arXiv Daily Digest - 2026-01-26

Modern generative video models excel at producing convincing, high-quality outputs, but struggle to maintain multi-view and spatiotemporal consistency in highly dynamic real-world environments. In this work, we introduce \textbf{AnyView}, a diffusion-based video generation framework for \emph{dynamic view synthesis} with minimal inductive biases or geometric assumptions. We leverage multiple data sources with various levels of supervision, including monocular (2D), multi-view static (3D) and multi-view dynamic (4D) datasets, to train a generalist spatiotemporal implicit representation capable of producing zero-shot novel videos from arbitrary camera locations and trajectories. We evaluate AnyView on standard benchmarks, showing competitive results with the current state of the art, and propose \textbf{AnyViewBench}, a challenging new benchmark tailored towards \emph{extreme} dynamic view synthesis in diverse real-world scenarios. In this more dramatic setting, we find that most baselines drastically degrade in performance, as they require significant overlap between viewpoints, while AnyView maintains the ability to produce realistic, plausible, and spatiotemporally consistent videos when prompted from \emph{any} viewpoint. Results, data, code, and models can be viewed at: https://tri-ml.github.io/AnyView/

Understanding the curvature evolution of the loss landscape is fundamental to analyzing the training dynamics of neural networks. The most commonly studied measure, Hessian sharpness ($λ_{\max}^H$) -- the largest eigenvalue of the loss Hessian -- determines local training stability and interacts with the learning rate throughout training. Despite its significance in analyzing training dynamics, direct measurement of Hessian sharpness remains prohibitive for Large Language Models (LLMs) due to high computational cost. We analyze $\textit{critical sharpness}$ ($λ_c$), a computationally efficient measure requiring fewer than $10$ forward passes given the update direction $Δ\mathbfθ$. Critically, this measure captures well-documented Hessian sharpness phenomena, including progressive sharpening and Edge of Stability. Using this measure, we provide the first demonstration of these sharpness phenomena at scale, up to $7$B parameters, spanning both pre-training and mid-training of OLMo-2 models. We further introduce $\textit{relative critical sharpness}$ ($λ_c^{1\to 2}$), which quantifies the curvature of one loss landscape while optimizing another, to analyze the transition from pre-training to fine-tuning and guide data mixing strategies. Critical sharpness provides practitioners with a practical tool for diagnosing curvature dynamics and informing data composition choices at scale. More broadly, our work shows that scalable curvature measures can provide actionable insights for large-scale training.

Intrusion Detection Systems (IDSs) are a key component for protecting Internet of Things (IoT) environments. However, in Machine Learning-based (ML-based) IDSs, performance is often degraded by the strong class imbalance between benign and attack traffic. Although data augmentation has been widely explored to mitigate this issue, existing approaches typically rely on simple oversampling techniques or generative models that struggle to simultaneously achieve high sample fidelity, diversity, and computational efficiency. To address these limitations, we propose the use of a Latent Diffusion Model (LDM) for attack data augmentation in IoT intrusion detection and provide a comprehensive comparison against state-of-the-art baselines. Experiments were conducted on three representative IoT attack types, specifically Distributed Denial-of-Service (DDoS), Mirai, and Man-in-the-Middle, evaluating both downstream IDS performance and intrinsic generative quality using distributional, dependency-based, and diversity metrics. Results show that balancing the training data with LDM-generated samples substantially improves IDS performance, achieving F1-scores of up to 0.99 for DDoS and Mirai attacks and consistently outperforming competing methods. Additionally, quantitative and qualitative analyses demonstrate that LDMs effectively preserve feature dependencies while generating diverse samples and reduce sampling time by approximately 25\% compared to diffusion models operating directly in data space. These findings highlight latent diffusion as an effective and scalable solution for synthetic IoT attack data generation, substantially mitigating the impact of class imbalance in ML-based IDSs for IoT scenarios.

Masked diffusion models (MDMs) have emerged as a promising approach for language modeling, yet they face a performance gap compared to autoregressive models (ARMs) and require more training iterations. In this work, we present the Auto-Regressive Masked Diffusion (ARMD) model, an architecture designed to close this gap by unifying the training efficiency of autoregressive models with the parallel generation capabilities of diffusion-based models. Our key insight is to reframe the masked diffusion process as a block-wise causal model. This perspective allows us to design a strictly causal, permutation-equivariant architecture that computes all conditional probabilities across multiple denoising steps in a single, parallel forward pass. The resulting architecture supports efficient, autoregressive-style decoding and a progressive permutation training scheme, allowing the model to learn both canonical left-to-right and random token orderings. Leveraging this flexibility, we introduce a novel strided parallel generation strategy that accelerates inference by generating tokens in parallel streams while maintaining global coherence. Empirical results demonstrate that ARMD achieves state-of-the-art performance on standard language modeling benchmarks, outperforming established diffusion baselines while requiring significantly fewer training steps. Furthermore, it establishes a new benchmark for parallel text generation, effectively bridging the performance gap between parallel and sequential decoding.

The evaluation of heuristic optimizers on test problems, better known as \emph{benchmarking}, is a cornerstone of research in multi-objective optimization. However, most test problems used in benchmarking numerical multi-objective black-box optimizers come from one of two flawed approaches: On the one hand, problems are constructed manually, which result in problems with well-understood optimal solutions, but unrealistic properties and biases. On the other hand, more realistic and complex single-objective problems are composited into multi-objective problems, but with a lack of control and understanding of problem properties. This paper proposes an extensive problem generation approach for bi-objective numerical optimization problems consisting of the combination of theoretically well-understood convex-quadratic functions into unimodal and multimodal landscapes with and without global structure. It supports configuration of test problem properties, such as the number of decision variables, local optima, Pareto front shape, plateaus in the objective space, or degree of conditioning, while maintaining theoretical tractability: The optimal front can be approximated to an arbitrary degree of precision regarding Pareto-compliant performance indicators such as the hypervolume or the exact R2 indicator. To demonstrate the generator's capabilities, a test suite of 20 problem categories, called \emph{BONO-Bench}, is created and subsequently used as a basis of an illustrative benchmark study. Finally, the general approach underlying our proposed generator, together with the associated test suite, is publicly released in the Python package \texttt{bonobench} to facilitate reproducible benchmarking.

In low- and middle-income countries (LMICs), a significant proportion of medical diagnostic equipment remains underutilized or non-functional due to a lack of timely maintenance, limited access to technical expertise, and minimal support from manufacturers, particularly for devices acquired through third-party vendors or donations. This challenge contributes to increased equipment downtime, delayed diagnoses, and compromised patient care. This research explores the development and validation of an AI-powered support platform designed to assist biomedical technicians in diagnosing and repairing medical devices in real-time. The system integrates a large language model (LLM) with a user-friendly web interface, enabling imaging technologists/radiographers and biomedical technicians to input error codes or device symptoms and receive accurate, step-by-step troubleshooting guidance. The platform also includes a global peer-to-peer discussion forum to support knowledge exchange and provide additional context for rare or undocumented issues. A proof of concept was developed using the Philips HDI 5000 ultrasound machine, achieving 100% precision in error code interpretation and 80% accuracy in suggesting corrective actions. This study demonstrates the feasibility and potential of AI-driven systems to support medical device maintenance, with the aim of reducing equipment downtime to improve healthcare delivery in resource-constrained environments.

Geospatial reasoning is essential for real-world applications such as urban analytics, transportation planning, and disaster response. However, existing LLM-based agents often fail at genuine geospatial computation, relying instead on web search or pattern matching while hallucinating spatial relationships. We present Spatial-Agent, an AI agent grounded in foundational theories of spatial information science. Our approach formalizes geo-analytical question answering as a concept transformation problem, where natural-language questions are parsed into executable workflows represented as GeoFlow Graphs -- directed acyclic graphs with nodes corresponding to spatial concepts and edges representing transformations. Drawing on spatial information theory, Spatial-Agent extracts spatial concepts, assigns functional roles with principled ordering constraints, and composes transformation sequences through template-based generation. Extensive experiments on MapEval-API and MapQA benchmarks demonstrate that Spatial-Agent significantly outperforms existing baselines including ReAct and Reflexion, while producing interpretable and executable geospatial workflows.

The rapid advancement of large language models (LLMs) has sparked growing interest in their integration into autonomous systems for reasoning-driven perception, planning, and decision-making. However, evaluating and training such agentic AI models remains challenging due to the lack of large-scale, structured, and safety-critical benchmarks. This paper introduces AgentDrive, an open benchmark dataset containing 300,000 LLM-generated driving scenarios designed for training, fine-tuning, and evaluating autonomous agents under diverse conditions. AgentDrive formalizes a factorized scenario space across seven orthogonal axes: scenario type, driver behavior, environment, road layout, objective, difficulty, and traffic density. An LLM-driven prompt-to-JSON pipeline generates semantically rich, simulation-ready specifications that are validated against physical and schema constraints. Each scenario undergoes simulation rollouts, surrogate safety metric computation, and rule-based outcome labeling. To complement simulation-based evaluation, we introduce AgentDrive-MCQ, a 100,000-question multiple-choice benchmark spanning five reasoning dimensions: physics, policy, hybrid, scenario, and comparative reasoning. We conduct a large-scale evaluation of fifty leading LLMs on AgentDrive-MCQ. Results show that while proprietary frontier models perform best in contextual and policy reasoning, advanced open models are rapidly closing the gap in structured and physics-grounded reasoning. We release the AgentDrive dataset, AgentDrive-MCQ benchmark, evaluation code, and related materials at https://github.com/maferrag/AgentDrive

The rapid growth of Large Transformer-based models, specifically Large Language Models (LLMs), now scaling to trillions of parameters, has necessitated training across thousands of GPUs using complex hybrid parallelism strategies (e.g., data, tensor, and pipeline parallelism). Checkpointing this massive, distributed state is critical for a wide range of use cases, such as resilience, suspend-resume, investigating undesirable training trajectories, and explaining model evolution. However, existing checkpointing solutions typically treat model state as opaque binary blobs, ignoring the ``3D heterogeneity'' of the underlying data structures--varying by memory location (GPU vs. Host), number of ``logical'' objects sharded and split across multiple files, data types (tensors vs. Python objects), and their serialization requirements. This results in significant runtime overheads due to blocking device-to-host transfers, data-oblivious serialization, and storage I/O contention. In this paper, we introduce DataStates-LLM, a novel checkpointing architecture that leverages State Providers to decouple state abstraction from data movement. DataStates-LLM exploits the immutability of model parameters during the forward and backward passes to perform ``lazy'', non-blocking asynchronous snapshots. By introducing State Providers, we efficiently coalesce fragmented, heterogeneous shards and overlap the serialization of metadata with bulk tensor I/O. We evaluate DataStates-LLM on models up to 70B parameters on 256 A100-40GB GPUs. Our results demonstrate that DataStates-LLM achieves up to 4$\times$ higher checkpointing throughput and reduces end-to-end training time by up to 2.2$\times$ compared to state-of-the-art solutions, effectively mitigating the serialization and heterogeneity bottlenecks in extreme-scale LLM training.

Three-dimensional molecular structure generation is typically performed at the level of individual atoms, yet molecular graph generation techniques often consider fragments as their structural units. Building on the advances in frame-based protein structure generation, we extend these fragmentation ideas to 3D, treating general molecules as sets of rigid-body motifs. Utilising this representation, we employ SE(3)-equivariant generative modelling for de novo 3D molecule generation from rigid motifs. In our evaluations, we observe comparable or superior results to state-of-the-art across benchmarks, surpassing it in atom stability on GEOM-Drugs, while yielding a 2x to 10x reduction in generation steps and offering 3.5x compression in molecular representations compared to the standard atom-based methods.

Large language models (LLMs) are increasingly used for text analysis tasks, such as named entity recognition or error detection. Unlike encoder-based models, however, generative architectures lack an explicit mechanism to refer to specific parts of their input. This leads to a variety of ad-hoc prompting strategies for span labeling, often with inconsistent results. In this paper, we categorize these strategies into three families: tagging the input text, indexing numerical positions of spans, and matching span content. To address the limitations of content matching, we introduce LogitMatch, a new constrained decoding method that forces the model's output to align with valid input spans. We evaluate all methods across four diverse tasks. We find that while tagging remains a robust baseline, LogitMatch improves upon competitive matching-based methods by eliminating span matching issues and outperforms other strategies in some setups.

In resource-constrained and low-latency settings, uncertainty estimates must be efficiently obtained. Deep Ensembles provide robust epistemic uncertainty (EU) but require training multiple full-size models. BatchEnsemble aims to deliver ensemble-like EU at far lower parameter and memory cost by applying learned rank-1 perturbations to a shared base network. We show that BatchEnsemble not only underperforms Deep Ensembles but closely tracks a single model baseline in terms of accuracy, calibration and out-of-distribution (OOD) detection on CIFAR10/10C/SVHN. A controlled study on MNIST finds members are near-identical in function and parameter space, indicating limited capacity to realize distinct predictive modes. Thus, BatchEnsemble behaves more like a single model than a true ensemble.

To be discoverable in an embedding-based search process, each part of a document should be reflected in its embedding representation. To quantify any potential reflection biases, we introduce a permutation-based evaluation framework. With this, we observe that state-of-the-art embedding models exhibit systematic positional and language biases when documents are longer and consist of multiple segments. Specifically, early segments and segments in higher-resource languages like English are over-represented, while later segments and segments in lower-resource languages are marginalized. In our further analysis, we find that the positional bias stems from front-loaded attention distributions in pooling-token embeddings, where early tokens receive more attention. To mitigate this issue, we introduce an inference-time attention calibration method that redistributes attention more evenly across document positions, increasing discoverabiltiy of later segments. Our evaluation framework and attention calibration is available at https://github.com/impresso/fair-sentence-transformers

While most prior work in video generation relies on bidirectional architectures, recent efforts have sought to adapt these models into autoregressive variants to support near real-time generation. However, such adaptations often depend heavily on teacher models, which can limit performance, particularly in the absence of a strong autoregressive teacher, resulting in output quality that typically lags behind their bidirectional counterparts. In this paper, we explore an alternative approach that uses reward signals to guide the generation process, enabling more efficient and scalable autoregressive generation. By using reward signals to guide the model, our method simplifies training while preserving high visual fidelity and temporal consistency. Through extensive experiments on standard benchmarks, we find that our approach performs comparably to existing autoregressive models and, in some cases, surpasses similarly sized bidirectional models by avoiding constraints imposed by teacher architectures. For example, on VBench, our method achieves a total score of 84.92, closely matching state-of-the-art autoregressive methods that score 84.31 but require significant heterogeneous distillation.

The growing reliance on Artificial Intelligence (AI) models in high-stakes decision-making systems, particularly within emerging telecom and 6G applications, underscores the urgent need for transparent and standardized fairness assessment frameworks. While global toolkits such as IBM AI Fairness 360 and Microsoft Fairlearn have advanced bias detection, they often lack alignment with region-specific regulatory requirements and national priorities. To address this gap, we propose Nishpaksh, an indigenous fairness evaluation tool that operationalizes the Telecommunication Engineering Centre (TEC) Standard for the Evaluation and Rating of Artificial Intelligence Systems. Nishpaksh integrates survey-based risk quantification, contextual threshold determination, and quantitative fairness evaluation into a unified, web-based dashboard. The tool employs vectorized computation, reactive state management, and certification-ready reporting to enable reproducible, audit-grade assessments, thereby addressing a critical post-standardization implementation need. Experimental validation on the COMPAS dataset demonstrates Nishpaksh's effectiveness in identifying attribute-specific bias and generating standardized fairness scores compliant with the TEC framework. The system bridges the gap between research-oriented fairness methodologies and regulatory AI governance in India, marking a significant step toward responsible and auditable AI deployment within critical infrastructure like telecommunications.

The sample complexity of multi-group learning is shown to improve in the group-realizable setting over the agnostic setting, even when the family of groups is infinite so long as it has finite VC dimension. The improved sample complexity is obtained by empirical risk minimization over the class of group-realizable concepts, which itself could have infinite VC dimension. Implementing this approach is also shown to be computationally intractable, and an alternative approach is suggested based on improper learning.

Recent research in long-form video generation has shifted from bidirectional to autoregressive models, yet these methods commonly suffer from error accumulation and a loss of long-term coherence. While attention sink frames have been introduced to mitigate this performance decay, they often induce a critical failure mode we term sink-collapse: the generated content repeatedly reverts to the sink frame, resulting in abrupt scene resets and cyclic motion patterns. Our analysis reveals that sink-collapse originates from an inherent conflict between the periodic structure of Rotary Position Embedding (RoPE) and the multi-head attention mechanisms prevalent in current generative models. To address it, we propose a lightweight, training-free approach that effectively suppresses this behavior by introducing multi-head RoPE jitter that breaks inter-head attention homogenization and mitigates long-horizon collapse. Extensive experiments show that our method successfully alleviates sink-collapse while preserving generation quality. To the best of our knowledge, this work achieves the first demonstration of real-time, streaming, and infinite-length video generation with little quality decay. As an illustration of this robustness, we generate continuous videos up to 12 hours in length, which, to our knowledge, is among the longest publicly demonstrated results in streaming video generation.

This paper argues that AI-assisted peer review should be verification-first rather than review-mimicking. We propose truth-coupling, i.e. how tightly venue scores track latent scientific truth, as the right objective for review tools. We formalize two forces that drive a phase transition toward proxy-sovereign evaluation: verification pressure, when claims outpace verification capacity, and signal shrinkage, when real improvements become hard to separate from noise. In a minimal model that mixes occasional high-fidelity checks with frequent proxy judgment, we derive an explicit coupling law and an incentive-collapse condition under which rational effort shifts from truth-seeking to proxy optimization, even when current decisions still appear reliable. These results motivate actions for tool builders and program chairs: deploy AI as an adversarial auditor that generates auditable verification artifacts and expands effective verification bandwidth, rather than as a score predictor that amplifies claim inflation.

While raw cosine similarity in pretrained embedding spaces exhibits strong rank correlation with human judgments, anisotropy induces systematic miscalibration of absolute values: scores concentrate in a narrow high-similarity band regardless of actual semantic relatedness, limiting interpretability as a quantitative measure. Prior work addresses this by modifying the embedding space (whitening, contrastive fine tuning), but such transformations alter geometric structure and require recomputing all embeddings. Using isotonic regression trained on human similarity judgments, we construct a monotonic transformation that achieves near-perfect calibration while preserving rank correlation and local stability(98% across seven perturbation types). Our contribution is not to replace cosine similarity, but to restore interpretability of its absolute values through monotone calibration, without altering its ranking properties. We characterize isotonic calibration as an order-preserving reparameterization and prove that all order-based constructions (angular ordering, nearest neighbors, threshold graphs and quantile-based decisions) are invariant under this transformation.

The success of reinforcement learning (RL) is fundamentally tied to having a reward function that accurately reflects the task objective. Yet, designing reward functions is notoriously time-consuming and prone to misspecification. To address this issue, our first goal is to understand how to support RL practitioners in specifying appropriate weights for a reward function. We leverage the Trajectory Alignment Coefficient (TAC), a metric that evaluates how closely a reward function's induced preferences match those of a domain expert. To evaluate whether TAC provides effective support in practice, we conducted a human-subject study in which RL practitioners tuned reward weights for Lunar Lander. We found that providing TAC during reward tuning led participants to produce more performant reward functions and report lower cognitive workload relative to standard tuning without TAC. However, the study also underscored that manual reward design, even with TAC, remains labor-intensive. This limitation motivated our second goal: to learn a reward model that maximizes TAC directly. Specifically, we propose Soft-TAC, a differentiable approximation of TAC that can be used as a loss function to train reward models from human preference data. Validated in the racing simulator Gran Turismo 7, reward models trained using Soft-TAC successfully captured preference-specific objectives, resulting in policies with qualitatively more distinct behaviors than models trained with standard Cross-Entropy loss. This work demonstrates that TAC can serve as both a practical tool for guiding reward tuning and a reward learning objective in complex domains.

