extending neurosymbolic methods beyond fully formalizable domains.

Existing Visual-Language Models (VLMs) have achieved significant progress by being trained on massive-scale datasets, typically under the assumption that data are independent and identically distributed (IID). However, in real-world scenarios, it is often impractical to expect that all data processed by an AI system satisfy this assumption. Furthermore, failure to appropriately handle out-of-distribution (OOD) objects may introduce safety risks in real-world applications (e.g., autonomous driving or medical assistance). Unfortunately, current research has not yet provided valid benchmarks that can comprehensively assess the performance of VLMs in response to OOD data. Therefore, we propose OODBench, a predominantly automated method with minimal human verification, for constructing new benchmarks and evaluating the ability of VLMs to process OOD data. OODBench contains 40K instance-level OOD instance-category pairs, and we show that current VLMs still exhibit notable performance degradation on OODBench, even when the underlying image categories are common. In addition, we propose a reliable automated assessment metric that employs a Basic-to-Advanced Progression of prompted questions to assess the impact of OOD data on questions of varying difficulty more fully. Lastly, we summarize substantial findings and insights to facilitate future research in the acquisition and evaluation of OOD data.

Conversational AI has been proposed as a scalable way to correct public misconceptions and spread misinformation. Yet its effectiveness may depend on perceptions of its political neutrality. As LLMs enter partisan conflict, elites increasingly portray them as ideologically aligned. We test whether these credibility attacks reduce LLM-based persuasion. In a preregistered U.S. survey experiment (N=2144), participants completed a three-round conversation with ChatGPT about a personally held economic policy misconception. Compared to a neutral control, a short message indicating that the LLM was biased against the respondent's party attenuated persuasion by 28%. Transcript analysis indicates that the warnings alter the interaction: respondents push back more and engage less receptively. These findings suggest that the persuasive impact of conversational AI is politically contingent, constrained by perceptions of partisan alignment.

Despite having hundreds of millions of speakers, handwritten Devanagari text remains severely underrepresented in publicly available benchmark datasets. Existing resources are limited in scale, focus primarily on isolated characters or short words, and lack controlled lexical content and writer level diversity, which restricts their utility for modern data driven handwriting analysis. As a result, they fail to capture the continuous, fused, and structurally complex nature of Devanagari handwriting, where characters are connected through a shared shirorekha (horizontal headline) and exhibit rich ligature formations. We introduce DohaScript, a large scale, multi writer dataset of handwritten Hindi text collected from 531 unique contributors. The dataset is designed as a parallel stylistic corpus, in which all writers transcribe the same fixed set of six traditional Hindi dohas (couplets). This controlled design enables systematic analysis of writer specific variation independent of linguistic content, and supports tasks such as handwriting recognition, writer identification, style analysis, and generative modeling. The dataset is accompanied by non identifiable demographic metadata, rigorous quality curation based on objective sharpness and resolution criteria, and page level layout difficulty annotations that facilitate stratified benchmarking. Baseline experiments demonstrate clear quality separation and strong generalization to unseen writers, highlighting the dataset's reliability and practical value. DohaScript is intended to serve as a standardized and reproducible benchmark for advancing research on continuous handwritten Devanagari text in low resource script settings.

Leveraging channel state information from multiple Wi-Fi bands can improve delay resolution for ranging and sensing when a wide contiguous spectrum is unavailable. However, frequency gaps shape the delay response, introducing sidelobes and secondary peaks that can obscure closely spaced multipath components. This paper examines multipath delay estimation for Wi-Fi-compliant multiband configurations using channel state information (CSI). For a two-path model with unknown complex gains and delays, the Cramér-Rao lower bound (CRLB) for delay separation is derived and analyzed, confirming the benefit of larger frequency aperture, while revealing pronounced, separation-dependent oscillations driven by gap geometry and inter-path coupling. Given the local nature of Cramér-Rao lower bound, the delay response is analyzed next. In the single-path case, the combined subband responses determine how delay-domain sidelobe levels are distributed. The dominant peak spacing is set primarily by the separation between subband center frequencies. In the two-path case, increased aperture sharpens the mainlobe but also intensifies sidelobes and leakage, yielding competing peaks and, in some regimes, a dominant peak shifted from the true delay. Finally, a normalized leakage metric is introduced to predict problematic separations and to identify regimes where local Cramér-Rao lower bound analysis does not capture practical peak-leakage behavior in delay estimation.

Equivariance is central to graph generative models, as it ensures the model respects the permutation symmetry of graphs. However, strict equivariance can increase computational cost due to added architectural constraints, and can slow down convergence because the model must be consistent across a large space of possible node permutations. We study this trade-off for graph generative models. Specifically, we start from an equivariant discrete flow-matching model, and relax its equivariance during training via a controllable symmetry modulation scheme based on sinusoidal positional encodings and node permutations. Experiments first show that symmetry-breaking can accelerate early training by providing an easier learning signal, but at the expense of encouraging shortcut solutions that can cause overfitting, where the model repeatedly generates graphs that are duplicates of the training set. On the contrary, properly modulating the symmetry signal can delay overfitting while accelerating convergence, allowing the model to reach stronger performance with $19\%$ of the baseline training epochs.

Accurate soil moisture (SM) estimation is critical for precision agriculture, water resources management and climate monitoring. Yet, existing satellite SM products are too coarse (>1km) for farm-level applications. We present a high-resolution (10m) SM estimation framework for vegetated areas across Europe, combining Sentinel-1 SAR, Sentinel-2 optical imagery and ERA-5 reanalysis data through machine learning. Using 113 International Soil Moisture Network (ISMN) stations spanning diverse vegetated areas, we compare modality combinations with temporal parameterizations, using spatial cross-validation, to ensure geographic generalization. We also evaluate whether foundation model embeddings from IBM-NASA's Prithvi model improve upon traditional hand-crafted spectral features. Results demonstrate that hybrid temporal matching - Sentinel-2 current-day acquisitions with Sentinel-1 descending orbit - achieves R^2=0.514, with 10-day ERA5 lookback window improving performance to R^2=0.518. Foundation model (Prithvi) embeddings provide negligible improvement over hand-crafted features (R^2=0.515 vs. 0.514), indicating traditional feature engineering remains highly competitive for sparse-data regression tasks. Our findings suggest that domain-specific spectral indices combined with tree-based ensemble methods offer a practical and computationally efficient solution for operational pan-European field-scale soil moisture monitoring.

In this work, we present HiAER-Spike, a modular, reconfigurable, event-driven neuromorphic computing platform designed to execute large spiking neural networks with up to 160 million neurons and 40 billion synapses - roughly twice the neurons of a mouse brain at faster than real time. This system, assembled at the UC San Diego Supercomputer Center, comprises a co-designed hard- and software stack that is optimized for run-time massively parallel processing and hierarchical address-event routing (HiAER) of spikes while promoting memory-efficient network storage and execution. The architecture efficiently handles both sparse connectivity and sparse activity for robust and low-latency event-driven inference for both edge and cloud computing. A Python programming interface to HiAER-Spike, agnostic to hardware-level detail, shields the user from complexity in the configuration and execution of general spiking neural networks with minimal constraints in topology. The system is made easily available over a web portal for use by the wider community. In the following, we provide an overview of the hard- and software stack, explain the underlying design principles, demonstrate some of the system's capabilities and solicit feedback from the broader neuromorphic community. Examples are shown demonstrating HiAER-Spike's capabilities for event-driven vision on benchmark CIFAR-10, DVS event-based gesture, MNIST, and Pong tasks.

Over the past four decades, distributed security has undergone a remarkable transformation -- from crash-fault tolerant protocols designed for controlled environments to sophisticated Byzantine-resilient architectures operating in open, adversarial settings. This vision paper examines this evolution and argues for a fundamental shift in how we approach distributed security: from studying individual security properties in isolation to understanding their synergistic combinations. We begin by conclude four foundational properties, \textit{agreement, consistency, privacy, verifiability, accountability}. We trace their theoretical origins and practical maturation. We then demonstrate how the frontier of research now lies at the intersection of these properties, where their fusion creates capabilities that neither property could achieve alone. Looking forward, we identify critical research challenges: discovering new security properties driven by emerging applications, developing systematic frameworks for property convergence, managing the computational overhead of cryptographic primitives in high-performance consensus layers, and addressing post-quantum and human-factor challenges. The future of distributed security lies not in improving individual properties, but in understanding and harnessing their synergies to build a singular fabric of trust.

Physics-informed deep learning models have emerged as powerful tools for learning dynamical systems. These models directly encode physical principles into network architectures. However, systematic benchmarking of these approaches across diverse physical phenomena remains limited, particularly in conservative and dissipative systems. In addition, benchmarking that has been done thus far does not integrate out full trajectories to check stability. In this work, we benchmark three prominent physics-informed architectures such as Hamiltonian Neural Networks (HNN), Lagrangian Neural Networks (LNN), and Symplectic Recurrent Neural Networks (SRNN) using the DeepChem framework, an open-source scientific machine learning library. We evaluate these models on six dynamical systems spanning classical conservative mechanics (mass-spring system, simple pendulum, double pendulum, and three-body problem, spring-pendulum) and non-conservative systems with contact (bouncing ball). We evaluate models by computing error on predicted trajectories and evaluate error both quantitatively and qualitatively. We find that all benchmarked models struggle to maintain stability for chaotic or nonconservative systems. Our results suggest that more research is needed for physics-informed deep learning models to learn robust models of classical mechanical systems.

Dual-to-Dual MLLMs refer to Multimodal Large Language Models, which can enable unified multimodal comprehension and generation through text and image modalities. Although exhibiting strong instantaneous learning and generalization capabilities, Dual-to-Dual MLLMs still remain deficient in lifelong evolution, significantly affecting continual adaptation to dynamic real-world scenarios. One of the challenges is that learning new tasks inevitably destroys the learned knowledge. Beyond traditional catastrophic forgetting, Dual-to-Dual MLLMs face other challenges, including hallucination, instruction unfollowing, and failures in cross-modal knowledge transfer. However, no standardized continual learning framework for Dual-to-Dual MLLMs has been established yet, leaving these challenges unexplored. Thus, in this paper, we establish Continual-NExT, a continual learning framework for Dual-to-Dual MLLMs with deliberately-architected evaluation metrics. To improve the continual learning capability of Dual-to-Dual MLLMs, we propose an efficient MAGE (Mixture and Aggregation of General LoRA and Expert LoRA) method to further facilitate knowledge transfer across modalities and mitigate forgetting. Extensive experiments demonstrate that MAGE outperforms other continual learning methods and achieves state-of-the-art performance.

Conditional Value-at-Risk (CVaR) is a widely used risk-sensitive objective for learning under rare but high-impact losses, yet its statistical behavior under heavy-tailed data remains poorly understood. Unlike expectation-based risk, CVaR depends on an endogenous, data-dependent quantile, which couples tail averaging with threshold estimation and fundamentally alters both generalization and robustness properties. In this work, we develop a learning-theoretic analysis of CVaR-based empirical risk minimization under heavy-tailed and contaminated data. We establish sharp, high-probability generalization and excess risk bounds under minimal moment assumptions, covering fixed hypotheses, finite and infinite classes, and extending to $β$-mixing dependent data; we further show that these rates are minimax optimal. To capture the intrinsic quantile sensitivity of CVaR, we derive a uniform Bahadur-Kiefer type expansion that isolates a threshold-driven error term absent in mean-risk ERM and essential in heavy-tailed regimes. We complement these results with robustness guarantees by proposing a truncated median-of-means CVaR estimator that achieves optimal rates under adversarial contamination. Finally, we show that CVaR decisions themselves can be intrinsically unstable under heavy tails, establishing a fundamental limitation on decision robustness even when the population optimum is well separated. Together, our results provide a principled characterization of when CVaR learning generalizes and is robust, and when instability is unavoidable due to tail scarcity.

City-scale person re-identification across distributed cameras must handle severe appearance changes from viewpoint, occlusion, and domain shift while complying with data protection rules that prevent sharing raw imagery. We introduce CityGuard, a topology-aware transformer for privacy-preserving identity retrieval in decentralized surveillance. The framework integrates three components. A dispersion-adaptive metric learner adjusts instance-level margins according to feature spread, increasing intra-class compactness. Spatially conditioned attention injects coarse geometry, such as GPS or deployment floor plans, into graph-based self-attention to enable projectively consistent cross-view alignment using only coarse geometric priors without requiring survey-grade calibration. Differentially private embedding maps are coupled with compact approximate indexes to support secure and cost-efficient deployment. Together these designs produce descriptors robust to viewpoint variation, occlusion, and domain shifts, and they enable a tunable balance between privacy and utility under rigorous differential-privacy accounting. Experiments on Market-1501 and additional public benchmarks, complemented by database-scale retrieval studies, show consistent gains in retrieval precision and query throughput over strong baselines, confirming the practicality of the framework for privacy-critical urban identity matching.

Deployed conformal predictors are long-lived decision infrastructure operating over finite operational windows. The real-world question is not only ``Does the true label lie in the prediction set at the target rate?'' (marginal coverage), but ``How often does the system commit versus defer? What error exposure does it induce when it acts? How do these rates trade off?'' Marginal coverage does not determine these deployment-facing quantities: the same calibrated thresholds can yield different operational profiles depending on score geometry. We provide a framework for operational certification and planning beyond coverage with three contributions. (1) Small-Sample Beta Correction (SSBC): we invert the exact finite-sample Beta/rank law for split conformal to map a user request $(α^\star,δ)$ to a calibrated grid point with PAC-style semantics, yielding explicit finite-window coverage guarantees. (2) Calibrate-and-Audit: since no distribution-free pivot exists for rates beyond coverage, we introduce a two-stage design in which an independent audit set produces a reusable region -- label table and certified finite-window envelopes (Binomial/Beta-Binomial) for operational quantities -- commitment frequency, deferral, decisive error exposure, and commit purity -- via linear projection. (3) Geometric characterization: we describe feasibility constraints, regime boundaries (hedging vs.\ rejection), and cost-coherence conditions induced by a fixed conformal partition, explaining why operational rates are coupled and how calibration navigates their trade-offs. The output is an auditable operational menu: for a fixed scoring model, we trace attainable operational profiles across calibration settings and attach finite-window uncertainty envelopes. We demonstrate the approach on Tox21 toxicity prediction (12 endpoints) and aqueous solubility screening using AquaSolDB.

Accurate, real-time, yet non-destructive estimation of internal states in lithium-ion batteries is critical for predicting degradation, optimizing usage strategies, and extending operational lifespan. Here, we introduce PINEAPPLE (Physics-Informed Neuro-Evolution Algorithm for Prognostic Parameter inference in Lithium-ion battery Electrodes), a novel framework that integrates physics-informed neural networks (PINNs) with an evolutionary search algorithm to enable rapid, scalable, and interpretable parameter inference with potential for application to next-generation batteries. The meta-learned PINN utilizes fundamental physics principles to achieve accurate zero-shot prediction of electrode behavior with test errors below 0.1$\%$ while maintaining an order-of-magnitude speed-up over conventional solvers. PINEAPPLE demonstrates robust parameter inference solely from voltage-time discharge curves across multiple batteries from the open-source CALCE repository, recovering the evolution of key internal state parameters such as Li-ion diffusion coefficients across usage cycles. Notably, the inferred cycle-dependent evolution of these parameters exhibit consistent trends across different batteries without any customized degradation physics-embedded heuristic, highlighting the effective regularizing effect and robustness that can be conferred through incorporation of fundamental physics in PINEAPPLE. By enabling computationally efficient, real-time parameter estimation, PINEAPPLE offers a promising route towards the non-destructive, physics-based characterization of inter-cell and intra-cell variability of battery modules and battery packs, thereby unlocking new opportunities for downstream on-the-fly needs in next-generation battery management systems such as individual cell-scale state-of-health diagnostics.

Reinforcement Learning from Human Feedback (RLHF) or Verifiable Rewards (RLVR) are two key steps in the post-training of modern Language Models (LMs). A common problem is reward hacking, where the policy may exploit inaccuracies of the reward and learn an unintended behavior. Most previous works address this by limiting the policy update with a Kullback-Leibler (KL) penalty towards a reference model. We propose a different framing: Train the LM in a way that biases policy updates towards regions in which the reward is more accurate. First, we derive a theoretical connection between the accuracy of a reward model and the flatness of an optimum at convergence. Gradient regularization (GR) can then be used to bias training to flatter regions and thereby maintain reward model accuracy. We confirm these results by showing that the gradient norm and reward accuracy are empirically correlated in RLHF. We then show that Reference Resets of the KL penalty implicitly use GR to find flatter regions with higher reward accuracy. We further improve on this by proposing to use explicit GR with an efficient finite-difference estimate. Empirically, GR performs better than a KL penalty across a diverse set of RL experiments with LMs. GR achieves a higher GPT-judged win-rate in RLHF, avoids overly focusing on the format in rule-based math rewards, and prevents hacking the judge in LLM-as-a-Judge math tasks.

Evaluation is no longer a final checkpoint in the machine learning lifecycle. As AI systems evolve from static models to compound, tool-using agents, evaluation becomes a core control function. The question is no longer "How good is the model?" but "Can we trust the system to behave as intended, under change, at scale?". Yet most evaluation practices remain anchored in assumptions inherited from the model-centric era: static benchmarks, aggregate scores, and one-off success criteria. This paper argues that such approaches are increasingly obscure rather than illuminating system behavior. We examine how evaluation pipelines themselves introduce silent failure modes, why high benchmark scores routinely mislead teams, and how agentic systems fundamentally alter the meaning of performance measurement. Rather than proposing new metrics or harder benchmarks, we aim to clarify the role of evaluation in the AI era, and especially for agents: not as performance theater, but as a measurement discipline that conditions trust, iteration, and governance in non-deterministic systems.

Mean-field reinforcement learning (MF-RL) scales multi-agent RL to large populations by reducing each agent's dependence on others to a single summary statistic -- the mean action. However, this reduction requires every agent to act at every time step; when some agents are idle, the mean action is simply undefined. Addressing asynchrony therefore requires a different summary statistic -- one that remains defined regardless of which agents act. The population distribution $μ\in Δ(\mathcal{O})$ -- the fraction of agents at each observation -- satisfies this requirement: its dimension is independent of $N$, and under exchangeability it fully determines each agent's reward and transition. Existing MF-RL theory, however, is built on the mean action and does not extend to $μ$. We therefore construct the Temporal Mean Field (TMF) framework around the population distribution $μ$ from scratch, covering the full spectrum from fully synchronous to purely sequential decision-making within a single theory. We prove existence and uniqueness of TMF equilibria, establish an $O(1/\sqrt{N})$ finite-population approximation bound that holds regardless of how many agents act per step, and prove convergence of a policy gradient algorithm (TMF-PG) to the unique equilibrium. Experiments on a resource selection game and a dynamic queueing game confirm that TMF-PG achieves near-identical performance whether one agent or all $N$ act per step, with approximation error decaying at the predicted $O(1/\sqrt{N})$ rate.

Scalable robot policy pre-training has been hindered by the high cost of collecting high-quality demonstrations for each platform. In this study, we address this issue by uniting offline reinforcement learning (offline RL) with cross-embodiment learning. Offline RL leverages both expert and abundant suboptimal data, and cross-embodiment learning aggregates heterogeneous robot trajectories across diverse morphologies to acquire universal control priors. We perform a systematic analysis of this offline RL and cross-embodiment paradigm, providing a principled understanding of its strengths and limitations. To evaluate this offline RL and cross-embodiment paradigm, we construct a suite of locomotion datasets spanning 16 distinct robot platforms. Our experiments confirm that this combined approach excels at pre-training with datasets rich in suboptimal trajectories, outperforming pure behavior cloning. However, as the proportion of suboptimal data and the number of robot types increase, we observe that conflicting gradients across morphologies begin to impede learning. To mitigate this, we introduce an embodiment-based grouping strategy in which robots are clustered by morphological similarity and the model is updated with a group gradient. This simple, static grouping substantially reduces inter-robot conflicts and outperforms existing conflict-resolution methods.

Training-free control over editing intensity is a critical requirement for diffusion-based image editing models built on the Diffusion Transformer (DiT) architecture. Existing attention manipulation methods focus exclusively on the Key space to modulate attention routing, leaving the Value space -- which governs feature aggregation -- entirely unexploited. In this paper, we first reveal that both Key and Value projections in DiT's multi-modal attention layers exhibit a pronounced bias-delta structure, where token embeddings cluster tightly around a layer-specific bias vector. Building on this observation, we propose Dual-Channel Attention Guidance (DCAG), a training-free framework that simultaneously manipulates both the Key channel (controlling where to attend) and the Value channel (controlling what to aggregate). We provide a theoretical analysis showing that the Key channel operates through the nonlinear softmax function, acting as a coarse control knob, while the Value channel operates through linear weighted summation, serving as a fine-grained complement. Together, the two-dimensional parameter space $(δ_k, δ_v)$ enables more precise editing-fidelity trade-offs than any single-channel method. Extensive experiments on the PIE-Bench benchmark (700 images, 10 editing categories) demonstrate that DCAG consistently outperforms Key-only guidance across all fidelity metrics, with the most significant improvements observed in localized editing tasks such as object deletion (4.9% LPIPS reduction) and object addition (3.2% LPIPS reduction).

In the literature, prior research on Security-oriented Video Understanding (SVU) has predominantly focused on detecting and localize the threats (e.g., shootings, robberies) in videos, while largely lacking the effective capability to generate and evaluate the threat causes. Motivated by these gaps, this paper introduces a new chat paradigm SVU task, i.e., In-depth Security-oriented Video Understanding (DeepSVU), which aims to not only identify and locate the threats but also attribute and evaluate the causes threatening segments. Furthermore, this paper reveals two key challenges in the proposed task: 1) how to effectively model the coarse-to-fine physical-world information (e.g., human behavior, object interactions and background context) to boost the DeepSVU task; and 2) how to adaptively trade off these factors. To tackle these challenges, this paper proposes a new Unified Physical-world Regularized MoE (UPRM) approach. Specifically, UPRM incorporates two key components: the Unified Physical-world Enhanced MoE (UPE) Block and the Physical-world Trade-off Regularizer (PTR), to address the above two challenges, respectively. Extensive experiments conduct on our DeepSVU instructions datasets (i.e., UCF-C instructions and CUVA instructions) demonstrate that UPRM outperforms several advanced Video-LLMs as well as non-VLM approaches. Such information.These justify the importance of the coarse-to-fine physical-world information in the DeepSVU task and demonstrate the effectiveness of our UPRM in capturing such information.

The dataset distributions in offline reinforcement learning (RL) often exhibit complex and multi-modal distributions, necessitating expressive policies to capture such distributions beyond widely-used Gaussian policies. To handle such complex and multi-modal datasets, in this paper, we propose Flow Actor-Critic, a new actor-critic method for offline RL, based on recent flow policies. The proposed method not only uses the flow model for actor as in previous flow policies but also exploits the expressive flow model for conservative critic acquisition to prevent Q-value explosion in out-of-data regions. To this end, we propose a new form of critic regularizer based on the flow behavior proxy model obtained as a byproduct of flow-based actor design. Leveraging the flow model in this joint way, we achieve new state-of-the-art performance for test datasets of offline RL including the D4RL and recent OGBench benchmarks.

The increasing use of language models in automated software testing raises concerns about their environmental impact, yet existing sustainability analyses focus almost exclusively on large language models. As a result, the energy and carbon characteristics of small language models (SLMs) during test generation remain largely unexplored. To address this gap, this work introduces the DeCEAT framework, which systematically evaluates the environmental and performance trade-offs of SLMs using the HumanEval benchmark and adaptive prompt variants (based on the Anthropic template). The framework quantifies emission and time-aware behavior under controlled conditions, with CodeCarbon measuring energy consumption and carbon emissions, and unit test coverage assessing the quality of generated tests. Our results show that different SLMs exhibit distinct sustainability strengths: some prioritize lower energy use and faster execution, while others maintain higher stability or accuracy under carbon constraints. These findings demonstrate that sustainability in the generation of SLM-driven tests is multidimensional and strongly shaped by prompt design. This work provides a focused sustainability evaluation framework specifically tailored to automated SLM-based test generation, clarifying how prompt structure and model choice jointly influence environmental and performance outcomes.

Mechanistic models encode scientific knowledge about dynamical systems and are widely used in downstream scientific and policy applications. Recent work has explored LLM-based agentic frameworks to automatically construct mechanistic models from data; however, existing problem settings substantially oversimplify real-world conditions, leaving it unclear whether LLM-generated mechanistic models are reliable in practice. To address this gap, we introduce the Neural-Integrated Mechanistic Modeling (NIMM) evaluation framework, which evaluates LLM-generated mechanistic models under realistic settings with partial observations and diversified task objectives. Our evaluation reveals fundamental challenges in current baselines, ranging from model effectiveness to code-level correctness. Motivated by these findings, we design NIMMgen, an agentic framework for neural-integrated mechanistic modeling that enhances code correctness and practical validity through iterative refinement. Experiments across three datasets from diversified scientific domains demonstrate its strong performance. We also show that the learned mechanistic models support counterfactual intervention simulation.

As large language models continue to scale, training demands on compute and system capacity grow rapidly, making single-vendor homogeneous clusters insufficient. This paper presents a technical solution for heterogeneous mixed training in AMD-NVIDIA environments. We first adopt a compatibility-oriented approach based on CPU-Forwarding Communication, with differentiated communication back-end selection across parallel groups and multi-NIC parallel data transfer. To achieve higher performance, we further propose another Device-Direct Communication approach, integrating a CPU-offloading P2P mechanism to enable direct cross-vendor GPU data transfer without host-memory staging. Experiments on LLaMA-8B and Qwen2-7B demonstrate that the proposed Device-Direct Communication approach achieves up to 98% of the throughput of an NVIDIA homogeneous system, while preserving training stability and correctness.

Heavy-tailed stochastic gradient noise, commonly observed in transformer models, can destabilize the optimization process. Recent works mainly focus on developing and understanding approaches to address heavy-tailed noise in the centralized or distributed, synchronous setting, leaving the interactions between such noise and asynchronous optimization underexplored. In this work, we investigate two communication schemes that handle stragglers with asynchronous updates in the presence of heavy-tailed gradient noise. We propose and theoretically analyze algorithmic modifications based on delay-aware learning rate scheduling and delay compensation to enhance the performance of asynchronous algorithms. Our convergence guarantees under heavy-tailed noise match the rate of the synchronous counterparts and improve delay tolerance compared with existing asynchronous approaches. Empirically, our approaches outperform prior synchronous and asynchronous methods in terms of accuracy/runtime trade-offs and are more robust to hyperparameters in both image and language tasks.

Academic advising in higher education is under severe strain, with advisor-to-student ratios commonly exceeding 300:1. These structural bottlenecks limit timely access to guidance, increase the risk of delayed graduation, and contribute to inequities in student support. We introduce Aurora, a modular neuro-symbolic advising agent that unifies retrieval-augmented generation (RAG), symbolic reasoning, and normalized curricular databases to deliver policy-compliant, verifiable recommendations at scale. Aurora integrates three components: (i) a Boyce-Codd Normal Form (BCNF) catalog schema for consistent program rules, (ii) a Prolog engine for prerequisite and credit enforcement, and (iii) an instruction-tuned large language model for natural-language explanations of its recommendations. To assess performance, we design a structured evaluation suite spanning common and edge-case advising scenarios, including short-term scheduling, long-term roadmapping, skill-aligned pathways, and out-of-scope requests. Across this diverse set, Aurora improves semantic alignment with expert-crafted answers from 0.68 (Raw LLM baseline) to 0.93 (+36%), achieves perfect precision and recall in nearly half of in-scope cases, and consistently produces correct fallbacks for unanswerable prompts. On commodity hardware, Aurora delivers sub-second mean latency (0.71s across 20 queries), approximately 83X faster than a Raw LLM baseline (59.2s). By combining symbolic rigor with neural fluency, Aurora advances a paradigm for accurate, explainable, and scalable AI-driven advising.

Real physical systems are dissipative -- a pendulum slows, a circuit loses charge to heat -- and forecasting their dynamics from partial observations is a central challenge in scientific machine learning. We address the \emph{position-only} (q-only) problem: given only generalized positions~$q_t$ at discrete times (momenta~$p_t$ latent), learn a structured model that (a)~produces stable long-horizon forecasts and (b)~recovers physically meaningful parameters when sufficient structure is provided. The port-Hamiltonian framework makes the conservative-dissipative split explicit via $\dot{x}=(J-R)\nabla H(x)$, guaranteeing $dH/dt\le 0$ when $R\succeq 0$. We introduce \textbf{PHAST} (Port-Hamiltonian Architecture for Structured Temporal dynamics), which decomposes the Hamiltonian into potential~$V(q)$, mass~$M(q)$, and damping~$D(q)$ across three knowledge regimes (KNOWN, PARTIAL, UNKNOWN), uses efficient low-rank PSD/SPD parameterizations, and advances dynamics with Strang splitting. Across thirteen q-only benchmarks spanning mechanical, electrical, molecular, thermal, gravitational, and ecological systems, PHAST achieves the best long-horizon forecasting among competitive baselines and enables physically meaningful parameter recovery when the regime provides sufficient anchors. We show that identification is fundamentally ill-posed without such anchors (gauge freedom), motivating a two-axis evaluation that separates forecasting stability from identifiability.

Whole-brain biological neural networks naturally support the learning and control of whole-body movements. However, the use of brain connectomes as neural network controllers in embodied reinforcement learning remains unexplored. We investigate using the exact neural architecture of an adult fruit fly's brain for the control of its body movement. We develop Fly-connectomic Graph Model (FlyGM), whose static structure is identical to the complete connectome of an adult Drosophila for whole-body locomotion control. To perform dynamical control, FlyGM represents the static connectome as a directed message-passing graph to impose a biologically grounded information flow from sensory inputs to motor outputs. Integrated with a biomechanical fruit fly model, our method achieves stable control across diverse locomotion tasks without task-specific architectural tuning. To verify the structural advantages of the connectome-based model, we compare it against a degree-preserving rewired graph, a random graph, and multilayer perceptrons, showing that FlyGM yields higher sample efficiency and superior performance. This work demonstrates that static brain connectomes can be transformed to instantiate effective neural policy for embodied learning of movement control.

Complex problems, whether in math, logic, or planning, are solved by humans through a sequence of steps where the result of one step informs the next. In this work, we adopt the perspective that the reasoning power of Transformers is fundamentally limited by a fixed maximum number of steps along any latent path of computation. To address this, we introduce Turbo Connection (TurboConn), a novel architecture that overcomes the fixed-depth constraint by routing multiple residual connections from the higher-layer hidden states of each token $t$ to the lower layers of token $t+1$. Fine-tuning pre-trained LLMs with our method not only yields accuracy gains of 0.9% to over 10% on benchmarks like GSM8K, Parity, and multi-step arithmetic, but also demonstrates that the density of these backward connections is critical; our dense interaction significantly outperforms "sparse" alternatives that only pass a single hidden state or vector. Notably, TurboConn can be integrated into pre-trained LLMs to overcome task-specific plateaus: while a fine-tuned Qwen-3-1.7B achieves only 53.78% on Parity, adding our architectural modification enables the model to reach 100% accuracy, all without the necessity to retrain the full model from scratch or sophisticated curriculum learning. Our results provide strong empirical evidence that the depth of the computational path is a key factor in reasoning ability, also offering a new mechanism to enhance LLMs without significantly affecting generation latency.

LLM-based systems increasingly generate structured workflows for complex tasks. In practice, automatic evaluation of these workflows is difficult, because metric scores are often not calibrated, and score changes do not directly communicate the severity of workflow degradation. We introduce WorkflowPerturb, a controlled benchmark for studying workflow evaluation metrics. It works by applying realistic, controlled perturbations to golden workflows. WorkflowPerturb contains 4,973 golden workflows and 44,757 perturbed variants across three perturbation types (Missing Steps, Compressed Steps, and Description Changes), each applied at severity levels of 10%, 30%, and 50%. We benchmark multiple metric families and analyze their sensitivity and calibration using expected score trajectories and residuals. Our results characterize systematic differences across metric families and support severity-aware interpretation of workflow evaluation scores. Our dataset will be released upon acceptance.

Machine learning is at the heart of managing the real-world problems associated with massive data. With the success of neural networks on such large-scale problems, more research in machine learning is being conducted now than ever before. This dissertation focuses on three different projects rooted in mathematical theory for machine learning applications. The first project deals with supervised learning and manifold learning. In theory, one of the main problems in supervised learning is that of function approximation: that is, given some data set $\mathcal{D}=\{(x_j,f(x_j))\}_{j=1}^M$, can one build a model $F\approx f$? We introduce a method which aims to remedy several of the theoretical shortcomings of the current paradigm for supervised learning. The second project deals with transfer learning, which is the study of how an approximation process or model learned on one domain can be leveraged to improve the approximation on another domain. We study such liftings of functions when the data is assumed to be known only on a part of the whole domain. We are interested in determining subsets of the target data space on which the lifting can be defined, and how the local smoothness of the function and its lifting are related. The third project is concerned with the classification task in machine learning, particularly in the active learning paradigm. Classification has often been treated as an approximation problem as well, but we propose an alternative approach leveraging techniques originally introduced for signal separation problems. We introduce theory to unify signal separation with classification and a new algorithm which yields competitive accuracy to other recent active learning algorithms while providing results much faster.

Retrieval-augmented generation is increasingly used for financial question answering over long regulatory filings, yet reliability depends on retrieving the exact context needed to justify answers in high stakes settings. We study a frequent failure mode in which the correct document is retrieved but the page or chunk that contains the answer is missed, leading the generator to extrapolate from incomplete context. Despite its practical significance, this within-document retrieval failure mode has received limited systematic attention in the Financial Question Answering (QA) literature. We evaluate retrieval at multiple levels of granularity, document, page, and chunk level, and introduce an oracle based analysis to provide empirical upper bounds on retrieval and generative performance. On a 150 question subset of FinanceBench, we reproduce and compare diverse retrieval strategies including dense, sparse, hybrid, and hierarchical methods with reranking and query reformulation. Across methods, gains in document discovery tend to translate into stronger page recall, yet oracle performance still suggests headroom for page and chunk level retrieval. To target this gap, we introduce a domain fine-tuned page scorer that treats pages as an intermediate retrieval unit between documents and chunks. Unlike prior passage-based hierarchical retrieval, we fine-tune a bi-encoder specifically for page-level relevance on financial filings, exploiting the semantic coherence of pages. Overall, our results demonstrate a significant improvement in page recall and chunk retrieval.

The paradigm of decision-making has been revolutionised by reinforcement learning and deep learning. Although this has led to significant progress in domains such as robotics, healthcare, and finance, the use of RL in practice is challenging, particularly when learning decision policies in high-stakes applications that may require guarantees. Traditional RL algorithms rely on a large number of online interactions with the environment, which is problematic in scenarios where online interactions are costly, dangerous, or infeasible. However, learning from offline datasets is hindered by the presence of hidden confounders. Such confounders can cause spurious correlations in the dataset and can mislead the agent into taking suboptimal or adversarial actions. Firstly, we address the problem of learning from offline datasets in the presence of hidden confounders. We work with instrumental variables (IVs) to identify the causal effect, which is an instance of a conditional moment restrictions (CMR) problem. Inspired by double/debiased machine learning, we derive a sample-efficient algorithm for solving CMR problems with convergence and optimality guarantees, which outperforms state-of-the-art algorithms. Secondly, we relax the conditions on the hidden confounders in the setting of (offline) imitation learning, and adapt our CMR estimator to derive an algorithm that can learn effective imitator policies with convergence rate guarantees. Finally, we consider the problem of learning high-level objectives expressed in linear temporal logic (LTL) and develop a provably optimal learning algorithm that improves sample efficiency over existing methods. Through evaluation on reinforcement learning benchmarks and synthetic and semi-synthetic datasets, we demonstrate the usefulness of the methods developed in this thesis in real-world decision making.

In active sequential testing, also termed pure exploration, a learner is tasked with the goal to adaptively acquire information so as to identify an unknown ground-truth hypothesis with as few queries as possible. This problem, originally studied by Chernoff in 1959, has several applications: classical formulations include Best-Arm Identification (BAI) in bandits, where actions index hypotheses, and generalized search problems, where strategically chosen queries reveal partial information about a hidden label. In many modern settings, however, the hypothesis space is continuous and naturally coincides with the query/action space: for example, identifying an optimal action in a continuous-armed bandit, localizing an $ε$-ball contained in a target region, or estimating the minimizer of an unknown function from a sequence of observations. In this work, we study pure exploration in such continuous spaces and introduce Continuous In-Context Pure Exploration for this regime. We introduce C-ICPE-TS, an algorithm that meta-trains deep neural policies to map observation histories to (i) the next continuous query action and (ii) a predicted hypothesis, thereby learning transferable sequential testing strategies directly from data. At inference time, C-ICPE-TS actively gathers evidence on previously unseen tasks and infers the true hypothesis without parameter updates or explicit hand-crafted information models. We validate C-ICPE-TS across a range of benchmarks, spanning continuous best-arm identification, region localization, and function minimizer identification.

This work formulates and solves optimization problems to generate input points that yield high errors between a neural network's predicted AC power flow solution and solutions to the AC power flow equations. We demonstrate this capability on an instance of the CANOS-PF graph neural network model, as implemented by the PF$Δ$ benchmark library, operating on a 14-bus test grid. Generated adversarial points yield errors as large as 3.4 per-unit in reactive power and 0.08 per-unit in voltage magnitude. When minimizing the perturbation from a training point necessary to satisfy adversarial constraints, we find that the constraints can be met with as little as an 0.04 per-unit perturbation in voltage magnitude on a single bus. This work motivates the development of rigorous verification and robust training methods for neural network surrogate models of AC power flow.

The increasing deployment of Federated Learning (FL) in Intrusion Detection Systems (IDS) introduces new challenges related to data privacy, centralized coordination, and susceptibility to poisoning attacks. While significant research has focused on protecting traditional FL-IDS with centralized aggregation servers, there remains a notable gap in addressing the unique challenges of decentralized FL-IDS (DFL-IDS). This study aims to address the limitations of traditional centralized FL-IDS by proposing a novel defense framework tailored for the decentralized FL-IDS architecture, with a focus on privacy preservation and robustness against poisoning attacks. We propose PenTiDef, a privacy-preserving and robust defense framework for DFL-IDS, which incorporates Distributed Differential Privacy (DDP) to protect data confidentiality and utilizes latent space representations (LSR) derived from neural networks to detect malicious updates in the decentralized model aggregation context. To eliminate single points of failure and enhance trust without a centralized aggregation server, PenTiDef employs a blockchain-based decentralized coordination mechanism that manages model aggregation, tracks update history, and supports trust enforcement through smart contracts. Experimental results on CIC-IDS2018 and Edge-IIoTSet demonstrate that PenTiDef consistently outperforms existing defenses (e.g., FLARE, FedCC) across various attack scenarios and data distributions. These findings highlight the potential of PenTiDef as a scalable and secure framework for deploying DFL-based IDS in adversarial environments. By leveraging privacy protection, malicious behavior detection in hidden data, and working without a central server, it provides a useful security solution against real-world attacks from untrust participants.

School congestion, where student enrollment exceeds school capacity, is a major challenge in low- and middle-income countries. It highly impacts learning outcomes and deepens inequities in education. While subsidy programs that transfer students from public to private schools offer a mechanism to alleviate congestion without capital-intensive construction, they often underperform due to fragmented data systems that hinder effective implementation. The Philippine Educational Service Contracting program, one of the world's largest educational subsidy programs, exemplifies these challenges, falling short of its goal to decongest public schools. This prevents the science-based and data-driven analyses needed to understand what shapes student enrollment flows, particularly how families respond to economic incentives and spatial constraints. We introduce a computational framework for modeling student flow patterns and simulating policy scenarios. By synthesizing heterogeneous government data across nearly 3,000 institutions, we employ a stochastic gravity model estimated via negative binomial regression to derive behavioral elasticities for distance, net tuition cost, and socioeconomic determinants. These elasticities inform a doubly constrained spatial allocation mechanism that simulates student redistribution under varying subsidy amounts while respecting both origin candidate pools and destination slot capacities. We find that geographic proximity constrains school choice four times more strongly than tuition cost and that slot capacity, not subsidy amounts, is the binding constraint. Our work demonstrates that subsidy programs alone cannot resolve systemic overcrowding, and computational modeling can empower education policymakers to make equitable, data-driven decisions by revealing the structural constraints that shape effective resource allocation, even when resources are limited.