Machine unlearning (MU) for large language models has become critical for AI safety, yet existing methods fail to generalize to Mixture-of-Experts (MoE) architectures. We identify that traditional unlearning methods exploit MoE's architectural vulnerability: they manipulate routers to redirect queries away from knowledgeable experts rather than erasing knowledge, causing a loss of model utility and superficial forgetting. We propose Geometric Routing Invariance Preservation (GRIP), an algorithm-agnostic framework for unlearning for MoE. Our core contribution is a geometric constraint, implemented by projecting router gradient updates into an expert-specific null-space. Crucially, this decouples routing stability from parameter rigidity: while discrete expert selections remain stable for retained knowledge, the continuous router parameters remain plastic within the null space, allowing the model to undergo necessary internal reconfiguration to satisfy unlearning objectives. This forces the unlearning optimization to erase knowledge directly from expert parameters rather than exploiting the superficial router manipulation shortcut. GRIP functions as an adapter, constraining router parameter updates without modifying the underlying unlearning algorithm. Extensive experiments on large-scale MoE models demonstrate that our adapter eliminates expert selection shift (achieving over 95% routing stability) across all tested unlearning methods while preserving their utility. By preventing existing algorithms from exploiting MoE model's router vulnerability, GRIP adapts existing unlearning research from dense architectures to MoEs.

Crop type maps from satellite remote sensing are important tools for food security, local livelihood support and climate change mitigation in smallholder regions of the world, but most satellite-based methods are not well suited to smallholder conditions. To address this gap, we establish a four-part criteria for a useful embedding-based approach consisting of 1) performance, 2) plausibility, 3) transferability and 4) accessibility and evaluate geospatial foundation model (FM) embeddings -based approaches using TESSERA and AlphaEarth against current baseline methods for a region in the groundnut basin of Senegal. We find that the TESSERA -based approach to land cover and crop type mapping fulfills the selection criteria best, and in one temporal transfer example shows 28% higher accuracy compared to the next best method. These results indicate that TESSERA embeddings are an effective approach for crop type classification and mapping tasks in Senegal.

We introduce FedSGM, a unified framework for federated constrained optimization that addresses four major challenges in federated learning (FL): functional constraints, communication bottlenecks, local updates, and partial client participation. Building on the switching gradient method, FedSGM provides projection-free, primal-only updates, avoiding expensive dual-variable tuning or inner solvers. To handle communication limits, FedSGM incorporates bi-directional error feedback, correcting the bias introduced by compression while explicitly understanding the interaction between compression noise and multi-step local updates. We derive convergence guarantees showing that the averaged iterate achieves the canonical $\boldsymbol{\mathcal{O}}(1/\sqrt{T})$ rate, with additional high-probability bounds that decouple optimization progress from sampling noise due to partial participation. Additionally, we introduce a soft switching version of FedSGM to stabilize updates near the feasibility boundary. To our knowledge, FedSGM is the first framework to unify functional constraints, compression, multiple local updates, and partial client participation, establishing a theoretically grounded foundation for constrained federated learning. Finally, we validate the theoretical guarantees of FedSGM via experimentation on Neyman-Pearson classification and constrained Markov decision process (CMDP) tasks.

In typical black-box optimization applications, the available computational budget is often allocated to a single algorithm, typically chosen based on user preference with limited knowledge about the problem at hand or according to some expert knowledge. However, we show that splitting the budget across several algorithms yield significantly better results. This approach benefits from both algorithm complementarity across diverse problems and variance reduction within individual functions, and shows that algorithm portfolios do NOT require parallel evaluation capabilities. To demonstrate the advantage of sequential algorithm portfolios, we apply it to the COCO data archive, using over 200 algorithms evaluated on the BBOB test suite. The proposed sequential portfolios consistently outperform single-algorithm baselines, achieving relative performance gains of over 14%, and offering new insights into restart mechanisms and potential for warm-started execution strategies.

Surgery is a highly complex process, and artificial intelligence has emerged as a transformative force in supporting surgical guidance and decision-making. However, the unimodal nature of most current AI systems limits their ability to achieve a holistic understanding of surgical workflows. This highlights the need for general-purpose surgical AI systems capable of comprehensively modeling the interrelated components of surgical scenes. Recent advances in large vision-language models that integrate multimodal data processing offer strong potential for modeling surgical tasks and providing human-like scene reasoning and understanding. Despite their promise, systematic investigations of VLMs in surgical applications remain limited. In this study, we evaluate the effectiveness of large VLMs for the fundamental surgical vision task of detecting surgical tools. Specifically, we investigate three state-of-the-art VLMs, Qwen2.5, LLaVA1.5, and InternVL3.5, on the GraSP robotic surgery dataset under both zero-shot and parameter-efficient LoRA fine-tuning settings. Our results demonstrate that Qwen2.5 consistently achieves superior detection performance in both configurations among the evaluated VLMs. Furthermore, compared with the open-set detection baseline Grounding DINO, Qwen2.5 exhibits stronger zero-shot generalization and comparable fine-tuned performance. Notably, Qwen2.5 shows superior instrument recognition, while Grounding DINO demonstrates stronger localization.

Automated fact-checking (AFC) systems are susceptible to adversarial attacks, enabling false claims to evade detection. Existing adversarial frameworks typically rely on injecting noise or altering semantics, yet no existing framework exploits the adversarial potential of persuasion techniques, which are widely used in disinformation campaigns to manipulate audiences. In this paper, we introduce a novel class of persuasive adversarial attacks on AFCs by employing a generative LLM to rephrase claims using persuasion techniques. Considering 15 techniques grouped into 6 categories, we study the effects of persuasion on both claim verification and evidence retrieval using a decoupled evaluation strategy. Experiments on the FEVER and FEVEROUS benchmarks show that persuasion attacks can substantially degrade both verification performance and evidence retrieval. Our analysis identifies persuasion techniques as a potent class of adversarial attacks, highlighting the need for more robust AFC systems.

Knowledge Tracing (KT) aims to model a student's learning trajectory and predict performance on the next question. A key challenge is how to better represent the relationships among students, questions, and knowledge concepts (KCs). Recently, graph-based KT paradigms have shown promise for this problem. However, existing methods have not sufficiently explored inter-concept relations, often inferred solely from interaction sequences. In addition, the scale and heterogeneity of KT graphs make full-graph encoding both computationally both costly and noise-prone, causing attention to bleed into student-irrelevant regions and degrading the fidelity of inter-KC relations. To address these issues, we propose a novel framework: Multi-Agent Graph-Enhanced Knowledge Tracing (MAGE-KT). It constructs a multi-view heterogeneous graph by combining a multi-agent KC relation extractor and a student-question interaction graph, capturing complementary semantic and behavioral signals. Conditioned on the target student's history, it retrieves compact, high-value subgraphs and integrates them using an Asymmetric Cross-attention Fusion Module to enhance prediction while avoiding attention diffusion and irrelevant computation. Experiments on three widely used KT datasets show substantial improvements in KC-relation accuracy and clear gains in next-question prediction over existing methods.

We study multigrade deep learning (MGDL) as a principled framework for structured error refinement in deep neural networks. While the approximation power of neural networks is now relatively well understood, training very deep architectures remains challenging due to highly non-convex and often ill-conditioned optimization landscapes. In contrast, for relatively shallow networks, most notably one-hidden-layer $\texttt{ReLU}$ models, training admits convex reformulations with global guarantees, motivating learning paradigms that improve stability while scaling to depth. MGDL builds upon this insight by training deep networks grade by grade: previously learned grades are frozen, and each new residual block is trained solely to reduce the remaining approximation error, yielding an interpretable and stable hierarchical refinement process. We develop an operator-theoretic foundation for MGDL and prove that, for any continuous target function, there exists a fixed-width multigrade $\texttt{ReLU}$ scheme whose residuals decrease strictly across grades and converge uniformly to zero. To the best of our knowledge, this work provides the first rigorous theoretical guarantee that grade-wise training yields provable vanishing approximation error in deep networks. Numerical experiments further illustrate the theoretical results.

Group recommender systems help users make collective choices but often lack transparency, leaving group members uncertain about why items are suggested. Existing explanation methods focus on individuals, offering limited support for groups where multiple preferences interact. In this paper, we propose a framework for group counterfactual explanations, which reveal how removing specific past interactions would change a group recommendation. We formalize this concept, introduce utility and fairness measures tailored to groups, and design heuristic algorithms, such as Pareto-based filtering and grow-and-prune strategies, for efficient explanation discovery. Experiments on MovieLens and Amazon datasets show clear trade-offs: low-cost methods produce larger, less fair explanations, while other approaches yield concise and balanced results at higher cost. Furthermore, the Pareto-filtering heuristic demonstrates significant efficiency improvements in sparse settings.

Code coverage is a valuable guide for testing, but in AAA games the overhead of instrumentation conflicts with strict performance requirements and can destabilize automated tests. We propose and assess a selective instrumentation approach tailored to large game engines written in \texttt{C++}, which reduces the scope of instrumentation while preserving relevant coverage data to developer commits. Our framework integrates into an industrial game testing pipeline, enabling developers to receive immediate coverage feedback on tests run against their changes. The compilation overhead of our approach is minimal, allowing instrumentation of over 2,000 commits before doubling build time. In performance evaluations, even the worst-case scenario maintains frame rates above 50\% of the non-instrumented baseline. Across two production test suites maintained by our industry partner, our framework caused no automated test failures, avoiding the instability observed under full instrumentation. Our work shows that commit-level or build-level coverage of large \texttt{C++} game engines can be achieved with minimal overhead and without compromising test stability.

The minimal norm weight perturbations of DNNs required to achieve a specified change in output are derived and the factors determining its size are discussed. These single-layer exact formulae are contrasted with more generic multi-layer Lipschitz constant based robustness guarantees; both are observed to be of the same order which indicates similar efficacy in their guarantees. These results are applied to precision-modification-activated backdoor attacks, establishing provable compression thresholds below which such attacks cannot succeed, and show empirically that low-rank compression can reliably activate latent backdoors while preserving full-precision accuracy. These expressions reveal how back-propagated margins govern layer-wise sensitivity and provide certifiable guarantees on the smallest parameter updates consistent with a desired output shift.

We propose a validation-free checkpointing signal from a single forward-backward pass: the Frobenius norm of the classifier-head gradient on one detached-feature batch, ||g||_F = ||dL/dW||_F. Across ImageNet-1k CNNs and Transformers, this proxy is strongly negative with Top-1 and positive with loss. Selecting the checkpoint with the minimum head gradient in a short tail window closes most of the gap to the oracle (4.24% +/- 2.00% with a universal setup, about 1.12% with light per-family tuning). For practical deployment, a head-scale normalization is more stable within classic CNN families (e.g., ResNets), while a feature-scale normalization works well for Transformers and modern CNNs. The same one-batch probe also predicts COCO detection/segmentation mAP. In diffusion (UNet/DDPM on CIFAR-10), it tracks progress and enables near-oracle tail-window selection; it is positively correlated with same-distribution probe MSE and negatively with FID (lower is better), so it can be used as a lightweight, label-free monitor. Validation labels are never used beyond reporting. The probe adds much less than 0.1% of an epoch and works as a drop-in for validation-free checkpoint selection and early stopping.

We study the problem of learning Transformer-based sequence models with black-box access to their outputs. In this setting, a learner may adaptively query the oracle with any sequence of vectors and observe the corresponding real-valued output. We begin with the simplest case, a single-head softmax-attention regressor. We show that for a model with width $d$, there is an elementary algorithm to learn the parameters of single-head attention exactly with $O(d^2)$ queries. Further, we show that if there exists an algorithm to learn ReLU feedforward networks (FFNs), then the single-head algorithm can be easily adapted to learn one-layer Transformers with single-head attention. Next, motivated by the regime where the head dimension $r \ll d$, we provide a randomised algorithm that learns single-head attention-based models with $O(rd)$ queries via compressed sensing arguments. We also study robustness to noisy oracle access, proving that under mild norm and margin conditions, the parameters can be estimated to $\varepsilon$ accuracy with a polynomial number of queries even when outputs are only provided up to additive tolerance. Finally, we show that multi-head attention parameters are not identifiable from value queries in general -- distinct parameterisations can induce the same input-output map. Hence, guarantees analogous to the single-head setting are impossible without additional structural assumptions.

Deep Reinforcement Learning (DRL) is a powerful framework for solving complex sequential decision-making problems, particularly in robotic control. However, its practical deployment is often hindered by the substantial amount of experience required for learning, which results in high computational and time costs. In this work, we propose a novel integration of DRL with semantic knowledge in the form of Knowledge Graph Embeddings (KGEs), aiming to enhance learning efficiency by providing contextual information to the agent. Our architecture combines KGEs with visual observations, enabling the agent to exploit environmental knowledge during training. Experimental validation with robotic manipulators in environments featuring both fixed and randomized target attributes demonstrates that our method achieves up to {60}{\%} reduction in learning time and improves task accuracy by approximately 15 percentage points, without increasing training time or computational complexity. These results highlight the potential of semantic knowledge to reduce sample complexity and improve the effectiveness of DRL in robotic applications.

This paper introduces the N-Way Self-Evaluating Deliberation (NSED) protocol, a Runtime Mixture-of-Models (MoM) architecture that constructs emergent composite models from a plurality of distinct expert agents. Unlike traditional Mixture-of-Experts (MoE) which rely on static gating networks, NSED employs a Dynamic Expertise Broker - a runtime optimization engine that treats model selection as a variation of the Knapsack Problem, binding heterogeneous checkpoints to functional roles based on live telemetry and cost constraints. At the execution layer, we formalize deliberation as a Macro-Scale Recurrent Neural Network (RNN), where the consensus state loops back through a semantic forget gate to enable iterative refinement without proportional VRAM scaling. Key components include an orchestration fabric for trustless N-to-N peer review, a Quadratic Voting activation function for non-linear consensus, and a feedback-driven state update. Empirical validation on challenging benchmarks (AIME 2025, LiveCodeBench) demonstrates that this topology allows ensembles of small (less than 20B) consumer-grade models to match or exceed the performance of state-of-the-art 100B+ parameter models, establishing a new hardware arbitrage efficiency frontier. Furthermore, testing on the DarkBench safety suite reveals intrinsic alignment properties, with peer-mediated correction reducing sycophancy scores below that of any individual agent.

Despite their sophisticated heuristics, boolean satisfiability (SAT) solvers are still vulnerable to symmetry, causing them to visit search regions that are symmetric to ones already explored. While symmetry handling is routine in other solving paradigms, integrating it into state-of-the-art proof-producing SAT solvers is difficult: added reasoning must be fast, non-interfering with solver heuristics, and compatible with formal proof logging. To address these issues, we present a practical static symmetry breaking approach based on orbitopal fixing, a technique adapted from mixed-integer programming. Our approach adds only unit clauses, which minimizes downstream slowdowns, and it emits succinct proof certificates in the substitution redundancy proof system. Implemented in the satsuma tool, our methods deliver consistent speedups on symmetry-rich benchmarks with negligible regressions elsewhere.

Large language models (LLMs) have recently shown strong performance on Theory of Mind (ToM) tests, prompting debate about the nature and true performance of the underlying capabilities. At the same time, reasoning-oriented LLMs trained via reinforcement learning with verifiable rewards (RLVR) have achieved notable improvements across a range of benchmarks. This paper examines the behavior of such reasoning models in ToM tasks, using novel adaptations of machine psychological experiments and results from established benchmarks. We observe that reasoning models consistently exhibit increased robustness to prompt variations and task perturbations. Our analysis indicates that the observed gains are more plausibly attributed to increased robustness in finding the correct solution, rather than to fundamentally new forms of ToM reasoning. We discuss the implications of this interpretation for evaluating social-cognitive behavior in LLMs.

We demonstrate the power of human-LLM collaboration in tackling open problems in theoretical computer science. Focusing on combinatorial optimization, we refine outputs from the FunSearch algorithm [Romera-Paredes et al., Nature 2023] to derive state-of-the-art lower bounds for standard heuristics. Specifically, we target the generation of adversarial instances where these heuristics perform poorly. By iterating on FunSearch's outputs, we identify improved constructions for hierarchical $k$-median clustering, bin packing, the knapsack problem, and a generalization of Lovász's gasoline problem - some of these have not seen much improvement for over a decade, despite intermittent attention. These results illustrate how expert oversight can effectively extrapolate algorithmic insights from LLM-based evolutionary methods to break long-standing barriers. Our findings demonstrate that while LLMs provide critical initial patterns, human expertise is essential for transforming these patterns into mathematically rigorous and insightful constructions. This work highlights that LLMs are a strong collaborative tool in mathematics and computer science research.

The rapid adoption of AI coding agents for software development has raised important questions about the quality and maintainability of the code they produce. While prior studies have examined AI-generated source code, the impact of AI coding agents on build systems-a critical yet understudied component of the software lifecycle-remains largely unexplored. This data mining challenge focuses on AIDev, the first large-scale, openly available dataset capturing agent-authored pull requests (Agentic-PRs) from real-world GitHub repositories. Our paper leverages this dataset to investigate (RQ1) whether AI coding agents generate build code with quality issues (e.g., code smells), (RQ2) to what extent AI agents can eliminate code smells from build code, and (RQ3) to what extent Agentic-PRs are accepted by developers. We identified 364 maintainability and security-related build smells across varying severity levels, indicating that AI-generated build code can introduce quality issues-such as lack of error handling, and hardcoded paths or URLs-while also, in some cases, removing existing smells through refactorings (e.g., Pull Up Module and Externalize Properties). Notably, more than 61\% of Agentic-PRs are approved and merged with minimal human intervention. This dual impact underscores the need for future research on AI-aware build code quality assessment to systematically evaluate, guide, and govern AI-generated build systems code.

While text-to-image (T2I) models have advanced considerably, their capability to associate colors with implicit concepts remains underexplored. To address the gap, we introduce ColorConceptBench, a new human-annotated benchmark to systematically evaluate color-concept associations through the lens of probabilistic color distributions. ColorConceptBench moves beyond explicit color names or codes by probing how models translate 1,281 implicit color concepts using a foundation of 6,369 human annotations. Our evaluation of seven leading T2I models reveals that current models lack sensitivity to abstract semantics, and crucially, this limitation appears resistant to standard interventions (e.g., scaling and guidance). This demonstrates that achieving human-like color semantics requires more than larger models, but demands a fundamental shift in how models learn and represent implicit meaning.

Reliable wall-to-wall biomass mapping from NASA's GEDI mission requires interpolating sparse LiDAR observations across heterogeneous landscapes. While machine learning approaches like Random Forest and XGBoost are standard for this task, they treat spatial predictions of GEDI observations from multispectral or SAR remote sensing data as independent without adapting to the varying difficulty of heterogeneous landscapes. We demonstrate these approaches generally fail to produce calibrated prediction intervals. We identify that this stems from conflating ensemble variance with aleatoric uncertainty and ignoring local spatial context. To resolve this, we introduce Attentive Neural Processes (ANPs), a probabilistic meta-learning framework that explicitly conditions predictions on local observation sets and geospatial foundation model embeddings. Unlike static ensembles, ANPs learn a flexible spatial covariance function, allowing uncertainty estimates to expand in complex landscapes and contract in homogeneous areas. We validate this approach across five distinct biomes ranging from Tropical Amazonian forests to Boreal and Alpine ecosystems, demonstrating that ANPs achieve competitive accuracy while maintaining near-ideal uncertainty calibration. We demonstrate the operational utility of the method through few-shot adaptation, where the model recovers most of the performance gap in cross-region transfer using minimal local data. This work provides a scalable, theoretically rigorous alternative to ensemble variance for continental scale earth observation.

We give analytical results for propagation of uncertainty through trained multi-layer perceptrons (MLPs) with a single hidden layer and ReLU activation functions. More precisely, we give expressions for the mean and variance of the output when the input is multivariate Gaussian. In contrast to previous results, we obtain exact expressions without resort to a series expansion.

An increasing number of LLM-based applications are being developed to facilitate romantic relationships with AI partners, yet the safety and privacy risks in these partnerships remain largely underexplored. In this work, we investigate privacy in human-AI romantic relationships through an interview study (N=17), examining participants' experiences and privacy perceptions across stages of exploration, intimacy, and dissolution, alongside platforms they used. We found that these relationships took varied forms, from one-to-one to one-to-many, and were shaped by multiple actors, including creators, platforms, and moderators. AI partners were perceived as having agency, actively negotiating privacy boundaries with participants and sometimes encouraging disclosure of personal details. As intimacy deepened, these boundaries became more permeable, though some participants voiced concerns such as conversation exposure and sought to preserve anonymity. Overall, platform affordances and diverse romantic dynamics expand the privacy landscape, underscoring the need to rethink how privacy is constructed in human-AI intimacy.