Clinical risk prediction models often fail to be generalized across cohorts because underlying data distributions differ by clinical site, region, demographics, and measurement protocols. This limitation is particularly pronounced in hip fracture risk prediction, where the performance of models trained on one cohort (the source cohort) can degrade substantially when deployed in other cohorts (target cohorts). We used a shared set of clinical and DXA-derived features across three large cohorts - the Study of Osteoporotic Fractures (SOF), the Osteoporotic Fractures in Men Study (MrOS), and the UK Biobank (UKB), to systematically evaluate the performance of three domain adaptation methods - Maximum Mean Discrepancy (MMD), Correlation Alignment (CORAL), and Domain - Adversarial Neural Networks (DANN) and their combinations. For a source cohort with males only and a source cohort with females only, domain-adaptation methods consistently showed improved performance than the no-adaptation baseline (source-only training), and the use of combinations of multiple domain adaptation methods delivered the largest and most stable gains. The method that combines MMD, CORAL, and DANN achieved the highest discrimination with the area under curve (AUC) of 0.88 for a source cohort with males only and 0.95 for a source cohort with females only), demonstrating that integrating multiple domain adaptation methods could produce feature representations that are less sensitive to dataset differences. Unlike existing methods that rely heavily on supervised tuning or assume known outcomes of samples in target cohorts, our outcome-free approaches enable the model selection under realistic deployment conditions and improve generalization of models in hip fracture risk prediction.

Online model selection in Bayesian bandits raises a fundamental exploration challenge: When an environment instance is sampled from a prior distribution, how can we design an adaptive strategy that explores multiple bandit learners and competes with the best one in hindsight? We address this problem by introducing a new Bayesian algorithm for online model selection in stochastic bandits. We prove an oracle-style guarantee of $O\left( d^* M \sqrt{T} + \sqrt{(MT)} \right)$ on the Bayesian regret, where $M$ is the number of base learners, $d^*$ is the regret coefficient of the optimal base learner, and $T$ is the time horizon. We also validate our method empirically across a range of stochastic bandit settings, demonstrating performance that is competitive with the best base learner. Additionally, we study the effect of sharing data among base learners and its role in mitigating prior mis-specification.

Artificial Intelligence (AI) increasingly automates various parts of the software development tasks. Although AI has enhanced the productivity of development tasks, it remains unstudied whether AI essentially outperforms humans in type-related programming tasks, such as employing type constructs properly for type safety, during its tasks. Moreover, there is no systematic study that evaluates whether AI agents overuse or misuse the type constructs under the complicated type systems to the same extent as humans. In this study, we present the first empirical analysis to answer these questions in the domain of TypeScript projects. Our findings show that, in contrast to humans, AI agents are 9x more prone to use the 'any' keyword. In addition, we observed that AI agents use advanced type constructs, including those that ignore type checks, more often compared to humans. Surprisingly, even with all these issues, Agentic pull requests (PRs) have 1.8x higher acceptance rates compared to humans for TypeScript. We encourage software developers to carefully confirm the type safety of their codebases whenever they coordinate with AI agents in the development process.

One-dimensional function approximation is a fundamental problem in scientific computing and engineering applications. While neural networks possess powerful universal approximation capabilities, their optimization process is often hindered by flat loss landscapes induced by parameter-space symmetries, leading to slow convergence and poor generalization, particularly for high-frequency components. Inspired by the principle of \emph{symmetry breaking} in physics, this paper proposes a hardware-friendly approach for function approximation through \emph{input-space expansion}. The core idea involves augmenting the original one-dimensional input (e.g., $x$) with constant values (e.g., $π$) to form a higher-dimensional vector (e.g., $[π, π, x, π, π]$), effectively breaking parameter symmetries without increasing the network's parameter count. We evaluate the method on ten representative one-dimensional functions, including smooth, discontinuous, high-frequency, and non-differentiable functions. Experimental results demonstrate that input-space expansion significantly accelerates training convergence (reducing LBFGS iterations by 12\% on average) and enhances approximation accuracy (reducing final MSE by 66.3\% for the optimal 5D expansion). Ablation studies further reveal the effects of different expansion dimensions and constant selections, with $π$ consistently outperforming other constants. Our work proposes a low-cost, efficient, and hardware-friendly technique for algorithm design.

Vision-Language-Action (VLA) models enable instruction-following robotic manipulation, but they are typically pretrained on 2D data and lack 3D spatial understanding. An effective approach is representation alignment, where a strong vision foundation model is used to guide a 2D VLA model. However, existing methods usually apply supervision at only a single layer, failing to fully exploit the rich information distributed across depth; meanwhile, naïve multi-layer alignment can cause gradient interference. We introduce ROCKET, a residual-oriented multi-layer representation alignment framework that formulates multi-layer alignment as aligning one residual stream to another. Concretely, ROCKET employs a shared projector to align multiple layers of the VLA backbone with multiple layers of a powerful 3D vision foundation model via a layer-invariant mapping, which reduces gradient conflicts. We provide both theoretical justification and empirical analyses showing that a shared projector is sufficient and outperforms prior designs, and further propose a Matryoshka-style sparse activation scheme for the shared projector to balance multiple alignment losses. Our experiments show that, combined with a training-free layer selection strategy, ROCKET requires only about 4% of the compute budget while achieving 98.5% state-of-the-art success rate on LIBERO. We further demonstrate the superior performance of ROCKET across LIBERO-Plus and RoboTwin, as well as multiple VLA models. The code and model weights can be found at https://github.com/CASE-Lab-UMD/ROCKET-VLA.

Background: Clinical named entity recognition tools commonly map free text to Unified Medical Language System (UMLS) Concept Unique Identifiers (CUIs). For many downstream tasks, however, the clinically meaningful unit is not a single CUI but a concept set comprising related synonyms, subtypes, and supertypes. Constructing such concept sets is labour-intensive, inconsistently performed, and poorly supported by existing tools, particularly for NLP pipelines that operate directly on UMLS CUIs. Methods We present CUICurate, a Graph-based retrieval-augmented generation (GraphRAG) framework for automated UMLS concept set curation. A UMLS knowledge graph (KG) was constructed and embedded for semantic retrieval. For each target concept, candidate CUIs were retrieved from the KG, followed by large language model (LLM) filtering and classification steps comparing two LLMs (GPT-5 and GPT-5-mini). The framework was evaluated on five lexically heterogeneous clinical concepts against a manually curated benchmark and gold-standard concept sets. Results Across all concepts, CUICurate produced substantially larger and more complete concept sets than the manual benchmarks whilst matching human precision. Comparisons between the two LLMs found that GPT-5-mini achieved higher recall during filtering, while GPT-5 produced classifications that more closely aligned with clinician judgements. Outputs were stable across repeated runs and computationally inexpensive. Conclusions CUICurate offers a scalable and reproducible approach to support UMLS concept set curation that substantially reduces manual effort. By integrating graph-based retrieval with LLM reasoning, the framework produces focused candidate concept sets that can be adapted to clinical NLP pipelines for different phenotyping and analytic requirements.

The trade-off between clean accuracy and adversarial robustness is a pervasive phenomenon in deep learning, yet its geometric origin remains elusive. In this work, we utilize Symmetry-Breaking Dimensional Expansion (SBDE) as a controlled probe to investigate the mechanism underlying this trade-off. SBDE expands input images by inserting constant-valued pixels, which breaks translational symmetry and consistently improves clean accuracy (e.g., from $90.47\%$ to $95.63\%$ on CIFAR-10 with ResNet-18) by reducing parameter degeneracy. However, this accuracy gain comes at the cost of reduced robustness against iterative white-box attacks. By employing a test-time \emph{mask projection} that resets the inserted auxiliary pixels to their training values, we demonstrate that the vulnerability stems almost entirely from the inserted dimensions. The projection effectively neutralizes the attacks and restores robustness, revealing that the model achieves high accuracy by creating \emph{sharp boundaries} (steep loss gradients) specifically along the auxiliary axes. Our findings provide a concrete geometric explanation for the accuracy-robustness paradox: the optimization landscape deepens the basin of attraction to improve accuracy but inevitably erects steep walls along the auxiliary degrees of freedom, creating a fragile sensitivity to off-manifold perturbations.

Gradient-based hyperparameter optimization (HPO) have emerged recently, leveraging bilevel programming techniques to optimize hyperparameter by estimating hypergradient w.r.t. validation loss. Nevertheless, previous theoretical works mainly focus on reducing the gap between the estimation and ground-truth (i.e., the bias), while ignoring the error due to data distribution (i.e., the variance), which degrades performance. To address this issue, we conduct a bias-variance decomposition for hypergradient estimation error and provide a supplemental detailed analysis of the variance term ignored by previous works. We also present a comprehensive analysis of the error bounds for hypergradient estimation. This facilitates an easy explanation of some phenomena commonly observed in practice, like overfitting to the validation set. Inspired by the derived theories, we propose an ensemble hypergradient strategy to reduce the variance in HPO algorithms effectively. Experimental results on tasks including regularization hyperparameter learning, data hyper-cleaning, and few-shot learning demonstrate that our variance reduction strategy improves hypergradient estimation. To explain the improved performance, we establish a connection between excess error and hypergradient estimation, offering some understanding of empirical observations.

Graph data is becoming increasingly prevalent due to the growing demand for relational insights in AI across various domains. Organizations regularly use graph data to solve complex problems involving relationships and connections. Causal learning is especially important in this context, since it helps to understand cause-effect relationships rather than mere associations. Since many real-world systems are inherently causal, graphs can efficiently model these systems. However, traditional graph machine learning methods including graph neural networks (GNNs), rely on correlations and are sensitive to spurious patterns and distribution changes. On the other hand, causal models enable robust predictions by isolating true causal factors, thus making them more stable under such shifts. Causal learning also helps in identifying and adjusting for confounders, ensuring that predictions reflect true causal relationships and remain accurate even under interventions. To address these challenges and build models that are robust and causally informed, we propose CCAGNN, a Confounder-Aware causal GNN framework that incorporates causal reasoning into graph learning, supporting counterfactual reasoning and providing reliable predictions in real-world settings. Comprehensive experiments on six publicly available datasets from diverse domains show that CCAGNN consistently outperforms leading state-of-the-art models.

We study an algorithm-independent, worst-case lower bound for the Gaussian process (GP) bandit problem in the frequentist setting, where the reward function is fixed and has a bounded norm in the known reproducing kernel Hilbert space (RKHS). Specifically, we focus on the squared exponential (SE) kernel, one of the most widely used kernel functions in GP bandits. One of the remaining open questions for this problem is the gap in the \emph{dimension-dependent} logarithmic factors between upper and lower bounds. This paper partially resolves this open question under a hyperspherical input domain. We show that any algorithm suffers $Ω(\sqrt{T (\ln T)^{d} (\ln \ln T)^{-d}})$ cumulative regret, where $T$ and $d$ represent the total number of steps and the dimension of the hyperspherical domain, respectively. Regarding the simple regret, we show that any algorithm requires $Ω(ε^{-2}(\ln \frac{1}ε)^d (\ln \ln \frac{1}ε)^{-d})$ time steps to find an $ε$-optimal point. We also provide the improved $O((\ln T)^{d+1}(\ln \ln T)^{-d})$ upper bound on the maximum information gain for the SE kernel. Our results guarantee the optimality of the existing best algorithm up to \emph{dimension-independent} logarithmic factors under a hyperspherical input domain.

Instruction optimization provides a lightweight, model-agnostic approach to enhancing the reasoning performance of large language models (LLMs). This paper presents the first systematic comparison of instruction optimization, based on the DSPy optimization framework, for tabular fact verification. We evaluate four out-of-the-box prompting techniques that cover both text-only prompting and code use: direct prediction, Chain-of-Thought (CoT), ReAct with SQL tools, and CodeAct with Python execution. We study three optimizers from the DSPy framework -- COPRO, MiPROv2, and SIMBA -- across four benchmarks and three model families. We find that instruction optimization consistently improves verification accuracy, with MiPROv2 yielding the most stable gains for CoT, and SIMBA providing the largest benefits for ReAct agents, particularly at larger model scales. Behavioral analyses reveal that SIMBA encourages more direct reasoning paths by applying heuristics, thereby improving numerical comparison abilities in CoT reasoning and helping avoid unnecessary tool calls in ReAct agents. Across different prompting techniques, CoT remains effective for tabular fact checking, especially with smaller models. Although ReAct agents built with larger models can achieve competitive performance, they require careful instruction optimization.

Graph data often contain noisy and spurious correlations that mask the true causal relationships, which are essential for enabling graph models to make predictions based on the underlying causal structure of the data. Dependence on spurious connections makes it challenging for traditional Graph Neural Networks (GNNs) to generalize effectively across different graphs. Furthermore, traditional aggregation methods tend to amplify these spurious patterns, limiting model robustness under distribution shifts. To address these issues, we propose Causal Neighbourhood Learning with Graph Neural Networks (CNL-GNN), a novel framework that performs causal interventions on graph structure. CNL-GNN effectively identifies and preserves causally relevant connections and reduces spurious influences through the generation of counterfactual neighbourhoods and adaptive edge perturbation guided by learnable importance masking and an attention-based mechanism. In addition, by combining structural-level interventions with the disentanglement of causal features from confounding factors, the model learns invariant node representations that are robust and generalize well across different graph structures. Our approach improves causal graph learning beyond traditional feature-based methods, resulting in a robust classification model. Extensive experiments on four publicly available datasets, including multiple domain variants of one dataset, demonstrate that CNL-GNN outperforms state-of-the-art GNN models.

In environments with sparse or delayed rewards, reinforcement learning (RL) incurs high sample complexity due to the large number of interactions needed for learning. This limitation has motivated the use of large language models (LLMs) for subgoal discovery and trajectory guidance. While LLMs can support exploration, frequent reliance on LLM calls raises concerns about scalability and reliability. We address these challenges by constructing a memory graph that encodes subgoals and trajectories from both LLM guidance and the agent's own successful rollouts. From this graph, we derive a utility function that evaluates how closely the agent's trajectories align with prior successful strategies. This utility shapes the advantage function, providing the critic with additional guidance without altering the reward. Our method relies primarily on offline input and only occasional online queries, avoiding dependence on continuous LLM supervision. Preliminary experiments in benchmark environments show improved sample efficiency and faster early learning compared to baseline RL methods, with final returns comparable to methods that require frequent LLM interaction.

Reinforcement learning (RL) agents often suffer from high sample complexity in sparse or delayed reward settings due to limited prior structure. Large language models (LLMs) can provide subgoal decompositions, plausible trajectories, and abstract priors that facilitate early learning. However, heavy reliance on LLM supervision introduces scalability constraints and dependence on potentially unreliable signals. We propose MIRA (Memory-Integrated Reinforcement Learning Agent), which incorporates a structured, evolving memory graph to guide early training. The graph stores decision-relevant information, including trajectory segments and subgoal structures, and is constructed from both the agent's high-return experiences and LLM outputs. This design amortizes LLM queries into a persistent memory rather than requiring continuous real-time supervision. From this memory graph, we derive a utility signal that softly adjusts advantage estimation to influence policy updates without modifying the underlying reward function. As training progresses, the agent's policy gradually surpasses the initial LLM-derived priors, and the utility term decays, preserving standard convergence guarantees. We provide theoretical analysis showing that utility-based shaping improves early-stage learning in sparse-reward environments. Empirically, MIRA outperforms RL baselines and achieves returns comparable to approaches that rely on frequent LLM supervision, while requiring substantially fewer online LLM queries. Project webpage: https://narjesno.github.io/MIRA/

Vision Transformers rely on positional embeddings and class tokens that encode fixed spatial priors. While effective for natural images, these priors may hinder generalization when spatial layout is weakly informative or inconsistent, a frequent condition in medical imaging and edge-deployed clinical systems. We introduce ZACH-ViT (Zero-token Adaptive Compact Hierarchical Vision Transformer), a compact Vision Transformer that removes both positional embeddings and the [CLS] token, achieving permutation invariance through global average pooling over patch representations. The term "Zero-token" specifically refers to removing the dedicated [CLS] aggregation token and positional embeddings; patch tokens remain unchanged and are processed normally. Adaptive residual projections preserve training stability in compact configurations while maintaining a strict parameter budget. Evaluation is performed across seven MedMNIST datasets spanning binary and multi-class tasks under a strict few-shot protocol (50 samples per class, fixed hyperparameters, five random seeds). The empirical analysis demonstrates regime-dependent behavior: ZACH-ViT (0.25M parameters, trained from scratch) achieves its strongest advantage on BloodMNIST and remains competitive with TransMIL on PathMNIST, while its relative advantage decreases on datasets with strong anatomical priors (OCTMNIST, OrganAMNIST), consistent with the architectural hypothesis. These findings support the view that aligning architectural inductive bias with data structure can be more important than pursuing universal benchmark dominance. Despite its minimal size and lack of pretraining, ZACH-ViT achieves competitive performance while maintaining sub-second inference times, supporting deployment in resource-constrained clinical environments. Code and models are available at https://github.com/Bluesman79/ZACH-ViT.

Manipulating deformable and fragile objects remains a fundamental challenge in robotics due to complex contact dynamics and strict requirements on object integrity. Existing approaches typically optimize either end-effector design or control strategies in isolation, limiting achievable performance. In this work, we present the first co-design framework that jointly optimizes end-effector morphology and manipulation control for deformable and fragile object manipulation. We introduce (1) a latent diffeomorphic shape parameterization enabling expressive yet tractable end-effector geometry optimization, (2) a stress-aware bi-level co-design pipeline coupling morphology and control optimization, and (3) a privileged-to-pointcloud policy distillation scheme for zero-shot real-world deployment. We evaluate our approach on challenging food manipulation tasks, including grasping and pushing jelly and scooping fillets. Simulation and real-world experiments demonstrate the effectiveness of the proposed method.

We study a sequential prediction problem in which an adversary is allowed to inject arbitrarily many adversarial instances in a stream of i.i.d.\ instances, but at each round, the learner may also \emph{abstain} from making a prediction without incurring any penalty if the instance was indeed corrupted. This semi-adversarial setting naturally sits between the classical stochastic case with i.i.d.\ instances for which function classes with finite VC dimension are learnable; and the adversarial case with arbitrary instances, known to be significantly more restrictive. For this problem, Goel et al. (2023) showed that, if the learner knows the distribution $μ$ of clean samples in advance, learning can be achieved for all VC classes without restrictions on adversary corruptions. This is, however, a strong assumption in both theory and practice: a natural question is whether similar learning guarantees can be achieved without prior distributional knowledge, as is standard in classical learning frameworks (e.g., PAC learning or asymptotic consistency) and other non-i.i.d.\ models (e.g., smoothed online learning). We therefore focus on the distribution-free setting where $μ$ is \emph{unknown} and propose an algorithm \textsc{AbstainBoost} based on a boosting procedure of weak learners, which guarantees sublinear error for general VC classes in \emph{distribution-free} abstention learning for oblivious adversaries. These algorithms also enjoy similar guarantees for adaptive adversaries, for structured function classes including linear classifiers. These results are complemented with corresponding lower bounds, which reveal an interesting polynomial trade-off between misclassification error and number of erroneous abstentions.

Polynomial functors model systems with interfaces: each polynomial specifies the outputs a system can produce and, for each output, the inputs it accepts. The bicategory $\mathbb{O}\mathbf{rg}$ of dynamic organizations \cite{spivak2021learners} gives a notion of state-driven interaction patterns that evolves over time, but each system's interface remains fixed throughout the interaction. Yet in many systems, the outputs sent and inputs received can reshape the interface itself: a cell differentiating in response to chemical signals gains or loses receptors; a sensor damaged by its input loses a channel; a neural network may grow its output resolution during training. Here we introduce *polynomial trees*, elements of the terminal $(u\triangleleft u)$-coalgebra where $u$ is the polynomial associated to a universe of sets, to model such systems: a polynomial tree is a coinductive tree whose nodes carry polynomials, and in which each round of interaction -- an output chosen and an input received -- determines a child tree, hence the next interface. We construct a monoidal closed category $\mathbf{PolyTr}$ of polynomial trees, with coinductively-defined morphisms, tensor product, and internal hom. We then build a bicategory $\mathbb{O}\mathbf{rgTr}$ generalizing $\mathbb{O}\mathbf{rg}$, whose hom-categories parametrize morphisms by state sets with coinductive action-and-update data. We provide a locally fully faithful functor $\mathbb{O}\mathbf{rg}\to\mathbb{O}\mathbf{rgTr}$ via constant trees, those for which the interfaces do not change through time. We illustrate the generalization by suggesting a notion of progressive generative adversarial networks, where gradient feedback determines when the image-generation interface grows to a higher resolution.

Long-horizon agents often compress interaction histories into write-time summaries. This creates a fundamental write-before-query barrier: compression decisions are made before the system knows what a future query will hinge on. As a result, summaries can cause unverifiable omissions -- decisive constraints (e.g., allergies) may be dropped, leaving the agent unable to justify an answer with traceable evidence. Retaining raw logs restores an authoritative source of truth, but grounding on raw logs by default is expensive: many queries are answerable from summaries, yet raw grounding still requires processing far longer contexts, inflating token consumption and latency. We propose TierMem, a provenance-linked framework that casts retrieval as an inference-time evidence allocation problem. TierMem uses a two-tier memory hierarchy to answer with the cheapest sufficient evidence: it queries a fast summary index by default, and a runtime sufficiency router Escalates to an immutable raw-log store only when summary evidence is insufficient. TierMem then writes back verified findings as new summary units linked to their raw sources. On LoCoMo, TierMem achieves 0.851 accuracy (vs.0.873 raw-only) while reducing input tokens by 54.1\% and latency by 60.7%.

Current biomedical question answering (QA) systems often assume that medical knowledge applies uniformly, yet real-world clinical reasoning is inherently conditional: nearly every decision depends on patient-specific factors such as comorbidities and contraindications. Existing benchmarks do not evaluate such conditional reasoning, and retrieval-augmented or graph-based methods lack explicit mechanisms to ensure that retrieved knowledge is applicable to given context. To address this gap, we propose CondMedQA, the first benchmark for conditional biomedical QA, consisting of multi-hop questions whose answers vary with patient conditions. Furthermore, we propose Condition-Gated Reasoning (CGR), a novel framework that constructs condition-aware knowledge graphs and selectively activates or prunes reasoning paths based on query conditions. Our findings show that CGR more reliably selects condition-appropriate answers while matching or exceeding state-of-the-art performance on biomedical QA benchmarks, highlighting the importance of explicitly modeling conditionality for robust medical reasoning.

Traditional AI alignment primarily focuses on individual model outputs; however, autonomous agents in long-horizon workflows require sustained reliability across entire interaction trajectories. We introduce APEMO (Affect-aware Peak-End Modulation for Orchestration), a runtime scheduling layer that optimizes computational allocation under fixed budgets by operationalizing temporal-affective signals. Instead of modifying model weights, APEMO detects trajectory instability through behavioral proxies and targets repairs at critical segments, such as peak moments and endings. Evaluation across multi-agent simulations and LLM-based planner--executor flows demonstrates that APEMO consistently enhances trajectory-level quality and reuse probability over structural orchestrators. Our results reframe alignment as a temporal control problem, offering a resilient engineering pathway for the development of long-horizon agentic systems.

Traditional neural topic models are typically optimized by reconstructing the document's Bag-of-Words (BoW) representations, overlooking contextual information and struggling with data sparsity. In this work, we propose a novel approach to construct semantically-grounded soft label targets using Language Models (LMs) by projecting the next token probabilities, conditioned on a specialized prompt, onto a pre-defined vocabulary to obtain contextually enriched supervision signals. By training the topic models to reconstruct the soft labels using the LM hidden states, our method produces higher-quality topics that are more closely aligned with the underlying thematic structure of the corpus. Experiments on three datasets show that our method achieves substantial improvements in topic coherence, purity over existing baselines. Additionally, we also introduce a retrieval-based metric, which shows that our approach significantly outperforms existing methods in identifying semantically similar documents, highlighting its effectiveness for retrieval-oriented applications.

Interactive systems such as chatbots and games are increasingly used to persuade and educate on sustainability-related topics, yet it remains unclear how different delivery formats shape learning and persuasive outcomes when content is held constant. Grounding on identical arguments and factual content across conditions, we present a controlled user study comparing three modes of information delivery: static essays, conversational chatbots, and narrative text-based games. Across subjective measures, the chatbot condition consistently outperformed the other modes and increased perceived importance of the topic. However, perceived learning did not reliably align with objective outcomes: participants in the text-based game condition reported learning less than those reading essays, yet achieved higher scores on a delayed (24-hour) knowledge quiz. Additional exploratory analyses further suggest that common engagement proxies, such as verbosity and interaction length, are more closely related to subjective experience than to actual learning. These findings highlight a dissociation between how persuasive experiences feel and what participants retain, and point to important design trade-offs between interactivity, realism, and learning in persuasive systems and serious games.

Large language models (LLMs) are increasingly used to automate scientific workflows, yet their integration with heterogeneous computational tools remains ad hoc and fragile. Current agentic approaches often rely on unstructured text to manage context and coordinate execution, generating often overwhelming volumes of information that may obscure decision provenance and hinder auditability. In this work, we present El Agente Gráfico, a single-agent framework that embeds LLM-driven decision-making within a type-safe execution environment and dynamic knowledge graphs for external persistence. Central to our approach is a structured abstraction of scientific concepts and an object-graph mapper that represents computational state as typed Python objects, stored either in memory or persisted in an external knowledge graph. This design enables context management through typed symbolic identifiers rather than raw text, thereby ensuring consistency, supporting provenance tracking, and enabling efficient tool orchestration. We evaluate the system by developing an automated benchmarking framework across a suite of university-level quantum chemistry tasks previously evaluated on a multi-agent system, demonstrating that a single agent, when coupled to a reliable execution engine, can robustly perform complex, multi-step, and parallel computations. We further extend this paradigm to two other large classes of applications: conformer ensemble generation and metal-organic framework design, where knowledge graphs serve as both memory and reasoning substrates. Together, these results illustrate how abstraction and type safety can provide a scalable foundation for agentic scientific automation beyond prompt-centric designs.

Attention-based regression models are often trained by jointly optimizing Mean Squared Error (MSE) loss and Pearson correlation coefficient (PCC) loss, emphasizing the magnitude of errors and the order or shape of targets, respectively. A common but poorly understood phenomenon during training is the PCC plateau: PCC stops improving early in training, even as MSE continues to decrease. We provide the first rigorous theoretical analysis of this behavior, revealing fundamental limitations in both optimization dynamics and model capacity. First, in regard to the flattened PCC curve, we uncover a critical conflict where lowering MSE (magnitude matching) can paradoxically suppress the PCC gradient (shape matching). This issue is exacerbated by the softmax attention mechanism, particularly when the data to be aggregated is highly homogeneous. Second, we identify a limitation in the model capacity: we derived a PCC improvement limit for any convex aggregator (including the softmax attention), showing that the convex hull of the inputs strictly bounds the achievable PCC gain. We demonstrate that data homogeneity intensifies both limitations. Motivated by these insights, we propose the Extrapolative Correlation Attention (ECA), which incorporates novel, theoretically-motivated mechanisms to improve the PCC optimization and extrapolate beyond the convex hull. Across diverse benchmarks, including challenging homogeneous data setting, ECA consistently breaks the PCC plateau, achieving significant improvements in correlation without compromising MSE performance.

Data collection is a critical component of modern statistical and machine learning pipelines, particularly when data must be gathered from multiple heterogeneous sources to study a target population of interest. In many use cases, such as medical studies or political polling, different sources incur different sampling costs. Observations often have associated group identities (for example, health markers, demographics, or political affiliations) and the relative composition of these groups may differ substantially, both among the source populations and between sources and target population. In this work, we study multi-source data collection under a fixed budget, focusing on the estimation of population means and group-conditional means. We show that naive data collection strategies (e.g. attempting to "match" the target distribution) or relying on standard estimators (e.g. sample mean) can be highly suboptimal. Instead, we develop a sampling plan which maximizes the effective sample size: the total sample size divided by $D_{χ^2}(q\mid\mid\overline{p}) + 1$, where $q$ is the target distribution, $\overline{p}$ is the aggregated source distribution, and $D_{χ^2}$ is the $χ^2$-divergence. We pair this sampling plan with a classical post-stratification estimator and upper bound its risk. We provide matching lower bounds, establishing that our approach achieves the budgeted minimax optimal risk. Our techniques also extend to prediction problems when minimizing the excess risk, providing a principled approach to multi-source learning with costly and heterogeneous data sources.

State space models (SSMs) have recently emerged for modeling long-range dependency in sequence data, with much simplified computational costs than modern alternatives, such as transformers. Advancing SMMs to graph structured data, especially for large graphs, is a significant challenge because SSMs are sequence models and the shear graph volumes make it very expensive to convert graphs as sequences for effective learning. In this paper, we propose COMBA to tackle large graph learning using state space models, with two key innovations: graph context gating and cross batch aggregation. Graph context refers to different hops of neighborhood for each node, and graph context gating allows COMBA to use such context to learn best control of neighbor aggregation. For each graph context, COMBA samples nodes as batches, and train a graph neural network (GNN), with information being aggregated cross batches, allowing COMBA to scale to large graphs. Our theoretical study asserts that cross-batch aggregation guarantees lower error than training GNN without aggregation. Experiments on benchmark networks demonstrate significant performance gains compared to baseline approaches. Code and benchmark datasets will be released for public access.

Maintaining and refactoring React web applications is challenging, as React code often becomes complex due to its core API called Hooks. For example, Hooks often lead developers to create complex dependencies among components, making code behavior unpredictable and reducing maintainability, i.e., anti-patterns. To address this challenge, we present HookLens, an interactive visual analytics system that helps developers understand howHooks define dependencies and data flows between components. Informed by an iterative design process with experienced React developers, HookLens supports users to efficiently understand the structure and dependencies between components and to identify anti-patterns. A quantitative user study with 12 React developers demonstrates that HookLens significantly improves participants' accuracy in detecting anti-patterns compared to conventional code editors. Moreover, a comparative study with state-of-the-art LLM-based coding assistants confirms that these improvements even surpass the capabilities of such coding assistants on the same task.

Artificial intelligence (AI) has increasingly transformed medical prognostics by enabling rapid and accurate analysis across imaging and pathology. However, the investigation of machine learning predictions applied to prospectively collected, standardized data from observational clinical intervention trials remains underexplored, despite its potential to reduce costs and improve patient outcomes. Chronic rhinosinusitis (CRS), a persistent inflammatory disease of the paranasal sinuses lasting more than three months, imposes a substantial burden on quality of life (QoL) and societal cost. Although many patients respond to medical therapy, others with refractory symptoms often pursue surgical intervention. Surgical decision-making in CRS is complex, as it must weigh known procedural risks against uncertain individualized outcomes. In this study, we evaluated supervised machine learning models for predicting surgical benefit in CRS, using the Sino-Nasal Outcome Test-22 (SNOT-22) as the primary patient-reported outcome. Our prospectively collected cohort from an observational intervention trial comprised patients who all underwent surgery; we investigated whether models trained only on preoperative data could identify patients who might not have been recommended surgery prior to the procedure. Across multiple algorithms, including an ensemble approach, our best model achieved approximately 85% classification accuracy, providing accurate and interpretable predictions of surgical candidacy. Moreover, on a held-out set of 30 cases spanning mixed difficulty, our model achieved 80% accuracy, exceeding the average prediction accuracy of expert clinicians (75.6%), demonstrating its potential to augment clinical decision-making and support personalized CRS care.

Web accessibility remains an unresolved issue for a large part of the web content. There are many tools to detect errors automatically, but fixing those issues is still mostly a manual, slow, and costly process in which it is easy for developers to overlook specific details. The situation becomes even more complex with modern Single-Page Applications (SPAs), whose dynamic nature makes traditional static analysis approaches inadequate. This work proposes a system that aims to address this challenge by using Large Language Models (LLMs) to automate accessibility fixes. The proposal presents a modular workflow applicable to both static websites and complex Angular projects. The framework actively implements corrections within the DOM of static web pages or the source code of SPAs. The system was tested on 12 static websites and 6 open-source Angular projects, fixing 80% of the accessibility issues on public websites and 86% of the issues on Angular applications. Our proposal also generates meaningful visual descriptions for images while preserving the application's design and stability. This work contributes to ensuring that accessibility stops being a technical debt deferred to the future and becomes a natural part of everyday development workflows.

Steering vectors are a lightweight method for controlling language model behavior by adding a learned bias to the activations at inference time. Although effective on average, steering effect sizes vary across samples and are unreliable for many target behaviors. In my thesis, I investigate why steering reliability differs across behaviors and how it is impacted by steering vector training data. First, I find that higher cosine similarity between training activation differences predicts more reliable steering. Second, I observe that behavior datasets where positive and negative activations are better separated along the steering direction are more reliably steerable. Finally, steering vectors trained on different prompt variations are directionally distinct, yet perform similarly well and exhibit correlated efficacy across datasets. My findings suggest that steering vectors are unreliable when the latent target behavior representation is not effectively approximated by the linear steering direction. Taken together, these insights offer a practical diagnostic for steering unreliability and motivate the development of more robust steering methods that explicitly account for non-linear latent behavior representations.

Stochastic gradient descent (SGD) is a cornerstone algorithm for high-dimensional optimization, renowned for its empirical successes. Recent theoretical advances have provided a deep understanding of how SGD enables feature learning in high-dimensional nonlinear models, most notably the \textit{single-index model} with i.i.d. data. In this work, we study the sequential learning problem for single-index models, also known as generalized linear bandits or ridge bandits, where SGD is a simple and natural solution, yet its learning dynamics remain largely unexplored. We show that, similar to the optimal interactive learner, SGD undergoes a distinct ``burn-in'' phase before entering the ``learning'' phase in this setting. Moreover, with an appropriately chosen learning rate schedule, a single SGD procedure simultaneously achieves near-optimal (or best-known) sample complexity and regret guarantees across both phases, for a broad class of link functions. Our results demonstrate that SGD remains highly competitive for learning single-index models under adaptive data.

We present MultiVer, a zero-shot multi-agent system for vulnerability detection that achieves state-of-the-art recall without fine-tuning. A four-agent ensemble (security, correctness, performance, style) with union voting achieves 82.7% recall on PyVul, exceeding fine-tuned GPT-3.5 (81.3%) by 1.4 percentage points -- the first zeroshot system to surpass fine-tuned performance on this benchmark. On SecurityEval, the same architecture achieves 91.7% detection rate, matching specialized systems. The recall improvement comes at a precision cost: 48.8% precision versus 63.9% for fine-tuned baselines, yielding 61.4% F1. Ablation experiments isolate component contributions: the multi-agent ensemble adds 17 percentage points recall over single-agent security analysis. These results demonstrate that for security applications where false negatives are costlier than false positives, zero-shot multi-agent ensembles can match and exceed fine-tuned models on the metric that matters most.

Vision-language models (VLMs) have made substantial progress across a wide range of visual question answering benchmarks, spanning visual reasoning, document understanding, and multimodal dialogue. These improvements are evident in a wide range of VLMs built on a variety of base models, alignment architectures, and training data. However, recent works show that these models trail behind in traditional image classification benchmarks, which test fine-grained visual knowledge. We test a large number of recent VLMs on fine-grained classification benchmarks and identify potential factors in the disconnect between fine-grained knowledge and other vision benchmarks. Through a series of ablation experiments, we find that using a better LLM improves all benchmark scores equally, while a better vision encoder disproportionately improves fine-grained classification performance. Furthermore, we find that the pretraining stage is also vital to fine-grained performance, particularly when the language model weights are unfrozen during pretraining. These insights pave the way for enhancing fine-grained visual understanding and vision-centric capabilities in VLMs.

Developing foundation models for time series classification is of high practical relevance, as such models can serve as universal feature extractors for diverse downstream tasks. Although early models such as Mantis have shown the promise of this approach, a substantial performance gap remained between frozen and fine-tuned encoders. In this work, we introduce methods that significantly strengthen zero-shot feature extraction for time series. First, we introduce Mantis+, a variant of Mantis pre-trained entirely on synthetic time series. Second, through controlled ablation studies, we refine the architecture and obtain MantisV2, an improved and more lightweight encoder. Third, we propose an enhanced test-time methodology that leverages intermediate-layer representations and refines output-token aggregation. In addition, we show that performance can be further improved via self-ensembling and cross-model embedding fusion. Extensive experiments on UCR, UEA, Human Activity Recognition (HAR) benchmarks, and EEG datasets show that MantisV2 and Mantis+ consistently outperform prior time series foundation models, achieving state-of-the-art zero-shot performance.

Understanding what features are encoded by learned directions in LLM activation space requires identifying inputs that strongly activate them. Feature visualization, which optimizes inputs to maximally activate a target direction, offers an alternative to costly dataset search approaches, but remains underexplored for LLMs due to the discrete nature of text. Furthermore, existing prompt optimization techniques are poorly suited to this domain, which is highly prone to local minima. To overcome these limitations, we introduce ADAPT, a hybrid method combining beam search initialization with adaptive gradient-guided mutation, designed around these failure modes. We evaluate on Sparse Autoencoder latents from Gemma 2 2B, proposing metrics grounded in dataset activation statistics to enable rigorous comparison, and show that ADAPT consistently outperforms prior methods across layers and latent types. Our results establish that feature visualization for LLMs is tractable, but requires design assumptions tailored to the domain.

Time-series forecasting is a critical task across many domains, from engineering to economics, where accurate predictions drive strategic decisions. However, applying advanced deep learning models in challenging, volatile domains like finance is difficult due to the inherent limitation and dynamic nature of financial time series data. This scarcity often results in sub-optimal model training and poor generalization. The fundamental challenge lies in determining how to reliably augment scarce financial time series data to enhance the predictive accuracy of deep learning forecasting models. Our main contribution is a demonstration of how Generative Adversarial Networks (GANs) can effectively serve as a data augmentation tool to overcome data scarcity in the financial domain. Specifically, we show that training a Long Short-Term Memory (LSTM) forecasting model on a dataset augmented with synthetic data generated by a transformer-based GAN (TTS-GAN) significantly improves the forecasting accuracy compared to using real data alone. We confirm these results across different financial time series (Bitcoin and S\&P500 price data) and various forecasting horizons. Furthermore, we propose a novel, time series specific quality metric that combines Dynamic Time Warping (DTW) and a modified Deep Dataset Dissimilarity Measure (DeD-iMs) to reliably monitor the training progress and evaluate the quality of the generated data. These findings provide compelling evidence for the benefits of GAN-based data augmentation in enhancing financial predictive capabilities.

JAX-Privacy is a library designed to simplify the deployment of robust and performant mechanisms for differentially private machine learning. Guided by design principles of usability, flexibility, and efficiency, JAX-Privacy serves both researchers requiring deep customization and practitioners who want a more out-of-the-box experience. The library provides verified, modular primitives for critical components for all aspects of the mechanism design including batch selection, gradient clipping, noise addition, accounting, and auditing, and brings together a large body of recent research on differentially private ML.

This paper investigates the enhancement of scientific literature chatbots through retrieval-augmented generation (RAG), with a focus on evaluating vector- and graph-based retrieval systems. The proposed chatbot leverages both structured (graph) and unstructured (vector) databases to access scientific articles and gray literature, enabling efficient triage of sources according to research objectives. To systematically assess performance, we examine two use-case scenarios: retrieval from a single uploaded document and retrieval from a large-scale corpus. Benchmark test sets were generated using a GPT model, with selected outputs annotated for evaluation. The comparative analysis emphasizes retrieval accuracy and response relevance, providing insight into the strengths and limitations of each approach. The findings demonstrate the potential of hybrid RAG systems to improve accessibility to scientific knowledge and to support evidence-based decision making.

Longitudinal low-dose CT follow-ups vary in noise, reconstruction kernels, and registration quality. These differences destabilize subtraction images and can trigger false new lesion alarms. We present TopoGate, a lightweight model that combines the follow-up appearance view with the subtraction view and controls their influence through a learned, quality-aware gate. The gate is driven by three case-specific signals: CT appearance quality, registration consistency, and stability of anatomical topology measured with topological metrics. On the NLST--New-Lesion--LongCT cohort comprising 152 pairs from 122 patients, TopoGate improves discrimination and calibration over single-view baselines, achieving an area under the ROC curve of 0.65 with a standard deviation of 0.05 and a Brier score of 0.14. Removing corrupted or low-quality pairs, identified by the quality scores, further increases the area under the ROC curve from 0.62 to 0.68 and reduces the Brier score from 0.14 to 0.12. The gate responds predictably to degradation, placing more weight on appearance when noise grows, which mirrors radiologist practice. The approach is simple, interpretable, and practical for reliable longitudinal LDCT triage.

Class imbalance induces systematic bias in deep neural networks by imposing a skewed effective class prior. This work introduces the Neural Prior Estimator (NPE), a framework that learns feature-conditioned log-prior estimates from latent representations. NPE employs one or more Prior Estimation Modules trained jointly with the backbone via a one-way logistic loss. Under the Neural Collapse regime, NPE is analytically shown to recover the class log-prior up to an additive constant, providing a theoretically grounded adaptive signal without requiring explicit class counts or distribution-specific hyperparameters. The learned estimate is incorporated into logit adjustment, forming NPE-LA, a principled mechanism for bias-aware prediction. Experiments on long-tailed CIFAR and imbalanced semantic segmentation benchmarks (STARE, ADE20K) demonstrate consistent improvements, particularly for underrepresented classes. NPE thus offers a lightweight and theoretically justified approach to learned prior estimation and imbalance-aware prediction.

Conversational agents increasingly mediate everyday digital interactions, yet the effects of their communication style on user experience and task success remain unclear. Addressing this gap, we describe the results of a between-subject user study where participants interact with one of two versions of a chatbot called NAVI which assists users in an interactive map-based 2D navigation task. The two chatbot versions differ only in communication style: one is friendly and supportive, while the other is direct and task-focused. Our results show that the friendly style increases subjective satisfaction and significantly improves task completion rates among female participants only, while no baseline differences between female and male participants were observed in a control condition without the chatbot. Furthermore, we find little evidence of users mimicking the chatbot's style, suggesting limited linguistic accommodation. These findings highlight the importance of user- and task-sensitive conversational agents and support that communication style personalization can meaningfully enhance interaction quality and performance.

Training and serving Large Language Models (LLMs) relies heavily on parallelization and collective operations, which are frequently bottlenecked by network bandwidth. Lossless compression using e.g., Huffman codes can alleviate the issue, however, Huffman codes suffer from slow, bit-sequential decoding and high hardware complexity due to deep tree traversals. Universal codes e.g., Exponential-Golomb codes are faster to decode but do not exploit the symbol frequency distributions. To address these limitations, this paper introduces Dual Length Codes, a hybrid approach designed to balance compression efficiency with decoding speed. Analyzing BFloat16 tensors from the Gemma model, we observed that the top 8 most frequent symbols account for approximately 50% of the cumulative probability. These 8 symbols are assigned a short 4 bit code. The remaining 248 symbols are assigned a longer 9 bit code. The coding scheme uses a single prefix bit to distinguish between the two code lengths. The scheme uses a small Look Up Table with only 8 entries for encoding and decoding. The scheme achieves a compressibility of 18.6% in comparison to 21.3% achieved by Huffman codes, but it significantly speeds up the decoding and simplifies the hardware complexity.

Recent work has shown that larger language models have better predictive power for eye movement and reading time data. While even the best models under-allocate probability mass to human responses, larger models assign higher-quality estimates of next tokens and their likelihood of production in cloze data because they are less sensitive to lexical co-occurrence statistics while being better aligned semantically to human cloze responses. The results provide support for the claim that the greater memorization capacity of larger models helps them guess more semantically appropriate words, but makes them less sensitive to low-level information that is relevant for word recognition.

Diffusion models generate high-quality samples but can also memorize training data, raising serious privacy concerns. Understanding the mechanisms governing when memorization versus generalization occurs remains an active area of research. In particular, it is unclear where along the noise schedule memorization is induced, how data geometry influences it, and how phenomena at different noise scales interact. We introduce a geometric framework that partitions the noise schedule into three regimes based on the coverage properties of training data by Gaussian shells and the concentration behavior of the posterior, which we argue are two fundamental objects governing memorization and generalization in diffusion models. This perspective reveals that memorization risk is highly non-uniform across noise levels. We further identify a danger zone at medium noise levels where memorization is most pronounced. In contrast, both the small and large noise regimes resist memorization, but through fundamentally different mechanisms: small noise avoids memorization due to limited training coverage, while large noise exhibits low posterior concentration and admits a provably near linear Gaussian denoising behavior. For the medium noise regime, we identify geometric conditions through which we propose a geometry-informed targeted intervention that mitigates memorization.

As developers increasingly rely on LLM-generated code summaries for documentation, testing, and review, it is important to study whether these summaries accurately reflect what the program actually does. LLMs often produce confident descriptions of what the code looks like it should do (intent), while missing subtle edge cases or logic changes that define what it actually does (behavior). We present a mutation-based evaluation methodology that directly tests whether a summary truly matches the code's logic. Our approach generates a summary, injects a targeted mutation into the code, and checks if the LLM updates its summary to reflect the new behavior. We validate it through three experiments totalling 624 mutation-summary evaluations across 62 programs. First, on 12 controlled synthetic programs with 324 mutations varying in type (statement, value, decision) and location (beginning, middle, end). We find that summary accuracy decreases sharply with complexity from 76.5% for single functions to 17.3% for multi-threaded systems, while mutation type and location exhibit weaker effects. Second, testing 150 mutated samples on 50 human-written programs from the Less Basic Python Problems (LBPP) dataset confirms the same failure patterns persist as models often describe algorithmic intent rather than actual mutated behavior with a summary accuracy rate of 49.3%. Furthermore, while a comparison between GPT-4 and GPT-5.2 shows a substantial performance leap (from 49.3% to 85.3%) and an improved ability to identify mutations as "bugs", both models continue to struggle with distinguishing implementation details from standard algorithmic patterns. This work establishes mutation analysis as a systematic approach for assessing whether LLM-generated summaries reflect program behavior rather than superficial textual patterns.

Large language models (LLMs) are increasingly deployed in safety and security critical applications, raising concerns about their robustness to model parameter fault injection attacks. Recent studies have shown that bit-flip attacks (BFAs), which exploit computer main memory (i.e., DRAM) vulnerabilities to flip a small number of bits in model weights, can severely disrupt LLM behavior. However, existing BFA on LLM largely induce un-targeted failure or general performance degradation, offering limited control over manipulating specific or targeted outputs. In this paper, we present TFL, a novel targeted bit-flip attack framework that enables precise manipulation of LLM outputs for selected prompts while maintaining almost no or minor degradation on unrelated inputs. Within our TFL framework, we propose a novel keyword-focused attack loss to promote attacker-specified target tokens in generative outputs, together with an auxiliary utility score that balances attack effectiveness against collateral performance impact on benign data. We evaluate TFL on multiple LLMs (Qwen, DeepSeek, Llama) and benchmarks (DROP, GSM8K, and TriviaQA). The experiments show that TFL achieves successful targeted LLM output manipulations with less than 50 bit flips and significantly reduced effect on unrelated queries compared to prior BFA approaches. This demonstrates the effectiveness of TFL and positions it as a new class of stealthy and targeted LLM model attack.

Supervised fine-tuning (SFT) relies critically on selecting training data that most benefits a model's downstream performance. Gradient-based data selection methods such as TracIn and Influence Functions leverage influence to identify useful samples, but their computational cost scales poorly, making them impractical for multi-billion-parameter large language models (LLMs). A common alternative is to use off-the-shelf smaller models as proxies, but they remain suboptimal since their learning dynamics are unclear, their sizes cannot be flexibly adjusted, and they cannot be further aligned with the target model in terms of gradient-based influence estimation. To address these challenges, we introduce Iprox, a two-stage framework that derives influence-preserving proxies directly from the target model. It first applies a low-rank compression stage to preserve influence information of the target model, and then an aligning stage to align both model gradients and logits, thereby constructing proxies that flexibly control computational cost while retaining the target model's influence. Experimental results across diverse LLM families and evaluation tasks show that Iprox consistently outperforms off-the-shelf proxies and baseline methods. On Qwen3-4B, a 1.5B proxy constructed with Iprox achieves stronger performance than the larger 1.7B off-the-shelf proxy. Notably, on Llama3.2, Iprox achieves better performance than baselines while reducing computational cost by more than half relative to the full 3B model. These results show that Iprox provides effective influence-preserving proxies, making gradient-based data selection more scalable for LLMs.

In the last decade, subgraph detection and enumeration have emerged as a central problem in distributed graph algorithms. This is largely due to the theoretical challenges and practical applications of these problems. In this paper, we initiate the systematic study of distributed sub-hypergraph enumeration in hypergraphs. To this end, we (1)~introduce several computational models for hypergraphs that generalize the CONGEST model for graphs and evaluate their relative computational power, (2)~devise algorithms for distributed triangle enumeration in our computational models and prove their optimality in two such models, (3)~introduce classes of sparse and ``everywhere sparse'' hypergraphs and describe efficient distributed algorithms for triangle enumeration in these classes, and (4)~describe general techniques that we believe to be useful for designing efficient algorithms in our hypergraph models.

Stochastic Optimal Control provides a unified mathematical framework for solving complex decision-making problems, encompassing paradigms such as maximum entropy reinforcement learning(RL) and imitation learning(IL). However, conventional parametric policies often struggle to represent the multi-modality of the solutions. Though diffusion-based policies are aimed at recovering the multi-modality, they lack an explicit probability density, which complicates policy-gradient optimization. To bridge this gap, we propose MePoly, a novel policy parameterization based on polynomial energy-based models. MePoly provides an explicit, tractable probability density, enabling exact entropy maximization. Theoretically, we ground our method in the classical moment problem, leveraging the universal approximation capabilities for arbitrary distributions. Empirically, we demonstrate that MePoly effectively captures complex non-convex manifolds and outperforms baselines in performance across diverse benchmarks.

We study the estimation of time-homogeneous drift functions in multivariate stochastic differential equations with known diffusion coefficient, from multiple trajectories observed at high frequency over a fixed time horizon. We formulate drift estimation as a denoising problem conditional on previous observations, and propose an estimator of the drift function which is a by-product of training a conditional diffusion model capable of simulating new trajectories dynamically. Across different drift classes, the proposed estimator was found to match classical methods in low dimensions and remained consistently competitive in higher dimensions, with gains that cannot be attributed to architectural design choices alone.

Evaluating the reasoning capabilities of Large Language Models is increasingly challenging as models improve. Human curation of hard questions is highly expensive, especially in recent benchmarks using PhD-level domain knowledge to challenge the most capable models. Even then, there is always a concern about whether these questions test genuine reasoning or if similar problems have been seen during training. Here, we take inspiration from 16th-century mathematical duels to design The Token Games (TTG): an evaluation framework where models challenge each other by creating their own puzzles. We leverage the format of Programming Puzzles - given a Python function that returns a boolean, find inputs that make it return True - to flexibly represent problems and enable verifying solutions. Using results from pairwise duels, we then compute Elo ratings, allowing us to compare models relative to each other. We evaluate 10 frontier models on TTG, and closely match the ranking from existing benchmarks such as Humanity's Last Exam, without involving any human effort in creating puzzles. We also find that creating good puzzles is still a highly challenging task for current models, not measured by previous benchmarks. Overall, our work suggests new paradigms for evaluating reasoning that cannot be saturated by design, and that allow testing models for other skills like creativity and task creation alongside problem solving.

Inferring causal relations in timeseries data with delayed effects is a fundamental challenge, especially when the underlying system exhibits complex dynamics that cannot be captured by simple functional mappings. Traditional approaches often fail to produce generalized and interpretable explanations, as multiple distinct input trajectories may yield nearly indistinguishable outputs. In this work, we present ruleXplain, a framework that leverages Large Language Models (LLMs) to extract formal explanations for input-output relations in simulation-driven dynamical systems. Our method introduces a constrained symbolic rule language with temporal operators and delay semantics, enabling LLMs to generate verifiable causal rules through structured prompting. ruleXplain relies on the availability of a principled model (e.g., a simulator) that maps multivariate input time series to output time series. Within ruleXplain, the simulator is used to generate diverse counterfactual input trajectories that yield similar target output, serving as candidate explanations. Such counterfactual inputs are clustered and provided as context to the LLM, which is tasked with the generation of symbolic rules encoding the joint temporal trends responsible for the patterns observable in the output times series. A closed-loop refinement process ensures rule consistency and semantic validity. We validate the framework using the PySIRTEM epidemic simulator, mapping testing rate inputs to daily infection counts; and the EnergyPlus building energy simulator, observing temperature and solar irradiance inputs to electricity needs. For validation, we perform three classes of experiments: (1) the efficacy of the ruleset through input reconstruction; (2) ablation studies evaluating the causal encoding of the ruleset; and (3) generalization tests of the extracted rules across unseen output trends with varying phase dynamics.

Generative Flow Networks (GFlowNets) are a flexible family of amortized samplers trained to generate discrete and compositional objects with probability proportional to a reward function. However, learning efficiency is constrained by the model's ability to rapidly explore diverse high-probability regions during training. To mitigate this issue, recent works have focused on incentivizing the exploration of unvisited and valuable states via curiosity-driven search and self-supervised random network distillation, which tend to waste samples on already well-approximated regions of the state space. In this context, we propose Adaptive Complementary Exploration (ACE), a principled algorithm for the effective exploration of novel and high-probability regions when learning GFlowNets. To achieve this, ACE introduces an exploration GFlowNet explicitly trained to search for high-reward states in regions underexplored by the canonical GFlowNet, which learns to sample from the target distribution. Through extensive experiments, we show that ACE significantly improves upon prior work in terms of approximation accuracy to the target distribution and discovery rate of diverse high-reward states.

Language models exhibit fundamental limitations -- hallucination, brittleness, and lack of formal grounding -- that are particularly problematic in high-stakes specialist fields requiring verifiable reasoning. I investigate whether formal domain ontologies can enhance language model reliability through retrieval-augmented generation. Using mathematics as proof of concept, I implement a neuro-symbolic pipeline leveraging the OpenMath ontology with hybrid retrieval and cross-encoder reranking to inject relevant definitions into model prompts. Evaluation on the MATH benchmark with three open-source models reveals that ontology-guided context improves performance when retrieval quality is high, but irrelevant context actively degrades it -- highlighting both the promise and challenges of neuro-symbolic approaches.

Collocating deep learning training tasks improves GPU utilization but causes drastic slowdowns due to resource contention and risks Out-of-Memory (OOM) failures. Accurate memory estimation is essential for robust collocation, while GPU utilization -- a key proxy for resource contention -- enables interference-aware scheduling to reduce slowdowns and improve throughput. Existing GPU memory estimators span three paradigms -- analytical models, CPU-side libraries, and ML-based estimators -- each with distinct limitations: dependence on detailed model specifications, intrusive integration, poor generalization, and varying latency overhead. GPU heterogeneity further complicates estimation, as identical tasks can exhibit markedly different memory footprints across hardware generations. GPU utilization remains comparatively understudied, further complicated by the non-additive nature of utilization metrics and hardware sensitivity. We conduct a systematic analysis of representative estimators from each paradigm -- Horus, PyTorch FakeTensor, and our lightweight ML-based estimator -- evaluating accuracy, generalizability, and practical overhead. We construct a synthetic dataset spanning MLPs, CNNs, and Transformers with controlled architectural variations, and train MLP- and Transformer-based estimators for memory prediction. We further experiment with utilization estimation on the same dataset. Our evaluation reveals key tradeoffs and validates estimators against real-world unseen models. Significant challenges remain: analytical models are hardware-dependent, CPU-side libraries impose intrusive integration costs, and ML-based estimators struggle with cross-architecture generalization. We release all datasets, tools, and artifacts to support further research.

Neuroscience has uncovered a fundamental mechanism of our social nature: human brain activity becomes synchronized with others in many social contexts involving interaction. Traditionally, social minds have been regarded as an exclusive property of living beings. Although large language models (LLMs) are widely accepted as powerful approximations of human behavior, with multi-LLM system being extensively explored to enhance their capabilities, it remains controversial whether they can be meaningfully compared to human social minds. In this work, we explore neural synchrony between socially interacting LLMs as an empirical evidence for this debate. Specifically, we introduce neural synchrony during social simulations as a novel proxy for analyzing the sociality of LLMs at the representational level. Through carefully designed experiments, we demonstrate that it reliably reflects both social engagement and temporal alignment in their interactions. Our findings indicate that neural synchrony between LLMs is strongly correlated with their social performance, highlighting an important link between neural synchrony and the social behaviors of LLMs. Our work offers a new perspective to examine the "social minds" of LLMs, highlighting surprising parallels in the internal dynamics that underlie human and LLM social interaction.

Query performance prediction (QPP) is an important and actively studied information retrieval task, having various applications, such as query reformulation, query expansion, and retrieval system selection, among many others. The task has been primarily studied in the context of text and image retrieval, whereas QPP for content-based video retrieval (CBVR) remains largely underexplored. To this end, we propose the first benchmark for video query performance prediction (VQPP), comprising two text-to-video retrieval datasets and two CBVR systems, respectively. VQPP contains a total of 56K text queries and 51K videos, and comes with official training, validation and test splits, fostering direct comparisons and reproducible results. We explore multiple pre-retrieval and post-retrieval performance predictors, creating a representative benchmark for future exploration of QPP in the video domain. Our results show that pre-retrieval predictors obtain competitive performance, enabling applications before performing the retrieval step. We also demonstrate the applicability of VQPP by employing the best performing pre-retrieval predictor as reward model for training a large language model (LLM) on the query reformulation task via direct preference optimization (DPO). We release our benchmark and code at https://github.com/AdrianLutu/VQPP.

A low out-degree orientation directs each edge of an undirected graph with the goal of minimizing the maximum out-degree of a vertex. In the parallel batch-dynamic setting, one can insert or delete batches of edges, and the goal is to process the entire batch in parallel with work per edge similar to that of a single sequential update and with span (or depth) for the entire batch that is polylogarithmic. In this paper we present faster parallel batch-dynamic algorithms for maintaining a low out-degree orientation of an undirected graph. All results herein achieve polylogarithmic depth, with high probability (whp); the focus of this paper is on minimizing the work, which varies across results. Our first result is the first parallel batch-dynamic algorithm to maintain an asymptotically optimal orientation with asymptotically optimal expected work bounds, in an amortized sense, improving over the prior best work bounds of Liu et al.~[SPAA~'22] by a logarithmic factor. Our second result is a $O(c \log n)$ orientation algorithm with expected worst-case $O(\sqrt{\log n})$ work per edge update, where $c$ is a known upper-bound on the arboricity of the graph. This matches the best-known sequential worst-case $O(c \log n)$ orientation algorithm given by Berglin and Brodal ~[Algorithmica~'18], albeit in expectation. Our final result is a $O(c + \log n)$-orientation algorithm with $O(\log^2 n)$ expected worst-case work per edge update. This algorithm significantly improves upon the recent result of Ghaffari and Koo~[SPAA~'25], which maintains a $O(c)$-orientation with $O(\log^9 n)$ worst-case work per edge whp.

Parameter-efficient fine-tuning methods such as LoRA enable practical adaptation of large language models but provide no principled uncertainty estimates, leading to poorly calibrated predictions and unreliable behavior under domain shift. We introduce Stiefel-Bayes Adapters (SBA), a Bayesian PEFT framework that places a Matrix Langevin prior over orthonormal adapter factors on the Stiefel manifold $\St$ and performs approximate posterior inference via tangent space Laplace approximation with geodesic retraction. Unlike Gaussian priors in flat space projected onto orthogonality constraints, our prior on the manifold naturally encodes the inductive bias that adapter subspaces should be well conditioned and orthogonal, while the posterior provides calibrated predictive uncertainty without recalibration. We prove formally that the tangent space approximation strictly avoids the structural variance inflation inherent in projecting from ambient space, establishing a rigorous theoretical advantage for intrinsic manifold inference. Across GLUE and SuperGLUE benchmarks on RoBERTa-large, LLaMA-2-7B, LLaMA-2-13B, Mistral-7B, and Qwen2.5-7B, domain shift evaluations, selective prediction protocols, and an abstractive summarization task, SBA achieves task performance comparable to LoRA and DoRA while reducing Expected Calibration Error by 18 to 34\% over deterministic baselines, improving selective prediction AUROC by 12 to 25\% under domain shift, and outperforming deep ensembles of five LoRA models on OOD detection at a fraction of the parameter cost. Our results demonstrate that where you place uncertainty, on the right geometric structure, matters more than simply adding any Bayesian treatment to adapters.

IoT applications are increasingly relying on on-device AI accelerators to ensure high performance, especially in limited connectivity and safety-critical scenarios. However, the limited on-chip memory of these accelerators forces inference runtimes to swap model segments between host and accelerator memory, substantially inflating latency. While collaborative processing by partitioning the model processing between CPU and accelerator resources can reduce accelerator memory pressure and latency, naive partitioning may worsen end-to-end latency by either shifting excessive computation to the CPU or failing to sufficiently curb swapping, a problem that is further amplified in multi-tenant and dynamic environments. To address these issues, we present SwapLess, a system for adaptive, multi-tenant TPU-CPU collaborative inference for memory-constrained Edge TPUs. SwapLess utilizes an analytic queueing model that captures partition-dependent CPU/TPU service times as well as inter- and intra-model swapping overheads across different workload mixes and request rates. Using this model, SwapLess continuously adjusts both the partition point and CPU core allocation online to minimize end-to-end response time with low decision overhead. An implementation on Edge TPU-equipped platforms demonstrates that SwapLess reduces mean latency by up to 63.8% for single-tenant workloads and up to 77.4% for multi-tenant workloads relative to the default Edge TPU compiler.

Mixture-of-Experts models rely on learned routers to assign tokens to experts, yet standard softmax gating provides no principled mechanism to control the tradeoff between sparsity and utilization. We propose Grassmannian MoE (GrMoE), a routing framework that operates on the Grassmannian manifold of subspaces, where gating weights arise from the concentration parameters of Matrix Bingham distributions. This construction yields a single, interpretable knob -- the concentration matrix $Λ$ -- that continuously controls routing entropy, replacing discrete top-$k$ selection with a smooth, geometrically principled sparsity mechanism. We further develop an amortized variational inference procedure for posterior routing distributions, enabling uncertainty-aware expert assignment that naturally resists expert collapse. We formally prove tight bounds relating the Bingham concentration spectrum to routing entropy, expected top-$k$ mass, and an exponential bound on expert collapse, establishing the first formal theory of concentration-controlled sparsity. On synthetic routing tasks, a 350M-parameter MoE language model with 8 experts, a 1.3B-parameter model with 16 experts, and a 2.7B-parameter model with 32 experts, GrMoE achieves 0\% routing collapse across all seeds, comparable or better perplexity with 15--30\% improved load balance, and a smooth monotonic relationship between concentration and effective sparsity that enables post-hoc sparsity tuning without retraining. Token-level analysis reveals that experts learn heterogeneous concentration values that correlate with linguistic specialization, providing interpretable routing behavior.

Skin cancer can be life-threatening if not diagnosed early, a prevalent yet preventable disease. Globally, skin cancer is perceived among the finest prevailing cancers and millions of people are diagnosed each year. For the allotment of benign and malignant skin spots, an area of critical importance in dermatological diagnostics, the application of two prominent deep learning models, VGG16 and DenseNet201 are investigated by this paper. We evaluate these CNN architectures for their efficacy in differentiating benign from malignant skin lesions leveraging enhancements in deep learning enforced to skin cancer spotting. Our objective is to assess model accuracy and computational efficiency, offering insights into how these models could assist in early detection, diagnosis, and streamlined workflows in dermatology. We used two deep learning methods DenseNet201 and VGG16 model on a binary class dataset containing 3297 images. The best result with an accuracy of 93.79% achieved by DenseNet201. All images were resized to 224x224 by rescaling. Although both models provide excellent accuracy, there is still some room for improvement. In future using new datasets, we tend to improve our work by achieving great accuracy.

Generative model ecosystems increasingly operate as competitive multi-platform markets, where platforms strategically select models from a shared pool and users with heterogeneous preferences choose among them. Understanding how platforms interact, when market equilibria exist, how outcomes are shaped by model-providers, platforms, and user behavior, and how social welfare is affected is critical for fostering a beneficial market environment. In this paper, we formalize a three-layer model-platform-user market game and identify conditions for the existence of pure Nash equilibrium. Our analysis shows that market structure, whether platforms converge on similar models or differentiate by selecting distinct ones, depends not only on models' global average performance but also on their localized attraction to user groups. We further examine welfare outcomes and show that expanding the model pool does not necessarily increase user welfare or market diversity. Finally, we design novel best-response training schemes that allow model providers to strategically introduce new models into competitive markets.

Mineral prospectivity mapping requires synthesizing heterogeneous geological knowledge, including textual deposit models and geospatial datasets, to identify regions likely to host specific mineral deposit types. This process is traditionally manual and knowledge-intensive. We present QueryPlot, a semantic retrieval and mapping framework that integrates large-scale geological text corpora with geologic map data using modern Natural Language Processing techniques. We curate descriptive deposit models for over 120 deposit types and transform the State Geologic Map Compilation (SGMC) polygons into structured textual representations. Given a user-defined natural language query, the system encodes both queries and region descriptions using a pretrained embedding model and computes semantic similarity scores to rank and spatially visualize regions as continuous evidence layers. QueryPlot supports compositional querying over deposit characteristics, enabling aggregation of multiple similarity-derived layers for multi-criteria prospectivity analysis. In a case study on tungsten skarn deposits, we demonstrate that embedding-based retrieval achieves high recall of known occurrences and produces prospective regions that closely align with expert-defined permissive tracts. Furthermore, similarity scores can be incorporated as additional features in supervised learning pipelines, yielding measurable improvements in classification performance. QueryPlot is implemented as a web-based system supporting interactive querying, visualization, and export of GIS-compatible prospectivity layers.To support future research, we have made the source code and datasets used in this study publicly available.

Machine learning (ML) has been increasingly used for topology optimization (TO). However, most existing ML-based approaches focus on simplified benchmark problems due to their high computational cost, spectral bias, and difficulty in handling complex physics. These limitations become more pronounced in multi-material, multi-physics problems whose objective or constraint functions are not self-adjoint. To address these challenges, we propose a framework based on physics-informed Gaussian processes (PIGPs). In our approach, the primary, adjoint, and design variables are represented by independent GP priors whose mean functions are parametrized via neural networks whose architectures are particularly beneficial for surrogate modeling of PDE solutions. We estimate all parameters of our model simultaneously by minimizing a loss that is based on the objective function, multi-physics potential energy functionals, and design-constraints. We demonstrate the capability of the proposed framework on benchmark TO problems such as compliance minimization, heat conduction optimization, and compliant mechanism design under single- and multi-material settings. Additionally, we leverage thermo-mechanical TO with single- and multi-material options as a representative multi-physics problem. We also introduce differentiation and integration schemes that dramatically accelerate the training process. Our results demonstrate that the proposed PIGP framework can effectively solve coupled multi-physics and design problems simultaneously -- generating super-resolution topologies with sharp interfaces and physically interpretable material distributions. We validate these results using open-source codes and the commercial software package COMSOL.

We consider the landscape of empirical risk minimization for high-dimensional Gaussian single-index models (generalized linear models). The objective is to recover an unknown signal $\boldsymbolθ^\star \in \mathbb{R}^d$ (where $d \gg 1$) from a loss function $\hat{R}(\boldsymbolθ)$ that depends on pairs of labels $(\mathbf{x}_i \cdot \boldsymbolθ, \mathbf{x}_i \cdot \boldsymbolθ^\star)_{i=1}^n$, with $\mathbf{x}_i \sim \mathcal{N}(0, I_d)$, in the proportional asymptotic regime $n \asymp d$. Using the Kac-Rice formula, we analyze different complexities of the landscape -- defined as the expected number of critical points -- corresponding to various types of critical points, including local minima. We first show that some variational formulas previously established in the literature for these complexities can be drastically simplified, reducing to explicit variational problems over a finite number of scalar parameters that we can efficiently solve numerically. Our framework also provides detailed predictions for properties of the critical points, including the spectral properties of the Hessian and the joint distribution of labels. We apply our analysis to the real phase retrieval problem for which we derive complete topological phase diagrams of the loss landscape, characterizing notably BBP-type transitions where the Hessian at local minima (as predicted by the Kac-Rice formula) becomes unstable in the direction of the signal. We test the predictive power of our analysis to characterize gradient flow dynamics, finding excellent agreement with finite-size simulations of local optimization algorithms, and capturing fine-grained details such as the empirical distribution of labels. Overall, our results open new avenues for the asymptotic study of loss landscapes and topological trivialization phenomena in high-dimensional statistical models.

Multi-turn interaction length is a dominant factor in the operational costs of conversational LLMs. In this work, we present a new failure mode in conversational LLMs: turn amplification, in which a model consistently prolongs multi-turn interactions without completing the underlying task. We show that an adversary can systematically exploit clarification-seeking behavior$-$commonly encouraged in multi-turn conversation settings$-$to scalably prolong interactions. Moving beyond prompt-level behaviors, we take a mechanistic perspective and identify a query-independent, universal activation subspace associated with clarification-seeking responses. Unlike prior cost-amplification attacks that rely on per-turn prompt optimization, our attack arises from conversational dynamics and persists across prompts and tasks. We show that this mechanism provides a scalable pathway to induce turn amplification: both supply-chain attacks via fine-tuning and runtime attacks through low-level parameter corruptions consistently shift models toward abstract, clarification-seeking behavior across prompts. Across multiple instruction-tuned LLMs and benchmarks, our attack substantially increases turn count while remaining compliant. We also show that existing defenses offer limited protection against this emerging class of failures.

Physics-governed models are increasingly paired with machine learning for accelerated predictions, yet most "physics--informed" formulations treat the governing equations as a penalty loss whose scale and meaning are set by heuristic balancing. This blurs operator structure, thereby confounding solution approximation error with governing-equation enforcement error and making the solving and learning progress hard to interpret and control. Here we introduce the Neural Basis Method, a projection-based formulation that couples a predefined, physics-conforming neural basis space with an operator-induced residual metric to obtain a well-conditioned deterministic minimization. Stability and reliability then hinge on this metric: the residual is not merely an optimization objective but a computable certificate tied to approximation and enforcement, remaining stable under basis enrichment and yielding reduced coordinates that are learnable across parametric instances. We use advective multiscale Darcian dynamics as a concrete demonstration of this broader point. Our method produce accurate and robust solutions in single solves and enable fast and effective parametric inference with operator learning.

We present a comprehensive characterization study of open-source message-oriented middleware (MOM) systems. We followed a rigorous methodology to select and study ten popular and diverse MOM systems. For each system, we examine 42 features with a total of 134 different options. We found that MOM systems have evolved to provide a framework for modern cloud applications through high flexibility and configurability and by offering core building blocks for complex applications including transaction support, active messaging, resource management, flow control, and native support for multi-tenancy. We also identify that there is an opportunity for the community to consolidate its efforts on fewer open-source projects. We have also created an annotated data set that makes it easy to verify our findings, which can also be used to help practitioners and developers understand and compare the features of different systems. For a wider impact, we make our data set publicly available.

We present a generative modeling framework for synthesizing physically feasible two-dimensional incompressible flows under arbitrary obstacle geometries and boundary conditions. Whereas existing diffusion-based flow generators either ignore physical constraints, impose soft penalties that do not guarantee feasibility, or specialize to fixed geometries, our approach integrates three complementary components: (1) a boundary-conditioned diffusion model operating on velocity fields; (2) a physics-informed training objective incorporating a divergence penalty; and (3) a projection-constrained reverse diffusion process that enforces exact incompressibility through a geometry-aware Helmholtz-Hodge operator. We derive the method as a discrete approximation to constrained Langevin sampling on the manifold of divergence-free vector fields, providing a connection between modern diffusion models and geometric constraint enforcement in incompressible flow spaces. Experiments on analytic Navier-Stokes data and obstacle-bounded flow configurations demonstrate significantly improved divergence, spectral accuracy, vorticity statistics, and boundary consistency relative to unconstrained, projection-only, and penalty-only baselines. Our formulation unifies soft and hard physical structure within diffusion models and provides a foundation for generative modeling of incompressible fields in robotics, graphics, and scientific computing.

Electroencephalography (EEG)-based P300 brain-computer interfaces (BCIs) enable communication without physical movement by detecting stimulus-evoked neural responses. Accurate and efficient decoding remains challenging due to high dimensionality, temporal dependence, and complex interactions across EEG channels. Most existing approaches treat channels independently or rely on black-box machine learning models, limiting interpretability and personalization. We propose a sparse Bayesian time-varying regression framework that explicitly models pairwise EEG channel interactions while performing automatic temporal feature selection. The model employs a relaxed-thresholded Gaussian process prior to induce structured sparsity in both channel-specific and interaction effects, enabling interpretable identification of task-relevant channels and channel pairs. Applied to a publicly available P300 speller dataset of 55 participants, the proposed method achieves a median character-level accuracy of 100\% using all stimulus sequences and attains the highest overall decoding performance among competing statistical and deep learning approaches. Incorporating channel interactions yields subgroup-specific gains of up to 7\% in character-level accuracy, particularly among participants who abstained from alcohol (up to 18\% improvement). Importantly, the proposed method improves median BCI-Utility by approximately 10\% at its optimal operating point, achieving peak throughput after only seven stimulus sequences. These results demonstrate that explicitly modeling structured EEG channel interactions within a principled Bayesian framework enhances predictive accuracy, improves user-centric throughput, and supports personalization in P300 BCI systems.

Hands play a central role in daily life, yet modeling natural hand motions remains underexplored. Existing methods that tackle text-to-hand-motion generation or hand animation captioning rely on studio-captured datasets with limited actions and contexts, making them costly to scale to "in-the-wild" settings. Further, contemporary models and their training schemes struggle to capture animation fidelity with text-motion alignment. To address this, we (1) introduce '3D Hands in the Wild' (3D-HIW), a dataset of 32K 3D hand-motion sequences and aligned text, and (2) propose CLUTCH, an LLM-based hand animation system with two critical innovations: (a) SHIFT, a novel VQ-VAE architecture to tokenize hand motion, and (b) a geometric refinement stage to finetune the LLM. To build 3D-HIW, we propose a data annotation pipeline that combines vision-language models (VLMs) and state-of-the-art 3D hand trackers, and apply it to a large corpus of egocentric action videos covering a wide range of scenarios. To fully capture motion in-the-wild, CLUTCH employs SHIFT, a part-modality decomposed VQ-VAE, which improves generalization and reconstruction fidelity. Finally, to improve animation quality, we introduce a geometric refinement stage, where CLUTCH is co-supervised with a reconstruction loss applied directly to decoded hand motion parameters. Experiments demonstrate state-of-the-art performance on text-to-motion and motion-to-text tasks, establishing the first benchmark for scalable in-the-wild hand motion modelling. Code, data and models will be released.

Agentic AI systems are increasingly capable of performing professional and personal tasks with limited human involvement. However, tracking these developments is difficult because the AI agent ecosystem is complex, rapidly evolving, and inconsistently documented, posing obstacles to both researchers and policymakers. To address these challenges, this paper presents the 2025 AI Agent Index. The Index documents information regarding the origins, design, capabilities, ecosystem, and safety features of 30 state-of-the-art AI agents based on publicly available information and email correspondence with developers. In addition to documenting information about individual agents, the Index illuminates broader trends in the development of agents, their capabilities, and the level of transparency of developers. Notably, we find different transparency levels among agent developers and observe that most developers share little information about safety, evaluations, and societal impacts. The 2025 AI Agent Index is available online at https://aiagentindex.mit.edu

Biodiversity loss poses a significant threat to humanity, making wildlife monitoring essential for assessing ecosystem health. Avian species are ideal subjects for this due to their popularity and the ease of identifying them through their distinctive songs. Traditionalavian monitoring methods require manual counting and are therefore costly and inefficient. In passive acoustic monitoring, soundscapes are recorded over long periods of time. The recordings are analyzed to identify bird species afterwards. Machine learning methods have greatly expedited this process in a wide range of species and environments, however, existing solutions require complex models and substantial computational resources. Instead, we propose running machine learning models on inexpensive microcontroller units (MCUs) directly in the field. Due to the resulting hardware and energy constraints, efficient artificial intelligence (AI) architecture is required. In this paper, we present our method for avian monitoring on MCUs. We trained and compressed models for various numbers of target classes to assess the detection of multiple bird species on edge devices and evaluate the influence of the number of species on the compressibility of neural networks. Our results demonstrate significant compression rates with minimal performance loss. We also provide benchmarking results for different hardware platforms and evaluate the feasibility of deploying energy-autonomous devices.

This work presents a practical benchmarking framework for optimizing artificial intelligence (AI) models on ARM Cortex processors (M0+, M4, M7), focusing on energy efficiency, accuracy, and resource utilization in embedded systems. Through the design of an automated test bench, we provide a systematic approach to evaluate across key performance indicators (KPIs) and identify optimal combinations of processor and AI model. The research highlights a nearlinear correlation between floating-point operations (FLOPs) and inference time, offering a reliable metric for estimating computational demands. Using Pareto analysis, we demonstrate how to balance trade-offs between energy consumption and model accuracy, ensuring that AI applications meet performance requirements without compromising sustainability. Key findings indicate that the M7 processor is ideal for short inference cycles, while the M4 processor offers better energy efficiency for longer inference tasks. The M0+ processor, while less efficient for complex AI models, remains suitable for simpler tasks. This work provides insights for developers, guiding them to design energy-efficient AI systems that deliver high performance in realworld applications.

A key problem of solid mechanics is the identification of the constitutive law of a material, that is, the relation between strain and stress. Machine learning has lead to considerable advances in this field lately. Here we introduce inelastic Constitutive Kolmogorov-Arnold Networks (iCKANs). This novel artificial neural network architecture can discover in an automated manner symbolic constitutive laws describing both the elastic and inelastic behavior of materials. That is, it can translate data from material testing into corresponding elastic and inelastic potential functions in closed mathematical form. We demonstrate the advantages of iCKANs using both synthetic data and experimental data of the viscoelastic polymer materials VHB 4910 and VHB 4905. The results demonstrate that iCKANs accurately capture complex viscoelastic behavior while preserving physical interpretability. It is a particular strength of iCKANs that they can process not only mechanical data but also arbitrary additional information available about a material (e.g., about temperature-dependent behavior). This makes iCKANs a powerful tool to discover in the future also how specific processing or service conditions affect the properties of materials.

A challenge in marine bioacoustic analysis is the detection of animal signals, like calls, whistles and clicks, for behavioral studies. Manual labeling is too time-consuming to process sufficient data to get reasonable results. Thus, an automatic solution to overcome the time-consuming data analysis is necessary. Basic mathematical models can detect events in simple environments, but they struggle with complex scenarios, like differentiating signals with a low signal-to-noise ratio or distinguishing clicks from echoes. Deep Learning Neural Networks, such as ANIMAL-SPOT, are better suited for such tasks. DNNs process audio signals as image representations, often using spectrograms created by Short-Time Fourier Transform. However, spectrograms have limitations due to the uncertainty principle, which creates a tradeoff between time and frequency resolution. Alternatives like the wavelet, which provides better time resolution for high frequencies and improved frequency resolution for low frequencies, may offer advantages for feature extraction in complex bioacoustic environments. This thesis shows the efficacy of CLICK-SPOT on Norwegian Killer whale underwater recordings provided by the cetacean biologist Dr. Vester. Keywords: Bioacoustics, Deep Learning, Wavelet Transformation

Predicting functional consequences of genetic variants in crop genes remains a critical bottleneck for precision breeding programs. We present AgriVariant, an end-to-end pipeline for variant-effect prediction in rice (Oryza sativa) that addresses the lack of crop-specific variant-interpretation tools and can be extended to any crop species with available reference genomes and gene annotations. Our approach integrates deep learning-based variant calling (DeepChem-Variant) with custom plant genomics annotation using RAP-DB gene models and database-independent deleteriousness scoring that combines the Grantham distance and the BLOSUM62 substitution matrix. We validate the pipeline through targeted mutations in stress-response genes (OsDREB2a, OsDREB1F, SKC1), demonstrating correct classification of stop-gained, missense, and synonymous variants with appropriate HIGH / MODERATE / LOW impact assignments. An exhaustive mutagenesis study of OsMT-3a analyzed all 1,509 possible single-nucleotide variants in 10 days, identifying 353 high-impact, 447 medium-impact, and 709 low-impact variants - an analysis that would have required 2-4 years using traditional wet-lab approaches. This computational framework enables breeders to prioritize variants for experimental validation across diverse crop species, reducing screening costs and accelerating development of climate-resilient crop varieties.

We propose Bayesian optimal sequential prediction as a new principle for understanding in-context learning (ICL). Unlike interpretations framing Transformers as performing implicit gradient descent, we formalize ICL as meta-learning over latent sequence tasks. For tasks governed by Linear Gaussian State Space Models (LG-SSMs), we prove a meta-trained selective SSM asymptotically implements the Bayes-optimal predictor, converging to the posterior predictive mean. We further establish a statistical separation from gradient descent, constructing tasks with temporally correlated noise where the optimal Bayesian predictor strictly outperforms any empirical risk minimization (ERM) estimator. Since Transformers can be seen as performing implicit ERM, this demonstrates selective SSMs achieve lower asymptotic risk due to superior statistical efficiency. Experiments on synthetic LG-SSM tasks and a character-level Markov benchmark confirm selective SSMs converge faster to Bayes-optimal risk, show superior sample efficiency with longer contexts in structured-noise settings, and track latent states more robustly than linear Transformers. This reframes ICL from "implicit optimization" to "optimal inference," explaining the efficiency of selective SSMs and offering a principled basis for architecture design.