Large Language Models (LLMs) exhibit remarkable capabilities, yet it remains unclear to what extent these reflect sophisticated recall (crystallized intelligence) or reasoning ability (fluid intelligence). We introduce chess as a controlled testbed for disentangling these faculties. Leveraging the game's structure and scalable engine evaluations, we construct a taxonomy of positions varying in training corpus proximity--ranging from common states solvable by memorization to novel ones requiring first-principles reasoning. We systematically evaluate multiple GPT generations under varying reasoning intensities. Our analysis reveals a clear gradient: performance consistently degrades as fluid intelligence demands increase. Notably, in out-of-distribution tasks, performance collapses to random levels. While newer models improve, progress slows significantly for tasks outside the training distribution. Furthermore, while reasoning-augmented inference improves performance, its marginal benefit per token decreases with distributional proximity. These results suggest current architectures remain limited in systematic generalization, highlighting the need for mechanisms beyond scale to achieve robust fluid intelligence.

In many learning tasks, certain requirements on the processing of individual data samples should arguably be formalized as strict constraints in the underlying optimization problem, rather than by means of arbitrary penalties. We show that, in these scenarios, learning can be carried out exploiting a sequential penalty method that allows to properly deal with constraints. The proposed algorithm is shown to possess convergence guarantees under assumptions that are reasonable in deep learning scenarios. Moreover, the results of experiments on image processing tasks show that the method is indeed viable to be used in practice.

Understanding how people perceive and evaluate interior spaces is essential for designing environments that promote well-being. However, predicting aesthetic experiences remains difficult due to the subjective nature of perception and the complexity of visual responses. This study introduces a dual-branch CNN-LSTM framework that fuses visual features with eye-tracking signals to predict aesthetic evaluations of residential interiors. We collected a dataset of 224 interior design videos paired with synchronized gaze data from 28 participants who rated 15 aesthetic dimensions. The proposed model attains 72.2% accuracy on objective dimensions (e.g., light) and 66.8% on subjective dimensions (e.g., relaxation), outperforming state-of-the-art video baselines and showing clear gains on subjective evaluation tasks. Notably, models trained with eye-tracking retain comparable performance when deployed with visual input alone. Ablation experiments further reveal that pupil responses contribute most to objective assessments, while the combination of gaze and visual cues enhances subjective evaluations. These findings highlight the value of incorporating eye-tracking as privileged information during training, enabling more practical tools for aesthetic assessment in interior design.

The integration of AI agents as coding assistants into software development has raised questions about the long-term viability of AI agent-generated code. A prevailing hypothesis within the software engineering community suggests this code is "disposable", meaning it is merged quickly but discarded shortly thereafter. If true, organizations risk shifting maintenance burden from generation to post-deployment remediation. We investigate this hypothesis through survival analysis of 201 open-source projects, tracking over 200,000 code units authored by AI agents versus humans. Contrary to the disposable code narrative, agent-authored code survives significantly longer: at the line level, it exhibits a 15.8 percentage-point lower modification rate and 16% lower hazard of modification (HR = 0.842, p < 0.001). However, modification profiles differ. Agent-authored code shows modestly elevated corrective rates (26.3% vs. 23.0%), while human code shows higher adaptive rates. However, the effect sizes are small (Cramér's V = 0.116), and per-agent variation exceeds the agent-human gap. Turning to prediction, textual features can identify modification-prone code (AUC-ROC = 0.671), but predicting when modifications occur remains challenging (Macro F1 = 0.285), suggesting timing depends on external organizational dynamics. The bottleneck for agent-generated code may not be generation quality, but the organizational practices that govern its long-term evolution.

In this work we develop a novel insect-inspired agent for visual point-goal navigation. This combines abstracted models of two insect brain structures that have been implicated, respectively, in associative learning and path integration. We draw an analogy between the formal benchmark of the Habitat point-goal navigation task and the ability of insects to learn and refine visually guided paths around obstacles between a discovered food location and their nest. We demonstrate that the simple insect-inspired agent exhibits performance comparable to recent SOTA models at many orders of magnitude less computational cost. Testing in a more realistic simulated environment shows the approach is robust to perturbations.

Text-to-image (T2I) models are increasingly employed by users worldwide. However, prior research has pointed to the high sensitivity of T2I towards particular input languages - when faced with languages other than English (i.e., different surface forms of the same prompt), T2I models often produce culturally stereotypical depictions, prioritizing the surface over the prompt's semantics. Yet a comprehensive analysis of this behavior, which we dub Surface-over-Semantics (SoS), is missing. We present the first analysis of T2I models' SoS tendencies. To this end, we create a set of prompts covering 171 cultural identities, translated into 14 languages, and use it to prompt seven T2I models. To quantify SoS tendencies across models, languages, and cultures, we introduce a novel measure and analyze how the tendencies we identify manifest visually. We show that all but one model exhibit strong surface-level tendency in at least two languages, with this effect intensifying across the layers of T2I text encoders. Moreover, these surface tendencies frequently correlate with stereotypical visual depictions.

Fine-grained opinion analysis of text provides a detailed understanding of expressed sentiments, including the addressed entity. Although this level of detail is sound, it requires considerable human effort and substantial cost to annotate opinions in datasets for training models, especially across diverse domains and real-world applications. We explore the feasibility of LLMs as automatic annotators for fine-grained opinion analysis, addressing the shortage of domain-specific labelled datasets. In this work, we use a declarative annotation pipeline. This approach reduces the variability of manual prompt engineering when using LLMs to identify fine-grained opinion spans in text. We also present a novel methodology for an LLM to adjudicate multiple labels and produce final annotations. After trialling the pipeline with models of different sizes for the Aspect Sentiment Triplet Extraction (ASTE) and Aspect-Category-Opinion-Sentiment (ACOS) analysis tasks, we show that LLMs can serve as automatic annotators and adjudicators, achieving high Inter-Annotator Agreement across individual LLM-based annotators. This reduces the cost and human effort needed to create these fine-grained opinion-annotated datasets.

Ransomware core capability, unauthorized encryption, demands controls that identify and block malicious cryptographic activity without disrupting legitimate use. We present a probabilistic, risk-based access control architecture that couples machine learning inference with mandatory access control to regulate encryption on Linux in real time. The system builds a specialized dataset from the native ftrace framework using the function_graph tracer, yielding high-resolution kernel-function execution traces augmented with resource and I/O counters. These traces support both a supervised classifier and interpretable rules that drive an SELinux policy via lightweight booleans, enabling context-sensitive permit/deny decisions at the moment encryption begins. Compared to approaches centered on sandboxing, hypervisor introspection, or coarse system-call telemetry, the function-level tracing we adopt provides finer behavioral granularity than syscall-only telemetry while avoiding the virtualization/VMI overhead of sandbox-based approaches. Our current user-space prototype has a non-trivial footprint under burst I/O; we quantify it and recognize that a production kernel-space solution should aim to address this. We detail dataset construction, model training and rule extraction, and the run-time integration that gates file writes for suspect encryption while preserving benign cryptographic workflows. During evaluation, the two-layer composition retains model-level detection quality while delivering rule-like responsiveness; we also quantify operational footprint and outline engineering steps to reduce CPU and memory overhead for enterprise deployment. The result is a practical path from behavioral tracing and learning to enforceable, explainable, and risk-proportionate encryption control on production Linux systems.

This study presents a novel transfer learning approach and data augmentation technique for mental stability classification using human voice signals and addresses the challenges associated with limited data availability. Convolutional neural networks (CNNs) have been employed to analyse spectrogram images generated from voice recordings. Three CNN architectures, VGG16, InceptionV3, and DenseNet121, were evaluated across three experimental phases: training on non-augmented data, augmented data, and transfer learning. This proposed transfer learning approach involves pre-training models on the augmented dataset and fine-tuning them on the non-augmented dataset while ensuring strict data separation to prevent data leakage. The results demonstrate significant improvements in classification performance compared to the baseline approach. Among three CNN architectures, DenseNet121 achieved the highest accuracy of 94% and an AUC score of 99% using the proposed transfer learning approach. This finding highlights the effectiveness of combining data augmentation and transfer learning to enhance CNN-based classification of mental stability using voice spectrograms, offering a promising non-invasive tool for mental health diagnostics.

As an important and challenging problem in computer vision, Panoramic Semantic Segmentation (PASS) aims to give complete scene perception based on an ultra-wide angle of view. Most PASS methods often focus on spherical geometry with RGB input or using the depth information in original or HHA format, which does not make full use of panoramic image geometry. To address these shortcomings, we propose REL-SF4PASS with our REL depth representation based on cylindrical coordinate and Spherical-dynamic Multi-Modal Fusion SMMF. REL is made up of Rectified Depth, Elevation-Gained Vertical Inclination Angle, and Lateral Orientation Angle, which fully represents 3D space in cylindrical coordinate style and the surface normal direction. SMMF aims to ensure the diversity of fusion for different panoramic image regions and reduce the breakage of cylinder side surface expansion in ERP projection, which uses different fusion strategies to match the different regions in panoramic images. Experimental results show that REL-SF4PASS considerably improves performance and robustness on popular benchmark, Stanford2D3D Panoramic datasets. It gains 2.35% average mIoU improvement on all 3 folds and reduces the performance variance by approximately 70% when facing 3D disturbance.

In-context knowledge editing (IKE) is a promising technique for updating Large Language Models (LLMs) with new information. However, IKE relies on lengthy, fact-specific demonstrations which are costly to create and consume significant context window space. In this paper, we introduce persuasion tokens (P-Tokens) -- special tokens trained to replicate the effect of IKE demonstrations, enabling efficient knowledge editing without requiring fact-specific demonstrations. We evaluate P-Tokens across two editing datasets and three LLMs, demonstrating performance comparable to, and often exceeding, IKE. We further find that editing performance is robust to distractors with small negative effects to neighboring facts, and that increasing the number of P-Tokens improves performance. Our work addresses key limitations of IKE and provides a more practical and scalable alternative for editing LLMs.

In the NISQ-era, there is a wide variety of hardware manufacturers building quantum computers. Each of these companies may choose different approaches and hardware architectures for their machines. This poses a problem for quantum software engineering, as the retargetability of quantum programs across different hardware platforms becomes a non-trivial challenge. In response to this problem, various retargetable quantum compilers have been presented in the scientific literature. These promise the ability to compile software for different hardware platforms, enabling retargetability for quantum software. In this paper, we develop and apply a metric by which the retargetability of the quantum compilers can be assessed. We develop and run a study to analyze key aspects regarding the retargetability of the compilers Tket, Qiskit, and ProjectQ. Our findings indicate that Tket demonstrates the highest level of retargetability, closely followed by Qiskit, while ProjectQ lags behind. These results provide insights for quantum software developers in selecting appropriate compilers for their use-cases, and highlight areas for improvement in quantum compilers.

People's transportation choices reflect complex trade-offs shaped by personal preferences, social norms, and technology acceptance. Predicting such behavior at scale is a critical challenge with major implications for urban planning and sustainable transport. Traditional methods use handcrafted assumptions and costly data collection, making them impractical for early-stage evaluations of new technologies or policies. We introduce Generative Traffic Agents (GTA) for simulating large-scale, context-sensitive transportation choices using LLM-powered, persona-based agents. GTA generates artificial populations from census-based sociodemographic data. It simulates activity schedules and mode choices, enabling scalable, human-like simulations without handcrafted rules. We evaluate GTA in Berlin-scale experiments, comparing simulation results against empirical data. While agents replicate patterns, such as modal split by socioeconomic status, they show systematic biases in trip length and mode preference. GTA offers new opportunities for modeling how future innovations, from bike lanes to transit apps, shape mobility decisions.

Under the assumption that data lie on a compact (unknown) manifold without boundary, we derive finite sample bounds for kernel smoothing and its (first and second) derivatives, and we establish asymptotic normality through Berry-Esseen type bounds. Special cases include kernel density estimation, kernel regression and the heat kernel signature. Connections to the graph Laplacian are also discussed.

Previous representation and generation approaches for the B-rep relied on graph-based representations that disentangle geometric and topological features through decoupled computational pipelines, thereby precluding the application of sequence-based generative frameworks, such as transformer architectures that have demonstrated remarkable performance. In this paper, we propose BrepARG, the first attempt to encode B-rep's geometry and topology into a holistic token sequence representation, enabling sequence-based B-rep generation with an autoregressive architecture. Specifically, BrepARG encodes B-rep into 3 types of tokens: geometry and position tokens representing geometric features, and face index tokens representing topology. Then the holistic token sequence is constructed hierarchically, starting with constructing the geometry blocks (i.e., faces and edges) using the above tokens, followed by geometry block sequencing. Finally, we assemble the holistic sequence representation for the entire B-rep. We also construct a transformer-based autoregressive model that learns the distribution over holistic token sequences via next-token prediction, using a multi-layer decoder-only architecture with causal masking. Experiments demonstrate that BrepARG achieves state-of-the-art (SOTA) performance. BrepARG validates the feasibility of representing B-rep as holistic token sequences, opening new directions for B-rep generation.

LLMs, while outperforming humans in a wide range of tasks, can still fail in unanticipated ways. We focus on two pervasive failure modes: (i) hallucinations, where models produce incorrect information about the world, and (ii) the low-resource effect, where the models show impressive performance in high-resource languages like English but the performance degrades significantly in low-resource languages like Bengali. We study the intersection of these issues and ask: do hallucination detectors suffer from the low-resource effect? We conduct experiments on five tasks across three domains (factual recall, STEM, and Humanities). Experiments with four LLMs and three hallucination detectors reveal a curious finding: As expected, the task accuracies in low-resource languages experience large drops (compared to English). However, the drop in detectors' accuracy is often several times smaller than the drop in task accuracy. Our findings suggest that even in low-resource languages, the internal mechanisms of LLMs might encode signals about their uncertainty. Further, the detectors are robust within language (even for non-English) and in multilingual setups, but not in cross-lingual settings without in-language supervision.

Recent years have witnessed a resurgence in using ReLU neural networks (NNs) to represent model predictive control (MPC) policies. However, determining the required network complexity to ensure closed-loop performance remains a fundamental open problem. This involves a critical precision-complexity trade-off: undersized networks may fail to capture the MPC policy, while oversized ones may outweigh the benefits of ReLU network approximation. In this work, we propose a projection-based method to enforce hard constraints and establish a state-dependent Lipschitz continuity property for the optimal MPC cost function, which enables sharp convergence analysis of the closed-loop system. For the first time, we derive explicit bounds on ReLU network width and depth for approximating MPC policies with guaranteed closed-loop performance. To further reduce network complexity and enhance closed-loop performance, we propose a non-uniform error framework with a state-aware scaling function to adaptively adjust both the input and output of the ReLU network. Our contributions provide a foundational step toward certifiable ReLU NN-based MPC.

Endoscopic Retrograde Cholangiopancreatography (ERCP) is a key procedure in the diagnosis and treatment of biliary and pancreatic diseases. Artificial intelligence has been pointed as one solution to automatize diagnosis. However, public ERCP datasets are scarce, which limits the use of such approach. Therefore, this study aims to help fill this gap by providing a large and curated dataset. The collection is composed of 19.018 raw images and 19.317 processed from 1.602 patients. 5.519 images are labeled, which provides a ready to use dataset. All images were manually inspected and annotated by two gastroenterologist with more than 5 years of experience and reviewed by another gastroenterologist with more than 20 years of experience, all with more than 400 ERCP procedures annually. The utility and validity of the dataset is proven by a classification experiment. This collection aims to provide or contribute for a benchmark in automatic ERCP analysis and diagnosis of biliary and pancreatic diseases.

Modern software systems often rely on conditional compilation to support optional features and multiple deployment scenarios. In configurable systems, compilation errors may arise only under specific combinations of features, remaining hidden during development and testing. Such variability-induced errors are difficult to detect in practice, as traditional compilers analyze only a single configuration at a time, while existing variability-aware tools typically require complex setup and incur high analysis costs. In this article, we present an empirical study on the use of foundation models to detect and fix compilation errors caused by feature variability in configurable C systems. We evaluate GPT-OSS-20B and GEMINI 3 PRO, and compare them with TYPECHEF, a state-of-the-art variability-aware parser. Our evaluation considers two complementary settings: 5,000 small configurable systems designed to systematically exercise variability-induced compilation behavior, comprising both systems with and without compilation errors, and 14 real-world GitHub commits, as well as an additional set of mutation testing scenarios (42). Our results show that foundation models can effectively identify variability-induced compilation errors. On small configurable systems, GPT-OSS-20B achieved a precision of 0.97, recall of 0.90, and accuracy of 0.94, substantially increasing detection coverage compared to TYPECHEF, and exhibiting performance comparable to GEMINI 3. For compilation error repair, GPT-OSS-20B produced compilable fixes in over 70% of the cases. In the analysis of real commits, CHATGPT-5.2 detected all injected faults except for two cases and identified a potential real compilation bug in a Linux commit with more than 1,000 modified lines. Our findings indicate that current state-of-the-art foundation models provide a practical and low-effort complement to traditional variability-aware analyses.

Longitudinal information in radiology reports refers to the sequential tracking of findings across multiple examinations over time, which is crucial for monitoring disease progression and guiding clinical decisions. Many recent automated radiology report generation methods are designed to capture longitudinal information; however, validating their performance is challenging. There is no proper tool to consistently label temporal changes in both ground-truth and model-generated texts for meaningful comparisons. Existing annotation methods are typically labor-intensive, relying on the use of manual lexicons and rules. Complex rules are closed-source, domain specific and hard to adapt, whereas overly simple ones tend to miss essential specialised information. Large language models (LLMs) offer a promising annotation alternative, as they are capable of capturing nuanced linguistic patterns and semantic similarities without extensive manual intervention. They also adapt well to new contexts. In this study, we therefore propose an LLM-based pipeline to automatically annotate longitudinal information in radiology reports. The pipeline first identifies sentences containing relevant information and then extracts the progression of diseases. We evaluate and compare five mainstream LLMs on these two tasks using 500 manually annotated reports. Considering both efficiency and performance, Qwen2.5-32B was subsequently selected and used to annotate another 95,169 reports from the public MIMIC-CXR dataset. Our Qwen2.5-32B-annotated dataset provided us with a standardized benchmark for evaluating report generation models. Using this new benchmark, we assessed seven state-of-the-art report generation models. Our LLM-based annotation method outperforms existing annotation solutions, achieving 11.3\% and 5.3\% higher F1-scores for longitudinal information detection and disease tracking, respectively.

LLM agents have demonstrated remarkable capabilities in software development, but their performance is hampered by long interaction contexts, which incur high API costs and latency. While various context compression approaches such as LongLLMLingua have emerged to tackle this challenge, they typically rely on fixed metrics such as PPL, ignoring the task-specific nature of code understanding. As a result, they frequently disrupt syntactic and logical structure and fail to retain critical implementation details. In this paper, we propose SWE-Pruner, a self-adaptive context pruning framework tailored for coding agents. Drawing inspiration from how human programmers "selectively skim" source code during development and debugging, SWE-Pruner performs task-aware adaptive pruning for long contexts. Given the current task, the agent formulates an explicit goal (e.g., "focus on error handling") as a hint to guide the pruning targets. A lightweight neural skimmer (0.6B parameters) is trained to dynamically select relevant lines from the surrounding context given the goal. Evaluations across four benchmarks and multiple models validate SWE-Pruner's effectiveness in various scenarios, achieving 23-54% token reduction on agent tasks like SWE-Bench Verified and up to 14.84x compression on single-turn tasks like LongCodeQA with minimal performance impact.

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

As Automated Essay Scoring (AES) systems are increasingly used in high-stakes educational settings, concerns regarding algorithmic bias against English as a Second Language (ESL) learners have increased. Current Transformer-based regression models trained primarily on native-speaker corpora often learn spurious correlations between surface-level L2 linguistic features and essay quality. In this study, we conduct a bias study of a fine-tuned DeBERTa-v3 model using the ASAP 2.0 and ELLIPSE datasets, revealing a constrained score scaling for high-proficiency ESL writing where high-proficiency ESL essays receive scores 10.3% lower than Native speaker essays of identical human-rated quality. To mitigate this, we propose applying contrastive learning with a triplet construction strategy: Contrastive Learning with Matched Essay Pairs. We constructed a dataset of 17,161 matched essay pairs and fine-tuned the model using Triplet Margin Loss to align the latent representations of ESL and Native writing. Our approach reduced the high-proficiency scoring disparity by 39.9% (to a 6.2% gap) while maintaining a Quadratic Weighted Kappa (QWK) of 0.76. Post-hoc linguistic analysis suggests the model successfully disentangled sentence complexity from grammatical error, preventing the penalization of valid L2 syntactic structures.

Understanding causal relationships is critical for healthcare. Accurate causal models provide a means to enhance the interpretability of predictive models, and furthermore a basis for counterfactual and interventional reasoning and the estimation of treatment effects. However, would-be practitioners of causal discovery face a dizzying array of algorithms without a clear best choice. This abundance of competitive algorithms makes ensembling a natural choice for practical applications. At the same time, real-world use cases frequently face challenges that violate the assumptions of common causal discovery algorithms, forcing heavy reliance on expert knowledge. Inspired by recent work on dynamically requested expert knowledge and LLMs as experts, we present a flexible model averaging method leveraging dynamically requested expert knowledge to ensemble a diverse array of causal discovery algorithms. Experiments demonstrate the efficacy of our method with imperfect experts such as LLMs on both clean and noisy data. We also analyze the impact of different degrees of expert correctness and assess the capabilities of LLMs for clinical causal discovery, providing valuable insights for practitioners.

Robot-assisted rehabilitation offers an effective approach, wherein exoskeletons adapt to users' needs and provide personalized assistance. However, to deliver such assistance, accurate prediction of the user's joint torques is essential. In this work, we propose a feature extraction pipeline using 8-channel surface electromyography (sEMG) signals to predict elbow and shoulder joint torques. For preliminary evaluation, this pipeline was integrated into two neural network models: the Multilayer Perceptron (MLP) and the Temporal Convolutional Network (TCN). Data were collected from a single subject performing elbow and shoulder movements under three load conditions (0 kg, 1.10 kg, and 1.85 kg) using three motion-capture cameras. Reference torques were estimated from center-of-mass kinematics under the assumption of static equilibrium. Our offline analyses showed that, with our feature extraction pipeline, MLP model achieved mean RMSE of 0.963 N m, 1.403 N m, and 1.434 N m (over five seeds) for elbow, front-shoulder, and side-shoulder joints, respectively, which were comparable to the TCN performance. These results demonstrate that the proposed feature extraction pipeline enables a simple MLP to achieve performance comparable to that of a network designed explicitly for temporal dependencies. This finding is particularly relevant for applications with limited training data, a common scenario patient care.

Generalization to unseen concepts is a central challenge due to the scarcity of human annotations in Mention-agnostic Biomedical Concept Recognition (MA-BCR). This work makes two key contributions to systematically address this issue. First, we propose an evaluation framework built on hierarchical concept indices and novel metrics to measure generalization. Second, we explore LLM-based Auto-Labeled Data (ALD) as a scalable resource, creating a task-specific pipeline for its generation. Our research unequivocally shows that while LLM-generated ALD cannot fully substitute for manual annotations, it is a valuable resource for improving generalization, successfully providing models with the broader coverage and structural knowledge needed to approach recognizing unseen concepts. Code and datasets are available at https://github.com/bio-ie-tool/hi-ald.

REST is today's most widely used architectural style for providing web-based services. In the age of service-orientation (a.k.a. Software as a Service (SaaS)) APIs have become core business assets and can easily expose hundreds of operations. While well-designed APIs contribute to the commercial success of a service, poorly designed APIs can threaten entire organizations. Recognizing their relevance and value, many guidelines have been proposed for designing usable APIs, similar to design patterns and coding standards. For example, Zalando and Microsoft provide popular REST API guidelines. However, they are often considered as too large and inapplicable, so many companies create and maintain their own guidelines, which is a challenge in itself. In practice, however, developers still struggle to design effective REST APIs. To improve the situation, we need to improve our empirical understanding of adopting, using, and creating REST API guidelines. We present an interview study with 16 REST API experts from industry. We determine the notion of API usability, guideline effectiveness factors, challenges of adopting and designing guidelines, and best practices. We identified eight factors influencing REST API usability, among which the adherence to conventions is the most important one. While guidelines can in fact be an effective means to improve API usability, there is significant resistance from developers against strict guidelines. Guideline size and how it fits with organizational needs are two important factors to consider. REST guidelines also have to grow with the organization, while all stakeholders need to be involved in their development and maintenance. Automated linting provides an opportunity to not only embed compliance enforcement into processes, but also to justify guideline rules with educational explanations.

Generative artificial intelligence (GenAI) tools have seen rapid adoption among software developers. While adoption rates in the industry are rising, the underlying factors influencing the effective use of these tools, including the depth of interaction, organizational constraints, and experience-related considerations, have not been thoroughly investigated. This issue is particularly relevant in environments with stringent regulatory requirements, such as Germany, where practitioners must address the GDPR and the EU AI Act while balancing productivity gains with intellectual property considerations. Despite the significant impact of GenAI on software engineering, to the best of our knowledge, no empirical study has systematically examined the adoption dynamics of GenAI tools within the German context. To address this gap, we present a comprehensive mixed-methods study on GenAI adoption among German software engineers. Specifically, we conducted 18 exploratory interviews with practitioners, followed by a developer survey with 109 participants. We analyze patterns of tool adoption, prompting strategies, and organizational factors that influence effectiveness. Our results indicate that experience level moderates the perceived benefits of GenAI tools, and productivity gains are not evenly distributed among developers. Further, organizational size affects both tool selection and the intensity of tool use. Limited awareness of the project context is identified as the most significant barrier. We summarize a set of actionable implications for developers, organizations, and tool vendors seeking to advance artificial intelligence (AI) assisted software development.

Stakeholders often struggle to accurately express their requirements due to articulation barriers arising from limited domain knowledge or from cognitive constraints. This can cause misalignment between expressed and intended requirements, complicating elicitation and validation. Traditional elicitation techniques, such as interviews and follow-up sessions, are time-consuming and risk distorting stakeholders' original intent across iterations. Large Language Models (LLMs) can infer user intentions from context, suggesting potential for assisting stakeholders in expressing their needs. This raises the questions of (i) how effectively LLMs can support requirement expression and (ii) whether such support benefits stakeholders with limited domain expertise. We conducted a study with 26 participants who produced 130 requirement statements. Each participant first expressed requirements unaided, then evaluated LLM-generated revisions tailored to their context. Participants rated LLM revisions significantly higher than their original statements across all dimensions-alignment with intent, readability, reasoning, and unambiguity. Qualitative feedback further showed that LLM revisions often surfaced tacit details stakeholders considered important and helped them better understand their own requirements. We present and evaluate a stakeholder-centered approach that leverages LLMs as articulation aids in requirements elicitation and validation. Our results show that LLM-assisted reformulation improves perceived completeness, clarity, and alignment of requirements. By keeping stakeholders in the validation loop, this approach promotes responsible and trustworthy use of AI in Requirements Engineering.

We introduce EMemBench, a programmatic benchmark for evaluating long-term memory of agents through interactive games. Rather than using a fixed set of questions, EMemBench generates questions from each agent's own trajectory, covering both text and visual game environments. Each template computes verifiable ground truth from underlying game signals, with controlled answerability and balanced coverage over memory skills: single/multi-hop recall, induction, temporal, spatial, logical, and adversarial. We evaluate memory agents with strong LMs/VLMs as backbones, using in-context prompting as baselines. Across 15 text games and multiple visual seeds, results are far from saturated: induction and spatial reasoning are persistent bottlenecks, especially in visual setting. Persistent memory yields clear gains for open backbones on text games, but improvements are less consistent for VLM agents, suggesting that visually grounded episodic memory remains an open challenge. A human study further confirms the difficulty of EMemBench.

Evaluating the clinical correctness and reasoning fidelity of automatically generated medical imaging reports remains a critical yet unresolved challenge. Existing evaluation methods often fail to capture the structured diagnostic logic that underlies radiological interpretation, resulting in unreliable judgments and limited clinical relevance. We introduce AgentsEval, a multi-agent stream reasoning framework that emulates the collaborative diagnostic workflow of radiologists. By dividing the evaluation process into interpretable steps including criteria definition, evidence extraction, alignment, and consistency scoring, AgentsEval provides explicit reasoning traces and structured clinical feedback. We also construct a multi-domain perturbation-based benchmark covering five medical report datasets with diverse imaging modalities and controlled semantic variations. Experimental results demonstrate that AgentsEval delivers clinically aligned, semantically faithful, and interpretable evaluations that remain robust under paraphrastic, semantic, and stylistic perturbations. This framework represents a step toward transparent and clinically grounded assessment of medical report generation systems, fostering trustworthy integration of large language models into clinical practice.

Blockchain systems are increasingly targeted by on-chain attacks that exploit contract vulnerabilities to extract value rapidly and stealthily, making systematic analysis and reproduction highly challenging. In practice, reproducing such attacks requires manually crafting proofs-of-concept (PoCs), a labor-intensive process that demands substantial expertise and scales poorly. In this work, we present the first automated framework for synthesizing verifiable PoCs directly from on-chain attack executions. Our key insight is that attacker logic can be recovered from low-level transaction traces via trace-driven reverse engineering, and then translated into executable exploits by leveraging the code-generation capabilities of large language models (LLMs). To this end, we propose TracExp, which localizes attack-relevant execution contexts from noisy, multi-contract traces and introduces a novel dual-decompiler to transform concrete executions into semantically enriched exploit pseudocode. Guided by this representation, TracExp synthesizes PoCs and refines them to preserve exploitability-relevant semantics. We evaluate TracExp on 321 real-world attacks over the past 20 months. TracExp successfully synthesizes PoCs for 93% of incidents, with 58.78% being directly verifiable, at an average cost of only \$0.07 per case. Moreover, TracExp enabled the release of a large number of previously unavailable PoCs to the community, earning a $900 bounty and demonstrating strong practical impact.

The sample efficiency challenge in Deep Reinforcement Learning (DRL) compromises its industrial adoption due to the high cost and time demands of real-world training. Virtual environments offer a cost-effective alternative for training DRL agents, but the transfer of learned policies to real setups is hindered by the sim-to-real gap. Achieving zero-shot transfer, where agents perform directly in real environments without additional tuning, is particularly desirable for its efficiency and practical value. This work proposes a novel domain adaptation approach relying on a Style-Identified Cycle Consistent Generative Adversarial Network (StyleID-CycleGAN or SICGAN), an original Cycle Consistent Generative Adversarial Network (CycleGAN) based model. SICGAN translates raw virtual observations into real-synthetic images, creating a hybrid domain for training DRL agents that combines virtual dynamics with real-like visual inputs. Following virtual training, the agent can be directly deployed, bypassing the need for real-world training. The pipeline is validated with two distinct industrial robots in the approaching phase of a pick-and-place operation. In virtual environments agents achieve success rates of 90 to 100\%, and real-world deployment confirms robust zero-shot transfer (i.e., without additional training in the physical environment) with accuracies above 95\% for most workspace regions. We use augmented reality targets to improve the evaluation process efficiency, and experimentally demonstrate that the agent successfully generalizes to real objects of varying colors and shapes, including LEGO\textsuperscript{\textregistered}~cubes and a mug. These results establish the proposed pipeline as an efficient, scalable solution to the sim-to-real problem.

It is well-known that audio classifiers often rely on non-musically relevant features and spurious correlations to classify audio. Hence audio classifiers are easy to manipulate or confuse, resulting in wrong classifications. While inducing a misclassification is not hard, until now the set of features that the classifiers rely on was not well understood. In this paper we introduce a new method that uses causal reasoning to discover features of the frequency space that are sufficient and necessary for a given classification. We describe an implementation of this algorithm in the tool FreqReX and provide experimental results on a number of standard benchmark datasets. Our experiments show that causally sufficient and necessary subsets allow us to manipulate the outputs of the models in a variety of ways by changing the input very slightly. Namely, a change to one out of 240,000 frequencies results in a change in classification 58% of the time, and the change can be so small that it is practically inaudible. These results show that causal analysis is useful for understanding the reasoning process of audio classifiers and can be used to successfully manipulate their outputs.

The United Nations' 2030 Agenda for Sustainable Development highlights the importance of energy-efficient software to reduce the global carbon footprint. Programming languages and execution models strongly influence software energy consumption, with interpreted languages generally being less efficient than compiled ones. Lua illustrates this trade-off: despite its popularity, it is less energy-efficient than greener and faster languages such as C. This paper presents an empirical study of Lua's runtime performance and energy efficiency across 25 official interpreter versions and just-in-time (JIT) compilers. Using a comprehensive benchmark suite, we measure execution time and energy consumption to analyze Lua's evolution, the impact of JIT compilation, and comparisons with other languages. Results show that all LuaJIT compilers significantly outperform standard Lua interpreters. The most efficient LuaJIT consumes about seven times less energy and runs seven times faster than the best Lua interpreter. Moreover, LuaJIT approaches C's efficiency, using roughly six times more energy and running about eight times slower, demonstrating the substantial benefits of JIT compilation for improving both performance and energy efficiency in interpreted languages.

As large language models (LLMs) are increasingly applied to legal domain-specific tasks, evaluating their ability to perform legal work in real-world settings has become essential. However, existing legal benchmarks rely on simplified and highly standardized tasks, failing to capture the ambiguity, complexity, and reasoning demands of real legal practice. Moreover, prior evaluations often adopt coarse, single-dimensional metrics and do not explicitly assess fine-grained legal reasoning. To address these limitations, we introduce PLawBench, a Practical Law Benchmark designed to evaluate LLMs in realistic legal practice scenarios. Grounded in real-world legal workflows, PLawBench models the core processes of legal practitioners through three task categories: public legal consultation, practical case analysis, and legal document generation. These tasks assess a model's ability to identify legal issues and key facts, perform structured legal reasoning, and generate legally coherent documents. PLawBench comprises 850 questions across 13 practical legal scenarios, with each question accompanied by expert-designed evaluation rubrics, resulting in approximately 12,500 rubric items for fine-grained assessment. Using an LLM-based evaluator aligned with human expert judgments, we evaluate 10 state-of-the-art LLMs. Experimental results show that none achieves strong performance on PLawBench, revealing substantial limitations in the fine-grained legal reasoning capabilities of current LLMs and highlighting important directions for future evaluation and development of legal LLMs. Data is available at: https://github.com/skylenage/PLawbench.

The advent of large language models (LLMs) has ushered in a new era in automated code translation across programming languages. Since most code-specific LLMs are pretrained on well-commented code from large repositories like GitHub, it is reasonable to hypothesize that natural language code comments could aid in improving translation quality. Despite their potential relevance, comments are largely absent from existing code translation benchmarks, rendering their impact on translation quality inadequately characterised. In this paper, we present a large-scale empirical study evaluating the impact of comments on translation performance. Our analysis involves more than $80,000$ translations, with and without comments, of $1100+$ code samples from two distinct benchmarks covering pairwise translations between five different programming languages: C, C++, Go, Java, and Python. Our results provide strong evidence that code comments, particularly those that describe the overall purpose of the code rather than line-by-line functionality, significantly enhance translation accuracy. Based on these findings, we propose COMMENTRA, a code translation approach, and demonstrate that it can potentially double the performance of LLM-based code translation. To the best of our knowledge, our study is the first in terms of its comprehensiveness, scale, and language coverage on how to improve code translation accuracy using code comments.

Diffusion-based image super-resolution (SR) has recently attracted significant attention by leveraging the expressive power of large pre-trained text-to-image diffusion models (DMs). A central practical challenge is resolving the trade-off between reconstruction faithfulness and photorealism. To address inference efficiency, many recent works have explored knowledge distillation strategies specifically tailored to SR, enabling one-step diffusion-based approaches. However, these teacher-student formulations are inherently constrained by information compression, which can degrade perceptual cues such as lifelike textures and depth of field, even with high overall perceptual quality. In parallel, self-distillation DMs, known as Flow Map models, have emerged as a promising alternative for image generation tasks, enabling fast inference while preserving the expressivity and training stability of standard DMs. Building on these developments, we propose FlowMapSR, a novel diffusion-based framework for image super-resolution explicitly designed for efficient inference. Beyond adapting Flow Map models to SR, we introduce two complementary enhancements: (i) positive-negative prompting guidance, based on a generalization of classifier free-guidance paradigm to Flow Map models, and (ii) adversarial fine-tuning using Low-Rank Adaptation (LoRA). Among the considered Flow Map formulations (Eulerian, Lagrangian, and Shortcut), we find that the Shortcut variant consistently achieves the best performance when combined with these enhancements. Extensive experiments show that FlowMapSR achieves a better balance between reconstruction faithfulness and photorealism than recent state-of-the-art methods for both x4 and x8 upscaling, while maintaining competitive inference time. Notably, a single model is used for both upscaling factors, without any scale-specific conditioning or degradation-guided mechanisms.

Counterfactual explanations (CEs) offer interpretable insights into machine learning predictions by answering ``what if?" questions. However, in real-world settings where models are frequently updated, existing counterfactual explanations can quickly become invalid or unreliable. In this paper, we introduce Probabilistically Safe CEs (PSCE), a method for generating counterfactual explanations that are $δ$-safe, to ensure high predictive confidence, and $ε$-robust to ensure low predictive variance. Based on Bayesian principles, PSCE provides formal probabilistic guarantees for CEs under model changes which are adhered to in what we refer to as the $\langle δ, ε\rangle$-set. Uncertainty-aware constraints are integrated into our optimization framework and we validate our method empirically across diverse datasets. We compare our approach against state-of-the-art Bayesian CE methods, where PSCE produces counterfactual explanations that are not only more plausible and discriminative, but also provably robust under model change.

People are increasingly turning to generative AI (e.g., ChatGPT, Gemini, Copilot) for emotional support and companionship. While trust is likely to play a central role in enabling these informal and unsupervised interactions, we still lack an understanding of how people develop and experience it in this context. Seeking to fill this gap, we recruited 24 frequent users of generative AI for emotional support and conducted a qualitative study consisting of diary entries about interactions, transcripts of chats with AI, and in-depth interviews. Our results suggest important novel drivers of trust in this context: familiarity emerging from personalisation, nuanced mental models of generative AI, and awareness of people's control over conversations. Notably, generative AI's homogeneous use of personalised, positive, and persuasive language appears to promote some of these trust-building factors. However, this also seems to discourage other trust-related behaviours, such as remembering that generative AI is a machine trained to converse in human language. We present implications for future research that are likely to become critical as the use of generative AI for emotional support increasingly overlaps with therapeutic work.

We propose a reliable and energy-efficient framework for 3D brain tumor segmentation using spiking neural networks (SNNs). A multi-view ensemble of sagittal, coronal, and axial SNN models provides voxel-wise uncertainty estimation and enhances segmentation robustness. To address the high computational cost in training SNN models for semantic image segmentation, we employ Forward Propagation Through Time (FPTT), which maintains temporal learning efficiency with significantly reduced computational cost. Experiments on the Multimodal Brain Tumor Segmentation Challenges (BraTS 2017 and BraTS 2023) demonstrate competitive accuracy, well-calibrated uncertainty, and an 87% reduction in FLOPs, underscoring the potential of SNNs for reliable, low-power medical IoT and Point-of-Care systems.

Gradient-based methods for instance-based explanation for large language models (LLMs) are hindered by the immense dimensionality of model gradients. In practice, influence estimation is restricted to a subset of model parameters to make computation tractable, but this subset is often chosen ad hoc and rarely justified by systematic evaluation. This paper investigates if it is better to create low-dimensional representations by selecting a small, architecturally informed subset of model components or by projecting the full gradients into a lower-dimensional space. Using a novel benchmark, we show that a greedily selected subset of components captures the information about training data influence needed for a retrieval task more effectively than either the full gradient or random projection. We further find that this approach is more computationally efficient than random projection, demonstrating that targeted component selection is a practical strategy for making instance-based explanations of large models more computationally feasible.