Large language models adapt to new tasks through in-context learning (ICL) without parameter updates. Current theoretical explanations for this capability assume test tasks are drawn from a distribution similar to that seen during pretraining. This assumption overlooks adversarial distribution shifts that threaten real-world reliability. To address this gap, we introduce a distributionally robust meta-learning framework that provides worst-case performance guarantees for ICL under Wasserstein-based distribution shifts. Focusing on linear self-attention Transformers, we derive a non-asymptotic bound linking adversarial perturbation strength ($ρ$), model capacity ($m$), and the number of in-context examples ($N$). The analysis reveals that model robustness scales with the square root of its capacity ($ρ_{\text{max}} \propto \sqrt{m}$), while adversarial settings impose a sample complexity penalty proportional to the square of the perturbation magnitude ($N_ρ- N_0 \propto ρ^2$). Experiments on synthetic tasks confirm these scaling laws. These findings advance the theoretical understanding of ICL's limits under adversarial conditions and suggest that model capacity serves as a fundamental resource for distributional robustness.

Genomic sequences span billions of base pairs (bp), posing a fundamental challenge for genome-scale foundation models. Existing approaches largely sidestep this barrier by either scaling relatively small models to long contexts or relying on heavy multi-GPU parallelism. Here we introduce GeneZip, a DNA compression model that leverages a key biological prior: genomic information is highly imbalanced. Coding regions comprise only a small fraction (about 2 percent) yet are information-dense, whereas most non-coding sequence is comparatively information-sparse. GeneZip couples HNet-style dynamic routing with a region-aware compression-ratio objective, enabling adaptive allocation of representation budget across genomic regions. As a result, GeneZip learns region-aware compression and achieves 137.6x compression with only 0.31 perplexity increase. On downstream long-context benchmarks, GeneZip achieves comparable or better performance on contact map prediction, expression quantitative trait loci prediction, and enhancer-target gene prediction. By reducing effective sequence length, GeneZip unlocks simultaneous scaling of context and capacity: compared to the prior state-of-the-art model JanusDNA, it enables training models 82.6x larger at 1M-bp context, supporting a 636M-parameter GeneZip model at 1M-bp context. All experiments in this paper can be trained on a single A100 80GB GPU.

Efficient stochastic optimization typically integrates an update direction that performs well in the deterministic regime with a mechanism adapting to stochastic perturbations. While Adam uses adaptive moment estimates to promote stability, Muon utilizes the weight layers' matrix structure via orthogonalized momentum, showing superior performance in large language model training. We propose a new optimizer and a diagonal extension, NAMO and NAMO-D, providing the first principled integration of orthogonalized momentum with norm-based Adam-type noise adaptation. NAMO scales orthogonalized momentum using a single adaptive stepsize, preserving orthogonality while improving upon Muon at negligible additional cost. NAMO-D instead right-multiplies orthogonalized momentum by a diagonal matrix with clamped entries. This design enables neuron-wise noise adaptation and aligns with the common near block-diagonal Hessian structure. Under standard assumptions, we establish optimal convergence rates for both algorithms in the deterministic setting and show that, in the stochastic setting, their convergence guarantees adapt to the noise level of stochastic gradients. Experiments on pretraining GPT-2 models demonstrate improved performance of both NAMO and NAMO-D compared to the AdamW and Muon baselines, with NAMO-D achieving further gains over NAMO via an additional clamping hyperparameter that balances the competing goals of maintaining a well-conditioned update direction and leveraging fine-grained noise adaptation.

Large Reasoning Models (LRMs) exhibit strong performance, yet often produce rationales that sound plausible but fail to reflect their true decision process, undermining reliability and trust. We introduce a formal framework for reasoning faithfulness, defined by two testable conditions: stance consistency (a coherent stance linking reasoning to answer) and causal influence (the stated reasoning causally drives the answer under output-level interventions), explicitly decoupled from accuracy. To operationalize this, we present RFEval, a benchmark of 7,186 instances across seven tasks that probes faithfulness via controlled, output-level counterfactual interventions. Evaluating twelve open-source LRMs, we find unfaithfulness in 49.7% of outputs, predominantly from stance inconsistency. Failures are concentrated in brittle, convergent domains such as math and code, and correlate more with post-training regimes than with scale: within-family ablations indicate that adding current RL-style objectives on top of supervised fine-tuning can reduce reasoning faithfulness, even when accuracy is maintained. Crucially, accuracy is neither a sufficient nor a reliable proxy for faithfulness: once controlling for model and task, the accuracy-faithfulness link is weak and statistically insignificant. Our work establishes a rigorous methodology for auditing LRM reliability and shows that trustworthy AI requires optimizing not only for correct outcomes but also for the structural integrity of the reasoning process. Our code and dataset can be found at project page: https://aidaslab.github.io/RFEval/}{https://aidaslab.github.io/RFEval/

Autonomous coding agents, powered by large language models (LLMs), are increasingly being adopted in the software industry to automate complex engineering tasks. However, these agents are prone to a wide range of misbehaviors, such as deviating from the user's instructions, getting stuck in repetitive loops, or failing to use tools correctly. These failures disrupt the development workflow and often require resource-intensive manual intervention. In this paper, we present a system for automatically recovering from agentic misbehaviors at scale. We first introduce a taxonomy of misbehaviors grounded in an analysis of production traffic, identifying three primary categories: Specification Drift, Reasoning Problems, and Tool Call Failures, which we find occur in about 30% of all agent trajectories. To address these issues, we developed a lightweight, asynchronous self-intervention system named Wink. Wink observes agent trajectories and provides targeted course-correction guidance to nudge the agent back to a productive path. We evaluated our system on over 10,000 real world agent trajectories and found that it successfully resolves 90% of the misbehaviors that require a single intervention. Furthermore, a live A/B test in our production environment demonstrated that our system leads to a statistically significant reduction in Tool Call Failures, Tokens per Session and Engineer Interventions per Session. We present our experience designing and deploying this system, offering insights into the challenges of building resilient agentic systems at scale.

Future 6G networks will interconnect not only devices, but autonomous machines that continuously sense, reason, and act. In such environments, communication can no longer be understood solely as delivering bits or even preserving semantic meaning. Even when two agents interpret the same information correctly, they may still behave inconsistently if their internal reasoning processes evolve differently. We refer to this emerging challenge as belief divergence. This article introduces reasoning native agentic communication, a new paradigm in which communication is explicitly designed to address belief divergence rather than merely transmitting representations. Instead of triggering transmissions based only on channel conditions or data relevance, the proposed framework activates communication according to predicted misalignment in agents internal belief states. We present a reasoning native architecture that augments the conventional communication stack with a coordination plane grounded in a shared knowledge structure and bounded belief modeling. Through enabling mechanisms and representative multi agent scenarios, we illustrate how such an approach can prevent coordination drift and maintain coherent behavior across heterogeneous systems. By reframing communication as a regulator of distributed reasoning, reasoning native agentic communication enables 6G networks to act as an active harmonizer of autonomous intelligence.

Mutual adaptation is a central challenge in human--AI teaming, as humans naturally adjust their strategies in response to a robot's policy. Existing approaches aim to improve diversity in training partners to approximate human behavior, but these partners are static and fail to capture adaptive behavior of humans. Exposing robots to adaptive behaviors is critical, yet when both agents learn simultaneously in a multi-agent setting, they often converge to opaque implicit coordination strategies that only work with the agents they were co-trained with. Such agents fail to generalize when paired with new partners. In order to capture the adaptive behavior of humans, we model the human-robot teaming scenario as an Interactive Partially Observable Markov Decision Process (I-POMDP), explicitly modeling human adaptation as part of the state. We propose a nested training regime to approximately learn the solution to a finite-level I-POMDP. In this framework, agents at each level are trained against adaptive agents from the level below. This ensures that the ego agent is exposed to adaptive behavior during training while avoiding the emergence of implicit coordination strategies, since the training partners are not themselves learning. We train our method in a multi-episode, required cooperation setup in the Overcooked domain, comparing it against several baseline agents designed for human-robot teaming. We evaluate the performance of our agent when paired with adaptive partners that were not seen during training. Our results demonstrate that our agent not only achieves higher task performance with these adaptive partners but also exhibits significantly greater adaptability during team interactions.

We introduce LLM-Wikirace, a benchmark for evaluating planning, reasoning, and world knowledge in large language models (LLMs). In LLM-Wikirace, models must efficiently navigate Wikipedia hyperlinks step by step to reach a target page from a given source, requiring look-ahead planning and the ability to reason about how concepts are connected in the real world. We evaluate a broad set of open- and closed-source models, including Gemini-3, GPT-5, and Claude Opus 4.5, which achieve the strongest results on the easy level of the task and demonstrate superhuman performance. Despite this, performance drops sharply on hard difficulty: the best-performing model, Gemini-3, succeeds in only 23\% of hard games, highlighting substantial remaining challenges for frontier models. Our analysis shows that world knowledge is a necessary ingredient for success, but only up to a point, beyond this threshold, planning and long-horizon reasoning capabilities become the dominant factors. Trajectory-level analysis further reveals that even the strongest models struggle to replan after failure, frequently entering loops rather than recovering. LLM-Wikirace is a simple benchmark that reveals clear limitations in current reasoning systems, offering an open arena where planning-capable LLMs still have much to prove. Our code and leaderboard available at https:/llmwikirace.github.io.

Task planning for robotic manipulation with large language models (LLMs) is an emerging area. Prior approaches rely on specialized models, fine tuning, or prompt tuning, and often operate in an open loop manner without robust environmental feedback, making them fragile in dynamic settings.MALLVi present a Multi Agent Large Language and Vision framework that enables closed loop feedback driven robotic manipulation. Given a natural language instruction and an image of the environment, MALLVi generates executable atomic actions for a robot manipulator. After action execution, a Vision Language Model (VLM) evaluates environmental feedback and decides whether to repeat the process or proceed to the next step Rather than using a single model, MALLVi coordinates specialized agents, Decomposer, Localizer, Thinker, and Reflector, to manage perception, localization, reasoning, and high level planning. An optional Descriptor agent provides visual memory of the initial state. The Reflector supports targeted error detection and recovery by reactivating only relevant agents, avoiding full replanning.Experiments in simulation and real world settings show that iterative closed loop multi agent coordination improves generalization and increases success rates in zero shot manipulation tasks.Code available at https://github.com/iman1234ahmadi/MALLVI.

Agile estimation techniques, particularly T-shirt sizing, are widely used in software development for their simplicity and utility in scoping work. However, when we apply these methods to artificial intelligence initiatives -- especially those involving large language models (LLMs) and multi-agent systems -- the results can be systematically misleading. This paper shares an evidence-backed analysis of five foundational assumptions we often make during T-shirt sizing. While these assumptions usually hold true for traditional software, they tend to fail in AI contexts: (1) linear effort scaling, (2) repeatability from prior experience, (3) effort-duration fungibility, (4) task decomposability, and (5) deterministic completion criteria. Drawing on recent research into multi-agent system failures, scaling principles, and the inherent unreliability of multi-turn conversations, we show how AI development breaks these rules. We see this through non-linear performance jumps, complex interaction surfaces, and "tight coupling" where a small change in data cascades through the entire stack. To help teams navigate this, we propose Checkpoint Sizing: a more human-centric, iterative approach that uses explicit decision gates where scope and feasibility are reassessed based on what we learn during development, rather than what we assumed at the start. This paper is intended for engineering managers, technical leads, and product owners responsible for planning and delivering AI initiatives.

Agent Skill framework, now widely and officially supported by major players such as GitHub Copilot, LangChain, and OpenAI, performs especially well with proprietary models by improving context engineering, reducing hallucinations, and boosting task accuracy. Based on these observations, an investigation is conducted to determine whether the Agent Skill paradigm provides similar benefits to small language models (SLMs). This question matters in industrial scenarios where continuous reliance on public APIs is infeasible due to data-security and budget constraints requirements, and where SLMs often show limited generalization in highly customized scenarios. This work introduces a formal mathematical definition of the Agent Skill process, followed by a systematic evaluation of language models of varying sizes across multiple use cases. The evaluation encompasses two open-source tasks and a real-world insurance claims data set. The results show that tiny models struggle with reliable skill selection, while moderately sized SLMs (approximately 12B - 30B) parameters) benefit substantially from the Agent Skill approach. Moreover, code-specialized variants at around 80B parameters achieve performance comparable to closed-source baselines while improving GPU efficiency. Collectively, these findings provide a comprehensive and nuanced characterization of the capabilities and constraints of the framework, while providing actionable insights for the effective deployment of Agent Skills in SLM-centered environments.

Machine learning can accelerate materials discovery. Models perform impressively on many benchmarks. However, strong benchmark performance does not imply that a model learned chemistry. I test a concrete alternative hypothesis: that property prediction can be driven by bibliographic confounding. Across five tasks spanning MOFs (thermal and solvent stability), perovskite solar cells (efficiency), batteries (capacity), and TADF emitters (emission wavelength), models trained on standard chemical descriptors predict author, journal, and publication year well above chance. When these predicted metadata ("bibliographic fingerprints") are used as the sole input to a second model, performance is sometimes competitive with conventional descriptor-based predictors. These results show that many datasets do not rule out non-chemical explanations of success. Progress requires routine falsification tests (e.g., group/time splits and metadata ablations), datasets designed to resist spurious correlations, and explicit separation of two goals: predictive utility versus evidence of chemical understanding.

Across academia, industry, and government, ``AI'' has become central in research and development, regulatory debates, and promises of ever faster and more capable decision-making and action. In numerous domains, especially safety-critical ones, there are significant concerns over how ``AI'' may affect decision-making, responsibility, or the likelihood of mistakes (to name only a few categories of critique). However, for most critiques, the target is generally ``AI'', a broad term admitting many (types of) systems used for a variety of tasks and each coming with its own set of limitations, challenges, and potential use cases. In this article, we focus on the military domain as a case study and present both a loose enumerative taxonomy of systems captured under the umbrella term ``military AI'', as well as discussion of the challenges of each. In doing so, we highlight that critiques of one (type of) system will not always transfer to other (types of) systems. Building on this, we argue that in order for debates to move forward fruitfully, it is imperative that the discussions be made more precise and that ``AI'' be excised from debates to the extent possible. Researchers, developers, and policy-makers should make clear exactly what systems they have in mind and what possible benefits and risks attend the deployment of those particular systems. While we focus on AI in the military as an exemplar for the overall trends in discussions of ``AI'', the argument's conclusions are broad and have import for discussions of AI across a host of domains.

In January 2026, torrential rains killed 200-300 people across Southern Africa, exposing a critical reality: 60% of the continent lacks effective early warning systems due to infrastructure costs. Traditional radar stations exceed USD 1 million each, leaving Africa with an 18x coverage deficit compared to the US and EU. We present a production-grade architecture for deploying NVIDIA Earth-2 AI weather models at USD 1,430-1,730/month for national-scale deployment - enabling coverage at 2,000-4,545x lower cost than radar. The system generates 15-day global atmospheric forecasts, cached in PostgreSQL to enable user queries under 200 milliseconds without real-time inference. Deployed in South Africa in February 2026, our system demonstrates three technical contributions: (1) a ProcessPoolExecutor-based event loop isolation pattern that resolves aiobotocore session lifecycle conflicts in async Python applications; (2) a database-backed serving architecture where the GPU writes global forecasts directly to PostgreSQL, eliminating HTTP transfer bottlenecks for high-resolution tensors; and (3) an automated coordinate management pattern for multi-step inference across 61 timesteps. Forecasts are delivered via WhatsApp, leveraging 80%+ market penetration. This architecture makes continent-scale early warning systems economically viable, supporting UNDRR findings that such systems reduce disaster death rates by 6x. All architectural details are documented inline for full reproducibility.

We show that training AI agents on high-fidelity reinforcement learning environments produces capabilities that generalize beyond the training distribution. We introduce CoreCraft, the first environment in EnterpriseBench, Surge AI's suite of agentic RL environments. CoreCraft is a fully operational enterprise simulation of a customer support organization, comprising over 2,500 entities across 14 entity types with 23 unique tools, designed to measure whether AI agents can perform the multi-step, domain-specific work that real jobs demand. Frontier models such as GPT-5.2 and Claude Opus 4.6 solve fewer than 30% of tasks when all expert-authored rubric criteria must be satisfied. Using this environment, we train GLM 4.6 with Group Relative Policy Optimization (GRPO) and adaptive clipping. After a single epoch of training, the model improves from 25.37% to 36.76% task pass rate on held-out evaluation tasks. More importantly, these gains transfer to out-of-distribution benchmarks: +4.5% on BFCL Parallel, +7.4% on Tau2-Bench Retail, and +6.8% on Tool Decathlon (Pass@1). We believe three environment properties are consistent with the observed transfer: task-centric world building that optimizes for diverse, challenging tasks; expert-authored rubrics enabling reliable reward computation; and enterprise workflows that reflect realistic professional patterns. Our results suggest that environment quality, diversity, and realism are key factors enabling generalizable agent capabilities.

Snapping instabilities in soft structures offer a powerful pathway to achieve rapid and energy-efficient actuation. In this study, an eccentric dome-shaped snapping actuator is developed to generate controllable asymmetric motion through geometry-induced instability. Finite element simulations and experiments reveal consistent asymmetric deformation and the corresponding pressure characteristics. By coupling four snapping actuators in a pneumatic network, a compact quadrupedal robot achieves coordinated wavelike locomotion using only a single pressure input. The robot exhibits frequency-dependent performance with a maximum speed of 72.78~mm/s at 7.5~Hz. These findings demonstrate the potential of asymmetric snapping mechanisms for physically controlled actuation and lay the groundwork for fully untethered and efficient soft robotic systems.

We study the phenomenon of the locking of the order parameter (or synchronization) in spin glasses at low temperatures. When two systems with independent disorders are coupled, their overlaps become similar. A crucial question is how this effect depends on the strength of the coupling between the two systems. Non-perturbative phenomena are present when $1 \ll ΔH \ll N$, being $ΔH$ the coupling Hamiltonian and $N$ the size of the system. In this intermediate-coupling region, the effect is related to finite-size free-energy corrections and to the correlations in the Dyson hierarchical spin glass, a model that mimics the physics of finite-dimensional systems. We study this phenomenon in the mean-field approach, both analytically and numerically, and we finally compute the critical exponents for finite-volume corrections in mean-field theory and for the decay of correlations in the Dyson hierarchical model.

This work introduces Ori-Sense, a compliant capacitive sensor inspired by the inverted Kresling origami pattern. The device translates torsional deformation into measurable capacitance changes, enabling proprioceptive feedback for soft robotic systems. Using dissolvable-core molding, we fabricated a monolithic silicone structure with embedded conductive TPU electrodes, forming an integrated soft capacitor. Mechanical characterization revealed low stiffness and minimal impedance, with torque values below 0.01 N mm for axial displacements between -15 mm and 15 mm, and up to 0.03 N mm at 30 degrees twist under compression. Finite-element simulations confirmed localized stresses along fold lines and validated the measured torque-rotation response. Electrical tests showed consistent capacitance modulation up to 30%, directly correlated with the twist angle, and maximal sensitivity of S_theta ~ 0.0067 pF/deg at 5 mm of axial deformation.

The forces that mixtures of motorized and passive crosslinking proteins collectively generate between cytoskeletal filaments within our cells are the key drivers of active cellular mechanics. Despite their importance, a unified theory to describe such crosslinking forces has so far been missing. In this paper, we derive a theory that predicts the forces generated collectively by crosslinking proteins linking two biopolymer filaments from measurable filament and crosslinker properties, using out-of-equilibrium thermodynamics. Our framework allows us to decompose the forces generated by crosslinkers into three separate components: entropic, active, and frictional. In doing so, it offers a clear physical interpretation of the fundamental mechanisms by which crosslinking proteins self-organize and collectively generate forces. We demonstrate the robustness and utility of this framework by applying it to four different experiments that probe the combined roles of passive and motorized crosslinkers. For each experiment, our theoretical approach allows us to disentangle the relative contributions of entropic, active, and frictional forces, clarifying how different physical processes underpin collective force production. In turn, this makes it possible to quantitatively compare and predict how various crosslinker combinations influence force generation between filaments, pattern formation along filaments, and the dynamics of filament pairs.

Next-generation spintronic sensors aim to overcome the limitations of traditional tunneling-magnetoresistance (TMR) devices, such as complex manufacturing, high $1/f$ noise, and significant offsets. This work presents a comprehensive modeling and experimental validation of a magnetic field sensor based on Spin Hall Magnetoresistance (SMR) in a Wheatstone bridge configuration. Utilizing a multiphysics approach, we simulate the interplay between SMR, Anisotropic Magnetoresistance (AMR), and Spin-Orbit Torque (SOT) using a Stoner-Wohlfarth model complemented by a Fuchs-Sondheimer analysis of current distribution. To account for the presence of magnetic domains, we incorporate a modified Stoner-Wohlfarth framework that considers non-uniform magnetization and domain wall motion through a "truncated astroid" approach, allowing for a statistical distribution of single-domain particles. The model is validated against experimental measurements of Pt/$\text{Fe}_{60}\text{Co}_{20}\text{B}_{20}$ and Ta/$\text{Fe}_{60}\text{Co}_{20}\text{B}_{20}$ bilayers patterned into Hall bars and Wheatstone bridges. The model provides critical design guidelines for optimizing material properties, layer thickness, and device layout to minimize power consumption and maximize sensitivity in SMR-based sensing applications.

Complex systems with nonlinear response mechanisms can be applied as reservoir computers for energy-efficient machine learning tasks. Historically explored at the macro- and meso-scale, physical reservoir computing has recently been extended to the atomic scale via chemical mixtures with strong and dynamic heterogeneity. Here we explore the possibility that configurational degeneracy within disordered materials might form the basis for solid-state atomic-scale reservoirs. Our proof-of-concept uses the disordered metal-organic framework DUT-8, which undergoes a series of disorder-disorder transitions on exposure to different guest species. We show that variations in X-ray diffuse scattering associated with these transitions function as suitable readouts for machine learning applications. A combination of nonlinearity and memory effects in the DUT-8 response allows the system to carry out both classification and time-series machine learning tasks with accuracies comparable to those of mesoscale physical reservoir computers. Our results suggest a new avenue for exploiting correlated disorder in solid phases whenever the nature of that disorder can be modulated through external perturbations-a phenomenon we term `responsive disorder'.

Transferring a physical system from an initial to a final state while minimizing energetic losses is an interdisciplinary control problem that bridges stochastic thermodynamics and optimal transport theory. Recent research typically considers problems in which the optimal solution is realized via conservative forces, but whether this situation applies depends on the problem's constraints. In systems with complex topologies like discrete networks, the optimal, dissipation-minimizing protocol involves applying nonconservative forces along cycles if the timescales of the transitions in the network are fixed. We show that although nonconservative driving is optimal in this setting, a conservative protocol exists whose dissipation is at most twice the optimal one. This finding is complemented with an example modeling transport across an energy barrier, which illustrates such improvements of order 1 explicitly. Qualitatively, conservative driving falls short of achieving optimality because direct transport across the barrier is avoided. We conclude with a discussion that the optimality of nonconservative driving might be a generic phenomenon: As fewer degrees of freedom can be optimized, additional degrees of freedom due to adding nonconservative forces become more significant.

Nodal-line semimetals (NLSs) can display unconventional quasiparticle dynamics and charge transport properties due to their extended band degeneracy and the peculiar geometry of their Fermi surface. We consider electron-acoustic phonon scattering as the dominant relaxation mechanism and compute the quasiparticle decay rate and dc conductivity by solving the linearized semiclassical Boltzmann equation in a minimal model of a doped circular NLS. We find that the toroidal geometry of the Fermi surface gives rise to two parametrically distinct Bloch-Grüneisen temperatures, associated with momentum transfers along the poloidal and toroidal directions, respectively. As a result, an intermediate temperature window opens between these two scales, in which the decay rate follows $Γ\propto T^2$, while the conductivity follows $σ\propto T^{-2}$. We also obtain the low- and high-temperature asymptotic behaviors, and discuss implications for ARPES and transport measurements in candidate NLS materials.

We investigate the operation of a qubit as a quantum thermal device within the repeated interaction framework, allowing for strong system-bath coupling and finite interaction times. We analyze two minimal models: an alternating-coupling setup, in which the qubit sequentially interacts with hot and cold baths, and a simultaneous-coupling setup, where both baths interact with the qubit during each collision. For the alternating model, we obtain an exact analytical solution for the limit-cycle state, valid for arbitrary coupling strengths and collision durations. Using this solution, we rigorously prove a no-go theorem for quantum refrigeration. We further demonstrate that, although work can be generated locally at individual system-bath contacts, the total work over a cycle is always nonpositive, precluding engine operation. In the absence of work, the model describes pure heat conduction, for which we derive a closed-form expression for the heat current and show that it exhibits a nonmonotonic turnover behavior. The simultaneous-coupling model is analyzed perturbatively. In the short-collision-time limit, it reproduces the same steady-state behavior as the alternating model, reinforcing the generality of the constraints identified. Our results establish fundamental limitations on qubit-based quantum thermal machines operating under Markovian repeated interactions and highlight the need for enriched models to realize functional quantum thermal devices.

Two-dimensional topological insulators are central to our understanding of the connection between topological symmetries in a material and its band electronics. Within this class of materials, a breadth of complex quantum behaviors, such as persistent spin-polarized current states in the presence of a broken time reversal symmetry, and temperature-independent topological protection of quantum states, are thought to exist. However, current studies using photoemission and spectroscopic analyses or transport experiments fail to provide insight into the interplay between the physical 2D manifold and the band topology itself, since they do not provide spatial resolution of the phenomena to be understood. In this work, we develop a methodology for applying magnetic force microscopy to such systems to address this issue. Using well-characterized 2D crystallites of bismuth telluride ($Bi_2$$Te_3$), we image the magnetic signal directly associated with topological edge states. The observed phase contrast is remarkably robust at a temperature of 25°C and occurs across crystallite sizes and shapes. A detailed analysis of the magnetic imaging suggests that the current observed is composed of two parts: the first is a persistent current ($I_{Persistent}$) as predicted by theory, and the second is due to Faraday induction, $I_{Faraday}$. Damping dynamics of the cantilever during imaging further suggest that this Faraday EMF is established by spin accumulation along the 1D edge channel of the crystal, which then converts to a charge current in the presence of time reversal symmetry breaking, creating a novel form of rectification in the channel. This unexpected result can prompt new ideas for topology-based circuit elements with extremely low losses and power consumption.

We study the two-dimensional equations of motion for a charged particle subjected to white, thermal, and active noises in uniform a magnetic field. By deriving the corresponding Fokker Planck equation, analytical solutions for the joint probability density are obtained in different time domains.

The steady-state quantum dynamics of three dipole-dipole coupled two-level emitters, fixed at the vertices of an equilateral triangle, and interacting via the environmental thermostat is investigated. We have analytically obtained the populations of the involved three-atom cooperative states as well as of the second- and third-order spatial photon correlation functions of the light scattered by the few-qubit sample. As a consequence, we have demonstrated that this incoherently excited system spontaneously generates streams of single photons possessing sub-Poissonian photon statistics. In analogy to the dipole-dipole blockade, one may expect that at smaller inter particle distances, compared to the photon emission wavelength, the reported phenomenon has the same origin. However, we have shown that the quantum photon features are due to the interaction's nature of the few symmetrically arranged two-level emitters with the surrounding thermal reservoir. Respectively, at larger atomic intervals the effect occurs because of high-order spatial interference phenomena. Sub-wavelength interference fringes can be observed too, via measurements of spatial higher-order photon correlation functions.

First-passage times are often the most relevant aspect of a complex Markovian network, because they signify when information processing has resulted in a definite decision. Previous studies have shown that for kinetic proofreading networks in the limit of large network size the first-passage time distribution converges either to a delta or to an exponential distribution. Remarkably, these two forms correspond to the two extreme distributions of minimal and maximal entropy for a fixed mean, respectively. Here we build on the connection between first-passage times and graph theory to show that these two limits are not model-specific, but arise generically in Markovian networks from the distribution of the eigenvalues of the generator matrix. A deterministic peak emerges when infinitely many eigenvalues contribute, while the exponential limit arises from a single dominant eigenvalue. We also show that the exponential limit emerges robustly for reversible networks when a backward bias exists. In contrast, the deterministic limit is not obtained from a simple reversal of this condition, but under structurally tighter conditions, revealing a fundamental asymmetry between both regimes. Our theoretical analysis is illustrated and validated by computer simulations of one-step master equations and random networks.

We investigate how controlled foaming alters the mechanical dissipation of liquid crystalline elastomers (LCEs). Using thermal expandable microspheres, we generate homogeneous foams with precisely tuned bubble volume fractions up to 13% and compare their behaviour with non-mesogenic silicon analogues. We show that microsphere expansion induces a particle-centred mesogenic interphase arising from local elastic distortion and preferential alignment of mesogenic units at the inclusion surface. At low bubble volume fraction (0.5 to 5%), these interfaces remain spatially isolated and produce a pronounced no-monotonic enhancement of damping, with the loss factor reaching tan-delta=0.2 even in the isotropic regime. At higher loading, interphase overlaps and mechanical constraints suppress this effect, and the dissipation returns towards baseline elastomeric values. Large-strain tensile tests and impact experiments exhibit the same non-monotonic trend, demonstrating that low density LCE forms achieve the highest mechanical energy absorption per unit mass. Compared with conventional high porosity polymer foams used for acoustic damping, these materials retain sufficient mechanical integrity to sustain impact loads, establishing a microstructural route to engineer high-performance damping in soft solids.

Swift electrons from highly focused beams produced in aberration-corrected scanning transmission electron microscopes offer a powerful route for probing and manipulating matter at the nanoscale. Although linear momentum transfer from swift electrons to nanoparticles has been investigated theoretically and experimentally, subsequent analyzes revealed that several earlier predictions relied on non-causal dielectric functions or insufficient numerical convergence, leading to spurious sign reversals in the transferred momentum. Here, we derive analytical expressions and develop a numerically efficient electrodynamic framework to compute the linear momentum transferred from a swift electron to an isolated spherical nanoparticle described by a fully causal, local dielectric response. We apply our framework to large nanoparticles with 50 nm radius and explicitly resolve the spectral density of linear momentum transfer across the full frequency domain. Using causal dielectric functions for aluminum and bismuth, we analyze the role of electron velocity, impact parameter, and material-specific resonances. We find that, when causality and full multipolar convergence are enforced, the net transverse linear momentum transferred to spherical nanoparticles remains attractive toward the electron trajectory for all nanoparticles considered, despite the presence of material-dependent sign changes in individual electric and magnetic contributions. These results contrast with earlier theoretical predictions of net repulsive behavior and indicate that additional physical mechanisms beyond the present isolated, local description are required to account for experimentally observed repulsion. Our work establishes a robust reference framework for momentum transfer calculations and provides quantitative benchmarks relevant for electron-beam-based nanoscale manipulation.

Understanding the fine structure of excitons is crucial for optoelectronic and quantum photonic applications of lead halide perovskites. It is demonstrated that polarization-sensitive photon echo spectroscopy in magnetic field provides a powerful method to access coherent exciton dynamics and reveal their energy level structure, which is hidden by inhomogeneous broadening. Exciton quantum beats observed in both Faraday and Voigt geometries offer a precise probe of the energy splittings among the four 1$s$ exciton states, enabling determination of the fine structure and bright-dark splittings. Application of this technique to bulk mixed halide perovskite crystals FA$_{0.9}$Cs$_{0.1}$PbI$_{2.8}$Br$_{0.2}$ reveals a bright-dark exciton splitting of $Δ_\mathrm{X}=0.46~$meV, along with electron and hole Landé $g$ factors $g_\mathrm{e}=3.38$ and $g_\mathrm{h}=-1.14$, respectively. The quantum beats persist on timescales of 20--50$~$ps, demonstrating remarkably robust spin and optical coherences at cryogenic temperature of 2$~$K. The decay of the quantum beats of the outer doublet is governed by dephasing due to dispersion of the bright-dark splitting of $\sim0.06~$meV caused by localization potential fluctuations, while dephasing in the bright exciton inner doublet originates from a small zero field splitting of $\sim0.035~$meV due to anisotropic potentials.

We investigate periodic integrable Markov models, constructed from set-theoretical solutions of the Yang-Baxter equation. We first focus on the simplest class of solutions, called Lyubashenko solutions. We show that the resulting models are equivalent to some twisted Symmetric Simple Exclusion Process (SSEP), which are usual periodic SSEP models where a twist is added on a bond of the ring. We also provide various possible interpretations for these Markov models. Then, we study the long time dynamics of the twisted SSEP, characterising its different stationary states and counting them. Allowing the twist to vary, we examine the possible transitions between the different stationary states. Finally, we extend our construction of Markov models to set-theoretical solutions that are more general than Lyubashenko solutions and show that such models are not equivalent to a twisted SSEP in general.

We study two-dimensional systems of interacting excitons with a moat dispersion, for which the ground-state energy manifold presents a ring of discrete or continuously degenerate minima around a single point in momentum space. At low densities and for an idealized, perfectly degenerate moat, we show that the excitons undergo statistical transmutation and stabilize a chiral spin liquid. At higher densities, the moat dispersion favors Bose-Einstein condensation into states occupying multiple momenta, leading to inhomogeneous condensate phases and potentially supersolidity. We discuss the impact of band-structure warping present in realistic systems and argue that it generically stabilizes Bose-condensed phases over the chiral spin liquid, and analyze the superfluid response of the former which is unconventional due to the moat band. We also demonstrate that a proper renormalization of the exciton-exciton interaction is essential for describing these phases, and show that even purely repulsive interactions can favor inhomogeneous condensates. To further explore inhomogeneous condensate phases, we employ a Gross-Pitaevskii framework with a pseudopotential approximation and map out the resulting phase diagram. We show that the presence of degenerate dispersion minima can drive supersolidity already at weak coupling, in contrast to systems with a standard parabolic dispersion. Finally, we discuss our results in the context of real excitonic systems and argue that moat-band-induced supersolidity can be within experimental reach for realistic values of the model parameters.

Resonant laser excitation of a two-level system with subsequent single-photon emission can be used to generate single photons with high indistinguishability or Hong-Ou-Mandel (HOM) visibility. However, spectral overlap between excitation laser and emitted photons generally poses significant challenges. Furthermore, emitter re-excitation intrinsically limits achievable single-photon purity. Established solutions mitigate these issues at significant cost to source efficiency and with increased source complexity. This motivates the use of few-level systems with spectral separation of excitation and emission pathways. One option is a three-level cascade. However, without targeted lifetime engineering of emitting states, the cascade naturally limits achievable photon indistinguishability. Here we study a semiconductor quantum dot with resonant and selective cavity-enhancement of biexciton-to-exciton transition. Following resonant two-photon excitation of the biexciton state, we collect the emitted single photon with the cavity. This approach circumvents emitter re-excitation and naturally introduces spectral separation of excitation laser and emitted single photon. Supported by first experimental results, we demonstrate theoretically that with selective Purcell enhancement, the observed quality quantifiers of single-photon emission (purity, equivalently $g^{(2)}(0)$, and HOM visibility $\mathcal{V}$, equivalently indistinguishability) are competitive with respect to high-quality deterministic quantum-dot single-photon sources. This is already achieved without systematic optimization or targeted system engineering, which firmly places the reported approach as a viable route to the next generation of highest-quality quantum-dot based deterministic single-photon sources.

We present a general electromagnetic description for dipoles confined to surfaces with oblique dipole moment orientations, extending the conventional in-plane (IP) and out-of-plane (OOP) treatments. This description is useful for describing localized polarization in, \textit{e.g.}, van der Waals heterostructures, thin films of molecular aggregates, and metal-dielectric interfaces. The theory is suitable for any material with vanishingly thin thickness relative to the light wavelength, independent of the geometry of the material and the media interfacing it. We apply the formalism to a uniaxial excitonic sheet, covering a large number of two-dimensional (2D) materials and organic thin films. Our theory reveals pairs of polaritonic resonances originating from the IP and OOP components of the excitonic dipole moment. The formalism suggests experimentally accessible signatures of dipole moment orientation, enhanced by near-field probes. This work proposes a unified language for the description of 2D materials, thin films and interfaces with anisotropic dipolar responses.

Fluids composed of chiral active components can exhibit odd viscosity, a property that breaks time-reversal and parity symmetries. We investigate the hydrodynamic response to monopole and dipole singularities in a compressible thin fluid layer with odd viscosity, supported by a conventional lubrication layer. Using the two-dimensional Green's function in Fourier space, we derive exact analytical solutions for the flow and pressure fields. These solutions provide a detailed description of the hydrodynamic interactions governing the motion of colloidal particles and microswimmers in confined chiral fluids, offering insight into the role of odd viscosity in modifying particle dynamics and collective behavior. The derived results are directly applicable to modeling transport, control, and self-organization phenomena in active and chiral microfluidic systems.

Quasisymmetry (QS) provides a novel route to understand and control near-degeneracies, Berry curvature, optical selection rules, and symmetry-protected phenomena in quantum materials. Here we give physical interpretations of the emergence of QS operators across multiple material families. Using density functional theory and the $\mathbf{k}\cdot\mathbf{p}$ formalism, we identify QS subspaces and calculate their representation matrices, quantifying the quasisymmetry via a metric $ε$ that measures subspace invariance. For Sn/SiC and transition-metal dichalcogenide monolayers, QS corresponds to an emergent mirror symmetry, whereas in wurtzite crystals it manifests as an emergent spatial inversion. By contrast, for AgLa the QS appearing in avoided crossings is inherited from a nearby high-symmetry point rather than being an emergent lattice symmetry. Combining group-theoretical analysis and $\mathbf{k}\cdot\mathbf{p}$ modeling, our results establish concrete physical pictures for QS and provide practical criteria to diagnose it in first-principles calculations.

The ability to establish coherent communication channels is key for scaling up quantum devices. Here, we engineer interactions between distant polaritons, hybrid spin-photon excitations formed at different lumped-element superconducting resonators within a chip. The chip consists of several resonator pairs, slightly detuned in frequency to make them addressable, capacitively coupled within each pair and inductively coupled to a common readout line. They interact locally with samples of PTMr and Tripak$^{-}$ organic free radicals, deposited onto their inductors, which provide model $S = 1/2$, $g \simeq 2$ spin ensembles. Frequency-dependent microwave transmission experiments, performed at very low temperatures, measure polariton frequencies as a function of magnetic field in different scenarios. When only one resonator within a pair hosts a molecular sample, the results evidence that spins couple remotely to the empty LER as well as to the local cavity mode. If both resonators interact with a spin ensemble, the magnetic field tunes the polariton frequencies relative to each other, on account of the different spin-photon interactions at each LER. When polaritons are brought into mutual resonance, an avoided level crossing emerges that gives direct spectroscopic evidence for a coherent polariton-polariton interaction mediated by the circuit. Pump-probe experiments reveal that the excitation of a polariton within a connected pair is felt, thus it can be read out, by the other one. These observations, backed by model calculations, illustrate the control and detection of distant photon-photon and spin-spin correlations and entanglement in a scalable modular chip.

We construct a contour function for the logarithmic negativity and the logarithm of the moments of the partial transpose of the reduced density matrix for multimode bosonic Gaussian states of a free lattice model. In one spatial dimension, numerical results are obtained for harmonic chains either in the ground state or at finite temperature, by considering, respectively, either a subsystem made by two adjacent or disjoint blocks on the line or a bipartition of the circle. The contour function of the logarithmic negativity diverges only at the entangling points, while the contour function for the logarithm of the moments of the partial transpose is divergent also at the boundary of the bipartite subsystem, as functions of the position. In a two-dimensional conformal field theory, analytic expressions that describe these divergencies are discussed. In one spatial dimension, we explore the partial derivative of the logarithmic negativity of two adjacent intervals with respect to the logarithm of the harmonic ratio of their lengths while their ratio and the other parameters are kept fixed. Considering the ground state of the harmonic chain on the line and in the massive regime, we report numerical results showing that this quantity displays a monotonically decreasing behaviour.