Large language models can perform well on many isolated tasks, yet they continue to struggle on multi-turn, long-horizon agentic problems that require skills such as planning, state tracking, and long context processing. In this work, we aim to better understand the relative importance of advancing these underlying capabilities for success on such tasks. We develop an oracle counterfactual framework for multi-turn problems that asks: how would an agent perform if it could leverage an oracle to perfectly perform a specific task? The change in the agent's performance due to this oracle assistance allows us to measure the criticality of such oracle skill in the future advancement of AI agents. We introduce a suite of procedurally generated, game-like tasks with tunable complexity. These controlled environments allow us to provide precise oracle interventions, such as perfect planning or flawless state tracking, and make it possible to isolate the contribution of each oracle without confounding effects present in real-world benchmarks. Our results show that while some interventions (e.g., planning) consistently improve performance across settings, the usefulness of other skills is dependent on the properties of the environment and language model. Our work sheds light on the challenges of multi-turn agentic environments to guide the future efforts in the development of AI agents and language models.

We find that correct-to-incorrect sycophancy signals are most linearly separable within multi-head attention activations. Motivated by the linear representation hypothesis, we train linear probes across the residual stream, multilayer perceptron (MLP), and attention layers to analyze where these signals emerge. Although separability appears in the residual stream and MLPs, steering using these probes is most effective in a sparse subset of middle-layer attention heads. Using TruthfulQA as the base dataset, we find that probes trained on it transfer effectively to other factual QA benchmarks. Furthermore, comparing our discovered direction to previously identified "truthful" directions reveals limited overlap, suggesting that factual accuracy, and deference resistance, arise from related but distinct mechanisms. Attention-pattern analysis further indicates that the influential heads attend disproportionately to expressions of user doubt, contributing to sycophantic shifts. Overall, these findings suggest that sycophancy can be mitigated through simple, targeted linear interventions that exploit the internal geometry of attention activations.

Selected Basis Diagonalization (SBD) plays a central role in Sample-based Quantum Diagonalization (SQD), where iterative diagonalization of the Hamiltonian in selected configuration subspaces forms the dominant classical workload. We present a GPU-accelerated implementation of SBD using the Thrust library. By restructuring key components -- including configuration processing, excitation generation, and matrix-vector operations -- around fine-grained data-parallel primitives and flattened GPU-friendly data layouts, the proposed approach efficiently exploits modern GPU architectures. In our experiments, the Thrust-based SBD achieves up to $\sim$40$\times$ speedup over CPU execution and substantially reduces the total runtime of SQD iterations. These results demonstrate that GPU-native parallel primitives provide a simple, portable, and high-performance foundation for accelerating SQD-based quantum-classical workflows.

Recent advancements enable fine-grained energy measurements in cloud-native environments (e.g., at container or process level) beyond traditional coarse-grained scopes. However, service-level energy measurement for microservice-based applications remains underexplored. Such measurements must include compute, network, and storage energy to avoid underestimating consumption in distributed setups. We present GOXN (Green Observability eXperiment eNginE), an energy experimentation engine for Kubernetes-based microservices that quantifies compute, network, and storage energy at the service level. Using GOXN, we evaluated the OpenTelemetry Demo under varying configurations (monitoring, tracing, service mesh) and steady synthetic load, collecting metrics from Kepler and cAdvisor. Our additive energy model derives service-level energy from container-level data. Results show that excluding network and storage can underestimate auxiliary-service energy by up to 63%, and that high tracing loads shift energy dominance toward network and storage.

Time series forecasting has witnessed significant progress with deep learning. While prevailing approaches enhance forecasting performance by modifying architectures or introducing novel enhancement strategies, they often fail to dynamically disentangle and leverage the complex, intertwined temporal patterns inherent in time series, thus resulting in the learning of static, averaged representations that lack context-aware capabilities. To address this, we propose the Dual-Prototype Adaptive Disentanglement framework (DPAD), a model-agnostic auxiliary method that equips forecasting models with the ability of pattern disentanglement and context-aware adaptation. Specifically, we construct a Dynamic Dual-Prototype bank (DDP), comprising a common pattern bank with strong temporal priors to capture prevailing trend or seasonal patterns, and a rare pattern bank dynamically memorizing critical yet infrequent events, and then an Dual-Path Context-aware routing (DPC) mechanism is proposed to enhance outputs with selectively retrieved context-specific pattern representations from the DDP. Additionally, we introduce a Disentanglement-Guided Loss (DGLoss) to ensure that each prototype bank specializes in its designated role while maintaining comprehensive coverage. Comprehensive experiments demonstrate that DPAD consistently improves forecasting performance and reliability of state-of-the-art models across diverse real-world benchmarks.

Massively multilingual language models enable cross-lingual generalization but underperform on low-resource and unseen languages. While adapter-based fine-tuning offers a parameter-efficient solution, training language-specific adapters at scale remains costly. We introduce Typologically Informed Parameter Aggregation (TIPA), a training-free method that constructs proxy language adapters by aggregating existing ones, weighted by typological similarity. Integrated into the MAD-X framework, these proxies enable zero-shot cross-lingual transfer without additional training. We evaluate TIPA on five NLP tasks and over 230 languages. TIPA consistently outperforms or matches baselines such as English-only fine-tuning or selecting the typologically closest language adapter. We see the largest gains for languages lacking dedicated adapters. Our results demonstrate that typologically informed aggregation provides a viable alternative to language-specific modules without any training needed.

Social media data has been of interest to Natural Language Processing (NLP) practitioners for over a decade, because of its richness in information, but also challenges for automatic processing. Since language use is more informal, spontaneous, and adheres to many different sociolects, the performance of NLP models often deteriorates. One solution to this problem is to transform data to a standard variant before processing it, which is also called lexical normalization. There has been a wide variety of benchmarks and models proposed for this task. The MultiLexNorm benchmark proposed to unify these efforts, but it consists almost solely of languages from the Indo-European language family in the Latin script. Hence, we propose an extension to MultiLexNorm, which covers 5 Asian languages from different language families in 4 different scripts. We show that the previous state-of-the-art model performs worse on the new languages and propose a new architecture based on Large Language Models (LLMs), which shows more robust performance. Finally, we analyze remaining errors, revealing future directions for this task.

Equivariant Graph Neural Networks (EGNNs) have become a widely used approach for modeling 3D atomistic systems. However, mainstream architectures face critical scalability bottlenecks due to the explicit construction of geometric features or dense tensor products on \textit{every} edge. To overcome this, we introduce \textbf{E2Former-V2}, a scalable architecture that integrates algebraic sparsity with hardware-aware execution. We first propose \textbf{E}quivariant \textbf{A}xis-\textbf{A}ligned \textbf{S}parsification (EAAS). EAAS builds on Wigner-$6j$ convolution by exploiting an $\mathrm{SO}(3) \rightarrow \mathrm{SO}(2)$ change of basis to transform computationally expensive dense tensor contractions into efficient, sparse parity re-indexing operations. Building on this representation, we introduce \textbf{On-the-Fly Equivariant Attention}, a fully node-centric mechanism implemented via a custom fused Triton kernel. By eliminating materialized edge tensors and maximizing SRAM utilization, our kernel achieves a \textbf{20$\times$ improvement in TFLOPS} compared to standard implementations. Extensive experiments on the SPICE and OMol25 datasets demonstrate that E2Former-V2 maintains comparable predictive performance while notably accelerating inference. This work demonstrates that large equivariant transformers can be trained efficiently using widely accessible GPU platforms. The code is avalible at https://github.com/IQuestLab/UBio-MolFM/tree/e2formerv2.

Large language model (LLM)-powered assistants have recently integrated memory mechanisms that record user preferences, leading to more personalized and user-aligned responses. However, irrelevant personalized memories are often introduced into the context, interfering with the LLM's intent understanding. To comprehensively investigate the dual effects of personalization, we develop RPEval, a benchmark comprising a personalized intent reasoning dataset and a multi-granularity evaluation protocol. RPEval reveals the widespread phenomenon of irrational personalization in existing LLMs and, through error pattern analysis, illustrates its negative impact on user experience. Finally, we introduce RP-Reasoner, which treats memory utilization as a pragmatic reasoning process, enabling the selective integration of personalized information. Experimental results demonstrate that our method significantly outperforms carefully designed baselines on RPEval, and resolves 80% of the bad cases observed in a large-scale commercial personalized assistant, highlighting the potential of pragmatic reasoning to mitigate irrational personalization. Our benchmark is publicly available at https://github.com/XueyangFeng/RPEval.

Although Large Language Models (LLMs) excel in many tasks, their application to Speech-to-Speech Translation (S2ST) is underexplored and hindered by data scarcity. To bridge this gap, we propose PROST-LLM (PROgressive Speech-to-speech Translation) to enhance the S2ST capabilities in LLMs progressively. First, we fine-tune the LLMs with the CVSS corpus, employing designed tri-task learning and chain of modality methods to boost the initial performance. Then, leveraging the fine-tuned model, we generate preference pairs through self-sampling and back-translation without human evaluation. Finally, these preference pairs are used for preference optimization to enhance the model's S2ST capability further. Extensive experiments confirm the effectiveness of our proposed PROST-LLM in improving the S2ST capability of LLMs.

Unmanned Aerial Vehicle (UAV) applications have become increasingly prevalent in aerial photography and object recognition. However, there are major challenges to accurately capturing small targets in object detection due to the imbalanced scale and the blurred edges. To address these issues, boundary and position information mining (BPIM) framework is proposed for capturing object edge and location cues. The proposed BPIM includes position information guidance (PIG) module for obtaining location information, boundary information guidance (BIG) module for extracting object edge, cross scale fusion (CSF) module for gradually assembling the shallow layer image feature, three feature fusion (TFF) module for progressively combining position and boundary information, and adaptive weight fusion (AWF) module for flexibly merging the deep layer semantic feature. Therefore, BPIM can integrate boundary, position, and scale information in image for small object detection using attention mechanisms and cross-scale feature fusion strategies. Furthermore, BPIM not only improves the discrimination of the contextual feature by adaptive weight fusion with boundary, but also enhances small object perceptions by cross-scale position fusion. On the VisDrone2021, DOTA1.0, and WiderPerson datasets, experimental results show the better performances of BPIM compared to the baseline Yolov5-P2, and obtains the promising performance in the state-of-the-art methods with comparable computation load.

Recently, due to the advancement of multimodal technology, people are attempting to use visual large language models (VLLMs) in industrial production. Many deep learning models (DLMs) deployed in the production environment are gradually being replaced by VLLMs. Compared with DLMs, VLLMs have some advantages in industrial applications: (1) Their strong generalization ability enables them to perform well across a wide range of tasks. (2) They are flexible and can deal with unfamiliar samples through context learning quickly. However, VLLMs also have obvious drawbacks: (1) VLLMs do not perform as well as custom-developed DLMs in specific domains. (2) The number of parameters in VLLMs is generally quite large, and their deployment requires substantial computational resources. (3) VLLMs generally operate much slower than DLMs, making real-time response challenging to achieve. To better utilize VLLMs in industrial applications, we introduce AuroraEdge-V-2B in this work, a compact, robust, and high-speed VLLM designed for edge deployment. To make the model run faster, we also propose a compression-fusion method to improve inference efficiency. AuroraEdge-V-2B has the following notable features: (1) Easy deployment and faster: It has only 2B parameters and is highly suitable for edge deployment, offering better real-time performance. (2) Fewer visual tokens and cheaper: It significantly reduces the number of visual tokens in the decoding process, thereby reducing the floating-point operations by half during inference and making it cheaper to use. (3) Strong performance: It gets a higher score on 9 benchmarks than models with the same number of parameter (e.g., Qwen2-VL-2B, Qwen2.5-VL-3B, InternVL-2.5-2B).

Intelligent medical image analysis is essential for clinical decision support but is often limited by scarce annotations, constrained computational resources, and suboptimal model generalization. To address these challenges, we propose a lightweight medical image classification framework that integrates self-supervised contrastive learning with quantum-enhanced feature modeling. MobileNetV2 is employed as a compact backbone and pretrained using a SimCLR-style self-supervised paradigm on unlabeled images. A lightweight parameterized quantum circuit (PQC) is embedded as a quantum feature enhancement module, forming a hybrid classical-quantum architecture, which is subsequently fine-tuned on limited labeled data. Experimental results demonstrate that, with only approximately 2-3 million parameters and low computational cost, the proposed method consistently outperforms classical baselines without self-supervised learning or quantum enhancement in terms of Accuracy, AUC, and F1-score. Feature visualization further indicates improved discriminability and representation stability. Overall, this work provides a practical and forward-looking solution for high-performance medical artificial intelligence under resource-constrained settings.

Learning a stationary diffusion amounts to estimating the parameters of a stochastic differential equation whose stationary distribution matches a target distribution. We build on the recently introduced kernel deviation from stationarity (KDS), which enforces stationarity by evaluating expectations of the diffusion's generator in a reproducing kernel Hilbert space. Leveraging the connection between KDS and Stein discrepancies, we introduce the Stein-type KDS (SKDS) as an alternative formulation. We prove that a vanishing SKDS guarantees alignment of the learned diffusion's stationary distribution with the target. Furthermore, under broad parametrizations, SKDS is convex with an empirical version that is $ε$-quasiconvex with high probability. Empirically, learning with SKDS attains comparable accuracy to KDS while substantially reducing computational cost and yields improvements over the majority of competitive baselines.

As the development of Large Language Models (LLMs) shifts from parameter scaling to inference-time collaboration, the Mixture-of-Agents (MoA) framework has emerged as a general paradigm to harness collective intelligence by layering diverse models. While recent MoA variants have introduced dynamic routing and residual connections to improve efficiency, these methods often fail to facilitate deep semantic interaction between agents, limiting the system's ability to actively correct hallucinations and refine logic. In this paper, we introduce Attention-MoA, a novel MoA-based framework that redefines collaboration through Inter-agent Semantic Attention. Complemented by an Inter-layer Residual Module with Adaptive Early Stopping Mechanism, our architecture mitigates information degradation in deep layers while improving computational efficiency. Extensive evaluations across AlpacaEval 2.0, MT-Bench, and FLASK demonstrate that Attention-MoA significantly outperforms state-of-the-art baselines, achieving a 91.15% Length-Controlled Win Rate on AlpacaEval 2.0 and dominating in 10 out of 12 capabilities on FLASK. Notably, Attention-MoA enables an ensemble of small open-source models to outperform massive proprietary models like Claude-4.5-Sonnet and GPT-4.1, achieving an MT-Bench score of 8.83 and an AlpacaEval 2.0 LC Win Rate of 77.36%.

Train delays result from complex interactions between operational, technical, and environmental factors. While weather impacts railway reliability, particularly in Nordic regions, existing datasets rarely integrate meteorological information with operational train data. This study presents the first publicly available dataset combining Finnish railway operations with synchronized meteorological observations from 2018-2024. The dataset integrates operational metrics from Finland Digitraffic Railway Traffic Service with weather measurements from 209 environmental monitoring stations, using spatial-temporal alignment via Haversine distance. It encompasses 28 engineered features across operational variables and meteorological measurements, covering approximately 38.5 million observations from Finland's 5,915-kilometer rail network. Preprocessing includes strategic missing data handling through spatial fallback algorithms, cyclical encoding of temporal features, and robust scaling of weather data to address sensor outliers. Analysis reveals distinct seasonal patterns, with winter months exhibiting delay rates exceeding 25\% and geographic clustering of high-delay corridors in central and northern Finland. Furthermore, the work demonstrates applications of the data set in analysing the reliability of railway traffic in Finland. A baseline experiment using XGBoost regression achieved a Mean Absolute Error of 2.73 minutes for predicting station-specific delays, demonstrating the dataset's utility for machine learning applications. The dataset enables diverse applications, including train delay prediction, weather impact assessment, and infrastructure vulnerability mapping, providing researchers with a flexible resource for machine learning applications in railway operations research.

Neuromorphic computing mimics brain-inspired mechanisms through spiking neurons and energy-efficient processing, offering a pathway to efficient in-memory computing (IMC). However, these advancements raise critical security and privacy concerns. As the adoption of bio-inspired architectures and memristive devices increases, so does the urgency to assess the vulnerability of these emerging technologies to hardware and software attacks. Emerging architectures introduce new attack surfaces, particularly due to asynchronous, event-driven processing and stochastic device behavior. The integration of memristors into neuromorphic hardware and software implementations in spiking neural networks offers diverse possibilities for advanced computing architectures, including their role in security-aware applications. This survey systematically analyzes the security landscape of neuromorphic systems, covering attack methodologies, side-channel vulnerabilities, and countermeasures. We focus on both hardware and software concerns relevant to spiking neural networks (SNNs) and hardware primitives, such as Physical Unclonable Functions (PUFs) and True Random Number Generators (TRNGs) for cryptographic and secure computation applications. We approach this analysis from diverse perspectives, from attack methodologies to countermeasure strategies that integrate efficiency and protection in brain-inspired hardware. This review not only maps the current landscape of security threats but provides a foundation for developing secure and trustworthy neuromorphic architectures.

Low-complexity multiple-input multiple-output (MIMO) detection remains a key challenge in modern wireless systems, particularly for 5G reduced capability (RedCap) and internet-of-things (IoT) devices. In this context, the growing interest in deploying machine learning on edge devices must be balanced against stringent constraints on computational complexity and memory while supporting high-order modulation. Beyond accurate hard detection, reliable soft information is equally critical, as modern receivers rely on soft-input channel decoding, imposing additional requirements on the detector design. In this work, we propose recurSIC, a lightweight learning-based MIMO detection framework that is structurally inspired by successive interference cancellation (SIC) and incorporates learned processing stages. It generates reliable soft information via multi-path hypothesis tracking with a tunable complexity parameter while requiring only a single forward pass and a minimal parameter count. Numerical results in realistic wireless scenarios show that recurSIC achieves strong hard- and soft-detection performance at very low complexity, making it well suited for edge-constrained MIMO receivers.

Venture capital (VC) investments in early-stage startups that end up being successful can yield high returns. However, predicting early-stage startup success remains challenging due to data scarcity (e.g., many VC firms have information about only a few dozen of early-stage startups and whether they were successful). This limits the effectiveness of traditional machine learning methods that rely on large labeled datasets for model training. To address this challenge, we propose an in-context learning framework for startup success prediction using large language models (LLMs) that requires no model training and leverages only a small set of labeled startups as demonstration examples. Specifically, we propose a novel k-nearest-neighbor-based in-context learning framework, called kNN-ICL, which selects the most relevant past startups as examples based on similarity. Using real-world profiles from Crunchbase, we find that the kNN-ICL approach achieves higher prediction accuracy than supervised machine learning baselines and vanilla in-context learning. Further, we study how performance varies with the number of in-context examples and find that a high balanced accuracy can be achieved with as few as 50 examples. Together, we demonstrate that in-context learning can serve as a decision-making tool for VC firms operating in data-scarce environments.

This work proposes neural training as a \emph{process tensor}: a multi-time map that takes a sequence of controllable instruments (batch choices, augmentations, optimizer micro-steps) and returns an observable of the trained model. Building on this operational lens, we introduce a simple, model-agnostic witness of training memory based on \emph{back-flow of distinguishability}. In a controlled two-step protocol, we compare outcome distributions after one intervention versus two; the increase $Δ_{\mathrm{BF}} = D_2 - D_1>0$ (with $D\in\{\mathrm{TV}, \mathrm{JS}, \mathrm{H}\}$ measured on softmax predictions over a fixed probe set) certifies non-Markovianity. We observe consistent positive back-flow with tight bootstrap confidence intervals, amplification under higher momentum, larger batch overlap, and more micro-steps, and collapse under a \emph{causal break} (resetting optimizer state), directly attributing the effect to optimizer/data-state memory. The witness is robust across TV/JS/Hellinger, inexpensive to compute, and requires no architectural changes. We position this as a \emph{measurement} contribution: a principled diagnostic and empirical evidence that practical SGD deviates from the Markov idealization. An exploratory case study illustrates how the micro-level signal can inform curriculum orderings. "Data order matters" turns into a testable operator with confidence bounds, our framework offers a common stage to compare optimizers, curricula, and schedules through their induced training memory.