Chiral fluids, for which the mobility tensor has antisymmetric, off-diagonal components, exhibit transport phenomena absent in conventional systems, including interaction-enhanced diffusion and negative mobility. While these effects have been predicted theoretically and observed in simulations, their microscopic origin has remained unclear. Here, we address this question using a mobility-based nonequilibrium approach, analysing the steady-state drift of a tracer driven through an interacting chiral fluid. We show that, under strong chirality, the tracer generates a reversed density wake, in which regions of particle accumulation and depletion are inverted compared to the achiral case. This structural inversion of the wake provides a unified physical mechanism underlying both enhanced diffusion and negative mobility. Furthermore, we demonstrate that these phenomena are robust to changes in the interaction potential, highlighting their generality as a consequence of odd mobility.

We demonstrate, by femtosecond time- and angle-resolved photoemission spectroscopy, that photoinduced interlayer charge transfer in a heterostructure consisting of Bernal-stacked bilayer graphene and a single atomic layer of silver on 6H-SiC(0001) transiently modulates the intrinsic potential landscape across the silver-graphene interface. This acts as an ultrafast optoelectronic gate that drives momentum-dependent band renormalizations, resulting in a transient band-gap opening on femtosecond timescales. Simultaneously, the photogenerated hot-carrier population enhances electronic screening, leading to subsequent closing of the band-gap beyond the thermal equilibrium value. These findings reveal two different mechanisms for photoinduced, reversible control of the electronic band structure in bilayer graphene -- interlayer charge transfer and hot-carrier-enhanced screening -- providing a general framework for the ultrafast control of electronic properties in graphene-based heterostructures. This opens up novel pathways for the realization of ultrafast optoelectronic devices and the exploration of correlated quantum phases in bilayer graphene under non-equilibrium conditions.

Atomically thin metallenes have emerged as a new member of the two-dimensional (2D) materials family. Recent experimental realization of metallenes in the Ångström limit has further intensified interest in this class of 2D materials. However, achieving sub-atomic insight into them demands the most detailed and systematic characterization of their electronic structure. Such understanding is essential for the rational design and exploitation of their properties in plasmonics, catalysis, and quantum optics. Existing electronic-structure studies are either scattered or focus on a few selected systems, and a comprehensive view of their band structures and Fermi surfaces remains missing. Here, we address this gap by studying 45 elemental metallenes in six monolayer lattices (honeycomb, square, hexagonal, and their buckled forms) using density-functional theory. We found that lattice type primarily fixes the shape and radial placement of the Fermi-lines, while out-of-plane buckling introduces controlled modifications: it shortens long straight Fermi-line segments, and occasionally creates, removes, or merges small Fermi-line pockets. The electronic configuration determines which orbital type dominates the Fermi level. We summarized Fermiology using a single score for each element, termed pocketness, derived from four descriptors that combine element properties (symmetry, coordination) with electronic characteristics (dispersion, Fermi-surface topology). This score enables targeted angle-resolved photoemission spectroscopy (ARPES) tests, controlled Lifshitz transitions, and provides a predictive basis for transport and device applications.

A central challenge in cellular signal processing is understanding how biochemical networks perform reliably despite molecular noise. Traditionally, mutual information has been widely used to quantify signaling fidelity, capturing how well outputs discriminate distinct input states. However, it fails to capture whether the output also faithfully mirrors the statistical structure of the input, a property crucial in processes like morphogen patterning, dose-dependent signaling, and cellular communication. To address this gap, we introduce the 2-Wasserstein distance from optimal transport theory, which provides a geometric basis for comparing input and output distributions. In our proposed framework, we define mutual information as informational fidelity and the inverse of the 2-Wasserstein distance as geometric fidelity. Applying this dual-fidelity framework to canonical regulatory motifs under Gaussian channel approximation reveals a topology-dependent trade-off: coherent feed-forward loops can achieve high performance in both dimensions, whereas feedback architectures typically sacrifice informational fidelity to enhance geometric fidelity. We demonstrate that theoretical predictions of feedback regulation are well supported by experimental data from tumor necrosis factor signaling. Our results demonstrate that maximizing information alone is not always advantageous and that reliable signaling arises from balancing information transmission with the geometric aspects of signaling. Thus, our analysis establishes geometric fidelity as a fundamental, yet previously unrecognized, dimension of signaling fidelity. It also provides a quantitative, experimentally accessible framework for dissecting natural networks and designing task-specific synthetic circuits.

Spin relaxometry using fluorescent nanodiamonds (FNDs) has been applied successfully to sense numerous paramagnetic target molecules such as free radicals and metalloproteins. However, despite their high sensitivity, T1 spin relaxation measurements are often hampered by their slow acquisition speed. Here, we demonstrate a method that allows for real-time monitoring of paramagnetic chemical reactions. We demonstrate T1 spin relaxometry from thousands of FNDs using an optimised cuvette-based system integrating an avalanche photodiode operated in linear mode, and a fast, fieldprogrammable gate array (FPGA) for data collation. We demonstrate chemical monitoring of the reduction of Cu(II) to Cu(I) ions in-solution with a 15 second integration using an optimised T1 sensing protocol. Our method achieves more than two orders of magnitude speed up with an order of magnitude reduction in cost when compared with traditional techniques. With further technical improvements, we believe this in-solution method could be extended to sense the sub-second chemical kinetics of paramagnetic molecules in solution.

We develop a framework to simulate quantum field theories (QFTs) with boundaries in $(1+1)$-dimenmsional curved spacetimes by employing open spin systems. Building upon our previous work that established a mapping from spin systems to QFTs in periodic geometries, we extend the correspondence to systems with boundaries, where boundary conditions play a crucial role in shaping the dynamics. Focusing on Majorana fermions, we derive the allowed boundary conditions from the requirement of inner product conservation and formulate their realization in spin systems. The corresponding spin model is shown to reproduce boundary conditions of QFT accurately when a free function in the spin model is appropriately chosen. As an explicit demonstration, we analyze a flat spacetime example, comparing spectra, mode functions, and linear responses between the continuum and lattice descriptions. Our findings confirm that open spin systems can successfully replicate QFT dynamics with boundaries.

We report local conductivity imaging of edge states in monolayer graphene by millikelvin microwave impedance microscopy (MIM). At the charge-neutrality point, as the magnetic field increases, the local conductivity at the edge drops to zero more slowly than in the bulk. This behavior is consistent with the calculated spatial profile of the charge gap in the canted antiferromagnetic phase. For comparison, we also perform microwave imaging of integer quantum Hall states away from neutrality, which host dissipationless chiral edge channels. The evolution of the edge signal as a function of the bulk gap is fundamentally different between the Landau level filling factor $ν= 0$ and $|ν| \ge 1$ integer quantum Hall states, which can be qualitatively explained by numerical simulations and theoretical analysis. Our results provide a comprehensive microscopic picture of the edge and bulk states as the Fermi level moves across the unique Landau-level spectrum of graphene.

Gate-defined semiconductor quantum dots utilize fast electrical control to manipulate spin and charge states of individual electrons. Electrical pulse distortions can limit control fidelities but are difficult to measure at the device level. Here, we use detuning-axis pulsed spectroscopy to characterize baseband pulse distortions in a silicon double quantum-dot. We extract the gate-voltage impulse response and apply a digital pre-distortion filter to eliminate pulse distortions on timescales longer than 1~ns. With the pre-distortion, we reduce the frequency chirp of coherent exchange oscillations in a singlet-triplet qubit. Our results suggest a scalable and tuning-efficient method for characterizing pulse distortions in quantum-dot spin qubits.

The time-dependent diffusion spreadability $\mathcal{S}(t)$ is a powerful dynamical probe of the microstructure of two-phase heterogeneous media across length scales [Torquato, S., \emph{Phys. Rev. E.}, 104 054102 (2021)]. It has been shown that when the spectral density takes the power-law form $\tildeχ_{_V}(\mathbf{k})\sim |\mathbf{k}|^α$ as the wavenumber $|\mathbf{k}|$ tends to zero, the normalized excess spreadability $\mathscr{s}^{ex}(t)$ [proportional to $\mathcal{S}(\infty)-\mathcal{S}(t)$] scales as $\mathscr{s}^{ex}(t)\sim t^{-\frac{d+α}{2}}$ in the long-time limit $t\to\infty$, enabling one to determine the infinite-wavelength scaling exponent $α$. An algorithm that allows one to reliably extract the exponent $α$ from long-time spreadability data was previously devised [Wang, H., Torquato, S., \emph{Phys. Rev. Appl.}, 17 034022 (2022)]. In this paper, we further improve this procedure to obtain $α$ even more accurately by incorporating higher-order correction terms to the long-time asymptotics and by utilizing analyticity properties of $\tildeχ_{_V}(k)$ at the origin. We illustrate our procedure by analyzing hyperuniform ($α> 0$), typical nonhyperuniform ($α=0$), and antihyperuniform ($-d < α<0$) models of two-phase media. In addition, by combining the large-$t$ asymptotic expansion of $\mathscr{s}^{ex}(t)$ with the small-$t$ expansion, we have devised a two-point Padé approximant to approximate $\mathscr{s}^{ex}(t)$ for all $t$ with just a few parameters. Our findings facilitate the characterization of the microstructure of two-phase media across length scales as obtained from numerical spreadability data or experimental data obtained from NMR relaxation measurements. Our work can also be applied in the inverse design of two-phase microstructures with targeted spreadability behaviors.

Density waves conventionally describe the periodic modulation of charge or spin, yet the spatial modulation of electronic topology has remained elusive. Here, we report the discovery of a Berry-curvature density wave in the noncollinear antiferromagnet Mn3NiN with compensated spins. Using high-precision Sagnac Kerr microscopy, we directly image micrometer-scale modulations of the Berry curvature. These topological ripples exhibit orientations unpinned to the crystal lattice, forming a nematic phase that spontaneously breaks rotational symmetry. We attribute this instability to field-induced spatial variations of the spin texture driven by competing magnetic interactions. This discovery unveils a new class of collective order in spin-compensated magnets arising from the geometric phase of the wavefunction itself and offering a tunable degree of freedom for topological spintronics based on antiferromagnets and altermagnets.

Scale invariance is a central organizing principle in physics, underlying phenomena that range from critical behaviour in statistical mechanics to transport and chaos in nonlinear dynamical systems. Here we present a unified and physically motivated exploration of scaling concepts, emphasizing how invariance under rescaling transformations emerges across systems of increasing dynamical complexity. Rather than adopting a purely abstract approach, we combine simple geometrical constructions, analytical arguments, and prototypical dynamical models to build physical intuition. We begin with elementary, easily reproducible examples governed by a single control parameter, showing how power-law behaviour naturally arises when characteristic scales are absent. We then extend the discussion to nonlinear dynamical systems exhibiting local bifurcations, where two scaling variables control the relaxation toward stationary states. In this context, scaling invariance manifests through critical exponents, crossover phenomena, and critical slowing down, allowing systems of different dimensionality to be grouped into universality classes. Finally, we address continuous phase transitions in chaotic dynamical systems, including transitions from integrability to non-integrability and from bounded to unbounded diffusion. By drawing on concepts traditionally associated with statistical mechanics, such as order parameters, susceptibilities, symmetry breaking, elementary excitations, and topological defects, we show how these transitions can be interpreted within a coherent scaling framework. Taken together, the examples discussed here demonstrate that scaling invariance provides a unifying language for understanding structure, transport, and criticality in nonlinear systems, bridging deterministic dynamics and nonequilibrium statistical physics in a transparent and physically intuitive manner.

Scaling laws in biological neural networks have long been investigated. From 1/f noise to neuronal avalanches, evidence of scaling in brain activity has been increasingly linked to tuning to or near criticality. The concept of scaling is intimately related to the renormalization group (RG), in essence providing coarse-grained, simplified descriptions that generalize to classes of diverse physical systems. Following the RG idea, a coarse-graining scheme has recently been proposed for populations of real neurons, and scaling behaviors in collective quantities have been reported in the hippocampus and in different areas of the rat cortex. To bridge the gap between neuronal population scales and species, here we consider large-scale, electrophysiological recordings of human brain activity in the awake resting-state. We demonstrate robust scaling behaviors of collective dynamics across coarse-graining scales, with exponents close to those measured in populations of spiking neurons. Further, we show that dynamics of neuronal avalanches, scale-free cascades of neural activity, are invariant under the proposed coarse-graining approach. Simulations of a non-equilibrium adaptive Ising model inferred from data and apt to reproduce a large repertoire of resting-state brain dynamics indicate that the scaling behaviors of the resting human brain activity emerge close to criticality and depend on the excitation/inhibition (E/I) balance of the network. While extending the range of validity of previous observations at small spatial scales and pointing to common scaling laws in mammals, the results open the way to a robust (currently missing) non-invasive approach to estimate the E/I balance, a key quantity in neuroscience research.

Boundary time crystals (BTCs) break time-translation symmetry and exhibit long-lived, robust oscillations insensitive to initial conditions. We show that collective spin BTCs can admit emergent topological winding numbers in operator space. Expanding the density operator in a spherical tensor basis, we map the Lindblad dynamics onto an effective local hopping problem, where collective degrees of freedom label sites of an emergent two-dimensional operator space lattice and identify topological obstructions that enforce the delocalization of operator modes on the lattice. The resulting spectral delocalization provides a natural explanation for the robust oscillatory dynamics observed in BTCs. When combined with non-reciprocal transport of operator weight across operator space, this mechanism moreover also leads to the universality of long-time dynamics across a broad class of initial states. Our results frame BTC dynamics as a form of topologically constrained operator space transport and suggest a close connection to non-Hermitian skin-effects.

We investigate how quantum localization encodes the hierarchical stickiness that governs transport in mixed classical phase spaces. Using the periodically driven kicked top, we show that the participation ratio (PR) of coherent states in the Floquet eigenbasis resolves the same layered structure that appears classically as a multimodal distribution of finite-time Lyapunov exponents (FTLEs). To establish a quantitative correspondence, we introduce a Gaussian coarse graining of the FTLE matched to the intrinsic semiclassical resolution of coherent states. Both local correlations and global comparisons of probability distributions demonstrate that quantum and classical indicators agree optimally within a finite window of evolution times, where sticky structures are most clearly resolved. Our results promote the participation ratio from a global measure of chaos to a sensitive probe of hierarchical transport and provide a practical route for diagnosing anomalous localization in driven quantum systems.

Accurate state estimation is a crucial requirement for the reliable operation and control of electric power systems. Here, we construct a data-driven, numerical method to infer missing power load values in large-scale power grids. Given partial observations of power demands, the method estimates the operational state using a linear regression algorithm, exploiting statistical correlations within synthetic training datasets. We evaluate the performance of the method on three synthetic transmission grid test systems. Numerical experiments demonstrate the high accuracy achieved by the method in reconstructing missing demand values under various operating conditions. We further apply the method to real data for the transmission power grid of Switzerland. Despite the restricted number of observations in this dataset, the method infers missing power loads rather accurately. Furthermore, Newton-Raphson power flow solutions show that deviations between true and inferred values for power loads result in smaller deviations between true and inferred values for flows on power lines. This ensures that the estimated operational state correctly captures potential line contingencies. Overall, our results indicate that simple data-based regression techniques can provide an efficient and reliable alternative for state estimation in modern power grids.

Adaptive coupling in networks of interacting neurons has gained recent attention due to the many applications both in biological and in artificial neural networks, where adaptive coupling or synaptic plasticity is considered as a key factor in learning processes. In the present study, we apply adaptive connectivity rules in networks of interacting FitzHugh-Nagumo oscillators. Adaptive coupling, here, is realized via Hebbian learning adjusted by the Oja rule to prevent the network link weights from growing without bounds. Numerical investigations demonstrate that during the adaptation process the FitzHugh-Nagumo network undergoes adaptive transitions realizing traveling waves, synchronized states and chimera states transiting through various multiplicities. These transitions become more evident when the time scales governing the coupling dynamics are much slower than the ones governing the nodal dynamics (nodal potentials). Namely, when the coupling time scales are slow, the network has the time to realize and demonstrate different synchronization regimes before reaching the final steady state. The transitions can be observed not only in the spacetime plots but also in the abrupt changes of the average coupling weights as the network evolves in time. Regarding the asymptotic coupling distributions, we show that the limiting average coupling strength follows an inverse power law with respect to the Oja parameter (also called "forgetting" parameter) which balances the learning growth. We also report abrupt transitions in the asymptotic coupling strengths when the parameter related to adaptive coupling crosses from fast to slow time scales. These findings are in line with previous studies on spiking neural networks.

We consider the Lie-algebraic notion of commutant in the setting of Poisson algebra. This provides a framework for deforming Hamiltonian differential equations. By taking a subalgebra of the algebra of integrals, and considering the set of functions that Poisson commute with that subalgebra, the Hamiltonian can be deformed, while retaining integrability. We deform Liouville integrable and superintegrable Lotka-Volterra systems studied in [19]. We present different explicit constructions considering Abelian and non-Abelian subalgebras of integrals. We obtain superintegrable systems for specific dimensions, and in arbitrary dimension. Polynomial systems are deformed to rational systems, some of which have non-rational integrals. Superintegrability seems to be preserved in this approach.

In the present work, we analyzed theoretically and experimentally the nonlinear dynamics of a magnetic pendulum excited through the interactions of a strong neodymium magnet and two coils placed symmetrically around the zero angular position. The forces between the magnet and coils and generated torques acting on the pendulum are derived using the magnetic charges interaction model and an experimentally fitted model. System equilibrium points are obtained, and their stability is investigated. It is found that when the currents in two coils are negative, the shape of the mechanical potential is bistable. The bistable potential might be symmetric if the currents have the same values and asymmetric when they are different. Asymmetric bistable potential is observed when coil currents have different signs. However, in the case of positive coil currents, a symmetric tristable potential is detected when the currents are the same, and an asymmetric tristable potential takes place when the positive currents have different values. Considering the sinusoidal coil current signals, analytical calculations using the harmonic balance method and numerical simulations are carried out for this electric-magneto-mechanical system. The obtained results are shown in terms of frequency-response diagrams, displacement time series, and phase portraits. The two-parameter bifurcation diagrams are plotted showing the different dynamical behaviors considering the current amplitudes and frequency as the control parameters. Amplitude jumps, hysteresis, and multistability are also observed. Some phase portraits and the coexistence of attractors are obtained numerically and confirmed experimentally. A good agreement between the numerical simulation and experimental measurement is achieved.

We present a comprehensive discussion of a transition from integrability to non-integrability in an oval billiard with a static boundary. This transition is controlled by a deformation parameter $ε$, which modifies the boundary shape from circular, corresponding to $ε=0$ and an integrable dynamics, to oval for $ε\neq 0$, where non-integrability emerges. The deformation of the circular billiard gives rise to a chaotic layer that develops along a well-defined stripe in phase space. By introducing a set of transformations that isolate this chaotic stripe, we characterise the diffusive spreading of ensembles of trajectories and identify an observable, $ω_{rms,{\rm sat}}$, which plays the role of an order parameter for the transition. For small deformations, the saturation value of the diffusion obeys the scaling law $ω_{rms,{\rm sat}}\proptoε^{\tildeα}$, with a critical exponent $\tildeα=0.507(2)$, vanishing continuously as $ε\rightarrow 0$. The associated susceptibility, $χ=dω_{rms,{\rm sat}}/dε$, diverges in the same limit, signalling the presence of critical behavior analogous to that observed in second-order (continuous) phase transitions in statistical mechanics.

We report a dynamical phase transition from integrability to non-integrability in a simple oval-like billiard with boundary $R(θ)=1+ε\cos(pθ)$. For $ε=0$, the phase space is {\it foliated} by invariant curves corresponding to periodic or quasiperiodic motion, whereas for small $ε$ a thin chaotic layer separates rotational and librational trajectories. As $ε$ increases, this layer grows according to a well-defined scaling law whose chaotic dispersion follows $ω_{\rm rms,sat}\simε^{\tildeα}$, where the exponent $\tildeα$ coincides with those of the Fermi-Ulam model, periodically corrugated waveguides, and a family of discrete mappings, revealing a universal mechanism for the onset of chaos in weakly perturbed integrable systems. The deviation of the reflection angle in the billiard, $ω_{\rm rms,sat}$, acts as an order parameter: it vanishes continuously as $ε\to 0$, signalling an ordered (integrable) phase, while its susceptibility $χ=dω_{\rm rms,sat}/dε$ diverges, indicating a second-order phase transition. A symmetry breaking and an analytically solvable diffusion process complete the near-critical phenomenology. These results establish a unified framework for the emergence of chaos from integrability.

We consider a Bose-Einstein condensate (BEC) loaded into a one-dimensional optical cavity under the combined action of an external potential and atom-cavity coupling with mutually incommensurate periods. Such configuration enables the localization of matter waves even in the absence of two-body interactions. We study families of localized modes within the mean-field approximation for red and blue detunings from atomic and cavity resonances in relatively shallow quasiperiodic lattices, beyond the validity of the tight-binding approximation. The parameter regimes supporting localization of atomic wave packets are identified. The system exhibits two types of bistability manifested as distinct photon numbers under otherwise identical conditions. One type arises from the coexistence of multiple families of localized modes, typical of conservative nonlinear systems, while the other stems from the multivalued dependence of the families on system parameters, characteristic of systems exhibiting hysteresis. BEC in a cavity may also display pseudodegeneracy, understood as the existence of two distinct atomic-density distributions corresponding to the same atomic and photon numbers (although different chemical potentials). The stability of the localized modes is analyzed. It is shown that, owing to the strong impact of long-range interactions on stability, a two-localized-mode configuration can operate as an XOR logic gate.

Phase matching is essential for efficient energy transfer in nonlinear wave-mixing processes. Traditional methods, such as birefringent and quasi-phase matching, have remained conceptually unchanged since their discovery over 60 years ago, each posing inherent constraints and limitations. Here, we demonstrate the concept of geometric phase matching as a new paradigm for tunable nonlinear wave mixing, based on a multilayered platform of nonlinear thin-film crystals. We leverage this concept to experimentally show reconfigurable and spin-controlled phase matching for second-harmonic generation (SHG), opening new avenues for real-time manipulation of nonlinear interactions in photonic devices. We specifically demonstrate full modulation of SHG from a bilayer structure, nearly perfect and tunable geometric phase matching from an eight-layer structure, and polarization tomography that reveals the evolution of the spin dependent interaction. This approach not only expands the design space for nonlinear optical processes but also paves the way for highly robust, tunable and efficient frequency conversion, for next-generation adaptive nonlinear photonic, quantum photonic and nonlinear optical metamaterial technologies based on thin-film crystals.

The transition matrix (T-matrix) is a complete description of an object's linear scattering response. As such, it has found wide adoption for the theoretical and computational description of multiple-scattering phenomena. In its original form, the T-matrix describes the interaction of a scatterer with a monochromatic source. In practice, however, information about the T-matrix is usually needed in an extended spectral domain. To access the frequency-dispersion, one might naively sample T-matrices over a finely resolved set of discrete frequencies and store one T-matrix per frequency. This approach has multiple drawbacks: it is computationally expensive, requires excessive memory, and it disregards the physical origin of the spectral features, weakening physical interpretability. To overcome these major limitations, we leverage a pole-expansion technique to represent the T-matrix with arbitrary frequency resolution within a selected frequency domain via a set of resonant contributions. A matrix-valued variant of the recently established adaptive Antoulas-Anderson (AAA) algorithm for rational approximation enables us to compute the pole-expansion at minimal computational cost using only a small number of direct evaluations. We demonstrate the benefits of such a representation with examples ranging from semi-analytically accessible scatterers to quasi-dual bound states in the continuum. To allow the wider community to capitalize on these findings, we provide open-source tools to perform the presented pole-expansion of the T-matrix.

Optimisation problems, which appear in numerous fields of science and industry, are challenging to solve even with modern supercomputers. Many such problems can be mapped onto ground-state searches of spin Hamiltonians, implemented on various physical platforms whose intrinsic dynamics are analogous to spin systems. However, the complex energy landscape of spin Hamiltonians often traps the system in local minima, preventing the system from reaching the ground-state (global minimum). We demonstrate an intrinsic feedback-driven annealing mechanism in class-B semiconductor laser arrays arising from the interplay of internal ($α$) and external ($η$) coupling. The instantaneous phase configuration self-modulates amplitude fluctuations, which act as an effective temperature, dynamically reshaping the potential and enabling the system to escape from local minima. Using a one-dimensional ring laser array, we analyze defect formation in the $α$-$η$ parameter space and identify an optimal regime achieving nearly 100% ground-state probability. Although both $α$ and $η$ are essential for the feedback loop, defect suppression results from modifying two competing timescales: amplitude stabilization (t_amp) and phase locking (t_phase), analogous to the Kibble-Zurek mechanism. These timescales can be tuned independently via $α$ or $η$. Identical timescale ratios yield identical defect probabilities, confirming that relative timescales, not specific parameters, govern defect formation. Our findings establish internal feedback-driven annealing as a practical route to ground-state search in semiconductor laser arrays, providing a foundation for efficient and scalable laser-based spin simulators for tackling hard optimization problems.

Magnetic flux (vortex) trapping remains a major obstacle to very large scale integration in superconducting electronics. Moats -- etched regions in circuit layers placed in ground planes and around critical circuitry -- offer a simple passive approach to sequester flux. Here, we systematically examine the effectiveness of moat arrays in superconducting niobium films as a function of geometry (size, shape, and density) and background magnetic field. By measuring the vortex expulsion field, we estimate the flux saturation number and flux trapping temperature for a range of geometries. We find that many moat designs effectively sequester flux in magnetically shielded environments (< 1 $μ$T), with high-aspect-ratio rectangular "slit" moats providing the strongest mitigation at minimal area cost. However, our measurements show that moats alone do not eliminate flux trapping in non-ideal films, as vortices can preferentially pin at material defects. These results provide design guidance for flux mitigation in superconducting integrated circuits and highlight the need for combined optimization of circuit geometries and materials.

We present a structured-illumination technique for full-field super-resolution transmission X-ray microscopy, which employs Fourier spectral decomposition inspired by established methods in visible-light microscopy. A 2D grating creating this illumination is stepped across one period to acquire a set of images at unique illumination positions. The Fourier domain of each image is described as a linear combination of replicated sample information at each frequency harmonic. As this superposition is created independently of detection, it contains spatial information exceeding native detector resolution. Recovering the encoded high-frequency components enables the population of an expanded frequency space. We demonstrate the presence of additional sample information in the Fourier spectrum and introduce a method to recover it. We achieve a resolution improvement by a factor of 2.1 for the projection image of a resolution test pattern. We further demonstrate seamless integration into standard X-ray tomography acquisition schemes. The acquisition is inherently multimodal, as phase-contrast and dark-field images can be computed from the same data using methods such as unified modulated pattern analysis, while providing an additional super-resolved transmission channel. These results indicate broad potential for non-destructive testing and biomedical imaging, as they alleviate pixel-size limitations in photon-counting detectors and sample-size restrictions imposed by optical magnification.

C.V. Raman (1888-1970) was one of the pioneering scientists to have emerged from India during the colonial era. His scientific explorations were driven by his curiosity to understand wave phenomena. Naturally, optics and related physical effects were at the heart of such an exploration. Apart from his Nobel prize-winning discovery of the Raman effect, his research included topics such as oblique diffraction, light scattering from liquids and amorphous solids, classical and quantum nature of light, acousto-optics, haloes and coronae (speckles), crystal dynamics and soft modes, optics of minerals, floral colors, physiology of vision and many other aspects related to light in natural settings. In this article, I give a historical overview of some of the work by C.V. Raman and his group that had a direct connection to optics and optical spectroscopy.

Photonic processors have emerged as an attractive platform for fast and energy-efficient matrix-vector multiplication. However, they are susceptible to error due to their analog nature. Here, we present an error-correction technique that implements a correction offset to the optical en-/decoders of photonic processors. Our proposed method is general-purpose, does not require introducing any additional components to the photonic network, and can address errors stemming from unbalanced losses, 50/50 beamsplitter deviations, digital-to-analog conversion inaccuracies, and any unknown sources. In particular, we show that our method is highly effective in mitigating unbalanced-loss errors, a problem that has not previously been addressed by any error-correction technique. Using this approach, we achieve over 90% error reduction in large triangular meshes, overcoming a key obstacle to highly accurate photonic processors for information processing.

Spontaneous Raman scattering (SpRS) is a weak non-linear effect, particularly relevant to classical-quantum coexistence transmission and sensing applications. In classical transmission, the relevant Raman effect is stimulated Raman scattering (SRS), and recent studies have examined it in space-division multiplexing (SDM) fibers. An intrinsic relation between SpRS and SRS allows previous SRS results to inform SpRS models. In this work, we extend SpRS models derived for single-mode fibers (SMFs) to SDM fibers with multiple mode groups of degenerate modes, covering both Stokes and anti-Stokes bands. The proposed model is a useful, fiber-design-independent tool for evaluating scattered noise in optical links, and it is validated through experimental measurements in field-deployed multi-core fibers (MCFs) and multi-mode fiber (MMF), showing good agreement.

Achieving near-unity fidelity in conventional continuous-variable quantum teleportation schemes based on two-mode squeezed vacuum states is fundamentally unattainable. To overcome this limitation, alternative approaches utilizing ensembles of two-dimensional entangled qubits have been proposed. In this work, we investigate continuous-variable quantum teleportation employing entangled qutrit resources under realistic noise effects. The results demonstrate that the proposed scheme performs well in both ideal and noisy conditions, enabling high-fidelity teleportation with a reasonable success probability.

Reservoir computing leverages nonlinear dynamics of physical systems to process temporal information with minimal training cost. Here, we demonstrate that cavity solitons sustained in a fiber optical cavity provide an optical platform for photonic reservoir computing. Our methodology exploits the use of a phase-modulated drive laser to encode the input, while the reservoir states are accessed through frequency-resolved readout. Numerical simulations indicate that the emission of Kelly waves enriches the dynamics and enhances performance for machine learning tasks. We evaluate the performance of the cavity-soliton reservoir computer on several standard benchmark tasks.

The diffusion of complex opinions is severely hindered in multilayer social networks by echo chambers and cognitive consistency mechanisms. We investigate Influence Maximization strategies within the 3-state multilayer q-voter model. Utilizing the mABCD benchmark, we simulate social environments ranging from integrated Open Worlds to segregated Fortress Worlds. Our results reveal a topological paradox that we term the "Fortress Trap". In highly modular networks, strategies maximizing local density such as Clique Influence Maximization (CIM) and k-Shell fail to trigger global cascades, creating isolated bunkers of consensus due to the Overkill Effect. Furthermore, we identify a Redundancy Trap in perfectly aligned Clan topologies, where the structural overlap of layers creates a "Perfect Prison," rendering it the most resistant environment to diffusion. We demonstrate that VoteRank, a strategy that prioritizes diversity of reach over local intensity, consistently outperforms structure-based methods. These findings suggest that, for complex contagion, maximizing topological entropy is more effective than reinforcing local clusters.

Full-wave numerical methods based on quasinormal modes (QNMs) offer valuable physical insights and computational efficiency for analyzing electromagnetic resonators. However, despite their advantages, many researchers in electromagnetism continue to favor real-frequency domain or time-domain approaches, often using finite element or finite-difference time-domain methods. This preference stems from various factors, including the perception that QNM theory is still developing or requires advanced mathematical tools from complex analysis. In this work, we combine numerical techniques with accurate ap-proximations to simplify the computation of QNMs and enable ultrafast reconstructions us-ing QNM expansions. The result is a new approach that is straightforwardly accessible to users familiar with real-frequency methods. We demonstrate the practicality of our ap-proach through an open-source package [Doi: 10.5281/zenodo.18708748] implemented within a widely-used commercial photonics software.

In orbital mechanics, Gauss's method for orbit determination (OD) is a popular, minimal assumption solution for obtaining the initial state estimate of a passing resident space object (RSO). Since much of the cislunar domain relies on three-body dynamics, a key assumption of Gauss's method is rendered incompatible, creating a need for a new, minimal assumption method for initial orbit determination (IOD). In this work, we present a framework for short and long term probabilistic target tracking in cislunar space which produces an initial state estimate with as few assumptions as possible. Specifically, we propose an IOD method involving the kinematic fitting of several series of noisy, consecutive ground-based observations. Once a probabilistic initial state estimate in the form of a particle cloud is formed, we apply the powerful Particle Gaussian Mixture (PGM) Filter to reduce the uncertainty of our state estimate over time. This combined IOD/OD framework is demonstrated for several classes of trajectories in cislunar space and compared to better-known filtering frameworks.

In heterogeneous disordered traffic, where various vehicle types operate without strict lane discipline, self-organized vehicle groups often emerge. While the formation of such groups has been recognized, their influence on macroscopic traffic dynamics remains unclear. This study investigates how the prevalence and composition of vehicle groups affect flow-density relationships in heterogeneous disordered traffic. Using trajectory data from real-world video observations, we apply three distinct Passenger Car Unit (PCU) estimation methods to construct flow-density diagrams that account for traffic heterogeneity. The analysis reveals that group proportions, i.e., the proportion of vehicles that are classified as belonging to groups, have a nonlinear and traffic-situation-dependent impact on flow characteristics. Specifically, moderate group proportions (30-60%) are associated with higher flow rates in medium- and high-density conditions, whereas proportions exceeding 50% correspond to skewed traffic distributions toward low- or high-density extremes. Comparisons between vehicle-count-based and PCU-based group proportions indicate that normalization methods significantly affect the interpretation of group dynamics, particularly when groups consist mainly of small-PCU vehicles such as motorcycles. Additionally, lower group proportions enhance flow under free-flow conditions, while the entropy-based analysis indicates that the association between entropy alone and speed is not consistently observed across traffic situations. By contrasting representative trends and extreme high-flow cases, the results further suggest that traffic under similar density and group-proportion conditions can exhibit low-efficiency and high-efficiency modes.

We investigate the utility of meta-optical encoders for generalizable image compression by leveraging their intrinsic shift-invariant point spread functions (PSFs). Compared with purely digital approaches, such optical encoders offer parallel and energy-efficient compression, enabling early data reduction prior to electronic processing and transmission, which is particularly attractive for resource-constrained and compact imaging systems. Although the operations realizable by a single passive optical layer remain fundamentally constrained, we systematically study several PSF encoding strategies combined with a total-variation (TV) digital reconstruction backend. Specifically, under identical compression ratios, we compare spatial binning, multi-channel random, and multi-channel orthogonal PSF based designs. Our results show that, at the same compression ratios, spatial binning achieves the highest reconstruction quality among all encoding strategies; however, it exhibits limited robustness to noise compared with multi-channel methods.

We present \textbf{warpax}, an open-source, GPU-accelerated Python toolkit for observer-robust energy condition analysis of warp drive spacetimes. Existing tools evaluate energy conditions for a finite sample of observer directions; \textbf{warpax} replaces discrete sampling with continuous, gradient-based optimization over the timelike observer manifold (rapidity and boost direction), backed by Hawking--Ellis algebraic classification. At Type~I stress-energy points, which comprise ${>}\,96$\% of all grid points across the tested metrics, an algebraic eigenvalue check determines energy-condition satisfaction \emph{exactly}, independent of any observer search or rapidity cap. At non-Type~I points the optimizer provides rapidity-capped diagnostics. Stress-energy tensors are computed from the ADM metric via forward-mode automatic differentiation, eliminating finite-difference truncation error. Geodesic integration with tidal-force and blueshift analysis is also included. We analyze five warp drive metrics (Alcubierre, Lentz, Van~Den~Broeck, Natário, Rodal) and one warp shell metric (used primarily as a numerical stress test). For the Rodal metric, the standard Eulerian-frame analysis misses violations at over $28\%$ of grid points (dominant energy condition) and over $15\%$ (weak energy condition). Even where the Eulerian frame identifies the correct violation set, observer optimization reveals that violation severity can be orders of magnitude larger (e.g.\ Alcubierre weak energy condition: ${\sim}\,90{,}000\times$ at rapidity cap $ζ_{\max} = 5$, scaling as $e^{2ζ_{\max}}$). These results demonstrate that single-frame evaluation can systematically underestimate both the spatial extent and the magnitude of energy condition violations in warp drive spacetimes. \textbf{warpax} is freely available at https://github.com/anindex/warpax.

Topological active materials have emerged as powerful paradigm bridging the discovery of exotic topological phases of matter with the development of functional topological devices. The recent extension of these material systems into dynamic regime, where topological properties can be actively manipulated at ultrafast timescales, promises unprecedented control over topological states and their functionalities. However, translating the static topological lasing signals into high-performance logic functions remain highly challenging, which imposes a far more stringent set of materials attributes. Here, leveraging the strong nonlinearity and pronounced spectral isolation of perovskite exciton-polaritons embedded in a Dirac vortex microcavity, we experimentally demonstrate the dynamic topological Majorana-like state polariton condensation with its ultrafast logic operations at room temperature. By actively coordinating pump and control beams in both spectral and temporal domain, we dynamically steer the topological polariton condensation process and demonstrate AND and NOT logic operations, achieving record extinction ratio (~20 dB), extremely low control fluence (~0.2 nJ/cm2) and sub-picosecond response time (~500 fs). Our results expand the frontier of dynamic topology and establish a novel pathway towards robust, ultrafast, and reconfigurable on-chip polaritonic logic circuits.

Air-to-air heat exchangers are vital for energy recovery and thermal management but often suffer from reduced effectiveness, high pressure losses, and increased pumping power in conventional designs. Advances in additive manufacturing have enabled nature-inspired geometries, such as lattice and triply periodic minimal surface (TPMS) structures, which enhance heat transfer through complex first-order surfaces but frequently cause excessive pressure drops. This study proposes an optimized higher-order harmonic heat-transfer surface tessellation developed through an optimization framework integrating analytical and numerical methods. The goal is to improve the overall thermal-hydraulic performance of the heat exchanger over a range of operating conditions. Results of sensitivity analysis show that secondary surface modification of this type can yield significant increase in the effectiveness reaching up to 70% although with associated increase in the pressure drop. The secondary surface wave frequency was found to be a more important control parameter than the amplitude in achieving high thermal-hydraulic performance. Additionally, we show that the optimized second order harmonic-type structure achieved relatively higher effectiveness and lower pressure-drop than the gyroid structure in the turbulent flow regime for Re>=7000. Although the gyroid TPMS structure had relatively higher effectiveness in the laminar and weakly turbulent flow regime, the associated pressure drop was found to be significantly higher than that of the harmonic-type structure.

Phase-only computer-generated holography (CGH) seeks a phase pattern for a spatial light modulator (SLM) whose propagated optical field reproduces a desired intensity distribution. In the far-field (Fraunhofer) regime, optical propagation reduces to a Fourier transform, such that each hologram pixel contributes to the entire reconstructed intensity distribution. When restricted to phase-only modulation, intensity must be shaped through global phase interference effects, making the inverse mapping from target intensity to phase highly non-linear and sensitive to local minima. We present a proof-of-concept physics-in-the-loop approach in which a transformer maps a target intensity image to a phase-only SLM field and is trained end-to-end through exact FFT-based propagation embedded directly within optimization. We further observe that patch tokenization strongly shapes the optimization geometry: coarse tokenization acts as an implicit spectral regularizer that stabilizes training and suppresses checkerboard-like attractors, while finer tokenization increases spatial degrees of freedom but benefits from curriculum or hierarchical refinement. Despite training on limited primitives and a single digit class (only digit 6), the learned generator exhibits out-of-distribution (OOD) generalization to unseen digits and hand-drawn target patterns. These results suggest that transformer architectures, whose self-attention enables global token interactions, are a natural fit for far-field holography and provide a viable foundation for scalable physics-grounded hologram generation.

This paper investigates a uncrewed aerial vehicles (UAV)-assisted cell-free architecture for vehicular networks in road-constrained environments. Roads are modeled using a Poisson Line Process (PLP), with multi-layer roadside access points (APs) deployed via 1-D Poisson Point Process (PPP). Each user forms a localized cell-free cluster by associating with the nearest AP in each layer along its corresponding road. This forms a road-constrained cell-free architecture. To enhance coverage, UAV act as an aerial tier, extending access from 1-D road-constrained layouts (embedded in 2-D) to 3-D. We employ a Matérn Hard-Core (MHC) point process to model the spatial distribution of UAV base stations, ensuring a minimum safety distance between them. In order to enable tractable analysis of the aggregate signal from multiple APs, a distance-based power control scheme is introduced. Leveraging tools from stochastic geometry, we have studied the coverage probability. Furthermore, we analyze the impact of key system parameters on coverage performance, providing useful insights into the deployment and optimization of UAV-assisted cell-free vehicular networks.

Robust cognitive radio development requires accurate 3D path loss models. Traditional empirical models often lack environment-awareness, while deep learning approaches are frequently constrained by the scarcity of large-scale training datasets. This work leverages the inherent sparsity of wireless propagation to model scenario-specific channels by identifying a discrete set of virtual signal sources. We propose a novel Greedy Sequential Sparse Bayesian Learning (GS-SBL) framework that bridges the gap between the computational efficiency of Orthogonal Matching Pursuit (OMP) and the robust uncertainty quantification of SBL. Unlike standard top-down SBL, which updates all source hyperparameters simultaneously, our approach employs a ``Micro-SBL'' architecture. We sequentially evaluate candidate source locations in isolation by executing localized, low-iteration SBL loops and selecting the source that minimizes the $L_2$ residual error. Once identified, the source and its corresponding power are added to the support set, and the process repeats on the signal residual to identify subsequent sources. Experimental results on real-world 3D propagation data demonstrate that the GS-SBL framework significantly outperforms OMP in terms of generalization. By utilizing SBL as a sequential source identifier rather than a global optimizer, the proposed method preserves Bayesian high-resolution accuracy while achieving the execution speeds necessary for real-time 3D path loss characterization.

Recent research has shown that unsourced massive access (UMA) is naturally well-suited for over-the-air computation (AirComp), as it does not require knowledge of each individual signal, as demonstrated by the massive digital AirComp (MD-AirComp) scheme proposed in prior work. The MD-AirComp scheme has proven effective in federated edge learning and is highly compatible with current digital wireless networks. However, it depends on channel pre-equalization, which may amplify computation errors in the presence of channel estimation inaccuracies, thus limiting its practical use. In this paper, we propose a blind MD-AirComp+ scheme, which takes advantage of the channel hardening effect in massive multiple-input multiple-output (MIMO) systems. We provide an upper bound on the computation mean square error, analyze the trade-off between computation accuracy and communication overhead, and determine the optimal quantization level. Additionally, we introduce a deep unfolding algorithm to reduce the computational complexity of solving the underdetermined detection problem formulated as a least absolute shrinkage and selection operator optimization problem. Simulation results confirm the effectiveness of the proposed MD-AirComp+ framework, the optimal quantization selection strategy, and the low-complexity detection algorithm.

Deploying multiple beyond diagonal reconfigurable intelligent surfaces (BD-RISs) can potentially improve the communication performance thanks to inter-element connections of each BD-RIS and inter-surface cooperative beamforming gain among BD-RISs. However, a major issue for multi-BD-RISassisted communication lies in the channel estimation overhead - the channel coefficients associated with the off-diagonal elements in each BD-RIS's scattering matrix as well as those associated with the reflection links among BD-RISs have to be estimated. In this paper, we propose an efficient channel estimation framework for double-BD-RIS-assisted multi-user multipleinput multiple-output (MIMO) systems. Specifically, we reveal that high-dimensional cascaded channels are characterized by five low-dimensional matrices by exploiting channel correlation properties. Based on this novel observation, in the ideal noiseless case, we develop a channel estimation scheme to recover these matrices sequentially and characterize the closed-form overhead required for perfect estimation as a function of the numbers of users and each BD-RIS's elements and channel ranks, which is with the same order as that in double-diagonal-RIS-aided communication systems. This exciting result implies the superiority of cooperative BD-RIS-aided communication over the diagonal- RIS counterpart even when channel estimation overhead is considered. We further extend the proposed scheme to practical noisy scenarios and provide extensive numerical simulations to validate its effectiveness.

This paper considers a millimeter wave (mmWave) integrated sensing and communication (ISAC) system, where a base station (BS) equipped with a large number of antennas but a small number of radio-frequency (RF) chains emits pencillike narrow beams for persistent tracking of multiple moving targets. Under this model, the tracking lost issue arising from the misalignment between the pencil-like beams and the true target positions is inevitable, especially when the trajectories of the targets are complex, and the conventional Kalman filter-based scheme does not work well. To deal with this issue, we propose a Transformer-based mmWave multi-target tracking framework, namely m3TrackFormer, with a novel re-acquisition mechanism, such that even if the echo signals from some targets are too weak to extract sensing information, we are able to re-acquire their locations quickly with small beam sweeping overhead. Specifically, the proposed framework can operate in two modes of normal tracking and target re-acquisition during the tracking procedure, depending on whether the tracking lost occurs. When all targets are hit by the swept beams, the framework works in the Normal Tracking Mode (N-Mode) with a Transformer encoder-based Normal Tracking Network (N-Net) to accurately estimate the positions of these targets and predict the swept beams in the next time block. While the tracking lost happens, the framework will switch to the Re-Acquisition Mode (R-Mode) with a Transformer decoder-based Re-Acquisition Network (RNet) to adjust the beam sweeping strategy for getting back the lost targets and maintaining the tracking of the remaining targets. Thanks to the ability of global trajectory feature extraction, the m3TrackFormer can achieve high beam prediction accuracy and quickly re-acquire the lost targets, compared with other tracking methods.

Malicious jamming presents a pervasive threat to the secure communications, where the challenge becomes increasingly severe due to the growing capability of the jammer allowing the adaptation to legitimate transmissions. This paper investigates the jamming mitigation by leveraging an active reconfigurable intelligent surface (ARIS), where the channel uncertainties are particularly addressed for robust anti-jamming design. Towards this issue, we adopt the Stackelberg game formulation to model the strategic interaction between the legitimate side and the adversary, acting as the leader and follower, respectively. We prove the existence of the game equilibrium and adopt the backward induction method for equilibrium analysis. We first derive the optimal jamming policy as the follower's best response, which is then incorporated into the legitimate-side optimization for robust anti-jamming design. We address the uncertainty issue and reformulate the legitimate-side problem by exploiting the error bounds to combat the worst-case jamming attacks. The problem is decomposed within a block successive upper bound minimization (BSUM) framework to tackle the power allocation, transceiving beamforming, and active reflection, respectively, which are iterated towards the robust jamming mitigation scheme. Simulation results are provided to demonstrate the effectiveness of the proposed scheme in protecting the legitimate transmissions under uncertainties, and the superior performance in terms of jamming mitigation as compared with the baselines.

This paper proposes a monostatic wideband system for integrated sensing and communication (ISAC) at millimeter-wave frequencies, based on multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM). The system operates in a hybrid near-/far-field regime. The transmitter (Tx) operates in the far field (FF) and uses low-complexity beam steering. The receiver (Rx), on the other hand, operates in a pervasive near field (NF), enabled by a very large effective array aperture. To enable a fully digital implementation, we introduce an extremely large antenna spacing (ELAS) design. This design attains the required aperture with only a few widely spaced antenna elements while avoiding grating lobes in the composite Tx-Rx response. We analytically characterize the NF range-angle response of this architecture and study the interplay between NF effects and waveform bandwidth. This leads to the definition of a super-resolution region, where NF propagation at the Rx dominates the achievable range resolution and surpasses the classical, bandwidth-limited resolution. As a case study, we consider an extended target modeled as a collection of scatterers and assess localization performance via maximum-likelihood estimation. Numerical results evaluated in terms of root mean square error (RMSE) and generalized optimal sub-pattern assignment (GOSPA) show that operating in NF conditions with the ELAS-based design yields significant gains compared to a conventional FF baseline at both the Tx and Rx.

This paper investigates the cell-free multi-user integrated sensing and communication (ISAC) system, where multiple base stations collaboratively track the users and detect their signals. Moreover, reconfigurable intelligent surfaces (RISs) are deployed to serve as additional reference nodes to overcome the line-of-sight blockage issue of mobile users for accomplishing seamless sensing. Due to the high-speed user mobility, the multi-user tracking and signal detection performance can be significantly deteriorated without elaborated online user kinematic state updating principles. To tackle this challenge, we first manage to establish a probabilistic signal model to comprehensively characterize the interdependencies among user states, transmit signals, and received signals during the tracking procedure. Based on the Bayesian problem formulation, we further propose a novel hybrid variational message passing (HVMP) algorithm to realize computationally efficient joint estimation of user states and transmit signals in an online manner, which integrates VMP and standard MP to derive the posterior probabilities of estimated variables. Furthermore, the Bayesian Cramer-Rao bound is provided to characterize the performance limit of the multi-user tracking problem, which is also utilized to optimize RIS phase profiles for tracking performance enhancement. Numerical results demonstrate that the proposed algorithm can significantly improve both tracking and signal detection performance over the representative Bayesian estimation counterparts.

To support sixth-generation (6G)-enabled intelligent transportation systems (ITSs), a multi-modal sensing residual-corrected graph neural network (MM-ResGNN) framework is proposed for millimeter-wave (mmWave) path loss prediction in vehicular communications for the first time. The propagation environment is formulated as an environment sensing path loss graph (ESPL-Graph), where nodes represent the transmitter (Tx) and receiver (Rx) entities and edges jointly describe Tx--Rx transmission links and Rx--Rx spatial correlation links. Meanwhile, a geometry-driven physical baseline is introduced to decouple deterministic attenuation trends from stochastic residual variations. A vehicular multi-modal path loss dataset (VMMPL) is constructed, which covers three representative scenarios, including the urban wide lane, urban crossroad, and suburban forking road environments, and achieves precise alignment between RGB images and global semantic information in the physical space, and link-level ray-tracing (RT)-based path loss data in the electromagnetic space. In MM-ResGNN, topology-aware graph representations and fine-grained visual semantics are synergistically integrated through a gated fusion mechanism to estimate the path loss residual relative to the physical baseline. Experimental results demonstrate that MM-ResGNN achieves significant improvements over empirical models and conventional data-driven baselines, with a normalized mean squared error (NMSE) of 0.0098, a mean absolute error (MAE) of 5.7991~dB, and a mean absolute percentage error (MAPE) of 5.0498\%. Furthermore, MM-ResGNN exhibits robust cross-scenario generalization through a few-shot fine-tuning strategy, enabling accurate path loss prediction in unseen vehicular environments with limited labeled data.

Short packets make channel learning expensive. In pilot-aided transmission (PAT), a non-negligible fraction of the packet is consumed by pilots, creating a direct pre-log loss and tightening the reliability margin needed for ultra-reliable low-latency communication. We propose a pilot-free polar-coded framework that replaces explicit pilots with \emph{coded pilots}. The message is carried by two polar-coded segments: a quadrature phase shift keying (QPSK) segment that is decodable without channel state information (CSI), and a higher-order quadrature amplitude modulation (QAM) segment that provides high spectral efficiency. The receiver employs \emph{hybrid decoding}: it first jointly infers CSI during successive-cancellation-based decoding of the QPSK segment by exploiting QPSK phase-rotation invariance together with polar frozen-bit constraints; the decoded QPSK symbols then act as \emph{implicit pilots} for coherent detection and decoding of the QAM segment. The split also makes rate adaptation practical by confining the symmetry/frozen-bit requirements for phase resolution to the QPSK segment, enabling puncturing and shortening without breaking the pilot-free mechanism. For multi-block fading, we optimize the split and code parameters via density evolution with Gaussian approximation (DEGA); for higher-order modulation, we use bit-interleaved coded modulation capacity approximation to obtain equivalent channel parameters. Incorporating channel-estimation error variance into the DEGA-based analysis, simulations over practical multi-block block-fading channels show gains up to $1.5$~dB over PAT in the short-blocklength regime.

Reconfigurable antennas (RAs) and movable antennas (MAs) have been recognized as promising technologies to enhance the performance of wireless communication and sensing systems by introducing additional degrees of freedom (DoFs) in tuning antenna radiation and/or placement. This paradigm shift from conventional non-reconfigurable/movable antennas offers tremendous new opportunities for realizing multi-functional, more adaptive, and efficient next-generation wireless networks. In this paper, we provide a comprehensive survey on the fundamentals, architectures, and applications of these two emerging antenna technologies. First, we provide a chronological overview of the parallel historical development of both RA and MA technologies. Next, we review and classify the state-of-the-art hardware architectures for implementing RAs and MAs, followed by a detailed comparison of their distinct mechanisms, performance metrics, and functionalities. Subsequently, we focus on various applications of RAs and MAs in wireless communication systems, analyzing their respective performance advantages and key design considerations such as mode selection, movement optimization, and channel acquisition. We also explore the significant roles of RAs and MAs in advancing wireless sensing and integrated sensing and communication (ISAC). Furthermore, we present numerical performance comparisons to illustrate the distinct characteristics and complementary advantages of RA and MA systems. Finally, we outline key challenges and identify promising future research directions to inspire further innovations in this burgeoning field.

Reliable odometry for legged robots without cameras or LiDAR remains challenging due to IMU drift and noisy joint velocity sensing. This paper presents a purely proprioceptive state estimator that uses only IMU and motor measurements to jointly estimate body pose and velocity, with a unified formulation applicable to biped, quadruped, and wheel-legged robots. The key idea is to treat each contacting leg as a kinematic anchor: joint-torque--based foot wrench estimation selects reliable contacts, and the corresponding footfall positions provide intermittent world-frame constraints that suppress long-term drift. To prevent elevation drift during extended traversal, we introduce a lightweight height clustering and time-decay correction that snaps newly recorded footfall heights to previously observed support planes. To improve foot velocity observations under encoder quantization, we apply an inverse-kinematics cubature Kalman filter that directly filters foot-end velocities from joint angles and velocities. The implementation further mitigates yaw drift through multi-contact geometric consistency and degrades gracefully to a kinematics-derived heading reference when IMU yaw constraints are unavailable or unreliable. We evaluate the method on four quadruped platforms (three Astrall robots and a Unitree Go2 EDU) using closed-loop trajectories. On Astrall point-foot robot~A, a $\sim$200\,m horizontal loop and a $\sim$15\,m vertical loop return with 0.1638\,m and 0.219\,m error, respectively; on wheel-legged robot~B, the corresponding errors are 0.2264\,m and 0.199\,m. On wheel-legged robot~C, a $\sim$700\,m horizontal loop yields 7.68\,m error and a $\sim$20\,m vertical loop yields 0.540\,m error. Unitree Go2 EDU closes a $\sim$120\,m horizontal loop with 2.2138\,m error and a $\sim$8\,m vertical loop with less than 0.1\,m vertical error. github.com/ShineMinxing/Ros2Go2Estimator.git

To enable fully automated driving of trains, numerous new technological components must be introduced into the railway system. Tasks that are nowadays carried out by the operating stuff, need to be taken over by automatic systems. Therefore, equipment for automatic train operation and observing the environment is needed. Here, an important task is the detection of collisions, including both (1) collisions with the front of the train as well as (2) collisions with the wheel, corresponding to an driving-over event. Technologies for detecting the driving-over events are barely investigated nowadays. Therefore, detailed driving-over experiments were performed to gather knowledge for fully automated rail operations, using a variety of objects made from steel, wood, stone and bones. Based on the captured test data, three methods were developed to detect driving-over events automatically. The first method is based on convolutional neural networks and the other two methods are classical threshold-based approaches. The neural network based approach provides an mean accuracy of 99.6% while the classical approaches show 85% and 88.6%, respectively.

This paper presents virtual upmixing of steering vectors captured by a fewer-channel spherical microphone array. This challenge has conventionally been addressed by recovering the directions and signals of sound sources from first-order ambisonics (FOA) data, and then rendering the higher-order ambisonics (HOA) data using a physics-based acoustic simulator. This approach, however, struggles to handle the mutual dependency between the spatial directivity of source estimation and the spatial resolution of FOA ambisonics data. Our method, named SIRUP, employs a latent diffusion model architecture. Specifically, a variational autoencoder (VAE) is used to learn a compact encoding of the HOA data in a latent space and a diffusion model is then trained to generate the HOA embeddings, conditioned by the FOA data. Experimental results showed that SIRUP achieved a significant improvement compared to FOA systems for steering vector upmixing, source localization, and speech denoising.

This paper investigates whether large-scale GNSS spoofing activity can be inferred from maritime Automatic Identification System (AIS) position reports. A data-processing framework, called SeaSpoofFinder, available here: seaspooffinder.github.io/ais_data, was developed to ingest and post-process global AIS streams and to detect candidate anomalies through a two-stage procedure. In Stage 1, implausible position jumps are identified using kinematic and data-quality filters; in Stage 2, events are retained only when multiple vessels exhibit spatially consistent source and target clustering, thereby reducing false positives from single-vessel artifacts. The resulting final potential spoofing events (FPSEs) reveal recurrent patterns in several regions, including the Baltic Sea, the Black Sea, Murmansk, Moscow, and the Haifa area, with affected footprints that can span large maritime areas. The analysis also highlights recurring non-spoofing artifacts (e.g., back-to-port jumps and data gaps) that can still pass heuristic filters in dense traffic regions. These results indicate that AIS-based monitoring can provide useful evidence for identifying and characterizing potential spoofing activity at scale, while emphasizing that AIS-only evidence does not provide definitive attribution.

We investigate qubit error bursts in 5- and 7-transmon processors of similar design, fabrication and packaging, but with different types of qubit Josephson junctions. Measurements for each are performed in two refrigerators to discern device-specific from refrigerator-dependent characteristics. The duration and rate of bursts are device specific but within the range of prior experiments and consistent with ionizing radiation. We observe two unforeseen signatures specifically in the processor with Dolan junctions. First, increasing the rate of $π$ pulsing in the detection scheme shortens the recovery time to equilibrium, which is explained by a quasiparticle pumping mechanism. The second signature is an anomalous time dependence in the burst rate: a surge happens days or weeks after cooldown, followed by a strong suppression that persists until thermal cycling.

The possibility of detecting solar oscillation modes using space-based gravitational-wave detectors has been investigated in the context of gravitational-wave interferometry, with Polnarev \cite{Polnarev:2009xf} demonstrating that low-frequency solar modes could, in principle, produce detectable signals in a LISA-type interferometer. Motivated by this work, I revisit the problem using current solar models, updated detector sensitivities, and improved theoretical and observational constraints on mode amplitudes. In this study, I compute the gravitational response of solar oscillation modes using standard solar models generated with \texttt{MESA}, and mode eigenfrequencies and eigenfunctions calculated with \texttt{GYRE}. I focus primarily on solar $g$ modes, evaluating their responses for degree $l=2$ and azimuthal orders $m=0$ and $m=2$. The analysis incorporates both the earlier proposed and the current updated LISA sensitivity curves, and I perform a comparative assessment with the TianQin mission in the relevant low-frequency band. To assess the robustness of the predicted signals, I estimate the gravitational responses using two different standard solar models based on the GS98 and AGSS09 abundance compilations. I find that the resulting signal responses are nearly identical for the two models, indicating that uncertainties in solar metallicity have a negligible impact on the detectability of solar $g$ modes by space-based interferometers.

Frustrated down conversion is a process in which a quantum superposition of emissions from two separate parametric down-conversion processes gives rise to observable interference. Depending on the phase relation between the probability amplitudes associated with emissions by the first and second crystal, the process can be enhanced or suppressed. This is achieved by aligning the setup so that the signal and idler modes from the first crystal are fed into the second and constitute its signal-idler modes. In Sci. Adv. 11, 1794 (2025), two-observer interwoven frustrated PDC processes produced interference effects based on path identity [Phys. Rev. Lett. 118, 080401 (2017)]. The signal and idler modes of source crystals I and II are arranged to fully overlap with the emission modes of crystals A and B, which serve as elements of measurement stations controlled by Alice and Bob. In the interwoven configuration, crystal A (B) receives the signal mode of crystal I (II) and the idler mode of crystal II (I), enabling interference between joint emission processes at the sources and at the measurement stations. It was conjectured that such interference may lead to new non-classical phenomena. In arXiv:2508.19207 it was shown that the process violates the standard Clauser-Horne Bell inequality without additional assumptions, provided suitable measurement settings are used. Here we extend the interference scheme to more than two measurement stations and demonstrate a violation of one of the WWWZB inequalities. This indicates that the proposed approach may provide a general method for revealing non-classicality in a range of phenomena discussed in [Rev. Mod. Phys. 94, 025007 (2022)]. We also present a GHZ/Hardy-type argument that further highlights the paradoxical character of the interference.

In this work, we introduce a general n-qubit formulation of control objectives that allows a control target to be specified in a diagonal frame, so that only the diagonal entries must be characterized, thus quadratically reducing the complexity of the cost functional in constrast to a full target matrix. We do so by representing any n-qubit unitary transformation as a diagonal phase map on the computational basis states, as they are naturally diagonalizable by unitarity. By using discrete derivative operators to analytically construct support-selective phase invariants, we enable to deterministically isolate and quantify any multi-qubit interactions encoded in the phase map. These phase invariants form a coordinate system for the formulation of specific control targets in terms of arbitrary desired multi-qubit interactions, without having to invert the diagonalization during the optimizatiion, solely relying on the experimentally accesible diagonal phases. To illustrate the framework, we synthesize two genuinely tripartite entangling gates, both, diagonal and non-diagonal. These are obtained with a single shaped microwave pulse, for a numerically simulated room-temperature nitrogen-vacancy center with a three qubit nuclear spin register, with durations of about a microsecond. These results represent a factor 10-100 reduction in operation time compared with the fastest existing NV-based entanglers that act on more than two qubits at once.

Deterministic single photon emission from a Rydberg ensemble coupled to an optical cavity requires high-fidelity preparation of collective single excitations. In such a setup imperfect Rydberg blockade can lead to unwanted double excitations, which degrade photon indistinguishability. In this work we adapt the Derivative Removal by Adiabatic Gate (DRAG) technique, originally developed for superconducting qubits, to shape optical pulses that suppress double excitations in this atomic platform. By combining analytical modeling with numerical optimization, DRAG provides an improvement over conventional sine-squared pulses. Further optimization of pulse duration and atomic ensemble size identifies a parameter regime, distinct from that used in [Nature Photonics 17, 688 (2023)], that enhances the single excitation probability from the previous theoretical benchmark of 77% to 91.9%, approaching the fundamental limits set by decoherence in the system. Benchmarking against GRAPE (Gradient Ascent Pulse Engineering) confirms that DRAG operates close to the optimal control limit, while maintaining smooth, experimentally feasible pulse shapes. These results demonstrate the effectiveness and cross platform adaptability of DRAG for a high-fidelity single photon source.

Recent advances in quantum photonics have enabled increasingly robust protocols in optical phase estimation, achieving precisions beyond the standard quantum limit and approaching the Heisenberg limit. While intrinsic losses hinder the realization of unconditional super-sensitivity, reaching quantum advantage, defined as sensitivity surpassing that of any classical counterpart with identical resources, remains achievable. Here we experimentally demonstrate such an advantage using a fully fibered Mach-Zehnder-type interferometer operating at telecom wavelengths, free of post-selection. The scheme relies on the conversion of polarization-entangled photon pairs, a degree of freedom commonly favored for experimental convenience, into energy-time entanglement, which is particularly well suited for scalable fiber-based sensors. All system imperfections, including asymmetric losses and detector inefficiencies, are accounted for in the Fisher information analysis, yielding a measured quantum advantage of 10%. This result highlights the practicality of compact, alignment-free quantum interferometers for real-world sensing applications.

Evaluating the reliability of noisy quantum circuits is essential for implementing quantum algorithms on noisy quantum devices. However, current quantum hardware exhibits diverse noise mechanisms whose compounded effects make accurate and efficient reliability evaluation challenging. While state fidelity is the most faithful indicator of circuit reliability, it is experimentally and computationally prohibitive to obtain. Alternative metrics, although easier to compute, often fail to accurately reflect circuit reliability, lack universality across circuit types, or offer limited interpretability. To address these challenges, we propose a fine-grained, scalable, and interpretable framework for efficient and accurate reliability evaluation of noisy quantum circuits. Our approach performs a state-independent analysis to model how circuit reliability progressively degrades during execution. We introduce the Noise Proxy Circuit (NPC), which removes all logical operations while preserving the complete sequence of noise channels, thereby providing an abstraction of cumulative noise effects. Based on the NPC, we define Proxy Fidelity, a reliability metric that quantifies both qubit-level and circuit-level reliability. We further develop an analytical algorithm to estimate Proxy Fidelity under depolarizing, thermal relaxation, and readout error channels. The proposed framework achieves fidelity-level reliability estimation while remaining execution-free, scalable, and interpretable. Experimental results show that our method accurately estimates circuit fidelity, with an average absolute difference (AAD) ranging from 0.031 to 0.069 across diverse circuits and devices.

The fragility of quantum coherence fundamentally limits the scalability of quantum technologies, as unavoidable environmental interactions induce decoherence and rapidly degrade quantum properties. The Quantum Zeno Effect offers a powerful route to suppress quantum evolution and protect coherence through frequent measurements, irrespective of the underlying dynamics. However, existing implementations require prior knowledge of the quantum state, severely restricting their applicability. Here we introduce a state- and dynamics-independent protection protocol embedding the system in a larger Hilbert space, temporarily swapping the quantum information from its original degree of freedom to a decoherence-free ancillary one. We experimentally validate the protocol on a quantum optical platform, demonstrating robust preservation of coherence and purity for arbitrary polarization qubits under decoherence, thereby enabling the universal safeguarding of unknown quantum states.

We consolidate coherence, athermality, and nonuniformity as sub-resources within an underlying quantum resource theory: instability. We formulate instability axiomatically as the transient information within a decaying physical system. Specifying a decay mechanism (e.g., dephasing, thermalization) recovers these familiar resources as specific manifestations of instability. We compute the one-shot distillation yield and dilution cost in various operational paradigms, and use them to pin down the extremal additive monotones. In the asymptotic regime, we show that all conversion rates are governed by a single additive monotone, and thereby we establish a universal second law for instability.

Stress distributions in the deuteron are related to form factors of the asymmetric energy-momentum tensor through three-dimensional Fourier transforms. There are eleven such form factors, which we calculate in an impulse approximation. We compare the obtained form factors to prior results for the six form factors that have been previously calculated. We then elaborate on the formalism for relating the form factors to internal distributions of mass, mass flux, momentum, stresses, and forces, and obtain results for all of these distributions. We obtain the principal stresses for the symmetric part of the stress tensor, and show that the antisymmetric part describes reorientation of fermion spin by torsion stress when the nucleon moves between the S- and D-waves. Force distributions in the nucleons depend on the so-called non-conserved form factors through the Cauchy momentum equation, and are non-radial owing to the presence of tensor forces and spin-orbit coupling.

We present general-relativistic radiation magnetohydrodynamics simulations of binary neutron star mergers performed with GR-Athena++. Neutrino transport is treated using a moment-based, energy-integrated scheme (M1), augmented by neutrino number density evolution (N0). Our implementation is validated through an extensive suite of standard tests and demonstrated to perform robustly under adaptive mesh refinement. As a first application, we simulate the gravitational collapse of a uniformly rotating, magnetized neutron star, demonstrating stable radiation evolution through apparent-horizon formation using a novel excision technique based on the tapering of state vector evolution inside the horizon. To further test robustness in highly dynamic environments, we apply our code to two demanding binary neutron star merger scenarios. We investigate a long-lived remnant with the DD2 equation of state, evolved with full general-relativistic magnetohydrodynamics and M1 neutrino transport. Following this, a gravitational collapse scenario with the SFHo equation of state is explored. We showcase long-term stable evolution on neutrino cooling time-scales, demonstrating robust handling of excision and stable evolution of the post-collapse accretion phase in three-dimensional mergers with magnetic fields and neutrino radiation.

We present updated constraints on cosmic string and superstring parameters derived from Cosmic Microwave Background (CMB) anisotropies. The constraints are obtained via Markov Chain Monte Carlo (MCMC) analyses of the full \textit{Planck} temperature and polarization data combined with the Atacama Cosmology Telescope (ACT) Data Release 6 (DR6). For ordinary cosmic strings, we constrain the string tension $Gμ$, the string wiggliness parameter $α$, and the self-chopping efficiency $\tilde{c}$. For cosmic superstrings, we constrain the fundamental string tension $Gμ_F$, the string coupling $g_s$, and a parameter $w$ describing the volume of the compact extra dimensions. In both cases, we find significantly tighter bounds on the string tension compared to previous analyses, obtaining $2σ$ upper limits of $Gμ< 3.66\times10^{-8}$ and $Gμ_F < 1.38\times10^{-8}$. We also discuss the significant prior-dependence of these results. The computational pipeline used in this work, including a modified version of \texttt{CAMB} capable of computing CMB anisotropies sourced by any active network described via unequal-time correlators, is released publicly as \texttt{CAMBactive} \cite{Raidal_CAMBactive_CAMB_extension_2026}.

We numerically investigate the dynamical evolution of spherically symmetric charge free wormholes. We concentrate on two specific examples, both of which exhibit wormhole expansion and wormhole collapse: the Ellis-Bronnikov wormhole, which is sourced by a real massless ghost scalar field, and the quantum corrected Schwarzschild black hole in semiclassical gravity (which has a wormhole structure and is not a true black hole), which is sourced by a renormalized energy-momentum tensor. Despite their very different sources, we demonstrate that the dynamics of these two wormholes are remarkably similar. Our analysis focuses on diagrams for the areal radius and components of the energy-momentum tensor. This work also serves as a review, offering a detailed description of how to perform a spherically symmetric dynamical evolution using double null coordinates as well as a review of the static solutions for our two examples.

We investigate the role of spatial geometry in controlling energy storage and work extraction in a non-Markovian quantum battery. The model consists of two identical two-level systems embedded in a structured waveguide environment, where one qubit acts as the charger and the other as the battery. The relative separation between the qubits introduces a geometry-dependent phase that governs collective interference effects and modulates.

Quantum metrology seeks to push the boundaries of measurement precision by harnessing quantum phenomena. Conventional methods often rely on maximally entangled resources, with states that are usually challenging to produce and sustain in practical setups. Here, we show that the maximally entangled constraint can be lifted by experimentally realizing a photonic two-qubit weighted graph state with an arbitrarily tunable graph weight. We use the generated state as a resource for quantum-enhanced phase sensing. We experimentally characterize the state and study its minimum estimator variance for two distinct local measurement bases as the graph weight varies from the maximally entangled to weakly entangled limit. We find excellent quantitative agreement with theoretical predictions, and observe a gain in precision beyond the classically attainable precision limit for graph weights substantially below the maximally entangled limit. This confirms that considerably less entanglement is required to achieve a quantum advantage. Albeit non-scalable in our test setup, this work represents the first experimental realization of weighted graph states with a tunable graph weight using linear optics. We expect more scalable versions of the model to be possible in an on-chip photonic platform.

We develop a self-consistent approach that provides an explicit potential for a Rydberg electron whose ionic core consists of a polarizable medium, typically realized with superfluid droplets. The electron's motion remains separable in spherical coordinates, but the radial force exerted by the droplet breaks degeneracy of the angular momentum states non-perturbatively. The ensuing electron spectrum reveals intriguing properties dependent on droplet size and electron excitation. Deviations of the polarizable medium from the continuous spherical distribution can be taken into account as a perturbation of this redefined Rydberg dynamics. We discuss specific but paradigmatic examples for superfluid helium and also propose a way to probe droplet properties including its possible crystallized fraction through stimulated transitions of the Rydberg electron.

Hong-Ou-Mandel (HOM) interference is a fundamental tool for assessing photon indistinguishability in quantum information processing. While the effect of chromatic dispersion on HOM interference has been widely studied, the interplay between dispersion and the finite detection window of realistic measurement devices remains under-explored. In this work, we demonstrate that the rectangular coincidence window inherent to modern time-tagging modules, which effectively acts as a temporal filter, breaks the standard dispersion cancellation condition and restores sensitivity to symmetric group velocity dispersion. We derive an analytical model for type-II SPDC processes that predicts a modification of the HOM dip shape, specifically the emergence of characteristic oscillations and dip broadening. We experimentally validate this theoretical framework using a ppKTP source and transmission through optical fibers of lengths up to 29 km. The experimental data show excellent agreement with the model, confirming the presence of window-induced oscillations and allowing for the precise extraction of the fiber dispersion parameter. These findings underscore the importance of accounting for finite timing resolution in the design and characterization of dispersive quantum communication links.

Clock synchronization is necessary for communication and distributed computing tasks. Previous schemes based on photon timing correlations use pulsed light or photon pairs for their strong timing correlations. In this work, we demonstrate successful synchronization of crystal clocks using weakly time-correlated photons of 180 ns coherence time from a bunched light source. A synchronization timing jitter of 10 ns is achieved over symmetric -102 dB optical channel loss between two parties, over a span of 25 hours. We also present a model that gives better estimates to the coherence peak finding success probabilities under low signal.

The study of highly charged electronic and muonic hydrogen-like ions, provides an intriguing way to probe the internal structure of their atomic nuclei. In this work, we use nuclear structure calculations to accurately calculate the hyperfine splitting of electronic and muonic hydrogen-like ions, focusing in particular on the incorporation of finite-volume corrections, such as Bohr-Weisskopf and Breit-Rosenthal, due to the penetration of the electron and muon wavefunction into the nuclear electric charge and magnetic dipole densities. These corrections are essential for refining our understanding of the nuclear magnetic dipole and electric quadrupole moments. Our simulations use a Skyrme-Hartree-Fock-BCS model known for its effectiveness in modeling well-deformed nuclei such as ${}^{159}\mathrm{Tb}^{64+}$ and ${}^{165}\mathrm{Ho}^{66+}$, with particular emphasis on ${}^{161,163}\mathrm{Dy}^{65+}$ isotopes. It can also be generalised to multi-electron ions by studying the hyperfine anomaly between two isotopes.

Universal control of bosonic degrees of freedom provides a hardware-efficient route for quantum information processing with high-dimensional systems. Bosonic circuit quantum electrodynamics (cQED), which leverages transmon ancillae to coherently control long-lived superconducting cavities, is well suited to this goal. However, the cavity transitions are nearly degenerate in the usual dispersive regime, which limits the direct addressability of individual excitation levels and increases the complexity of engineered gates. Here, we integrate an on-chip flux-control architecture with a long-lived bosonic memory housed in a 3D superconducting cavity to dynamically access resonant Jaynes-Cummings (JC) interactions, and realize efficient arbitrary rotations between any pair of Fock levels in the memory. This on-demand access to JC interactions offers a versatile toolbox for implementing robust Fock-basis qudits and harnessing the rich dynamics of high-dimensional bosonic elements for quantum information processing.

This work investigates the geometrical properties of self-gravitating $N$-body systems from the perspective established by Henri Poincaré and Albert Einstein concerning the operational nature of measured geometry. Utilizing recent numerical analyses of central configurations--special equilibrium solutions to the Newtonian $N$-body problem--we uncover systematic spatial variations in nearest-neighbor particle separations correlated with the radial distance from the system's center of mass. We argue that these variations reflect a context-dependent, emergent effective geometry shaped by gravitational interactions, in accordance with Poincaré's assertion that measured geometry depends on the forces influencing measuring devices, and Einstein's view that rods and clocks define physical geometry through their local dynamics. By revisiting these foundational insights within a modern computational framework, we provide evidence that geometry in self-gravitating Newtonian systems is not a fixed background, but an emergent construct arising from internal physical interactions.

The dynamical consequences of the non-Hermitian skin effect (NHSE) remain largely unexplored despite extensive studies of its static properties. Here we address this gap by applying quantum Liang information flow (QLIF) an inherently directional measure of causal influence to the nonHermitian Su Schrieffer Heeger model with non reciprocal hopping. Unlike symmetric correlation functions, QLIF directly captures the directional asymmetry characteristic of non reciprocal systems. We demonstrate a scissors effect where the asymmetry varies approximately linearly with the non-reciprocity parameter gamma for small gamma, and exhibits non-monotonic dependence on the skin length, with optimal asymmetry at moderate skin localization. The velocity ordering reveals NHSE-induced blocking of information flow against the skin direction. Three distinct temporal regimes emerge: light-cone-bounded spreading, gamma-dependent stabilization, and coherent oscillations. These results establish the first quantitative connection between static skin localization and directional information dynamics, offering new insights into information propagation in non-reciprocal quantum systems.

Defects in solid-state systems constitute a promising platform for the realization of deterministic quantum emitters. Among many candidate materials and emitters, point defects in hexagonal Boron Nitride (hBN) have recently emerged as particularly promising. In this work, we probe the coherence of an individual B center with a zero phonon line at 436 nm, under pulsed resonant excitation. We observe power-dependent Rabi oscillations up to 5π, demonstrating optical coherent control of the transition. We achieve an excellent single photon purity of 93% at π-pulse. Furthermore, we probe the coherence of the two-level system using Ramsey interferometry, revealing an inhomogeneous coherence time of T_2*=0.60 ns. These results establish B centers in hBN as viable candidates for triggered, coherent quantum emitters and represent an important step towards their integration into quantum photonic platforms.

The real-time evolution of strongly interacting matter remains a frontier of fundamental physics, as classical simulations are hampered by exponential Hilbert space growth and entanglement-driven bottlenecks in tensor networks. This study reports the quantum simulation of hadron dynamics within a $(1+1)$-dimensional SU(2) lattice gauge theory using a 156-qubit IBM superconducting processor. Leveraging a hardware-efficient Loop-String-Hadron (LSH) encoding, we simulate the dynamics of the physical degrees of freedom on a $60$-site lattice in the weak-coupling regime, as a crucial step toward the continuum limit. We successfully observe the light-cone propagation of a confined meson and internal oscillations indicative of early-time hadronic breathing modes. Notably, these high-fidelity results were obtained directly from the quantum data via a differential measurement protocol, together with measurement error mitigation, demonstrating a robust pathway for large-scale simulations even on noisy hardware. To validate the results, we benchmarked the quantum algorithm and outcome from the quantum processor against state-of-the-art approximated classical algorithms using CPU -- based on tensor network methods and Pauli propagation method, respectively. Furthermore, we provide a quantitative comparison demonstrating that as the system approaches the weak-coupling or the continuum limit, the quantum processor maintains a consistent structural robustness where classical tensor networks and Pauli propagation methods encounter an onset of exponential complexity or symmetry violations as an artifact of approximation in the algorithm. These results establish a scalable pathway for simulating non-Abelian dynamics on near-term quantum hardware and mark a critical step toward achieving a practical quantum advantage in high-energy physics.

After a concise overview of Einstein spacetimes of type II (or more special) in four and five dimensions, we summarize recent results in the six-dimensional case. We assume the optical matrix to be non-degenerate and ``generic'', and the Weyl tensor to fall off sufficiently rapidly at infinity. As it turns out, the most general metric is characterized by one discrete (normalized) and three continuous parameters, is of type D and belongs to the Kerr-Schild class. Its relation to the previously known Kerr-(A)dS and Kerr-NUT-(A)dS metrics is clarified.

Standard quantum state tomography assumes sufficient control of a system to measure an informationally complete set of observables. Dynamical quantum state tomography (DQST) presents an alternative: given a system with known dynamics and a single fixed observable, it almost always suffices to control only the time at which each i.i.d. copy of the system is measured. This work presents an analogous scheme for tomography of multi-mode Bosonic Gaussian states undergoing Gaussian evolution, using a fixed single-mode homodyne measurement and only assuming control of the time of measurement. I prove that the scheme enables tomography for all discrete homogenous Gaussian evolutions and Gaussian quantum dynamical semigroups except for a null set which includes unitary evolution. When the state is known to be pure, a smaller number of measurement times is shown to be sufficient.

We establish a new SU(N) symmetric model of interaction between the fields of four types: the multiplet of vector fields, which describes the so-called color aether, the multiplet of pseudoscalar fields, which is associated with the multi-component cosmic dark matter, the gauge and gravitational fields. The extended Lagrangian of the model contains a new constructive element, which is based on the covariant SU(N) symmetric divergence of the multiplet of vector fields; this new element, being the multiplet of scalars from the point of view of spacetime transformations and the color vector from the point of view of the SU(N) group space, gives us the possibility to formulate properly the multi-axionic extension of the Peccei-Quinn theory. The hypothesis of a spontaneous polarization of the multi-axionic color aether in the early Universe is presented. The set of self-consistent master equations of the model is derived. An application to cosmology is considered: the obtained master equations are solved for the truncated test model based on the Bianchi-I spacetime platform.

Mahadev [SIAM J. Comput. 2022] introduced the first protocol for classical verification of quantum computation based on the Learning-with-Errors (LWE) assumption, achieving a 4-message interactive scheme. This breakthrough naturally raised the question of whether fewer messages are possible in the plain model. Despite its importance, this question has remained unresolved. In this work, we prove that there is no quantum black-box reduction of non-interactive classical verification of quantum computation of $\textsf{QMA}$ to any falsifiable assumption. Here, "non-interactive" means that after an instance-independent setup, the protocol consists of a single message. This constitutes a strong negative result given that falsifiable assumptions cover almost all standard assumptions used in cryptography, including LWE. Our separation holds under the existence of a $\textsf{QMA} \text{-} \textsf{QCMA}$ gap problem. Essentially, these problems require a slightly stronger assumption than $\textsf{QMA}\neq \textsf{QCMA}$. To support the existence of such problems, we present a construction relative to a quantum unitary oracle.