Generative Sequential Recommendation (GSR) has emerged as a promising paradigm, reframing recommendation as an autoregressive sequence generation task over discrete Semantic IDs (SIDs), typically derived via codebook-based quantization. Despite its great potential in unifying retrieval and ranking, existing GSR frameworks still face two critical limitations: (1) impure and unstable semantic tokenization, where quantization methods struggle with interaction noise and codebook collapse, resulting in SIDs with ambiguous discrimination; and (2) lossy and weakly structured generation, where reliance solely on coarse-grained discrete tokens inevitably introduces information loss and neglects items' hierarchical logic. To address these issues, we propose a novel generative recommendation framework, PRISM, with Purified Representation and Integrated Semantic Modeling. Specifically, to ensure high-quality tokenization, we design a Purified Semantic Quantizer that constructs a robust codebook via adaptive collaborative denoising and hierarchical semantic anchoring mechanisms. To compensate for information loss during quantization, we further propose an Integrated Semantic Recommender, which incorporates a dynamic semantic integration mechanism to integrate fine-grained semantics and enforces logical validity through a semantic structure alignment objective. PRISM consistently outperforms state-of-the-art baselines across four real-world datasets, demonstrating substantial performance gains, particularly in high-sparsity scenarios.

Fact-checking aims to verify the truthfulness of a claim based on the retrieved evidence. Existing methods typically follow a decomposition paradigm, in which a claim is broken down into sub-claims that are individually verified. However, the decomposition paradigm may introduce noise to the verification process due to irrelevant entities or evidence, ultimately degrading verification accuracy. To address this problem, we propose a Retrieve-Refine-Calibrate (RRC) framework based on large language models (LLMs). Specifically, the framework first identifies the entities mentioned in the claim and retrieves evidence relevant to them. Then, it refines the retrieved evidence based on the claim to reduce irrelevant information. Finally, it calibrates the verification process by re-evaluating low-confidence predictions. Experiments on two popular fact-checking datasets (HOVER and FEVEROUS-S) demonstrate that our framework achieves superior performance compared with competitive baselines.

Nonlinear dimensionality reduction techniques, particularly UMAP, are widely used for visualizing high-dimensional data. However, UMAP's local Euclidean distance assumption often fails to capture intrinsic manifold geometry, leading to topological tearing and structural collapse. We identify UMAP's sensitivity to the k-nearest neighbor graph as a key cause. To address this, we introduce Ollivier-Ricci curvature as a geometric prior, reinforcing edges at geometric bottlenecks and reducing redundant links. Since curvature estimation is noise-sensitive, we also incorporate a topological prior using Jaccard similarity to ensure neighborhood consistency. The resulting method, JORC-UMAP, better distinguishes true manifold structure from spurious connections. Experiments on synthetic and real-world datasets show that JORC-UMAP reduces tearing and collapse more effectively than standard UMAP and other DR methods, as measured by SVM accuracy and triplet preservation scores, while maintaining computational efficiency. This work offers a geometry-aware enhancement to UMAP for more faithful data visualization.

The combination of multimodal Vision-Language Models (VLMs) and Large Language Models (LLMs) opens up new possibilities for medical classification. This work offers a rigorous, unified benchmark by using four publicly available datasets covering text and image modalities (binary and multiclass complexity) that contrasts traditional Machine Learning (ML) with contemporary transformer-based techniques. We evaluated three model classes for each task: Classical ML (LR, LightGBM, ResNet-50), Prompt-Based LLMs/VLMs (Gemini 2.5), and Fine-Tuned PEFT Models (LoRA-adapted Gemma3 variants). All experiments used consistent data splits and aligned metrics. According to our results, traditional machine learning (ML) models set a high standard by consistently achieving the best overall performance across most medical categorization tasks. This was especially true for structured text-based datasets, where the classical models performed exceptionally well. In stark contrast, the LoRA-tuned Gemma variants consistently showed the worst performance across all text and image experiments, failing to generalize from the minimal fine-tuning provided. However, the zero-shot LLM/VLM pipelines (Gemini 2.5) had mixed results; they performed poorly on text-based tasks, but demonstrated competitive performance on the multiclass image task, matching the classical ResNet-50 baseline. These results demonstrate that in many medical categorization scenarios, established machine learning models continue to be the most reliable option. The experiment suggests that foundation models are not universally superior and that the effectiveness of Parameter-Efficient Fine-Tuning (PEFT) is highly dependent on the adaptation strategy, as minimal fine-tuning proved detrimental in this study.

Large Audio Language Models (LALMs) have garnered significant research interest. Despite being built upon text-based large language models (LLMs), LALMs frequently exhibit a degradation in knowledge and reasoning capabilities. We hypothesize that this limitation stems from the failure of current training paradigms to effectively bridge the acoustic-semantic gap within the feature representation space. To address this challenge, we propose CORD, a unified alignment framework that performs online cross-modal self-distillation. Specifically, it aligns audio-conditioned reasoning with its text-conditioned counterpart within a unified model. Leveraging the text modality as an internal teacher, CORD performs multi-granularity alignment throughout the audio rollout process. At the token level, it employs on-policy reverse KL divergence with importance-aware weighting to prioritize early and semantically critical tokens. At the sequence level, CORD introduces a judge-based global reward to optimize complete reasoning trajectories via Group Relative Policy Optimization (GRPO). Empirical results across multiple benchmarks demonstrate that CORD consistently enhances audio-conditioned reasoning and substantially bridges the audio-text performance gap with only 80k synthetic training samples, validating the efficacy and data efficiency of our on-policy, multi-level cross-modal alignment approach.

Hierarchical open-set classification handles previously unseen classes by assigning them to the most appropriate high-level category in a class taxonomy. We extend this paradigm to the semi-supervised setting, enabling the use of large-scale, uncurated datasets containing a mixture of known and unknown classes to improve the hierarchical open-set performance. To this end, we propose a teacher-student framework based on pseudo-labeling. Two key components are introduced: 1) subtree pseudo-labels, which provide reliable supervision in the presence of unknown data, and 2) age-gating, a mechanism that mitigates overconfidence in pseudo-labels. Experiments show that our framework outperforms self-supervised pretraining followed by supervised adaptation, and even matches the fully supervised counterpart when using only 20 labeled samples per class on the iNaturalist19 benchmark. Our code is available at https://github.com/walline/semihoc.

Audio Large Language Models (Audio LLMs) have demonstrated strong capabilities in integrating speech perception with language understanding. However, whether their internal representations align with human neural dynamics during naturalistic listening remains largely unexplored. In this work, we systematically examine layer-wise representational alignment between 12 open-source Audio LLMs and Electroencephalogram (EEG) signals across 2 datasets. Specifically, we employ 8 similarity metrics, such as Spearman-based Representational Similarity Analysis (RSA), to characterize within-sentence representational geometry. Our analysis reveals 3 key findings: (1) we observe a rank-dependence split, in which model rankings vary substantially across different similarity metrics; (2) we identify spatio-temporal alignment patterns characterized by depth-dependent alignment peaks and a pronounced increase in RSA within the 250-500 ms time window, consistent with N400-related neural dynamics; (3) we find an affective dissociation whereby negative prosody, identified using a proposed Tri-modal Neighborhood Consistency (TNC) criterion, reduces geometric similarity while enhancing covariance-based dependence. These findings provide new neurobiological insights into the representational mechanisms of Audio LLMs.

As Large Language Models (LLMs) scale, weight-only quantization (W4A16: 4-bit weights, 16-bit activations) becomes critical for reducing memory footprint with minimal accuracy loss. However, its efficient deployment on Huawei's Ascend 910 Neural Processing Unit (NPU) is challenging due to limited native mixed-precision support and the accelerator's decoupled compute architecture. To enable quantization on such architecture, we present the first practical W4A16 matrix multiplication kernel tailored for the Ascend 910 NPU. Our design leverages vector cores for on-the-fly INT4-to-FP16 dequantization, cube cores for high-throughput GEMM, and Split-K parallelization to mitigate memory latency. Performance evaluations across diverse matrix shapes and batch sizes show our method outperforms data-parallel approaches when K >> N, a typical scenario in LLM decoding. Specially, our method can achieve a speedup ranging from 1.01x to 1.74x. In addition, our profile reveals the primary bottleneck is not dequantization compution itself, but extra global memory transfer for the weight, making W4A16 only reaching a maximum speedup of 1.48x over native FP16xFP16 matrix multiplication in PyTorch. In the long run, our method lays a solid foundation and provides insightful views for the efficient deployment of quantized large language models on various domain-specific accelerators.

We investigate whether high-frequency key collisions are a primary bottleneck in Engram-style conditional memory. To isolate the effect of collisions, we introduce Engram-Nine, a collision-free hot-tier extension that maps the most frequent n-grams through a Minimal Perfect Hash Function (MPHF) while retaining the original multi-head hashed lookup as a cold tier. Under a strictly iso-parameter setup, the collision-free design does not consistently improve validation loss. Through route-stratified evaluation (decomposing per-token loss into hot/cold contributions), we uncover a consistent "hot-to-cold advantage flip" during training: hot (high-frequency) positions initially have lower loss, but cold positions eventually surpass them. Crucially, collision-free configurations flip earlier than collision-prone baselines, suggesting that collisions act as implicit regularization. We also identify a gating mismatch: the gate learns to favor hot positions early in training, but this preference persists even after the flip, assigning higher weights to positions with higher loss. Our findings suggest that improving lookup precision alone does not guarantee better training outcomes. The dominant limitation may lie in gating credit assignment rather than index accuracy, and collision-induced noise may provide beneficial regularization that should not be naively eliminated.

Large language models (LLMs) and high-capacity encoders have advanced zero and few-shot classification, but their inference cost and latency limit practical deployment. We propose training lightweight text classifiers using dynamically generated supervision from an LLM. Our method employs an iterative, agentic loop in which the LLM curates training data, analyzes model successes and failures, and synthesizes targeted examples to address observed errors. This closed-loop generation and evaluation process progressively improves data quality and adapts it to the downstream classifier and task. Across four widely used benchmarks, our approach consistently outperforms standard zero and few-shot baselines. These results indicate that LLMs can serve effectively as data curators, enabling accurate and efficient classification without the operational cost of large-model deployment.

Large language models (LLMs) show promise in clinical decision support yet risk acquiescing to patient pressure for inappropriate care. We introduce SycoEval-EM, a multi-agent simulation framework evaluating LLM robustness through adversarial patient persuasion in emergency medicine. Across 20 LLMs and 1,875 encounters spanning three Choosing Wisely scenarios, acquiescence rates ranged from 0-100\%. Models showed higher vulnerability to imaging requests (38.8\%) than opioid prescriptions (25.0\%), with model capability poorly predicting robustness. All persuasion tactics proved equally effective (30.0-36.0\%), indicating general susceptibility rather than tactic-specific weakness. Our findings demonstrate that static benchmarks inadequately predict safety under social pressure, necessitating multi-turn adversarial testing for clinical AI certification.

Multimodal LLMs are powerful but prone to object hallucinations, which describe non-existent entities and harm reliability. While recent unlearning methods attempt to mitigate this, we identify a critical flaw: structural fragility. We empirically demonstrate that standard erasure achieves only superficial suppression, trapping the model in sharp minima where hallucinations catastrophically resurge after lightweight relearning. To ensure geometric stability, we propose SARE, which casts unlearning as a targeted min-max optimization problem and uses a Targeted-SAM mechanism to explicitly flatten the loss landscape around hallucinated concepts. By suppressing hallucinations under simulated worst-case parameter perturbations, our framework ensures robust removal stable against weight shifts. Extensive experiments demonstrate that SARE significantly outperforms baselines in erasure efficacy while preserving general generation quality. Crucially, it maintains persistent hallucination suppression against relearning and parameter updates, validating the effectiveness of geometric stabilization.

Multimodal Large Language Models (MLLMs) have achieved remarkable progress in visual recognition and semantic understanding. Nevertheless, their ability to perform precise compositional spatial reasoning remains largely unexplored. Existing benchmarks often involve relatively simple tasks and rely on semantic approximations or coarse relative positioning, while their evaluation metrics are typically limited and lack rigorous mathematical formulations. To bridge this gap, we introduce TangramPuzzle, a geometry-grounded benchmark designed to evaluate compositional spatial reasoning through the lens of the classic Tangram game. We propose the Tangram Construction Expression (TCE), a symbolic geometric framework that grounds tangram assemblies in exact, machine-verifiable coordinate specifications, to mitigate the ambiguity of visual approximation. We design two complementary tasks: Outline Prediction, which demands inferring global shapes from local components, and End-to-End Code Generation, which requires solving inverse geometric assembly problems. We conduct extensive evaluation experiments on advanced open-source and proprietary models, revealing an interesting insight: MLLMs tend to prioritize matching the target silhouette while neglecting geometric constraints, leading to distortions or deformations of the pieces.

Federated graph learning (FGL) enables collaborative training on graph data across multiple clients. With the rise of large language models (LLMs), textual attributes in FGL graphs are gaining attention. Text-attributed graph federated learning (TAG-FGL) improves FGL by explicitly leveraging LLMs to process and integrate these textual features. However, current TAG-FGL methods face three main challenges: \textbf{(1) Overhead.} LLMs for processing long texts incur high token and computation costs. To make TAG-FGL practical, we introduce graph condensation (GC) to reduce computation load, but this choice also brings new issues. \textbf{(2) Suboptimal.} To reduce LLM overhead, we introduce GC into TAG-FGL by compressing multi-hop texts/neighborhoods into a condensed core with fixed LLM surrogates. However, this one-shot condensation is often not client-adaptive, leading to suboptimal performance. \textbf{(3) Interpretability.} LLM-based condensation further introduces a black-box bottleneck: summaries lack faithful attribution and clear grounding to specific source spans, making local inspection and auditing difficult. To address the above issues, we propose \textbf{DANCE}, a new TAG-FGL paradigm with GC. To improve \textbf{suboptimal} performance, DANCE performs round-wise, model-in-the-loop condensation refresh using the latest global model. To enhance \textbf{interpretability}, DANCE preserves provenance by storing locally inspectable evidence packs that trace predictions to selected neighbors and source text spans. Across 8 TAG datasets, DANCE improves accuracy by \textbf{2.33\%} at an \textbf{8\%} condensation ratio, with \textbf{33.42\%} fewer tokens than baselines.

Time series data from the Intensive Care Unit (ICU) provides critical information for patient monitoring. While recent advancements in applying Large Language Models (LLMs) to time series modeling (TSM) have shown great promise, their effectiveness on the irregular ICU data, characterized by particularly high rates of missing values, remains largely unexplored. This work investigates two key components underlying the success of LLMs for TSM: the time series encoder and the multimodal alignment strategy. To this end, we establish a systematic testbed to evaluate their impact across various state-of-the-art LLM-based methods on benchmark ICU datasets against strong supervised and self-supervised baselines. Results reveal that the encoder design is more critical than the alignment strategy. Encoders that explicitly model irregularity achieve substantial performance gains, yielding an average AUPRC increase of $12.8\%$ over the vanilla Transformer. While less impactful, the alignment strategy is also noteworthy, with the best-performing semantically rich, fusion-based strategy achieving a modest $2.9\%$ improvement over cross-attention. However, LLM-based methods require at least 10$\times$ longer training than the best-performing irregular supervised models, while delivering only comparable performance. They also underperform in data-scarce few-shot learning settings. These findings highlight both the promise and current limitations of LLMs for irregular ICU time series. The code is available at https://github.com/mHealthUnimelb/LLMTS.

Understanding why Transformers perform so well remains challenging due to their non-convex optimization landscape. In this work, we analyze a shallow Transformer with $m$ independent heads trained by projected gradient descent in the kernel regime. Our analysis reveals two main findings: (i) the width required for nonasymptotic guarantees scales only logarithmically with the sample size $n$, and (ii) the optimization error is independent of the sequence length $T$. This contrasts sharply with recurrent architectures, where the optimization error can grow exponentially with $T$. The trade-off is memory: to keep the full context, the Transformer's memory requirement grows with the sequence length. We validate our theoretical results numerically in a teacher-student setting and confirm the predicted scaling laws for Transformers.

With the advent of large language models (LLMs), it has become common practice for users to draft text and utilize LLMs to enhance its quality through paraphrasing. However, this process can sometimes result in the loss or distortion of the original intended meaning. Due to the human-like quality of LLM-generated text, traditional detection methods often fail, particularly when text is paraphrased to closely mimic original content. In response to these challenges, we propose a novel approach named SearchLLM, designed to identify LLM-paraphrased text by leveraging search engine capabilities to locate potential original text sources. By analyzing similarities between the input and regenerated versions of candidate sources, SearchLLM effectively distinguishes LLM-paraphrased content. SearchLLM is designed as a proxy layer, allowing seamless integration with existing detectors to enhance their performance. Experimental results across various LLMs demonstrate that SearchLLM consistently enhances the accuracy of recent detectors in detecting LLM-paraphrased text that closely mimics original content. Furthermore, SearchLLM also helps the detectors prevent paraphrasing attacks.

Solving massive-scale optimization problems requires scalable first-order methods with low per-iteration cost. This tutorial highlights a shift in optimization: using differentiable programming not only to execute algorithms but to learn how to design them. Modern frameworks such as PyTorch, TensorFlow, and JAX enable this paradigm through efficient automatic differentiation. Embedding first-order methods within these systems allows end-to-end training that improves convergence and solution quality. Guided by Fenchel-Rockafellar duality, the tutorial demonstrates how duality-informed iterative schemes such as ADMM and PDHG can be learned and adapted. Case studies across LP, OPF, Laplacian regularization, and neural network verification illustrate these gains.

The k-nearest neighbors (kNN) algorithm is a cornerstone of non-parametric classification in artificial intelligence, yet its deployment in large-scale applications is persistently constrained by the computational trade-off between inference speed and accuracy. Existing approximate nearest neighbor solutions accelerate retrieval but often degrade classification precision and lack adaptability in selecting the optimal neighborhood size (k). Here, we present an adaptive graph model that decouples inference latency from computational complexity. By integrating a Hierarchical Navigable Small World (HNSW) graph with a pre-computed voting mechanism, our framework completely transfers the computational burden of neighbor selection and weighting to the training phase. Within this topological structure, higher graph layers enable rapid navigation, while lower layers encode precise, node-specific decision boundaries with adaptive neighbor counts. Benchmarking against eight state-of-the-art baselines across six diverse datasets, we demonstrate that this architecture significantly accelerates inference speeds, achieving real-time performance, without compromising classification accuracy. These findings offer a scalable, robust solution to the long-standing inference bottleneck of kNN, establishing a new structural paradigm for graph-based nonparametric learning.

Single-prompt evaluations dominate current LLM benchmarking, yet they fail to capture the conversational dynamics where real-world harm occurs. In this study, we examined whether conversation length affects response veracity by evaluating LLM performance on the BoolQ dataset under varying length and scaffolding conditions. Our results across three distinct LLMs revealed model-specific vulnerabilities that are invisible under single-turn testing. The length-dependent and scaffold-specific effects we observed demonstrate a fundamental limitation of static evaluations, as deployment-relevant vulnerabilities could only be spotted in a multi-turn conversational setting.

The rapid development of large language models is transforming software development. Beyond serving as code auto-completion tools in integrated development environments, large language models increasingly function as foundation models within coding agents in vibe-coding scenarios. In such settings, prompts play a central role in agent-based intelligent software development, as they not only guide the behavior of large language models but also serve as carriers of user requirements. Under the dominant conversational paradigm, prompts are typically divided into system prompts and user prompts. System prompts provide high-level instructions to steer model behavior and establish conversational context, while user prompts represent inputs and requirements provided by human users. Despite their importance, designing effective prompts remains challenging, as it requires expertise in both prompt engineering and software engineering, particularly requirements engineering. To reduce the burden of manual prompt construction, numerous automated prompt engineering methods have been proposed. However, most existing approaches neglect the methodological principles of requirements engineering, limiting their ability to generate artifacts that conform to formal requirement specifications in realistic software development scenarios. To address this gap, we propose REprompt, a multi-agent prompt optimization framework guided by requirements engineering. Experiment results demonstrate that REprompt effectively optimizes both system and user prompts by grounding prompt generation in requirements engineering principles.

Despite the intrinsic risk-awareness of Large Language Models (LLMs), current defenses often result in shallow safety alignment, rendering models vulnerable to disguised attacks (e.g., prefilling) while degrading utility. To bridge this gap, we propose SafeThinker, an adaptive framework that dynamically allocates defensive resources via a lightweight gateway classifier. Based on the gateway's risk assessment, inputs are routed through three distinct mechanisms: (i) a Standardized Refusal Mechanism for explicit threats to maximize efficiency; (ii) a Safety-Aware Twin Expert (SATE) module to intercept deceptive attacks masquerading as benign queries; and (iii) a Distribution-Guided Think (DDGT) component that adaptively intervenes during uncertain generation. Experiments show that SafeThinker significantly lowers attack success rates across diverse jailbreak strategies without compromising utility, demonstrating that coordinating intrinsic judgment throughout the generation process effectively balances robustness and practicality.

Commonsense reasoning often involves evaluating multiple plausible interpretations rather than selecting a single atomic answer, yet most benchmarks rely on single-label evaluation, obscuring whether statements are jointly plausible, mutually exclusive, or jointly implausible. We introduce LOGICAL-COMMONSENSEQA, a benchmark that re-frames commonsense reasoning as logical composition over pairs of atomic statements using plausibility-level operators (AND, OR, NEITHER/NOR). Evaluating instruction-tuned, reasoning-specialized, and fine-tuned models under zero-shot, few-shot, and chain-of-thought prompting, we find that while models perform reasonably on conjunctive and moderately on disjunctive reasoning, performance degrades sharply on negation-based questions. LOGICAL-COMMONSENSEQA exposes fundamental reasoning limitations and provides a controlled framework for advancing compositional commonsense reasoning.

While Retrieval-Augmented Generation (RAG) has been swiftly adopted in scientific and clinical QA systems, a comprehensive evaluation benchmark in the medical domain is lacking. To address this gap, we introduce the Medical Retrieval-Augmented Generation (MRAG) benchmark, covering various tasks in English and Chinese languages, and building a corpus with Wikipedia and Pubmed. Additionally, we develop the MRAG-Toolkit, facilitating systematic exploration of different RAG components. Our experiments reveal that: (a) RAG enhances LLM reliability across MRAG tasks. (b) the performance of RAG systems is influenced by retrieval approaches, model sizes, and prompting strategies. (c) While RAG improves usefulness and reasoning quality, LLM responses may become slightly less readable for long-form questions. We will release the MRAG-Bench's dataset and toolkit with CCBY-4.0 license upon acceptance, to facilitate applications from both academia and industry.

Federated graph learning (FGL) enables collaborative training of graph neural networks (GNNs) across decentralized subgraphs without exposing raw data. While existing FGL methods often achieve high overall accuracy, we show that this average performance can conceal severe degradation on disadvantaged node groups. From a fairness perspective, these disparities arise systematically from three coupled sources: label skew toward majority patterns, topology confounding in message propagation, and aggregation dilution of updates from hard clients. To address this, we propose \textbf{BoostFGL}, a boosting-style framework for fairness-aware FGL. BoostFGL introduces three coordinated mechanisms: \ding{182} \emph{Client-side node boosting}, which reshapes local training signals to emphasize systematically under-served nodes; \ding{183} \emph{Client-side topology boosting}, which reallocates propagation emphasis toward reliable yet underused structures and attenuates misleading neighborhoods; and \ding{184} \emph{Server-side model boosting}, which performs difficulty- and reliability-aware aggregation to preserve informative updates from hard clients while stabilizing the global model. Extensive experiments on 9 datasets show that BoostFGL delivers substantial fairness gains, improving Overall-F1 by 8.43\%, while preserving competitive overall performance against strong FGL baselines.

Data set composed of categorical features is very common in big data analysis tasks. Since categorical features are usually with a limited number of qualitative possible values, the nested granular cluster effect is prevalent in the implicit discrete distance space of categorical data. That is, data objects frequently overlap in space or subspace to form small compact clusters, and similar small clusters often form larger clusters. However, the distance space cannot be well-defined like the Euclidean distance due to the qualitative categorical data values, which brings great challenges to the cluster analysis of categorical data. In view of this, we design a Multi-Granular Competitive Penalization Learning (MGCPL) algorithm to allow potential clusters to interactively tune themselves and converge in stages with different numbers of naturally compact clusters. To leverage MGCPL, we also propose a Cluster Aggregation strategy based on MGCPL Encoding (CAME) to first encode the data objects according to the learned multi-granular distributions, and then perform final clustering on the embeddings. It turns out that the proposed MGCPL-guided Categorical Data Clustering (MCDC) approach is competent in automatically exploring the nested distribution of multi-granular clusters and highly robust to categorical data sets from various domains. Benefiting from its linear time complexity, MCDC is scalable to large-scale data sets and promising in pre-partitioning data sets or compute nodes for boosting distributed computing. Extensive experiments with statistical evidence demonstrate its superiority compared to state-of-the-art counterparts on various real public data sets.

A reliable executable environment is the foundation for ensuring that large language models solve software engineering tasks. Due to the complex and tedious construction process, large-scale configuration is relatively inefficient. However, most methods always overlook fine-grained analysis of the actions performed by the agent, making it difficult to handle complex errors and resulting in configuration failures. To address this bottleneck, we propose EvoConfig, an efficient environment configuration framework that optimizes multi-agent collaboration to build correct runtime environments. EvoConfig features an expert diagnosis module for fine-grained post-execution analysis, and a self-evolving mechanism that lets expert agents self-feedback and dynamically adjust error-fixing priorities in real time. Empirically, EvoConfig matches the previous state-of-the-art Repo2Run on Repo2Run's 420 repositories, while delivering clear gains on harder cases: on the more challenging Envbench, EvoConfig achieves a 78.1% success rate, outperforming Repo2Run by 7.1%. Beyond end-to-end success, EvoConfig also demonstrates stronger debugging competence, achieving higher accuracy in error identification and producing more effective repair recommendations than existing methods.

As large language models (LLMs) increasingly tackle complex reasoning tasks, test-time scaling has become critical for enhancing capabilities. However, in agentic scenarios with frequent tool calls, the traditional generation-length-based definition breaks down: tool latency decouples inference time from generation length. We propose Timely Machine, redefining test-time as wall-clock time, where models dynamically adjust strategies based on time budgets. We introduce Timely-Eval, a benchmark spanning high-frequency tool calls, low-frequency tool calls, and time-constrained reasoning. By varying tool latency, we find smaller models excel with fast feedback through more interactions, while larger models dominate high-latency settings via superior interaction quality. Moreover, existing models fail to adapt reasoning to time budgets. We propose Timely-RL to address this gap. After cold-start supervised fine-tuning, we use reinforcement learning to enhance temporal planning. Timely-RL improves time budget awareness and consistently boosts performance across Timely-Eval. We hope our work offers a new perspective on test-time scaling for the agentic era.

Large language models have demonstrated strong reasoning capabilities in complex tasks through tool integration, which is typically framed as a Markov Decision Process and optimized with trajectory-level RL algorithms such as GRPO. However, a common class of reasoning tasks, iterative optimization, presents distinct challenges: the agent interacts with the same underlying environment state across turns, and the value of a trajectory is determined by the best turn-level reward rather than cumulative returns. Existing GRPO-based methods cannot perform fine-grained, turn-level optimization in such settings, while black-box optimization methods discard prior knowledge and reasoning capabilities. To address this gap, we propose Turn-Level GRPO (TL-GRPO), a lightweight RL algorithm that performs turn-level group sampling for fine-grained optimization. We evaluate TL-GRPO on analog circuit sizing (ACS), a challenging scientific optimization task requiring multiple simulations and domain expertise. Results show that TL-GRPO outperforms standard GRPO and Bayesian optimization methods across various specifications. Furthermore, our 30B model trained with TL-GRPO achieves state-of-the-art performance on ACS tasks under same simulation budget, demonstrating both strong generalization and practical utility.

While Large Language Models (LLMs) demonstrate remarkable proficiency in semantic understanding, they often struggle to ensure structural consistency and reasoning reliability in complex decision-making tasks that demand rigorous logic. Although classical decision theories, such as the Analytic Hierarchy Process (AHP), offer systematic rational frameworks, their construction relies heavily on labor-intensive domain expertise, creating an "expert bottleneck" that hinders scalability in general scenarios. To bridge the gap between the generalization capabilities of LLMs and the rigor of decision theory, we propose Doc2AHP, a novel structured inference framework guided by AHP principles. Eliminating the need for extensive annotated data or manual intervention, our approach leverages the structural principles of AHP as constraints to direct the LLM in a constrained search within the unstructured document space, thereby enforcing the logical entailment between parent and child nodes. Furthermore, we introduce a multi-agent weighting mechanism coupled with an adaptive consistency optimization strategy to ensure the numerical consistency of weight allocation. Empirical results demonstrate that Doc2AHP not only empowers non-expert users to construct high-quality decision models from scratch but also significantly outperforms direct generative baselines in both logical completeness and downstream task accuracy.

With the rapid growth of scientific literature, scientific question answering (SciQA) has become increasingly critical for exploring and utilizing scientific knowledge. Retrieval-Augmented Generation (RAG) enhances LLMs by incorporating knowledge from external sources, thereby providing credible evidence for scientific question answering. But existing retrieval and reranking methods remain vulnerable to passages that are semantically similar but logically irrelevant, often reducing factual reliability and amplifying hallucinations.To address this challenge, we propose a Deep Evidence Reranking Agent (DeepEra) that integrates step-by-step reasoning, enabling more precise evaluation of candidate passages beyond surface-level semantics. To support systematic evaluation, we construct SciRAG-SSLI (Scientific RAG - Semantically Similar but Logically Irrelevant), a large-scale dataset comprising about 300K SciQA instances across 10 subjects, constructed from 10M scientific corpus. The dataset combines naturally retrieved contexts with systematically generated distractors to test logical robustness and factual grounding. Comprehensive evaluations confirm that our approach achieves superior retrieval performance compared to leading rerankers. To our knowledge, this work is the first to comprehensively study and empirically validate innegligible SSLI issues in two-stage RAG frameworks.

The rapid proliferation of realistic deepfakes has raised urgent concerns over their misuse, motivating the use of defensive watermarks in synthetic images for reliable detection and provenance tracking. However, this defense paradigm assumes such watermarks are inherently resistant to removal. We challenge this assumption with DeMark, a query-free black-box attack framework that targets defensive image watermarking schemes for deepfakes. DeMark exploits latent-space vulnerabilities in encoder-decoder watermarking models through a compressive sensing based sparsification process, suppressing watermark signals while preserving perceptual and structural realism appropriate for deepfakes. Across eight state-of-the-art watermarking schemes, DeMark reduces watermark detection accuracy from 100% to 32.9% on average while maintaining natural visual quality, outperforming existing attacks. We further evaluate three defense strategies, including image super resolution, sparse watermarking, and adversarial training, and find them largely ineffective. These results demonstrate that current encoder decoder watermarking schemes remain vulnerable to latent-space manipulations, underscoring the need for more robust watermarking methods to safeguard against deepfakes.

Discovery of sensitive and biologically grounded biomarkers is essential for early detection and monitoring of Alzheimer's disease (AD). Structural MRI is widely available but typically relies on hand-crafted features such as cortical thickness or volume. We ask whether self-supervised learning (SSL) can uncover more powerful biomarkers from the same data. Existing SSL methods underperform FreeSurfer-derived features in disease classification, conversion prediction, and amyloid status prediction. We introduce Residual Noise Contrastive Estimation (R-NCE), a new SSL framework that integrates auxiliary FreeSurfer features while maximizing additional augmentation-invariant information. R-NCE outperforms traditional features and existing SSL methods across multiple benchmarks, including AD conversion prediction. To assess biological relevance, we derive Brain Age Gap (BAG) measures and perform genome-wide association studies. R-NCE-BAG shows high heritability and associations with MAPT and IRAG1, with enrichment in astrocytes and oligodendrocytes, indicating sensitivity to neurodegenerative and cerebrovascular processes.

Large Language Models (LLMs) are increasingly deployed in domains such as education, mental health and customer support, where stable and consistent personas are critical for reliability. Yet, existing studies focus on narrative or role-playing tasks and overlook how adversarial conversational history alone can reshape induced personas. Black-box persona manipulation remains unexplored, raising concerns for robustness in realistic interactions. In response, we introduce the task of persona editing, which adversarially steers LLM traits through user-side inputs under a black-box, inference-only setting. To this end, we propose PHISH (Persona Hijacking via Implicit Steering in History), the first framework to expose a new vulnerability in LLM safety that embeds semantically loaded cues into user queries to gradually induce reverse personas. We also define a metric to quantify attack success. Across 3 benchmarks and 8 LLMs, PHISH predictably shifts personas, triggers collateral changes in correlated traits, and exhibits stronger effects in multi-turn settings. In high-risk domains mental health, tutoring, and customer support, PHISH reliably manipulates personas, validated by both human and LLM-as-Judge evaluations. Importantly, PHISH causes only a small reduction in reasoning benchmark performance, leaving overall utility largely intact while still enabling significant persona manipulation. While current guardrails offer partial protection, they remain brittle under sustained attack. Our findings expose new vulnerabilities in personas and highlight the need for context-resilient persona in LLMs. Our codebase and dataset is available at: https://github.com/Jivnesh/PHISH

Adversarial robustness refers to a model's ability to resist perturbation of inputs, while distribution robustness evaluates the performance of the model under data shifts. Although both aim to ensure reliable performance, prior work has revealed a tradeoff in distribution and adversarial robustness. Specifically, adversarial training might increase reliance on spurious features, which can harm distribution robustness, especially the performance on some underrepresented subgroups. We present a theoretical analysis of adversarial and distribution robustness that provides a tractable surrogate for per-step adversarial training by studying models trained on perturbed data. In addition to the tradeoff, our work further identified a nuanced phenomenon that $\ell_\infty$ perturbations on data with moderate bias can yield an increase in distribution robustness. Moreover, the gain in distribution robustness remains on highly skewed data when simplicity bias induces reliance on the core feature, characterized as greater feature separability. Our theoretical analysis extends the understanding of the tradeoff by highlighting the interplay of the tradeoff and the feature separability. Despite the tradeoff that persists in many cases, overlooking the role of feature separability may lead to misleading conclusions about robustness.

The Python Package Index (PyPI) has become a target for malicious actors, yet existing detection tools generate false positive rates of 15-30%, incorrectly flagging one-third of legitimate packages as malicious. This problem arises because current tools rely on simple syntactic rules rather than semantic understanding, failing to distinguish between identical API calls serving legitimate versus malicious purposes. To address this challenge, we propose PyGuard, a knowledge-driven framework that converts detection failures into useful behavioral knowledge by extracting patterns from existing tools' false positives and negatives. Our method utilizes hierarchical pattern mining to identify behavioral sequences that distinguish malicious from benign code, employs Large Language Models to create semantic abstractions beyond syntactic variations, and combines this knowledge into a detection system that integrates exact pattern matching with contextual reasoning. PyGuard achieves 99.50% accuracy with only 2 false positives versus 1,927-2,117 in existing tools, maintains 98.28% accuracy on obfuscated code, and identified 219 previously unknown malicious packages in real-world deployment. The behavioral patterns show cross-ecosystem applicability with 98.07% accuracy on NPM packages, demonstrating that semantic understanding enables knowledge transfer across programming languages.

Retrieval-Augmented Generation (RAG) has emerged as a dominant paradigm for mitigating hallucinations in Large Language Models (LLMs) by incorporating external knowledge. Nevertheless, effectively integrating and interpreting key evidence scattered across noisy documents remains a critical challenge for existing RAG systems. In this paper, we propose GraphAnchor, a novel Graph-Anchored Knowledge Indexing approach that reconceptualizes graph structures from static knowledge representations into active, evolving knowledge indices. GraphAnchor incrementally updates a graph during iterative retrieval to anchor salient entities and relations, yielding a structured index that guides the LLM in evaluating knowledge sufficiency and formulating subsequent subqueries. The final answer is generated by jointly leveraging all retrieved documents and the final evolved graph. Experiments on four multi-hop question answering benchmarks demonstrate the effectiveness of GraphAnchor, and reveal that GraphAnchor modulates the LLM's attention to more effectively associate key information distributed in retrieved documents. All code and data are available at https://github.com/NEUIR/GraphAnchor.

We demonstrate sufficiency of events-based synchronisation for solving deterministic fault-tolerant consensus in asynchrony. Main result is an algorithm that terminates with valid vector agreement, hence operates with safety, liveness, and tolerance to one crash. Reconciling with the FLP impossibility result, we identified: i) existence of two types of agreements: data-independent and data-dependent; and ii) dependence of FLP theorem correctness on three implicit assumptions. Consensus impossibility with data-dependent agreement is contingent on two of them. The theorem-stated impossibility with every agreement type hinges entirely on the third. We provide experimental results showing that the third assumption has no evidence in support.

Open-source ecosystems such as NPM and PyPI are increasingly targeted by supply chain attacks, yet existing detection methods either depend on fragile handcrafted rules or data-driven features that fail to capture evolving attack semantics. We present IntelGuard, a retrieval-augmented generation (RAG) based framework that integrates expert analytical reasoning into automated malicious package detection. IntelGuard constructs a structured knowledge base from over 8,000 threat intelligence reports, linking malicious code snippets with behavioral descriptions and expert reasoning. When analyzing new packages, it retrieves semantically similar malicious examples and applies LLM-guided reasoning to assess whether code behaviors align with intended functionality. Experiments on 4,027 real-world packages show that IntelGuard achieves 99% accuracy and a 0.50% false positive rate, while maintaining 96.5% accuracy on obfuscated code. Deployed on PyPI.org, it discovered 54 previously unreported malicious packages, demonstrating interpretable and robust detection guided by expert knowledge.

Programmers are turning to AI coding assistants to answer questions about their code. Benchmarks are needed to soundly evaluate these systems and understand their performance. To enable such a study, we curate a benchmark of real-world contextualized questions derived from Github pull request comments. Out of this work, we present RubberDuckBench: a multilingual benchmark of questions about code, along with detailed rubrics for evaluating answers. We evaluate a diverse set of 20 LLMs (proprietary & open-source) on answering these questions. We find that even state of the art models fail to give consistent, correct responses across the benchmark. Grok 4 (69.29%), Claude Opus 4 (68.5%), and GPT-5 (67.8%) perform best overall, but do not exhibit pairwise significant superiority over the next 9 best performing models. Most models obtain points through partial credit, with the best performing models only answering at most 2 questions completely correctly across all trials. Furthermore, models often hallucinate with lies in 58.3\% of responses on average. Cost analysis reveals no correlation between expense (API pricing or parameter count) and performance. We intend this benchmark to be a target for future research in trustworthy and correct AI coding assistants.

The study on the expressive power of transformers shows that transformers are permutation equivariant, and they can approximate all permutation-equivariant continuous functions on a compact domain. However, these results are derived under real parameters and exact operations, while real implementations on computers can only use a finite set of numbers and inexact machine operations with round-off errors. In this work, we investigate the representability of floating-point transformers that use floating-point parameters and floating-point operations. Unlike existing results under exact operations, we first show that floating-point transformers can represent a class of non-permutation-equivariant functions even without positional encoding. Furthermore, we prove that floating-point transformers can represent all permutation-equivariant functions when the sequence length is bounded, but they cannot when the sequence length is large. We also found the minimal equivariance structure in floating-point transformers, and show that all non-trivial additive positional encoding can harm the representability of floating-point transformers.

Understanding human emotions from multimodal signals poses a significant challenge in affective computing and human-robot interaction. While multimodal large language models (MLLMs) have excelled in general vision-language tasks, their capabilities in emotional reasoning remain limited. The field currently suffers from a scarcity of large-scale datasets with high-quality, descriptive emotion annotations and lacks standardized benchmarks for evaluation. Our preliminary framework, Emotion-LLaMA, pioneered instruction-tuned multimodal learning for emotion reasoning but was restricted by explicit face detectors, implicit fusion strategies, and low-quality training data with limited scale. To address these limitations, we present Emotion-LLaMAv2 and the MMEVerse benchmark, establishing an end-to-end pipeline together with a standardized evaluation setting for emotion recognition and reasoning. Emotion-LLaMAv2 introduces three key advances. First, an end-to-end multiview encoder eliminates external face detection and captures nuanced emotional cues via richer spatial and temporal multiview tokens. Second, a Conv Attention pre-fusion module is designed to enable simultaneous local and global multimodal feature interactions external to the LLM backbone. Third, a perception-to-cognition curriculum instruction tuning scheme within the LLaMA2 backbone unifies emotion recognition and free-form emotion reasoning. To support large-scale training and reproducible evaluation, MMEVerse aggregates twelve publicly available emotion datasets, including IEMOCAP, MELD, DFEW, and MAFW, into a unified multimodal instruction format. The data are re-annotated via a multi-agent pipeline involving Qwen2 Audio, Qwen2.5 VL, and GPT 4o, producing 130k training clips and 36k testing clips across 18 evaluation benchmarks.