We discuss the possibility of a simultaneous cosmic variation of two fundamental entities: the Newtonian gravitational coupling $G$ and the electron mass $m_e$. We show that this variation can account for the late-time cosmic acceleration without invoking a cosmological constant or an explicit dark-energy fluid. We compare the derived $m_e$ variation with laboratory bounds found from Quasar absorption Spectra. Our results indicate that late-time cosmic acceleration could be a manifestation of evolving fundamental couplings, establishing a direct bridge between precision tests of gravity, particle physics and the origin of cosmic acceleration.

We present a method to conduct both quantum state reconstruction and entanglement purification simultaneously that is advantageous in several respects over previous work in this direction, showing that the number of Bell pairs necessary to boot a quantum network can be significantly reduced compared to an existing method. The existing method requires at least $10^5$ Bell pairs for the state reconstruction phase to estimate that the state is of fidelity $0.99$ within the error range of $10^{-2}$, whereas our approach only requires around $2,841$ to be certain with $99.7\%$ of confidence that the estimated fidelity lies within $[0.99-0.01, 0.99+0.01]$. In addition, in our approach we can start with a lower fidelity Bell pair and purify it multiple times, estimating at the same time the resultant fidelity with guarantee of $99.7\%$ that the fidelity estimate lies within a certain range. Moreover, the existing method cannot correct both bit-flip and phase-flip errors at the same time and can only correct one of these, whereas our approach can correct both bit-flip and phase-flip errors simultaneously. This research produces numerical estimates for the number of Bell pairs actually needed to guarantee a certain threshold fidelity $F$. The research can support the functioning real-world quantum networking by providing the information of the time needed for the bootstrapping of a quantum network to finish.

Adiabatic quantum computation with Rydberg atoms provides a natural route for solving combinatorial optimization problems such as the maximum independent set (MIS). However, its performance is fundamentally limited by the reduction of the spectral gap with increasing system size and connectivity, which induces population leakage from the ground state during finite-time evolution. Here we introduce the Adjusted Detuning for Ground-Energy Leakage Blockade (ADGLB), a spectral-gap-guided schedule engineering method that modifies the laser detuning profile to suppress leakage without introducing additional Hamiltonian terms or iterative optimization loops. We experimentally benchmark ADGLB on a quasi-one-dimensional chain of $N=10$ atoms, and the MIS preparation probability increases substantially compared with the standard adiabatic schedule. Furthermore, we show that the schedule optimized for smaller instances can be directly applied to larger two-dimensional triangular lattices with $N=25$ and $N=37$. With a small heuristic offset, the method also remains effective for instances with higher hardness parameters. These findings demonstrate that spectral-gap-guided schedule engineering offers a scalable and hardware-efficient strategy for enhancing adiabatic quantum optimization on neutral-atom platforms.

The Hadamard product of two tensors in the tensor-train (TT) format is a fundamental operation across various applications, such as TT-based function multiplication for nonlinear differential equations or convolutions. However, conventional methods for computing this product typically scale as at least $\mathcal{O}(χ^4)$ with respect to the TT bond dimension (TT-rank) $χ$, creating a severe computational bottleneck in practice. By combining randomized tensor-train sketching with slice selection via interpolative decomposition, we introduce Recursive Sketched Interpolation (RSI), a ``scale product'' algorithm that computes the Hadamard product of TTs at a computational cost of $\mathcal{O}(χ^3)$. Benchmarks across various TT scenarios demonstrate that RSI offers superior scalability compared to traditional methods while maintaining comparable accuracy. We generalize RSI to compute more complex operations, including Hadamard products of multiple TTs and other element-wise nonlinear mappings, without increasing the complexity beyond $\mathcal{O}(χ^3)$.

Distributing qubits across quantum processing units (QPUs) connected by shared entanglement enables scaling beyond monolithic architectures. Hyperbolic Floquet codes use only weight-2 measurements and are good candidates for distributed quantum error correcting codes. We construct hyperbolic and semi-hyperbolic Floquet codes from $\{8,3\}$, $\{10,3\}$, and $\{12,3\}$ tessellations via the Wythoff kaleidoscopic construction with the Low-Index Normal Subgroups (LINS) algorithm and distribute them across QPUs via spectral bisection. The $\{10,3\}$ and $\{12,3\}$ families are new to hyperbolic Floquet codes. We simulate these distributed codes under four noise models: depolarizing, SDEM3, correlated EM3, and erasure. With depolarizing noise ($p_{\text{local}} = 0.03\%$), fine-grained codes achieve non-local pseudo-thresholds up to 3.0\% for $\{8,3\}$, 3.0\% for $\{10,3\}$, and 1.75\% for $\{12,3\}$. Correlated EM3 yields pseudo-thresholds up to 0.75\% for $\{8,3\}$, 0.75\% for $\{10,3\}$, and 0.50\% for $\{12,3\}$; crossing-based thresholds from same-$k$ families are ${\sim}1.75$--$2.9\%$ across all tessellations. Using the SDEM3 model, fine-grained codes achieve distributed pseudo-thresholds of 1.75\% for $\{8,3\}$, 1.25\% for $\{10,3\}$, and 1.00\% for $\{12,3\}$. Under erasure noise motivated by spin-optical architectures, thresholds at 1\% local loss are 35--40\% for $\{8,3\}$, 30--35\% for $\{10,3\}$, and 25--30\% for $\{12,3\}$.

A new generation of global communications technology has been emerging. These systems, which utilize established device technologies and quantum effect devices, require ultra-high speeds, low cost, and strong security. In recent years, global communication systems have faced various practical security challenges depending on their configurations, and research efforts are underway to address these issues. In particular, the issue of the security of physical layer security from microwave wireless systems to quantum optical communication systems is urgent problem. However, concepts of cryptographic schemes have also been diversifying. Typical examples are mathematical ciphers, the Wyner scheme and QKD. Then, the Y-00 protocol has recently emerged as a third pillar cryptographic technology in the optical quantum domain. These security principles differ significantly from one another. This makes it difficult for different fields to understand each other. At this stage, comparative explanations of the security principles underlying these various cryptographic technologies are likely to promote mutual understanding among researchers across different fields. As the first trial, this lecture note explains the security mechanism of the third pillar (Y-00), comparing it with the principles of other mechanisms.

We investigate the cosmological dynamics of FHDE within $f(T)$ gravity by employing the dynamical system approach in a spatially flat FRW background. By introducing appropriate dimensionless variables, the field equations are reformulated as a closed system, which allows a systematic phase-space analysis. The resulting system admits four critical points, including two saddle points corresponding to radiation and matter-dominated epochs, and two stable points associated with a DE-dominated phase and a de Sitter solution. The radiation- and matter-dominated critical points are found to possess a saddle character in phase space, ensuring their transient nature and enabling the cosmological evolution to naturally progress toward a stable late-time accelerated attractor. The stable critical points describe accelerated expansion with effective equations of state compatible with DE and de Sitter regimes. Overall, the analysis indicates that $f(T)$ gravity is capable of reproducing the standard cosmological sequence within a consistent dynamical framework.

We report an updated constraint on the Solar System quadrupole parameter $Q_2$, which encodes the external field effect predicted by modified gravity versions of the MOND paradigm. Using the dataset employed to compute the DE440 planetary ephemerides, and estimating it simultaneously with other parameters included in the planetary ephemerides, we find $Q_2 = (1.6 \pm 1.8) \times 10^{-27}\,\mathrm{s}^{-2}$ (1-$σ$), representing an improvement of 40% over previous estimates. We also show explicitly that the contribution to the MOND prediction of $Q_2$ from the Solar System's largest planet, Jupiter, is at the 0.05% level, validating the approximation of retaining only the Sun in theoretical calculations. With this new constraint on $Q_2$, we update previously acknowledged tensions with external galaxy rotation curves, now leading to discrepancies at the $3$-$15σ$ level depending on the detailed mass modeling or the subset of galaxies considered. Within the Milky Way itself, the $Q_2$ constraint imposes an upper bound of only 2% (at 95% confidence) on the MOND boost to the galactic radial acceleration (i.e., the ratio of the observed over baryonic Newtonian acceleration) at the position of the Sun, in strong tension with current observational limits. The updated $Q_2$ posterior finally confirms that Solar System measurements provide stronger constraints than current wide-binary data on classical modified gravity versions of MOND.

Quantum metrology of an incoherent signal is a canonical sensing problem related to superresolution and noise spectroscopy. We show that quantum computing can accelerate searches for a weak incoherent signal when the signal and noise are not precisely known. In particular, we consider weak Schur sampling, density matrix exponentiation, and quantum signal processing for testing the rank, purity, and spectral gap of the unknown quantum state to detect the incoherent signal. We show that these algorithms are faster than full-state tomography, which scales with the dimension of the Hilbert space. We apply our results to detecting exoplanets, stochastic gravitational waves, ultralight dark matter, geontropic quantum gravity, and Pauli noise.

Quantum simulation of many-body systems offers a powerful approach to exploring collective quantum dynamics beyond classical computational reach. Although spin and fermionic models have been extensively simulated on digital quantum computers, the simulation of bosonic systems on programmable quantum processors is often hindered by the intrinsically large Hilbert space of bosonic modes. In this work, we study the digital quantum simulation of bosonic modes using the Holstein-Primakoff (HP) transformation and implement this protocol on a cloud-based superconducting quantum processor. Two representative models are realized on quantum hardware: (i) the driven harmonic oscillator and (ii) the Jaynes-Cummings model. Using data obtained from the quantum simulations, we systematically examine the interplay between algorithmic and hardware-induced errors to identify optimal simulation parameters. The dominant algorithmic errors arise from the finite number of qubits used in the HP mapping and the finite number of Trotter steps in the time evolution, while hardware errors mainly originate from gate infidelity, decoherence, and readout errors. This study advances the digital quantum simulation of many-body systems involving bosonic degrees of freedom on currently available cloud quantum processors and provides a framework that can be extended to more complex spin-boson and multimode cavity models.

Quantum information processing relies on a variety of resources, including entanglement, coherence, non-Gaussianity, and magic. In realistic settings, protocols run on networks of parties with heterogeneous local resource constraints, so different resources coexist and interact. Yet, resource theories have mostly treated each resource in isolation, and a general theory for manipulation in such distributed settings has been lacking. We develop a unified framework for composite quantum resource theories that describes distributed networks of locally constrained parties. We formulate natural axioms a composite theory should satisfy to respect the local structure, and from these axioms derive fundamental bounds on resource manipulation that hold universally, independent of the particular network characteristics. We apply our results to central operational tasks, including resource conversion and assisted distillation, and introduce new methods to construct new resource monotones from this setup. Our framework further reveals previously unexplored phenomena in the remote certification of quantum resources. Together, these results establish foundational laws for distributed quantum resource manipulation across diverse physical platforms.

We consider the quantum Zeno dynamics arising from monitoring a time-dependent projector. Starting from a stroboscopic measurement protocol, it is shown that the effective Hamiltonian for Zeno dynamics involves a nonadiabatic geometric connection that takes the form of the Kato-Avron Hamiltonian for parallel transport, stirring the evolution within the time-dependent Zeno subspace. The latter reduces to counterdiabatic driving when projective measurements are performed in the instantaneous energy eigenbasis of the quantum system. The effective Zeno Hamiltonian can also be derived in the context of continuous quantum measurements of a time-dependent observable and the non-Hermitian evolution with a complex absorbing potential varying in time. Our results thus provide a unified framework for realizing shortcuts to adiabaticity via adaptive quantum Zeno measurements.

A decision diagram (DD) is a graph-like data structure for homomorphic compression of Boolean and pseudo-Boolean functions. Over the past decades, decision diagrams have been successfully applied to verification, linear algebra, stochastic reasoning, and quantum circuit analysis. Floating-point errors have, however, significantly slowed down practical implementations of real- and complex-valued decision diagrams. In the context of quantum computing, attempts to mitigate this numerical instability have thus far lacked theoretical scaling guarantees and have had only limited success in practice. Here, we focus on the analysis of quantum circuits consisting of Clifford gates and $T$ gates (a common universal gate set). We first hand-craft an algebraic representation for complex numbers, which replace the floating point coefficients in a decision diagram. Then, we prove that the sizes of these algebraic representations are linearly bounded in the number of $T$ gates and qubits, and constant in the number of Clifford gates. Furthermore, we prove that both the runtime and the number of nodes of decision diagrams are upper bounded as $2^t \cdot poly(g, n)$, where $t$ ($g$) is the number of $t$ gates (Clifford gates) and $n$ the number of qubits. Our proofs are based on a $T$-count dependent characterization of the density matrix entries of quantum states produced by circuits with Clifford+$T$ gates, and uncover a connection between a quantum state's stabilizer nullity and its decision diagram width. With an open source implementation, we demonstrate that our exact method resolves the inaccuracies occurring in floating-point-based counterparts and can outperform them due to lower node counts. Our contributions are, to the best of our knowledge, the first scaling guarantees on the runtime of (exact) quantum decision diagram simulation for a universal gate set.

Gravitational wave (GW) spectral sirens offer a promising method for measuring cosmological parameters using GW data only - without relying on external redshift information such as electromagnetic counterparts or galaxy catalogs - by exploiting distributional features in the population of GW sources. The advent of third-generation detectors like the Einstein Telescope (ET) will provide catalogs three orders of magnitudes larger than current ones, raising questions about the scalability and robustness of existing inference pipelines. We present a blinded mock data challenge that tests three public pipelines with distinct numerical implementations, namely, $\texttt{ICAROGW}$, $\texttt{CHIMERA}$, and $\texttt{pymcpop-gw}$, on simulated ET observations containing the best $\mathcal{O}(10^4)$ binary black hole mergers that can be observed in 1 year. We assess their computational performance, validate their agreement in a blinded setting, and forecast cosmological constraints. We find that, thanks to GPU acceleration, these pipelines can process the events expected from ET within a manageable timeframe. All pipelines recover consistent cosmological and population parameters. Assuming a flat $Λ$CDM model, we measure $H(z)$ at $z\sim1.5$ with 2.4% precision, and achieve a mean precision on $H(z)$ of 2.8% across $0.7<z<1.8$ with a catalog of $\sim 12,000$ high-S/N events. This corresponds to joint constraints of $\sim 10%$ on $H_0$ and $\sim 26%$ on $Ω_{\rm m,0}$. We also identify the events that contribute mostly to constraining cosmological parameters, showing that low-distance sources near population features drive the constraining power on all cosmological parameters, while higher-distance events primarily constrain $Ω_{\rm m,0}$. Our results establish a validated, performance-tested framework for spectral siren cosmology in the era of third-generation GW observatories.

Quantum optimization has gained increasing attention as advances in quantum hardware enable the exploration of problem instances approaching real-world scale. Among existing approaches, variational quantum algorithms and quantum annealing dominate current research; however, both typically rely on one-hot encodings that severely limit scalability. Pauli Correlation Encoding (PCE) was recently introduced as an alternative paradigm that reduces qubit requirements by embedding problem variables into Pauli correlations. Despite its promise, PCE has not yet been studied in the context of constrained optimization. In this work, we extend the PCE framework to constrained combinatorial optimization problems and evaluate its performance across multiple problem sizes. Our results show that the standard PCE formulation struggles to reliably enforce constraints, which motivates the introduction of the Iterative-$α$ PCE. This iterative strategy significantly improves solution quality, achieving consistent constraint satisfaction while yielding better cut sizes across a wide range of instances. These findings highlight both the limitations of current PCE formulations for constrained problems and the effectiveness of iterative strategies for advancing quantum optimization in the NISQ era.

We prove that there exists an absolute constant $α<1$ such that for every finite dimension $d$ and every quantum channel $T$ on $\mathsf{L}(\mathbb{C}^d)$, $\left\|Θ\circ(\mathrm{id}-T)\right\|_\diamond \le α\,\left\|Θ\right\|_\diamond\,\left\|\mathrm{id}-T\right\|_\diamond$, where $Θ$ is the transposition map. In fact we show the explicit choice $α=1/\sqrt{2}$ works.

Many computational methods in ab initio quantum chemistry are formulated in terms of high-order tensor contractions, whose cost determines the size of system that can be studied. We introduce stochastic tensor contraction to perform such operations with greatly reduced cost, and present its application to the gold-standard quantum chemistry method, coupled cluster theory with up to perturbative triples. For total energy errors more stringent than chemical accuracy, we reduce the computational scaling to that of mean-field theory, while starting to approach the mean-field absolute cost, thereby challenging the existing cost-to-accuracy landscape. Benchmarks against state-of-the-art local correlation approximations further show that we achieve an order-of-magnitude improvement in both total computation time and error, with significantly reduced sensitivity to system dimensionality and electron delocalization. We conclude that stochastic tensor contraction is a powerful computational primitive to accelerate a wide range of quantum chemistry.

In this Letter, we show that, in contrast to Dyson shells surrounding uncharged compact objects, which are generally unstable, a neutral Dyson shell enclosing a charged compact object described by the Reissner-Nordstrom spacetime can attain a stable equilibrium configuration. We analytically derive the conditions for stability, determine the equilibrium radius and the corresponding minimum asymptotic energy, and show that small perturbations about this equilibrium lead to a stable oscillatory motion of the shell. The oscillation frequency is obtained explicitly and shown to increase with the shell mass and decrease with the charge of the central object. When the shell itself carries charge, its stability depends on the sign of this charge. Shells with the same sign as the central charge become progressively less stable, while oppositely charged shells exhibit enhanced stability due to the electrostatic attraction. These findings highlight the stabilizing role of electromagnetic interactions in Dyson-type thin-shell configurations within general relativity.

Environmental effects such as hierarchical triple motion can introduce cumulative phase modulation in gravitational-wave signals through time-dependent line-of-sight acceleration. Whether such smooth time-warp distortions are observable depends jointly on deformation strength and signal-to-noise ratio (SNR), yet this relationship has not been quantified in a template-free framework. We study the detectability of these distortions using time-frequency representations derived from the continuous wavelet transform. Instead of reconstruction error alone, we examine trajectory-based statistics, in particular the evolution of the power-weighted frequency centroid. We find that environmental modulation can be detected using a single-sample statistic referenced to an isolated-binary distribution, without requiring matched templates. Across a grid of cumulative phase distortions and SNR, detection performance collapses onto a single scaling parameter defined as the product of phase distortion and SNR. The ROC-AUC follows a sigmoid transition in this parameter. Moderate distortions are detectable at low SNR, whereas smaller distortions require higher SNR. These results indicate that smooth environmental phase modulation is not generically absorbed by intrinsic waveform variability; instead, detectability is governed by a simple scaling between cumulative phase distortion and signal strength.

Modified gravity is often approached in the context of effective-field theory (EFT), with the view that the EFT corrections permit a more desirable theory. In this paper, we posit that this should extend to the causal structure of curved spacetime in addition to the standard demands such that of flat spacetime positivity and unitarity. We propose a new guiding principle for modified-gravity theories, namely that closed timelike curves should always be {\it harder} to obtain than in General Relativity. By demanding this, one can place powerful constraints on modified gravity. To elucidate this claim, we investigate modified-gravity EFTs on rotating black-hole backgrounds, focusing on the appearance/disappearance of closed timelike curves, and provide parameter bounds which only partly overlap with other approaches based on time delay. We construct perturbative rotating black-hole solutions in modified-gravity EFTs based on the Horndeski class and provide parameter bounds necessary to preserve causality and stability. Finally, we present a novel probe for the existence of closed timelike curves through quasinormal modes and black-hole echoes. This can be used to diagnose spacetime causality once next-generation gravitational-wave data becomes available.

The leading singularities of one-loop scattering amplitudes in planar $\mathcal{N}=4$ super Yang-Mills theory are known to factorise into products of tree-level amplitudes, and this can be seen from a number of different perspectives e.g. generalised unitarity or on-shell diagrams. Here we investigate the leading singularities from the perspective of the Wilson loop expectation values to which these amplitudes are dual, in particular making use of the twistor Wilson loop formalism. We show that the factorisation of one-loop leading singularities of a null Wilson loop's expectation value into a product of tree-level objects is manifest at the level of twistor Wilson loop diagrams, and is a simple consequence of planarity, without appeal to e.g. unitarity on the amplitude side of the duality. We then use the same approach to derive compact formulae for the one-loop leading singularities of correlators of multiple light-like Wilson loop operators in terms of tree-level objects. Via the chiral box expansion, these formulae provide a simple route to writing down the $O(g^2)$ correlation function of any number of Wilson loops at any MHV degree.

We perform a systematic study of lepton-number-violating (LNV) dimension-9 operators in the Standard Model Effective Field Theory (SMEFT) that can mediate neutrinoless double beta decay ($0νββ$) at tree level, and map them to their possible tree-level ultraviolet completions. Using a diagram-based classification, we enumerate all such completions and isolate minimal two-particle models that avoid generating the dimension-5 Weinberg operator or dimension-7 LNV operators at tree level. We then chart how these minimal models populate the operator landscape and organise them by the loop order at which they radiatively induce lower-dimensional LNV operators, highlighting scenarios in which the tree-level dimension-9 contribution can compete with or dominate loop-suppressed neutrino-mass (dimension-5) effects. Representative one-loop and two-loop classes are matched onto the SMEFT, and their implications for neutrino masses, charged-lepton flavour violation, and the relative size of dimension-9 versus dimension-5 contributions to $0νββ$ are analysed, delineating regions of parameter space where upcoming experiments can be sensitive to genuinely short-range LNV dynamics.

Flavor hierarchies are organized by a single parameter $B\simeq 5.357$ in a single-flavon Froggatt--Nielsen (FN) framework, in which each effective Yukawa entry arises from the sum of \emph{three} unit-magnitude messenger chains. We present benchmark complex $O(1)$ Yukawa matrices that reproduce quark and charged-lepton masses at $M_Z$ as powers of $ε\equiv 1/B$. The organizing principle is a two-over-two (2/2) lattice of quadrilateral mass ratios, which maps directly to a rational lattice of FN exponents. Sequential dominance preserves the leading-power exponent matrices, while subleading messenger chains generate entry-dependent complex $O(1)$ coefficients and provide a UV-friendly origin for CP violation. Neutrino masses are discussed at the level of eigenvalues within the same $B^n$ counting.

We report first results from RadioAxion, an underground experiment searching for axion dark matter through periodic modulations of radioisotope decays. We monitor the $α$ decay of $^{241}\mathrm{Am}$ via its $59.5$ keV $γ$ line using a NaI detector installed at the Gran Sasso Laboratory, where cosmic-ray-induced systematics are strongly suppressed. We present the measured spectra and the corresponding time-series analysis. No evidence for a periodic modulation is observed. From these data we derive constraints on the axion decay constant in the axion mass range from $10^{-21}$ to $10^{-9}$ eV.

Modern measurements in flavour physics rely on accurate simulations of signal and background processes, provided by a wide range of general-purpose and specialised Monte-Carlo event generators. Due to the inclusion of a larger amount of specialised decays of heavy hadrons, EvtGen is often the tool of choice for many scenarios. We investigate the phase-space sampling algorithm of EvtGen and demonstrate that it generates unphysical features in kinematic distributions of semileptonic $B$ decays involving resonances, originating from neglected phase-space factors. We provide a short-term solution to correct the affected simulated samples through reweighting of the hadronic invariant mass distribution.

We study the phase structure of a model containing two flavors of massless staggered fermions interacting through two independent four-fermion couplings, UI and UB, formulated on a three-dimensional Euclidean space-time lattice. At UB = 0, this model is known to exhibit a direct second-order quantum phase transition between a massless fermion (MF) phase and a phase in which fermions acquire masses through the mechanism commonly referred to as symmetric mass generation (SMG). We demonstrate that introducing a small nonzero value of UB qualitatively alters this structure: the single exotic transition at UB = 0 splits into two distinct, conventional transitions, separated by an intermediate phase in which fermion masses arise through the standard mechanism of spontaneous symmetry breaking (SSB). The first of these is a Gross-Neveu transition separating the MF phase from the SSB-induced massive phase, while the second is a three-dimensional XY transition between the SSB phase and the SMG phase. Using the fermion-bag Monte Carlo method, we verify that the critical exponents associated with both transitions are consistent with the literature, thereby yielding a quantitative characterization of the resulting phase structure of the model.

We study the influence of the temperature-dependent interaction between dark matter (DM) and neutrinos on the measurement of cosmological parameters. We pay attention to the neutrino mass effects, so that the derivation of Boltzmann equations needs to specify the concrete form of interaction. We work in a model in which the DM-neutrino scatterings are induced by a dimension-six operator, and present the details for deriving the full Boltzmann hierarchy for DM and neutrinos, including a novel method to obtain the fluid approximation for modes entering the horizon. It is shown that our interaction can induce the dark acoustic oscillation in the DM-neutrino fluid, leaving distinct signatures on the CMB and matter power spectra. By using the latest CMB and BAO datasets from Planck, DESI and ACT, the constraint on today's DM-neutrino interaction parameter for the normal neutrino mass ordering reaches $u^0_{χ-ν} \lesssim {\cal O}(10^{-13})$, nearly nine orders stronger than that for temperature-independent case in the literature. This can be understood by noting that the scattering cross section increases nearly quadratically with cosmological temperature in the early universe, leading to enhanced effects. We have investigated alternative scenarios with different neutrino mass assumptions. In particular, models with degenerate neutrino masses give rise to weaker constraint of $u^0_{χ-ν} \lesssim {\cal O}(10^{-11})$, showing the importance to incorporate the realistic neutrino mass ordering in the fits. Finally, when employing the logarithmic flat prior for $u^0_{χ-ν}$, we have shown hints to a nonzero interaction at $95\%$ CL by combining Planck, DESI and ACT data.

Scalar field theory is studied by constructing interacting saddle point expansions in the symmetric and broken phase, respectively. Focusing on analytically tractable saddle expansions, it is found that broken and symmetric phases are related by sign flip of the quartic coupling. Applications to dimensions $d<4$ recover previous results for the phase diagram, whereas $d=4$ is possibly new.

The first measurement of the jet axis decorrelation in events with jets recoiling from an isolated photon is presented for lead-lead (PbPb) and proton-proton (pp) collisions at a nucleon-nucleon center-of-mass energy of 5.02 TeV. The jet axis decorrelation is the angular difference ($Δ{j}$) between two definitions of the jet axis. This quantity is expected to be sensitive to the scattering of jet constituents in the quark-gluon plasma (QGP). Events which have a leading isolated photon with transverse momentum 60 $\lt$ $p_{\mathrm{T}}^γ$ $\lt$ 200 GeV and recoiling jets with 30 $\lt$ $p_{\mathrm{T}}^{\text{jet}}$ $\lt$ 100 GeV are selected for the analysis. The PbPb result is reported as a function of collision centrality and compared to pp reference data. Jets with $p_{\mathrm{T}}^{\text{jet}}$ $\lt$ 60 GeV have consistent $Δ{j}$ shapes for pp and PbPb collisions. However, a narrowing is observed for jets with $p_{\mathrm{T}}^{\text{jet}}$ $\gt$ 60 GeV in central PbPb collisions. The results are compared to predictions from the JEWEL, PYQUEN, and HYBRID theoretical models, which include different descriptions of parton energy loss in the QGP.

We consider the two- and three-flavor QMD models as renormalizable low-energy models for QCD at finite quark chemical potentials with quarks, mesons, and diquarks as effective degrees of freedom. Using the on-shell scheme the parameters in the scalar sector can be fixed and expressed in terms of observed meson masses and decay constants. The remaining parameters can be varied. In the QMD models, all the symmetries are global, including the $SU(N_c)$ symmetry. The breaking of the global symmetries gives rise to a number of Goldstone bosons depending on the symmetry-breaking pattern, i.e. whether the system is in the 2SC phase or the color-flavor-locked (CFL) phase. This is in contrast to perturbative QCD, where some of the gauge bosons become massive via the Higgs mechanism. We classify the Goldstone bosons and show that their type and number are in accordance with general counting rules. The thermodynamic potential $Ω$ is calculated in the mean-field approximation, where we include quark loops, while mesons and diquarks are treated at tree level. As important applications, we study the properties of the pion-condensed phase at finite isospin chemical potential, and the 2SC and CFL phases at finite baryon chemical potential. We present a few numerical results focusing on the speed of sound, gaps, and condensates. It is shown that the BCS gaps approaches a constant for large isospin and baryon chemical potentials and that the speed of sound approaches the conformal value from above in the same limit.

In this work, we propose a new three-dimensional nonlocal spinor superfield model. This theory is constructed by introducing form factors in the local spinor superfield action. Then, we couple it minimally to a scalar superfield, for which we calculate the one-loop effective potential as a first constructive example of perturbative calculations in this new theory.

Double parton distributions at small distances between the two partons are dominated by a mechanism in which the two observed partons originate from the splitting of a single parton. This contribution can be computed in terms of single-parton distributions and perturbative splitting kernels. We demonstrate that two-loop corrections to these kernels can have a substantial quantitative impact and considerably improve the stability of predictions for double parton scattering. We also consider the impact of heavy quark masses in the two-loop splitting kernels in an approximate manner.

In the Standard Model, the couplings between gauge bosons are tightly constrained by the principles of gauge symmetry and renormalizability. However, the presence of anomalous couplings suggests the possibility of new physics beyond the Standard Model (BSM). In this study, we focus on the sensitivities of anomalous quartic gauge couplings (aQGCs), specially the dimension-8 operators associated with field-strength tensor structures within the effective field theory (EFT) framework, at future Muon Colliders. Our analysis targets the neutral aQGC-sensitive processes $μ^{+}μ^{-} \to μ^+ γγμ^-$ and $μ^{+} μ^{-} \to μ^+ Z γμ^-$, simulated at center-of-mass energies of 3 TeV and 10 TeV. Signal and background events are generated using {\sc MadGraph5\_aMC@NLO}, interfaced with Pythia8 for parton showering and hadronization, and Delphes for fast detector simulation. A multivariate analysis based on Boosted Decision Trees (BDTs) is employed to enhance signal-to-background discrimination, utilizing a comprehensive set of kinematic and reconstructed observables from the final-state particles. Unitarity is preserved through the application of an energy-dependent clipping procedure within the EFT validity regime. Our findings indicate that future muon colliders offer significant sensitivity improvements over current experimental constraints on aQGCs. Furthermore, a comparison with other future collider scenarios shows that the 10 TeV Muon Collider, even with a 10\% systematic uncertainty, provides substantially stronger projected limits at 95\% confidence level than those currently reported by the ATLAS collaboration at the LHC as well as projected limits by future hadron colliders. These results underscore the enhanced potential of high-energy muon collider to probe new physics in the electroweak sector through precision measurements of aQGCs.

A search is performed for a new resonance X decaying into either a pair of Higgs bosons (HH) or into a Higgs boson and a new scalar boson Y (HY), using proton-proton collision data collected at $\sqrt{s}$ = 13 TeV, corresponding to an integrated luminosity of 138 fb$^{-1}$. This study performs a comprehensive exploitation of the bbZZ events, encompassing the following decay topologies. One H candidate is identified through its decay into a bottom quark-antiquark pair, while the other H or the Y candidate is selected through its decay into a pair of Z bosons. One Z boson is required to decay leptonically and the other, to decay into a pair of quarks or neutrinos. Events of interest are categorized based on the Lorentz boosts of the hadronically decaying H and Z bosons. Machine-learning-based discriminants, together with the reconstructed resonance mass, are employed across the different categories to separate signal from backgrounds, and their corresponding distributions are included in a simultaneous fit. No significant deviations from the standard model predictions are observed. Upper limits at the 95% confidence level are set on the HH and HY production cross sections. For resonant HH production, the upper limit on the cross section of pp $\to$ HH production is 1 pb for a high-mass resonance. For HY production, the upper limit on the cross section of the process pp $\to$ X $\to$ bbZZ is approximately 5 fb for a high-mass resonance. This is comparable to the sensitivity achieved in other analyses, which focus on H decays to $γγ$ or $ττ$ and Y decays into a pair of bottom quarks or massive vector bosons.

We compute for the first time the production rate for four jets in electron-positron annihilation at next-to-next-to-leading order. Our calculation exhibits the highest final-state jet multiplicity considered at this perturbative accuracy to date. The cancellation of infrared singularities is achieved in the antenna subtraction scheme, relying particularly on generalized antenna functions. The evaluation of the two-loop virtual corrections is enabled by the construction of a new basis of transcendental special functions tailored to four-particle decay kinematics. Our results are compared with LEP data, finding improved agreement with respect to the next-to-leading order calculation. In the region where perturbative predictions are most reliable, we observe a significant reduction of theory uncertainties, which now fall below the experimental ones.

We propose a computational strategy for NLO+PS simulations in the MC@NLO framework that starts from Born-accurate (LO) events and reweights them to the full MC@NLO S-event weight, while generating H-events separately. We validate the approach on two representative LHC processes and compare to direct NLO event generation for both standard MC@NLO and MC@NLO-Delta matching. Employing large folding values in the radiative variables stabilizes the S-event integral, reduces weight variance, and significantly lowers the fraction of negative weights compared to S-event generation without folding. At fixed precision, this pipeline has comparable wall-clock times relative to standard S-event generation and unweighting, with room for further optimisation.

We use a diffusion Monte Carlo (DMC) algorithm to solve the Schrödinger equation that describes $udsc\bar c$ pentaquarks within the framework of a non-relativistic constituent quark model. We considered only multiquark states with defined values of parity, color, spin and isospin, selected to be compatible with the experimentally favored assignment $J^P=1/2^-$ for one of the candidates, and assumed $I=0$. However, we found that, to explain the existence of the $P_{cs}(4338)$ and $P_{cs}(4459)$ pentaquarks, we need the total wavefunction to be also an eigenvector of the SU(3) {\em flavor} operator. When we impose that condition, we obtain two structures compatible with the masses extracted from the $J/ψΛ$ spectrum. In addition, two states are predicted below the $J/ψΛ$ threshold but above the $η_cΛ$ one that would not appear in that channel. If we only impose the $I=0$ condition, we obtain a {\em single} (not two) structure compatible with the experimental quantum numbers, with a mass below the $J/ψΛ$ threshold.

We study one-loop logarithmic corrections to the entropy of static hyperbolic BPS black holes in asymptotically AdS$_4$ spacetime. Our analysis is carried out in a consistent real-scalar truncation of ${\cal N}=2$ Fayet-Iliopoulos gauged supergravity specified by the prepotential $F=-i X^0 X^1$, which corresponds to an Einstein-Dilaton-Maxwell theory with a nontrivial scalar potential. In this model, the classical BPS attractor mechanism exhibits flat directions, leaving scalar moduli on the black hole horizon unfixed, while the Bekenstein-Hawking entropy depends only on the charges. We show that the resulting one-loop correction to the black hole entropy acquires a nontrivial dependence on the horizon moduli and induces an effective quantum potential that dynamically stabilizes them at a preferred value. Our results provide an explicit and concrete realization of quantum lifting of classical attractor flat directions in gauged supergravity.

We give a phenomenological analysis of the behavior of inelastic, elastic and total cross-sections of the high energy $\textit{pp}$ interaction. In particular, we argue that the universal picture of behavior of cross-sections and their ratios is a consequence of the rapid increase of inelastic cross-section with energy and its large value compared to $σ_{\mathrm{el} }(s)$. We observed that the value of the fundamental ratio $(m_{π^{\mathrm{0}}}/m_{p}) $, the minimum value of the ratio $ (σ_{\mathrm{el}}/σ_{\mathrm{tot}})$, and some other quantities are determined by the roots of the equation $ (9\,x^{2}+4\,\sqrt{2}\,x-1)=0 $.

We provide a systematic discussion of the phenomenology of Dark Photon iDM (A$^{\prime}$iDM) where the Standard Model (SM) is extended by a dark sector containing an additional $U(1)_D$ gauge symmetry under which all SM particles are neutral, and that couples to the SM hypercharge gauge boson through a kinetic mixing parameter $ε$. The model contains two Majorana states $χ_1$ and $χ_2$ with $δ=M_{χ_2}-M_{χ_1}>0$ and $χ_1$ the dark matter candidate, and a dark photon $A^{\prime}$ with mass $M_{A^{\prime}}$. Our analysis represents an integration of existing ones, where only specific benchmarks of the A$^{\prime}$iDM scenario have been discussed. In particular, we fix the $U(1)_D$ coupling $α_D$ equal to the electromagnetic one $α_{EM}$ and $ε$ to its experimental upper bound, and perform a complete scan of the remaining parameters $(M_{χ_1},δ, M_{A^{\prime}})$, discussing the $χ_1$ relic abundance, its direct and indirect searches, as well as potential signals from astrophysics and accelerators. Our systematic scan shows that $α_D$ = $α_{EM}$ is not disfavored, as some previous analyses, limited to specific benchmarks, may suggest. We also find that when the $χ_1$ relic density matches observation direct and indirect searches are not kinematically accessible. On the other hand we find that the projected luminosity of FASER, a detector searching for Long Lived Particles (LLP) decay at the LHC, can probe or rule out the parameters space of the model for $M_{χ_1}\lesssim$ 7 GeV, 100 MeV $\lesssim δ\lesssim$ 300 MeV and $M_{A^{\prime}}\lesssim$ 25 GeV. This range of parameter could be significantly extended by the FASER 2 upgrade proposed for the High-Luminosity phase at the LHC. The parameter space probed by LLP seaches partially overlaps with that probed by $χ_1$ capture in neutron stars.

In this study, we compute the radiative corrections to the $Λ_b \to p$ transition form factors at next-to-leading logarithmic accuracy, employing the framework of QCD light-cone sum rules with the light-cone distribution amplitudes of the $Λ_b$ baryon. The factorization formulae of the vacuum-to-$Λ_b$ correlation function, constructed from the interpolating current for the proton, are derived at leading power in $m_p / m_{Λ_b}$, using the method of regions. With our specific choice of interpolating current, only the twist-4 distribution amplitude of the $Λ_b$ baryon contributes to the form factors. Numerically, we find that the next-to-leading order QCD perturbative corrections reduce the tree-level form factors to approximately 65$\%$ of their original value, with the next-to-leading-order jet function providing the dominant contribution. In the large-energy limit ($E_p \to \infty$), the form factors exhibit a clear $1/E_p^3$ scaling, consistent with the expected power-counting behavior. By applying the $z$-series parameterization to perform a combined fit of the form factors from our results and available lattice QCD simulations, we further investigate the decay rate of $Λ_b \to p \ell^- \barν_{\ell}$ and extract the CKM matrix element $|V_{ub}| = (3.33\pm 0.43 ) \times 10^{-3}$.

We study holographic pseudoentropy for subregions in non-unitary Euclidean conformal field theories (CFTs) within the framework of the de Sitter/conformal field theory (dS/CFT) correspondence. Pseudoentropy, defined as the von Neumann entropy of a transition matrix, is computed holographically from codimension-two extremal surfaces in dS space and is divergent due to the asymptotic bulk volume at future infinity. We show that a finite and regulator-independent definition follows from the on-shell action of conformal gravity in four and six dimensions, implemented through the replica construction. We illustrate the formalism for spherical entangling surfaces and small shape deformations thereof. The renormalized pseudoentropy isolates the universal contribution, which for a spherical entangling surface is proportional to the complex-valued central charge $a^\star$ of the non-unitary CFT. On an equal footing, for infinitesimal deformations away from the sphere, we recover, at quadratic order in the deformation parameter, an analytic continuation of the Mezei-like formula in its anti-de Sitter counterpart.

We examine the Lüscher quantization condition to high order for the scattering of a spinless particle and a spin-1/2 particle in a periodic box. First, we derive the quantization conditions in a non-relativistic framework up to total angular momentum $J=11/2$ in both cubic and elongated geometries, and for both rest and moving frames. Then, we introduce a method to transparently cross-check their convergence, using both quantized energy levels in the box and infinite-volume phase shifts for the same potential. We clarify how to incorporate spin-orbit coupling into the formalism and show in detail how the quantization conditions converge order by order in the various irreducible representations. In all, we validated 19 quantization conditions (12 in cubic box, 7 in elongated box). This is a necessary step in applying the method in precision studies of systems in finite volume with half-integer spin, such as meson-baryon scattering.

The Electron--Ion Collider (EIC) offers a unique environment to study kinematically controlled lepton--nucleus (e+A) reactions, where a primary hard scattering is followed by an intranuclear cascade and the subsequent statistical de-excitation of the nuclear remnant. Utilizing the BeAGLE model, we demonstrate that event-by-event fluctuations in nucleon removal and energy deposition populate a diverse ensemble of excited remnants. Furthermore, we show that varying the target mass systematically shifts the distribution of these remnants across the (N, Z) plane. Although this excited prefragment remnant is not directly observable, its properties are shown to be strongly correlated with final-state fragments; specifically, the largest nuclear residue and the intensity of evaporation activity serve as effective experimental proxies for event-level remnant characterization. We also evaluate photon observables essential for nuclear spectroscopy. While various photon sources overlap significantly in pseudorapidity, we find that in the nucleus-rest frame, the low-energy spectrum is dominated by de-excitation $γ$ rays and exhibits distinct discrete structures. These findings motivate an EIC research program that correlates rare-isotope production and de-excitation radiation with well-defined initial conditions, providing a collider-based approach to nuclear spectroscopy that is complementary to existing fixed-target facilities.

Existence of solutions to the field equations of the gauged Chern-Simons-O(3)-Sigma model on a compact Riemann surface is proved by a topological method. Existence of a minimal deformation constant $κ_{*} > 0$ is proved, such that for any prescribed configuration of vortices and antivortices, at least one solution exists for $|κ| \leq κ_{*}$. For small values of the Chern-Simons deformation parameter $κ$, it is proved that the field equations admit multiple solutions, provided the total number of vortices and antivortices are different. The Maxwell limit is computed for solutions of the field equations. In contrast, if the number of vortices equals the number of antivortices, it is proved that the field equations admit at least one solution for any value of $κ$ and the limit $κ\to \infty$ is proved. dependence of the fields on the deformation parameter is investigated numerically on the sphere.

Sterile neutrinos in the keV mass range have long been considered a well-motivated dark matter (DM) candidate. In this work, we explore a sterile neutrino production mechanism through active neutrino self-interactions in the early universe, assuming that they constitute the full DM abundance. We implement a self-consistent treatment of the sterile-neutrino free streaming and the active-neutrino self-interactions on structure formation, which yield a unique scale-dependent modification to the linear matter power spectrum. We then set bounds on this scenario using a combination of the cosmic microwave background and Ly$α$ forest constraints. Specifically, we utilize the two recent likelihoods derived from eBOSS data: (i) an effective field theory (EFT) based full-shape likelihood and (ii) a compressed likelihood obtained from the PRIYA-simulation emulator. We produce some of the most stringent observational constraints to date on sterile neutrino DM, comparable to the bounds from the most stringent laboratory constraints.

We study junctions between confining strings. These junctions arise in Yang-Mills theories, and we focus on their universal low-energy dynamics. Using open-closed duality, we map junctions with nonlinear corrections to the $s$-wave scattering amplitudes between confining string loops. In $(3+1)$ dimensions, we uncover an accidental $\mathbb{Z}_2$ symmetry. This symmetry implies novel selection rules for loop scattering amplitudes and is broken by the junction mass at subleading order. We determine the total mass of baryons up to order $(\text{baryon size})^{-3}$, providing new, testable predictions for lattice simulations.

Minimal electroweak dark matter models are compelling due to their simplicity, though calculations of their freezeout abundance are complicated by nonperturbative effects due to Sommerfeld enhancement and bound-state formation. It has been shown that all individual multiplet scenarios beyond the doublet lead to direct-detection signals above the neutrino floor and thus within the reach of next-generation experiments. If no signals are found, would minimal dark matter be excluded? Yes for the simplest models, but it has been unknown for the important extension of two multiplets coupled by Higgs interactions (Higgs-coupled minimal dark matter). We present a generalized framework for calculating nonperturbative effects for such models that also covers the case of individual multiplets. In this framework, we calculate nonperturbative effects on freezeout as well as the prospects for direct detection, correcting shortcomings and omissions in the literature. Importantly, for the mixed Majorana (odd) and Dirac (even) multiplet combinations 3M2D, 5M4D, and 7M6D, we find that the predicted direct-detection signals can extend below the neutrino floor. Fully testing minimal dark matter will thus require more than direct-detection experiments.

The Belle~II experiment operates at high luminosity, where an increasing beam-induced background imposes stringent demands on the hardware Level-1 trigger system, which must operate under tight latency and bandwidth constraints. To achieve online data reduction within the Level-1 trigger system, we have developed a hit-filtering algorithm based on the lightweight Interaction Network architecture. In this work, we present a hardware-aware model-compression workflow for this hit-filtering algorithm targeting deployment on FPGA devices within the Belle~II trigger system. The network is adapted to the detector and trigger conditions through model-size and graph-size reduction, low-precision (4 bit) fixed-point arithmetic, and unstructured pruning. We assess the resulting design using the total number of bit operations as a hardware-aware computational complexity metric. Using this metric, we identify a configuration that decreases this cost by more than two orders of magnitude relative to the full-precision reference implementation. This reduction is achieved while preserving performance close to the reference model in terms of hit efficiency and background rejection, as indicated by only a modest decrease in the AUC score from 97.4 to 96.8, evaluated on Belle~II collision data.

We consider the dynamics of confined strings embedded in a gapless four-dimensional theory. To this end, we examine finite-tension string-like solutions to the equations of motion of the $\mathbb{C}\mathbb{P}^1$ non-linear sigma model. We present a comprehensive analysis of the quantum fluctuations around these solutions and derive the corresponding spectrum. These results allow us to determine the quantum corrections to the closed string ground state energy in both the finite- and infinite-size limits. Furthermore, we analyze quantum corrections to the string's effective width. We find that these observables generically depart from the universal predictions of standard Effective String Theory (EST), and we identify specific limits in which the bulk dynamics decouple and EST is recovered. Finally, we discuss the connection between these string configurations and stable electric and magnetic fluxes arising in certain ultraviolet completions of the $\mathbb{C}\mathbb{P}^1$ model.

We perform a systematic holographic study of Krylov complexity for a wide class of confining quantum field theories. Using the geometric prescription that identifies the time derivative of the complexity with the proper momentum of a massive probe, we analyse radial geodesics in several top-down gravity duals exhibiting confinement and a mass gap. In all geometries with a smooth infrared end-of-space we uncover a robust and universal qualitative feature: Krylov complexity exhibits oscillatory behaviour. The oscillation frequency is controlled by the confinement scale, while the amplitude depends on both the ultraviolet cutoff and the infrared scale. Additional conserved charges modify these patterns without altering their qualitative structure. We further compare our results with the Krylov complexity of the longitudinally perturbed Ising model. The qualitative agreement suggests that oscillatory behaviour of Krylov complexity constitutes a universal signature of confinement and provides a sensitive probe of infrared reorganisation in strongly coupled quantum field theories.

We propose a class of theories to generate quark and lepton mass matrices where the scale of new physics is at the TeV scale, without inducing the large flavor and CP violating processes that are often thought to relegate the origin of flavor to energies above $\sim 100$ TeV. The models have new vector-like leptons and quarks, with mass mixings to each other and Yukawa couplings to light Standard Model fields encoded in "chains" reminiscent of dimensional deconstruction. Locality in the chains both generates the hierarchical Standard Model Yukawa matrices, and ensures that CP and flavor violating effects are small, even with the vector-like particles at the TeV scale. A simple extension also generates neutrino masses, whose tiny size is parametrically related to the square of the electron Yukawa coupling. We outline the essential features of these models, explain how fermion mass hierarchies and mixing angles emerge, and explore their phenomenological implications. This framework can be tested both in the final run of the LHC as well as at possible future colliders operating at the 10 TeV scale, and we identify some of the distinctive experimental signatures associated with the production and decay of the new vector-like fermions.

The dynamic enhancement of symmetry-breaking effects in neutron-nucleus resonances provides a sensitive testing ground for Time-Reversal Invariance Violation (TRIV). Exploiting this mechanism, the Neutron Optics Parity and Time Reversal Experiment (NOPTREX) seeks to elucidate the origin of the universe's baryon asymmetry. Critical to this effort is the precise measurement of Parity Violation (PV) asymmetries, which is essential to calibrate the nuclear parameters required for future TRIV experiments. To facilitate these studies, we developed an eV polarized neutron at the Back-n white neutron beamline of the China Spallation Neutron Source (CSNS). Neutron polarization is generated by an in-situ Spin-Exchange Optical Pumping (SEOP) $^3$He filter. Spin manipulation is performed by an adiabatic spin flipper, while spin polarization is preserved over the flight path by a vacuum transport system equipped with a solenoidal guide field. Experiments successfully measured an asymmetry of approximately $7.8 \pm 2.4$ (stat.) $\pm 0.3$ (sys.) % at the 0.747 eV p-wave resonance of $^{139}$La. These results are in agreement with previous results on this resonance and validate the system's capability for PV measurements.

We present X-ray proper motion (PM) measurements of 19 pulsars using new and archival data from the Chandra X-ray Observatory, including pulsar wind trails and X-ray filaments. Precise X-ray PMs are often limited by uncertainties in aligning observations to a common reference frame. Our analysis uses unresolved X-ray flux from stars in the Gaia catalog in addition to X-ray bright point sources for alignment, improving uncertainties. We obtain absolute positions referenced to Gaia with typical astrometric precision $\sim$10 mas and PM statistical uncertainties down to 1.3 mas yr$^{-1}$, the most precise X-ray PM achieved to date. With our improved frame alignment, PM accuracies are now limited by the pulsar flux in most cases. These results reveal a new X-ray filament and illuminate the wind nebula structures and origins of several of these pulsars.

Integral Field Spectroscopy (IFS) surveys offer a unique new landscape in which to learn in both spatial and spectroscopic dimensions and could help uncover previously unknown insights into galaxy evolution. In this work, we demonstrate a new unsupervised deep learning framework using Convolutional Long-Short Term Memory Network Autoencoders to encode generalized feature representations across both spatial and spectroscopic dimensions spanning $19$ optical emission lines (3800A $< λ<$ 8000A) among a sample of $\sim 9000$ galaxies from the MaNGA IFS survey. As a demonstrative exercise, we assess our model on a sample of $290$ Active Galactic Nuclei (AGN) and highlight scientifically interesting characteristics of some highly anomalous AGN.

Previous numerical simulations have shown that cold clumps can form within hot accretion flows, offering insights into the detailed processes of the state transition in black hole X-ray binaries. However, the evolution of the cold clumps has not been investigated in detail yet. In this paper, we conduct hydrodynamic simulations to investigate the evolution of the cold clumps. In addition to previous result that when the accretion rate is high enough the cold clumps emerge within the hot accretion flow, we found that instead of directly moving toward to the black hole, the clumps moves outward when they initially form. The reason should be the combination of viscous torque and the condensation of hot gas from larger radii, which lead to the slightly super-Keplerian angular momentum of the clumps. After reaching the equilibrium position, the clumps begin to fragment at the inner edge with each fragment moving inward sequentially. Generally, the azimuthal movement of the clumps are quasi-Keplerian, being closer to the outer detached Keplerian cold disk rather than the surrounding sub-Keplerian hot accretion flow, which agrees well with the semi-analytical results for weak coupling case in Wang et al. (2012).

Terrestrial double-$Λ$ hypernuclear data and astronomical observations of neutron stars provide complementary constraints on the $ΛΛ$ interaction. In this work, we investigate the $ΛΛ$ interaction within a Skyrme energy density functional framework based on the KIDS (Korea-IBS-Daegu-SKKU) models. We employ a Skyrme-type $ΛΛ$ interaction that includes the standard $s$- and $p$-wave terms, as well as a density-dependent term that effectively represents an $NΛΛ$ three-body force. The $s$-wave terms are constrained using data on double-$Λ$ hypernuclei supplemented by pseudodata obtained from core + $2Λ$ three-body model calculations including heavier hypernuclei. We show that the data on heavier systems are essential to simultaneously constrain the two $s$-wave parameters. We further explore the impact of the $p$-wave and $NΛΛ$ components on the neutron-star properties and find that appropriate repulsive contributions of these terms yield consistency with current neutron-star mass-radius observations. These results indicate that the present framework provides phenomenologically acceptable equations of state for dense $(N,Λ)$ matter over a wide range of densities and highlight the importance of future experimental data on heavier double-$Λ$ hypernuclei.

Modern spectroscopic surveys obtain spectra for millions of stars. However, classical spectroscopic methods can often be computationally expensive, rendering them impractical for the analysis of large datasets. We introduce a novel simulation-based deep-learning approach for the efficient analysis of high-resolution stellar spectra to be obtained with the upcoming high-resolution 4MOST spectrograph. We used a suite of synthetic non-local thermodynamic equilibrium (NLTE) spectra generated with Turbospectrum to mimic 4MOST observations and trained a conditional invertible neural network (cINN) for the purpose of predicting self-consistently stellar surface parameters and chemical abundances. The cINN is a neural network architecture that estimates full posterior distributions for the target stellar properties, providing an intrinsic uncertainty estimate. We evaluated the predictive performance of the trained cINN model on both synthetic data and observed spectra of stars. We found that our new cINN trained on NLTE synthetic spectra is capable of recovering stellar parameters with average errors ($σ$) of $33$ K for $T_\mathrm{eff}$, $0.16$ dex for $\log(g)$, and $0.12$ dex for [Fe/H], $0.1$ dex for [Ca/Fe], $0.11$ for [Mg/Fe], and $0.51$ dex for [Li/Fe], respectively, at a signal to noise ratio of 250 per Angstrom. From the analysis of the observed spectra of Gaia-ESO / 4MOST / PLATO benchmark stars, we verified that our NLTE estimates for stellar parameters and abundances are consistent with results obtained with the independent code TSFitPy. We conclude that the NLTE cINN is robust and can, theoretically, evaluate 4 million high-resolution 4MOST spectra in less than a day, using GPU acceleration.

A dynamic view of mass assembly is essential for understanding the formation of massive stars and clusters. Interpreting evolutionary diagnostics from Galactic-wide surveys, however, requires careful control of distance and environmental variations. The G316.8 filament provides an ideal laboratory: a 14-pc nearly linear structure composed of three contiguous subregions with comparable molecular gas reservoirs (~10,000 $M_\odot$ each) but spanning a clear evolutionary sequence from an infrared dark cloud (young) through a massive young stellar object (intermediate) to an HII region (evolved). As part of the Linear filament and nested cluster evolution tomography (LANCET) project, we mapped the full filament with the Atacama Compact Array at 1.3 mm, achieving 0.08 pc resolution over 17.1 pc$^2$. Combined with Herschel and APEX/ArTéMiS data, we derived high-resolution temperature and column-density maps. We quantify structural evolution using dense-fragment statistics, column-density PDFs, and $Δ$-variance analysis. From young to evolved regions, the maximum fragment mass increases from 8 to 490 $M_\odot$, while the dense-gas mass fraction ($>0.5$ g cm$^{-2}$) rises from 0.4% to 9.6%. The N-PDF develops a secondary power-law tail and the $Δ$-variance slope becomes progressively shallower, indicating ongoing assembly of dense sub-parsec structures. Our further ALMA 12m continuum and spectral line data will extend this dynamic scenarios down to 800 AU scale.

Deep radio surveys are now producing catalogues with millions of radio sources. Radio sources can have complex morphologies that depend both on the production mechanisms and the local environment. Recently, an unusual class of circular radio sources have been identified that were named "odd radio circles" (ORC). They have sizes of about 1', a circular/elliptical shape and appear edge brightened. Subsequent observations have suggested that this class may comprise a variety of sources. Despite various attempts to explain these sources, their origin remains unclear. The main goal of this work is to increase the number of known ORCs and derive common characteristics that can help identify the origin of these sources. We searched the LOFAR Two Metre Sky Survey (LoTSS) data release 3 (DR3) at 144 MHz for ORCs using a combination of parameter filtering on catalogue entries and visual inspection. We then identified possible optical counterparts and derived information such as redshift, physical size, and spectral index using further radio data at 54 and 1400 MHz. We isolated 18 sources with ORC structures. Four of these are double headed ORCs (ORCs with two rings), and two are new discoveries. We also found 5 new single headed ORCs and 9 candidate ORCs. With this work we significantly expanded the population of known ORCs. Our findings confirms that ORCs are a rare and heterogenous population of radio sources. We confirm the association with large ellipticals in most cases and we note a relation between the ORCs physical size and their integrated spectral index with small ORCs avoiding steep spectra.

This paper presents the first results from the Dandelion experiment, a directional detection, which searches for 1 meV dark photon dark matter. We use a spherical mirror to convert dark photons into standard millimeter-wavelength photons that can then be detected with an array of 221 Kinetic Inductance Detectors (KIDs) cooled down to 150 mK within the KISS (KIDs Interferometric Spectral Surveyor) camera and operating between 150 and 350 GHz. We used 1480 minutes of data to search for the signal of dark photons in the KID detectors, which is expected to be modulated due to the Earth's rotation. Our main challenge was to deal with a large background from room temperature and stray-light fluctuations. We used a de-correlation analysis to remove these background fluctuations. Templates of the background fluctuations were constructed from a Principal Component Analysis decomposition of detector measurements outside the expected Field of View trajectory of dark photons. We found that the dark photon signal was consistent with zero, giving a new upper limit on the dark photon's kinetic mixing, $χ$, with masses between 0.6 meV and 1.4 meV. These are the first constraints on dark photons as a dark matter candidate using an array of KIDs at millimeter wavelength.

Ionised regions around OB-type stars are formed at an early stage of their evolution and are important to investigate the formation process of these objects. However, so far only few observations of their physical structure and interaction with the parental molecular cloud have been made. The high resolution and sensitivity of new instruments such as ALMA and the upgraded VLA allow us to fill this gap in our knowledge. We investigate the well known core-halo ultracompact HII region G31.41+0.31 and the surrounding molecular clump with the aim to determine the density and temperature of both the ionised and neutral gas, and possibly obtain a 3D picture of their spacial distribution. We take advantage of the full-band frequency coverage at 3 mm obtained with ALMA for the GUAPOS project to image the emission of a plethora of hydrogen recombination lines towards the G31.41+0.31 HII region as well as several molecular transitions which are tracers of medium-density ($\sim$$10^4$--$10^6$ cm$^{-3}$) gas. The line data are complemented by continuum measurements obtained with the VLA at 1 cm and 7 mm. By fitting these lines also using a model that takes into account non-LTE effects we can investigate the density and temperature structure and the velocity field of the region. Our findings, based on a model fit accounting for non-LTE effects, indicate that the electron temperature of the HII region is mostly spanning a range between 5000 and 6000 K, while the density varies between 2500 and 7500 cm$^{-3}$. All in all, the distribution of these parameters as well as the corresponding velocity field hint at a cometary shaped HII region expanding away from the observer to the NW. The molecular gas appears to be still infalling towards the peak of the UC HII region, and its density and temperature are consistent with pressure confinement of the ionised gas to the SE.

The existence of a distinct mass boundary between the heaviest neutron stars and the lightest black holes remains in question. It is an artefact of our ignorance of the properties of matter at supra-nuclear densities, which exist in the cores of neutron stars. The study addresses these problems with a physics-informed machine learning approach, guided by astrophysical observations. The Transformer model is trained on an agnostically generated ensemble of equations of state. Two geometric parameters are defined on the mass-radius sequence of a neutron star--the front bending and the back bending. The transformer provides a two-step solution: first, the model predicts the maximum mass and radius using the bending parameters. Second, it predicts the square of the sound speed profile, completing the inverse mapping. The prediction is that massive neutron stars form when the sound speed peaks at low density, leading to strong back-bending and an early phase transition to quark matter. Massive stars favour a stiff equation of state at low density, and the density of matter at the star's core is sufficiently small. The maximum mass for a neutron star predicted by the astrophysical constrained transformer model is $2.477$ solar masses, and a minimum radius of about $11.498$ km for a neutron star of $1.4$ solar masses.

Untargeted neutral hydrogen (HI) surveys are well suited to identifying low surface brightness galaxies (LSBGs) that are gas rich, and they offer a complementary view to optically selected populations. We examined the LSBG population as identified via stellar and gaseous content using the MIGHTEE HI XMM-LSS early science data and the publicly available catalogs of optically identified LSBGs. There is currently little overlap between these datasets, with only three galaxies commonly detected. We performed surface brightness photometry of selected MIGHTEE HI detections to find 29 LSBGs, and 26 of these meet the size requirement (R_eff > 1.5 kpc) to be ultra-diffuse galaxies (UDGs). Furthermore, we extracted HI spectra at the location of all optically identified galaxies, placing upper limits on the HI-to-stellar mass ratio in these systems. While the HI-identified population overall tends toward bluer colors, the HI-identified and the optically selected samples mostly overlap in mean effective surface brightness, effective radii, and color. Although it is not straightforward to discern why the HI-identified LSBGs were missed in optical searches, this work highlights the utility of HI surveys in finding these faint systems. The HI-identified LSBGs are gas rich compared to the general HI-selected population. Furthermore, three out of four HI-selected UDGs with available kinematics show no systematic offset from the baryonic Tully-Fisher relation, although we are biased away from sources with low rotational velocities due to the low spectral resolution of the data. This work demonstrates the utility of HI observations for finding and characterizing the low surface brightness Universe.

The mass composition of ultra-high-energy cosmic rays (UHECRs) is commonly inferred from the first two moments of the depth of shower maximum, $X_{\rm max}$, measured by fluorescence and hybrid detectors. Such analyses require fast and accurate mappings between the moments of $X_{\rm max}$ and those of the logarithmic mass, $\ln A$, based on realistic air-shower simulations. In this work we provide updated parametrizations of the $X_{\rm max}$ moments and distributions for air showers initiated by nuclei from proton to iron, simulated with CONEX for three state-of-the-art hadronic interaction models: Epos LHC-R, Sibyll 2.3e, and QGSJet-III-01. We parametrize the mean depth $\langle X_{\rm max}\rangle$ and the variance $σ^2(X_{\rm max})$ as functions of energy and mass. For the variance we compare a second-order polynomial model with an exponential model. In addition, we model the full $X_{\rm max}$ distributions with a three-parameter generalized Gumbel function. The Gumbel parameters are fitted using an unbinned likelihood and are validated by comparing the implied mean and variance with the raw CONEX samples and with the moment parametrizations. Across the full energy range considered, residuals between the parametrizations (or the Gumbel representation) and the simulations are at the level of a few g cm$^{-2}$ for the mean and a few (g cm$^{-2}$)$^2$ for the variance, making these parametrizations suitable for precision UHECR composition studies and forward-folding analyses of $X_{\rm max}$ distributions.

The Equation of State (EoS) of dense nuclear matter remains one of the most compelling open questions in high-energy astrophysics. While static EoS models are increasingly well-constrained by observations of binary neutron star (BNS) inspirals, the possibility of a dynamic phase transition occurring during the coalescence has been thus far deferred from standard gravitational-wave (GW) analyses. In this work, we investigate the detectability of such a phase transition, manifesting as a macroscopic shift in the tidal deformability parameter $Λ$, using GWs from Neutron Star-Black Hole (NSBH) coalescences. We argue that NSBH systems serve as a cleaner laboratory for this phenomenology than BNS systems due to the absence of the $\tildeΛ(Λ_1,Λ_2)$ degeneracy, allowing for the isolation of single-body tidal evolution. We introduce a phenomenological waveform model, TROYE (Transitional Representation Of varYing Equation-of-state), which stitches together two waveform approximants in the time domain to simulate a smooth but rapid transition between two equations of state during the late inspiral. We perform a comprehensive Bayesian injection and recovery campaign on 100 simulated events using the bilby inference library. Our results demonstrate that a phase transition corresponding to a tidal shift of $|ΔΛ| \gtrsim 400$ is detectable with Advanced LIGO design sensitivity, yielding decisive statistical evidence ($\ln B > 5$). We further identify a "V-shape" asymmetry in detectability, where "softening" transitions (decreasing $Λ$) are systematically easier to detect than "stiffening" ones due to the specific phase evolution of the tidal sector. Finally, we present "stress tests" showing that the transition remains recoverable even when marginalized over uncertainties in the stitching time and binary mass ratio.

Galaxy mergers play a crucial role in shaping the morphology, the star formation, and the mass growth of galaxies across cosmic time. While mergers have been extensively investigated in the local Universe, the evolution of their frequency and physical properties in the early Universe has yet to be fully understood. We investigate the role of mergers in a large spectroscopic sample of 1233 galaxies in the range 5<z<14 with good detection (S/N-pixel > 3) in JWST imaging, covering six different extragalactic fields. We identify mergers from rest-frame optical disturbances in F444W, using a combination of Gini, M-20, and Asymmetry parameters. We find a morphological merger fraction f_m that does not strongly evolve with redshift from z=0 to z ~ 8. The average f_m of our primary major merger condition (Gini+0.14xM-20 > 0.33, A>0.35) is ~ 5 %, which increases to ~13 % for major+minor merger tracers. Accounting for the evolving observability timescale of each tracer, we find that the merger rate is strongly increasing from z=1 to 7 by more than 1 dex, averaging ~ 2 merger/galaxy/Gyr at 5<z<10 for major mergers (in agreement with photometric pair studies), and a factor of 3 higher for minor+major mergers. We also perform SED modeling using available HST+JWST photometry to infer stellar masses and SFRs, using a non parametric star-formation history. We find that mergers at z > 5 have a significant impact, although significantly lower than at z<1, on the SFR of galaxies. When averaged over 10 Myr (comparable to the observability timescale of morphological disturbances), their SFRs are a factor of 1.7 higher than a mass and redshift matched sample of non-mergers, suggesting that mergers trigger new star-formation through short-lived powerful bursty episodes. Despite this, mergers contribute only by 5% - 10% to the mass build-up of galaxies in the redshift range explored.

To date, there has been significant interest in globular cluster (GC)-rich ultra-diffuse galaxies (UDGs) and the evidence that they have formed via an unexpected, ``failed galaxy'' formation pathway. The majority of the evidence for ``failed galaxy'' UDGs originates from spectroscopic observations targeting passive GC-rich UDGs, with a focus on those residing in galaxy clusters. In this work, we study the gas-rich, GC-rich group UDG MATLAS-42 and derive its stellar population properties using the Keck Cosmic Web Imager. We measure a redshift for the galaxy ($V_{\rm R, \star}=2433\pm8$~km s$^{-1}$), confirming the previous assumptions that it is both part of the NGC~502 group and has an associated HI-reservoir ($V_{\rm R,HI}=2423\pm 15$~km s$^{-1}$). We measure integrated stellar populations and find the galaxy to be both young (mass-weighted age $=3.2^{+2.6}_{-1.5}$Gyr) and of average-to-low metallicity ($[M/H]=-1.19^{+0.42}_{-0.30}$ dex). When considering these properties in the context of the galaxy's formation, we note it likely does not follow the ``failed galaxy'' formation pathway commonly attributed to GC-rich, cluster UDGs, as it has experienced recent star formation. At most it started failed, however, it has recently rejuvenated its star formation. Finally, we build a toy model of the passive evolution of this galaxy, finding that its relative GC-richness (i.e., $M_{\rm GC}/M_\star$) will likely decrease with time as GCs slowly evaporate/disrupt to contribute to the stellar mass of the galaxy. Due to this, we hypothesise that it is likely not a low redshift analogue of the progenitor to a ``failed galaxy'' UDGs.

We present a comprehensive radio polarization study of the Perseus molecular cloud using wideband L-band observations from the Karl G. Jansky Very Large Array. Our survey covers $\sim13.8$~deg$^2$ with a mean Stokes~$I$ sensitivity of $\sim80~μ$Jy~beam$^{-1}$, enabling the detection of 1410 compact radio sources. From this population, we construct a catalog of source properties, including positions, integrated flux densities, and spectral indices measured across nine spectral windows. The majority of sources exhibit negative spectral indices, consistent with non-thermal synchrotron emission. Using RM Synthesis and RM CLEAN techniques, we detect 205 polarized background sources above an $8σ$ threshold. This corresponds to a sampling density of $\sim14.8$~deg$^{-2}$, representing more than an order-of-magnitude increase compared to previous NVSS-based measurements. The resulting rotation measures exhibit coherent large-scale variations across the surveyed region, with additional small-scale structure superimposed. The enhanced sensitivity, frequency coverage, and sampling density of our observations enable a substantially improved mapping of the line-of-sight magnetic field component toward the Perseus molecular cloud compared to previous surveys.

We investigate the kinematics of the potential ultra-faint dwarf galaxy (UFD) UMa III/U1 using Bayesian inference to search for the signal of any potential intrinsic rotation. The magnitude of rotation is relevant to estimating the total mass of UMa III/U1, which is critical in determining whether or not UMa III/U1 is in fact a UFD, or possibly a star cluster home to a significant binary fraction. A non-rotating model and a rotational model are fitted for the current total population of member stars of UMa III/U1, finding that a non-rotating model was preferred by a factor of $\sim 5-12 \times$. This was repeated on a reduced population of UMa III/U1, where potential contaminant stars were removed. A similar preference for non-rotation was found for these reduced populations. We calculate a lower-bound rotational mass estimate for UMa III/U1 and a corresponding lower bound mass-to-light ratio of $ 734.4^{+339.0}_{-176.2} \mathrm{M_\odot} / \mathrm{L_\odot} $ for the total population. We conclude that UMa III/U1 still remains an ambiguous object with viable arguments for both the UFD and self-gravitating star cluster scenarios, however under both, UMa III/U1 is unlikely to be supported by rotational pressure.

Astronomical images often show sharp features that are caused by cosmic ray (CR) hits, hot pixels, or non-Gaussian noise. L.A.Cosmic (van Dokkum 2001) is a widely used edge detection algorithm that identifies and replaces such features. Here we describe dfcosmic, a direct python port of L.A.Cosmic utilizing PyTorch and C++ to enable efficient performance on both CPUs and GPUs. The code was developed for the MOTHRA array, which is projected to produce more than 1000 large format CMOS images every 15 minutes. Compared to previous python implementations, dfcosmic achieves a speed gain of at least 20%.

To gain deeper insights into galaxy evolution and the mechanisms driving transitions between different galaxy morphologies, we analyse the connection between star formation rate and stellar mass for 231 galaxies spanning Hubble types E1-Sdm from the Calar Alto Legacy Integral Field Spectroscopy Area survey using three complementary classification schemes. The Hubble classification provides structural information, the circular velocity curve classification$-$based on principal component analysis$-$ traces the total gravitational potential, and the Quenching classification$-$derived from H$_α$ equivalent width maps$-$indicates the spatial extent of quenched regions relative to star-forming areas. We find a clear separation of galaxy populations on the star formation rate-stellar mass plane. Late-type spirals with slow-rising circular velocity curves, represented by star-forming and quiescent-nuclear-ring galaxies, dominate the blue cloud. Early-type spirals with flat or round-peaked circular velocity curves belonging to centrally quiescent or mixed class populate the green valley, representing a transitional stage. Elliptical and lenticular galaxies with round- or sharp-peaked circular velocity curves from nearly retired or fully retired QSs reside on the red sequence. Furthermore, our results indicate that the morphological groups Sc-Scd, Sd-Sdm, and E1-E3 are characterized by a unique set of QSs and circular velocity curves, while galaxies with morphologies such as Sa-Sbc spread over multiple QSs and circular velocity curves. The distribution of the classification classes shows a tight link between galaxy structure, gravitational potential, and suppression of star-formation in the galaxies.

Neutral gas outflows play a crucial role in the baryon cycle of galaxies, their properties provide key insights into the transition from star formation to quiescence. In this work, we investigate the neutral gas outflow of 23 massive ($M_\star = 10^{10.1-11.6}\,\rm M_\odot$) quiescent galaxies (QGs) at z=2.82--4.61, selected from the JWST NIRSpec (R~1000) and NIRCam program DeepDive. We trace the neutral gas outflows using the NaI Doublet absorption lines and detect excess NaI D in 13/23 (57%) targets, of which 7/23 (30%) show blueshifted absorption with velocity offsets $|Δv|$ >~ 150 km/s. The z ~ 3.5 targets have $Δv$ similar to those of their local counterparts; they are also equivalent when compared in SFR--$Δv$ space. We derive mass outflow rates and identify the most extreme neutral gas outflow rate $\log(\dot M_{\rm out} / \mathrm{M_\odot \, yr}^{-1})=2.68\pm0.27$ beyond the local Universe, coincident with an X-ray AGN. For all NaI D detected systems, the inferred mass outflow rate can, in principle, suppress ongoing star formation; however, the outflows are unlikely to escape their hosts, suggesting recycling on relatively short timescales (~3--180 Myr), depending on the assumed potential and launching radius. All NaI D detected targets occupy the LI(N)ER region of the BPT diagram and/or are X-ray detected, but we find no strong correlation between ongoing AGN and the neutral outflow: 2/4 broad-line/X-ray AGNs are NaI D undetected -- yet, the outflows can be powered by fossil/episodic AGNs, and one broad-line target shows a possible P-Cygni profile that indicates strong outflows. As neutral outflows alone are not able to permanently quench star formation by removing gas in our sample at z ~ 3.5, the presence of gas cycling in and out of massive passive systems may instead be the signature of feedback-regulated quenching-maintenance processes.

Many post-starburst galaxies at $z\sim0.7$ have been shown to retain substantial molecular gas reservoirs yet host low ongoing star formation, suggesting that the remaining gas may be inefficient at forming stars during the early post-burst phase. We present new Atacama Large Millimeter/submillimeter Array CO(5-4) observations of nine gas-rich post-starburst galaxies at $z\sim0.7$ from the Studying Quenching in Intermediate-z Galaxies: Gas, angu$\vec{L}$ar momentum, and Evolution (SQuIGG$\vec{L}$E) survey, providing a view of the molecular gas excitation in these systems. Combined with existing CO(2-1) data, we detect CO(5-4) in 8/9 targets and find that most have moderate CO excitation with $r_{52}\equiv L'_{\rm CO(5-4)}/L'_{\rm CO(2-1)}\approx0.1-0.3$. These systems show no clear trend between $r_{52}$ and either total or surface-density of star formation. Specifically, all objects have $Σ_{\mathrm{SFR}} \sim 0.01-1\ \text{M}_\odot\ \text{yr}^{-1}\ \text{kpc}^{-2}$, consistent with compact, modest star formation, even when allowing for buried activity, as these galaxies decline from their peak. One object J1448+1010, which has clear optical, mid-infrared, and radio indicators of an active galactic nucleus, is an outlier with $r_{52}\approx0.6$; its elevated excitation likely requires significant non-stellar heating, with a contribution from potentially obscured star formation. Together, most gas-rich SQuIGG$\vec{L}$E post-starbursts have moderately excited molecular gas alongside little to modest star-forming activity, indicating that the remaining gas hosts relatively suppressed star formation efficiencies instead of strong buried starburst activity.

Star formation within filaments may arise due to the growth of cores according to linear perturbation theory. This implies a minimum core separation, as shorter modes would not be able to grow. While many observations agree with core separations by theoretical predictions, some observations also show star forming cores which lie closer together than the minimum wavelength given by perturbation theory. We explore whether non-linear effects during the late stages of core growth can explain the discrepancy between theory and observations. We perform three-dimensional hydrodynamical simulations with the Ramses code to follow the evolution of initial perturbations within filaments and compare the measured growth rates to expectations from theoretical models. Non-linear evolution sets in as soon as the core mass reaches a value where the gravitational potential is not any longer dominated by the cylindrical potential of the filament but by the spherical potential of the Bonnor-Ebert sphere. Consequently, core collapse is not triggered by the loss of hydrostatic stability of the filament but by the loss of hydrostatic stability of the Bonnor-Ebert sphere. As the core is embedded in the filament, the maximum core mass is given by the pressure within the filament which results in a constant line-mass threshold for core collapse. As core collapse is triggered as soon as overdensities reach a certain line-mass, cores which form as large line-mass perturbations during filament formation can go into direct collapse even if their separation is closer than predicted by linear perturbation theory. Therefore, our result can explain the discrepancy between theory and observations.

Young massive clusters (YMCs) provide favorable environments for frequent stellar collisions, potentially leading to the formation of very massive stars (VMSs) and seeds of intermediate-mass black holes (IMBHs). We investigate the role of repeated stellar collisions in YMCs using TITANS, a new suite of 18 direct $N$-body simulations. Our models span cluster masses $10^5 - 10^6\,\rm M_\odot$, half-mass densities $ρ_{\rm h}=100 - 10^5\,\rm M_\odot\,pc^{-3}$, and include high primordial binary fractions, consistent with observations of massive stars in young clusters. Overall, our simulations assume cluster properties that are typical of YMCs in the low-redshift Universe. We find that repeated stellar collisions are efficient only in the densest clusters with short relaxation times and are absent in systems with $ρ_{\rm h}<500\,\rm M_\odot\,pc^{-3}$ and $t_{\rm rh}>1.3\,\rm Gyr$. Rapid mass segregation allows massive stars to sink to the cluster center, merge, and undergo subsequent collisions, even in clusters with long core-collapse times. However, collision chains are typically triggered by primordial binary mergers and usually involve only two collisions. In our simulations, only three VMSs form through repeated collisions and reach $m_*>330\,\rm M_\odot$, while most VMSs have $m_*<300\,\rm M_\odot$ and form through primordial binary mergers. None constitute viable IMBH seeds, as their helium cores fall in the (pulsational) pair-instability regime. We form five IMBHs from stellar collisions involving stars at different evolutionary stages, while the dominant channel is the merger of stellar-mass black holes, producing twelve IMBHs. For properties typical of local YMCs, stellar collision chains are therefore inefficient in producing IMBHs more massive than $140\,\rm M_\odot$, as most collisionally formed VMSs attain masses that fall in the pair-instability regime.

We perform a comprehensive analysis of cosmic-ray propagation using the time-dependent AMS-02 flux measurements covering a full solar cycle, with particular emphasis on the role of solar modulation. We fit two representative Galactic propagation scenarios, convection- and re-acceleration-dominated models, in combination with three solar modulation prescriptions: the standard force-field approximation, an extended force-field model with a rigidity break, and the heliospheric propagation code $\texttt{HelMod}$. The inclusion of time-resolved antiproton data provides a unique probe of charge-sign-dependent modulation effects and low-energy systematics. We find that the force-field approximation can describe positively charged nuclei reasonably well outside the solar maximum in convection-dominated models, but fails during periods of high solar activity and for antiprotons at all times. In re-acceleration scenarios, strong degeneracies between solar modulation and low-energy propagation lead to unphysical results when simple modulation models are employed. Across all models, we identify systematic uncertainties of order 10-15% in the reconstructed local interstellar spectra and propagation parameters, driven by limitations in current solar modulation modelling. Compared to the percent level error of current measurements, these uncertainties significantly limit the precision of cosmic-ray studies. Future time-dependent measurements spanning a full 22-year solar cycle will be crucial to reduce these uncertainties.

Gradual crustal deformation can generate strongly twisted magnetic fields around magnetars, potentially triggering giant flares with total energies exceeding $10^{44}\,\mathrm{erg}$. In this Letter, we present the first relativistic magnetohydrodynamic simulation of a surface shear-driven magnetar eruption, capturing reconnection-driven plasma heating, the ejection of relativistically hot plasma, and the formation of a hot fireball confined within the inner magnetosphere. We find that magnetic reconnection in the equatorial current sheet launches a hot trailing outflow capable of powering the initial spike observed in giant flares, while simultaneously leaving behind a thermally stratified fireball with sufficient thermal energy to produce the pulsating, decaying tail. Together, these features provide a self-consistent physical framework for understanding the observed energetics of magnetar giant flares. The eruption also expels a magnetically dominated giant plasmoid carrying up to $\sim 9\%$ of the magnetosphere's total magnetic energy. Furthermore, our simulation demonstrates how the plasmoid drives the formation of a blast wave -- an important ingredient in models linking magnetar eruptions to fast radio bursts.

We use the high angular resolution NIRCam images from the PDRs4All program, combined with those of GTO program 1256, to extract key properties of disks in the Orion Nebula Cluster. We measure disk radii in silhouette against the bright background, identify dissociation fronts (DFs) and ionization fronts (IFs), determine Paschen $α$ intensities, and derive near-infrared spectral energy distributions (SEDs). From these diagnostics we define a typology of ONC disks. Type I sources show merged IFs and DFs close to the disk surface. Type II sources have DFs at the disk surface and IFs located tens of astronomical units away. Type III sources show a DF at the disk surface but no IF. For all types, PAH emission traces the PDR. We find that the disk radius $r_{\rm disk}$ increases with projected distance to the ionizing source $d_{\rm proj}$, following $r_{\rm disk} \propto d_{\rm proj}^{0.30}$, consistent with disk truncation by photoevaporation. Disk radii measured in the infrared are larger than those measured at millimeter wavelengths, suggesting radial dust segregation within the disks. In agreement with PDR models, the thermal pressure in the disk PDR increases with the FUV radiation field $G_0$, but with a flatter slope. Finally, the SEDs of candidate Jupiter Mass Binary Objects (JuMBOs) are similar to those of Type III disks, except for JuMBO24, which resembles a Type I or Type II source. Its SED is consistent with a young low-mass binary hosting an unresolved ionized disk.