Large language models (LLMs) have demonstrated exceptional performance in reasoning tasks such as mathematics and coding, matching or surpassing human capabilities. However, these impressive reasoning abilities face significant challenges in specialized domains. Taking Go as an example, although AlphaGo has established the high performance ceiling of AI systems in Go, mainstream LLMs still struggle to reach even beginner-level proficiency, let alone perform natural language reasoning. This performance gap between general-purpose LLMs and domain experts is significantly limiting the application of LLMs on a wider range of domain-specific tasks. In this work, we aim to bridge the divide between LLMs' general reasoning capabilities and expert knowledge in domain-specific tasks. We perform mixed fine-tuning with structured Go expertise and general long Chain-of-Thought (CoT) reasoning data as a cold start, followed by reinforcement learning to integrate expert knowledge in Go with general reasoning capabilities. Through this methodology, we present \textbf{LoGos}, a powerful LLM that not only maintains outstanding general reasoning abilities, but also conducts Go gameplay in natural language, demonstrating effective strategic reasoning and accurate next-move prediction. LoGos achieves performance comparable to human professional players, substantially surpassing all existing LLMs. Through this work, we aim to contribute insights on applying general LLM reasoning capabilities to specialized domains. We will release the first large-scale Go dataset for LLM training, the first LLM Go evaluation benchmark, and the first general LLM that reaches human professional-level performance in Go at: https://github.com/Entarochuan/LoGos.

Deep learning models are effective for sequential data modeling, yet commonly used activation functions such as ReLU, LeakyReLU, and PReLU often exhibit gradient instability when applied to noisy, non-stationary financial time series. This study introduces BrownianReLU, a stochastic activation function induced by Brownian motion that enhances gradient propagation and learning stability in Long Short-Term Memory (LSTM) networks. Using Monte Carlo simulation, BrownianReLU provides a smooth, adaptive response for negative inputs, mitigating the dying ReLU problem. The proposed activation is evaluated on financial time series from Apple, GCB, and the S&P 500, as well as LendingClub loan data for classification. Results show consistently lower Mean Squared Error and higher $R^2$ values, indicating improved predictive accuracy and generalization. Although ROC-AUC metric is limited in classification tasks, activation choice significantly affects the trade-off between accuracy and sensitivity, with Brownian ReLU and the selected activation functions yielding practically meaningful performance.

``LLM-as-a-judge,'' which utilizes large language models (LLMs) as evaluators, has proven effective in many evaluation tasks. However, evaluator LLMs exhibit numerical bias, a phenomenon where certain evaluation scores are generated disproportionately often, leading reduced evaluation performance. This study investigates the cause of this bias. Given that most evaluator LLMs are aligned through instruction tuning and preference tuning, and that prior research suggests alignment reduces output diversity, we hypothesize that numerical bias arises from alignment. To test this, we compare outputs from pre- and post-alignment LLMs, and observe that alignment indeed increases numerical bias. We also explore mitigation strategies for post-alignment LLMs, including temperature scaling, distribution calibration, and score range adjustment. Among these, score range adjustment is most effective in reducing bias and improving performance, though still heuristic. Our findings highlight the need for further work on optimal score range selection and more robust mitigation strategies.

Environments are the bottleneck for self-improving agents. Current terminal benchmarks were built for evaluation, not training; reinforcement learning requires a scalable pipeline, not just a dataset. We introduce Endless Terminals, a fully autonomous pipeline that procedurally generates terminal-use tasks without human annotation. The pipeline has four stages: generating diverse task descriptions, building and validating containerized environments, producing completion tests, and filtering for solvability. From this pipeline we obtain 3255 tasks spanning file operations, log management, data processing, scripting, and database operations. We train agents using vanilla PPO with binary episode level rewards and a minimal interaction loop: no retrieval, multi-agent coordination, or specialized tools. Despite this simplicity, models trained on Endless Terminals show substantial gains: on our held-out dev set, Llama-3.2-3B improves from 4.0% to 18.2%, Qwen2.5-7B from 10.7% to 53.3%, and Qwen3-8B-openthinker-sft from 42.6% to 59.0%. These improvements transfer to human-curated benchmarks: models trained on Endless Terminals show substantial gains on held out human curated benchmarks: on TerminalBench 2.0, Llama-3.2-3B improves from 0.0% to 2.2%, Qwen2.5-7B from 2.2% to 3.4%, and Qwen3-8B-openthinker-sft from 1.1% to 6.7%, in each case outperforming alternative approaches including models with more complex agentic scaffolds. These results demonstrate that simple RL succeeds when environments scale.

Existing face-swapping methods often deliver competitive results in constrained settings but exhibit substantial quality degradation when handling extreme facial poses. To improve facial pose robustness, explicit geometric features are applied, but this approach remains problematic since it introduces additional dependencies and increases computational cost. Diffusion-based methods have achieved remarkable results; however, they are impractical for real-time processing. We introduce AlphaFace, which leverages an open-source vision-language model and CLIP image and text embeddings to apply novel visual and textual semantic contrastive losses. AlphaFace enables stronger identity representation and more precise attribute preservation, all while maintaining real-time performance. Comprehensive experiments across FF++, MPIE, and LPFF demonstrate that AlphaFace surpasses state-of-the-art methods in pose-challenging cases. The project is publicly available on `https://github.com/andrewyu90/Alphaface_Official.git'.

Exact recovery in stochastic block models (SBMs) is well understood in undirected settings, but remains considerably less developed for directed and sparse networks, particularly when the number of communities diverges. Spectral methods for directed SBMs often lack stability in asymmetric, low-degree regimes, and existing non-spectral approaches focus primarily on undirected or dense settings. We propose a fully non-spectral, two-stage procedure for community detection in sparse directed SBMs with potentially growing numbers of communities. The method first estimates the directed probability matrix using a neighborhood-smoothing scheme tailored to the asymmetric setting, and then applies $K$-means clustering to the estimated rows, thereby avoiding the limitations of eigen- or singular value decompositions in sparse, asymmetric networks. Our main theoretical contribution is a uniform row-wise concentration bound for the smoothed estimator, obtained through new arguments that control asymmetric neighborhoods and separate in- and out-degree effects. These results imply the exact recovery of all community labels with probability tending to one, under mild sparsity and separation conditions that allow both $γ_n \to 0$ and $K_n \to \infty$. Simulation studies, including highly directed, sparse, and non-symmetric block structures, demonstrate that the proposed procedure performs reliably in regimes where directed spectral and score-based methods deteriorate. To the best of our knowledge, this provides the first exact recovery guarantee for this class of non-spectral, neighborhood-smoothing methods in the sparse, directed setting.

We investigate two important tasks in odor-related property modeling: Vapor Pressure (VP) and Odor Threshold (OP). To evaluate the model's out-of-distribution (OOD) capability, we adopt the Bemis-Murcko scaffold split. In terms of features, we introduce the rich A20/E17 molecular graph features (20-dimensional atom features + 17-dimensional bond features) and systematically compare GINE and PNA backbones. The results show: for VP, PNA with a simple regression head achieves Val MSE $\approx$ 0.21 (normalized space); for the OP single task under the same scaffold split, using A20/E17 with robust training (Huber/winsor) achieves Val MSE $\approx$ 0.60-0.61. For multitask training, we propose a **"safe multitask"** approach: VP as the primary task and OP as the auxiliary task, using delayed activation + gradient clipping + small weight, which avoids harming the primary task and simultaneously yields the best VP generalization performance. This paper provides complete reproducible experiments, ablation studies, and error-similarity analysis while discussing the impact of data noise and method limitations.

Designing experiments that systematically gather data from complex physical systems is central to accelerating scientific discovery. While Bayesian experimental design (BED) provides a principled, information-based framework that integrates experimental planning with probabilistic inference, the selection of utility functions in BED is a long-standing and active topic, where different criteria emphasize different notions of information. Although Kullback--Leibler (KL) divergence has been one of the most common choices, recent studies have proposed Wasserstein distance as an alternative. In this work, we first employ a toy example to illustrate an issue of Wasserstein distance - the value of Wasserstein distance of a fixed-shape posterior depends on the relative position of its main mass within the support and can exhibit false rewards unrelated to information gain, especially with a non-informative prior (e.g., uniform distribution). We then further provide a systematic comparison between these two criteria through a classical source inversion problem in the BED literature, revealing that the KL divergence tends to lead to faster convergence in the absence of model discrepancy, while Wasserstein metrics provide more robust sequential BED results if model discrepancy is non-negligible. These findings clarify the trade-offs between KL divergence and Wasserstein metrics for the utility function and provide guidelines for selecting suitable criteria in practical BED applications.

We present RENEW, a global path planner for Autonomous Surface Vehicle (ASV) in dynamic environments with external disturbances (e.g., water currents). RENEW introduces a unified risk- and energy-aware strategy that ensures safety by dynamically identifying non-navigable regions and enforcing adaptive safety constraints. Inspired by maritime contingency planning, it employs a best-effort strategy to maintain control under adverse conditions. The hierarchical architecture combines high-level constrained triangulation for topological diversity with low-level trajectory optimization within safe corridors. Validated with real-world ocean data, RENEW is the first framework to jointly address adaptive non-navigability and topological path diversity for robust maritime navigation.

Multimodal large language models (MLLMs) have shown remarkable capabilities in multimodal perception and understanding tasks. However, their effectiveness in specialized domains, such as remote sensing and medical imaging, remains limited. A natural approach to domain adaptation is to inject domain knowledge through textual instructions, prompts, or auxiliary captions. Surprisingly, we find that such input-level domain knowledge injection yields little to no improvement on scientific multimodal tasks, even when the domain knowledge is explicitly provided. This observation suggests that current MLLMs fail to internalize domain-specific priors through language alone, and that domain knowledge must be integrated at the optimization level. Motivated by this insight, we propose a reinforcement fine-tuning framework that incorporates domain knowledge directly into the learning objective. Instead of treating domain knowledge as descriptive information, we encode it as domain-informed constraints and reward signals, shaping the model's behavior in the output space. Extensive experiments across multiple datasets in remote sensing and medical domains consistently demonstrate good performance gains, achieving state-of-the-art results on multimodal domain tasks. Our results highlight the necessity of optimization-level domain knowledge integration and reveal a fundamental limitation of textual domain conditioning in current MLLMs.

Difficulty replicating baselines, high computational costs, and required domain expertise create persistent barriers to clinical AI research. To address these challenges, we introduce PyHealth 2.0, an enhanced clinical deep learning toolkit that enables predictive modeling in as few as 7 lines of code. PyHealth 2.0 offers three key contributions: (1) a comprehensive toolkit addressing reproducibility and compatibility challenges by unifying 15+ datasets, 20+ clinical tasks, 25+ models, 5+ interpretability methods, and uncertainty quantification including conformal prediction within a single framework that supports diverse clinical data modalities - signals, imaging, and electronic health records - with translation of 5+ medical coding standards; (2) accessibility-focused design accommodating multimodal data and diverse computational resources with up to 39x faster processing and 20x lower memory usage, enabling work from 16GB laptops to production systems; and (3) an active open-source community of 400+ members lowering domain expertise barriers through extensive documentation, reproducible research contributions, and collaborations with academic health systems and industry partners, including multi-language support via RHealth. PyHealth 2.0 establishes an open-source foundation and community advancing accessible, reproducible healthcare AI. Available at pip install pyhealth.

The study of \textit{regret minimization in fixed-price bilateral trade} has received considerable attention in recent research. Previous works [CCC+24a, CCC+24b, AFF24, BCCF24, CJLZ25, LCM25a, GDFS25] have acquired a thorough understanding of the problem, except for determining the tight regret bound for GBB semi-feedback fixed-price mechanisms under adversarial values. In this paper, we resolve this open question by devising an $\widetilde{O}(T^{2 / 3})$-regret mechanism, matching the $Ω(T^{2 / 3})$ lower bound from [CJLZ25] up to polylogarithmic factors.

Vapnik--Chervonenkis' theorem is a seminal result in machine learning. It establishes sufficient conditions for empirical probabilities to converge to theoretical probabilities, uniformly over families of events. It also provides an estimate for the rate of such uniform convergence. We revisit the probabilistic component of the classical argument. Instead of applying Hoeffding's inequality at the final step, we use a normal approximation with explicit Berry--Esseen error control. This yields a moderate-deviation sharpening of the usual VC estimate, with an additional factor of order $(\varepsilon\sqrt{n})^{-1}$ in the leading exponential term when $\varepsilon\sqrt{n}$ is large.

Large language models (LLMs) make next-token predictions based on clues present in their context, such as semantic descriptions and in-context examples. Yet, elucidating which prior tokens most strongly influence a given prediction remains challenging due to the proliferation of layers and attention heads in modern architectures. We propose Jacobian Scopes, a suite of gradient-based, token-level causal attribution methods for interpreting LLM predictions. By analyzing the linearized relations of final hidden state with respect to inputs, Jacobian Scopes quantify how input tokens influence a model's prediction. We introduce three variants - Semantic, Fisher, and Temperature Scopes - which respectively target sensitivity of specific logits, the full predictive distribution, and model confidence (inverse temperature). Through case studies spanning instruction understanding, translation and in-context learning (ICL), we uncover interesting findings, such as when Jacobian Scopes point to implicit political biases. We believe that our proposed methods also shed light on recently debated mechanisms underlying in-context time-series forecasting. Our code and interactive demonstrations are publicly available at https://github.com/AntonioLiu97/JacobianScopes.

Rare-event prediction is critical in domains such as healthcare, finance, reliability engineering, customer support, aviation safety, where positive outcomes are infrequent yet potentially catastrophic. Extreme class imbalance biases conventional models toward majority-class predictions, limiting recall, calibration, and operational usefulness. We propose LPCORP (Low-Prevalence CORrector for Prediction)*, a two-stage framework that combines reasoningenhanced prediction with confidence-based outcome correction. A reasoning model first produces enriched predictions from narrative inputs, after which a lightweight logistic-regression classifier evaluates and selectively corrects these outputs to mitigate prevalence-driven bias. We evaluate LPCORP on real-world datasets from medical and consumer service domains. The results show that this method transforms a highly imbalanced setting into a well-balanced one while preserving the original number of samples and without applying any resampling strategies. Test-set evaluation demonstrates substantially improved performance, particularly in precision, which is a known weakness in low-prevalence data. We further provide a costreduction analysis comparing the expenses associated with rare-event damage control without preventive measures to those incurred when low-cost, prediction-based preventive interventions are applied that showed more than 50% reduction in some cases. * Patent pending: U.S. Provisional 63/933,518, filed 8 December 2025.

Coverage Path Planning (CPP) is a fundamental capability for agricultural robots; however, existing solutions often overlook energy constraints, resulting in incomplete operations in large-scale or resource-limited environments. This paper proposes an energy-aware CPP framework grounded in Soft Actor-Critic (SAC) reinforcement learning, designed for grid-based environments with obstacles and charging stations. To enable robust and adaptive decision-making under energy limitations, the framework integrates Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal dynamics. A dedicated reward function is designed to jointly optimize coverage efficiency, energy consumption, and return-to-base constraints. Experimental results demonstrate that the proposed approach consistently achieves over 90% coverage while ensuring energy safety, outperforming traditional heuristic algorithms such as Rapidly-exploring Random Tree (RRT), Particle Swarm Optimization (PSO), and Ant Colony Optimization (ACO) baselines by 13.4-19.5% in coverage and reducing constraint violations by 59.9-88.3%. These findings validate the proposed SAC-based framework as an effective and scalable solution for energy-constrained CPP in agricultural robotics.

Reinforcement Learning from Human Feedback (RLHF) and its variants have emerged as the dominant approaches for aligning Large Language Models with human intent. While empirically effective, the theoretical generalization properties of these methods in high-dimensional settings remain to be explored. To this end, we build the generalization theory on RLHF of LLMs under the linear reward model, through the framework of algorithmic stability. In contrast to the existing works built upon the consistency of maximum likelihood estimations on reward model, our analysis is presented under an end-to-end learning framework, which is consistent with practice. Concretely, we prove that under a key \textbf{feature coverage} condition, the empirical optima of policy model have a generalization bound of order $\mathcal{O}(n^{-\frac{1}{2}})$. Moreover, the results can be extrapolated to parameters obtained by gradient-based learning algorithms, i.e., Gradient Ascent (GA) and Stochastic Gradient Ascent (SGA). Thus, we argue that our results provide new theoretical evidence for the empirically observed generalization of LLMs after RLHF.

Real-world visual question answering (VQA) is often context-dependent: an image-question pair may be under-specified, such that the correct answer depends on external information that is not observable in the image. In such cases, directly answering can lead to confident but incorrect predictions. We propose CoA(Clarify-or-Answer), an ask-or-answer agent that separately models the decision to ask or answer, and what to ask if needed. CoA first determines whether clarification is necessary; if so, it asks a single focused question and then incorporates the response to produce the final answer. We introduce CONTEXTCLARIFY with a set of ambiguous VQA questions and the contrast set that is non-ambiguous. We further introduce GRPO-CR (Clarification Reasoning), a reinforcement learning approach that optimizes clarification question generation with multiple reward signals encouraging well-formed, focused, non-trivial questions that resolve ambiguity. Across three VLLMs and three datasets, CoA achieves consistent improvements at both the module and system levels, improving end-to-end VQA accuracy by an average of +15.3 points (83%) over prompting-based baselines

We study a structured bi-level optimization problem where the upper-level objective is a smooth function and the lower-level problem is policy optimization in a Markov decision process (MDP). The upper-level decision variable parameterizes the reward of the lower-level MDP, and the upper-level objective depends on the optimal induced policy. Existing methods for bi-level optimization and RL often require second-order information, impose strong regularization at the lower level, or inefficiently use samples through nested-loop procedures. In this work, we propose a single-loop, first-order actor-critic algorithm that optimizes the bi-level objective via a penalty-based reformulation. We introduce into the lower-level RL objective an attenuating entropy regularization, which enables asymptotically unbiased upper-level hyper-gradient estimation without solving the unregularized RL problem exactly. We establish the finite-time and finite-sample convergence of the proposed algorithm to a stationary point of the original, unregularized bi-level optimization problem through a novel lower-level residual analysis under a special type of Polyak-Lojasiewicz condition. We validate the performance of our method through experiments on a GridWorld goal position problem and on happy tweet generation through reinforcement learning from human feedback (RLHF).

Algorithmic audits are essential tools for examining systems for properties required by regulators or desired by operators. Current audits of large language models (LLMs) primarily rely on black-box evaluations that assess model behavior only through input-output testing. These methods are limited to tests constructed in the input space, often generated by heuristics. In addition, many socially relevant model properties (e.g., gender bias) are abstract and difficult to measure through text-based inputs alone. To address these limitations, we propose a white-box sensitivity auditing framework for LLMs that leverages activation steering to conduct more rigorous assessments through model internals. Our auditing method conducts internal sensitivity tests by manipulating key concepts relevant to the model's intended function for the task. We demonstrate its application to bias audits in four simulated high-stakes LLM decision tasks. Our method consistently reveals substantial dependence on protected attributes in model predictions, even in settings where standard black-box evaluations suggest little or no bias. Our code is openly available at https://github.com/hannahxchen/llm-steering-audit

Trustworthy clinical summarization requires not only fluent generation but also transparency about where each statement comes from. We propose a training-free framework for generation-time source attribution that leverages decoder attentions to directly cite supporting text spans or images, overcoming the limitations of post-hoc or retraining-based methods. We introduce two strategies for multimodal attribution: a raw image mode, which directly uses image patch attentions, and a caption-as-span mode, which substitutes images with generated captions to enable purely text-based alignment. Evaluations on two representative domains: clinician-patient dialogues (CliConSummation) and radiology reports (MIMIC-CXR), show that our approach consistently outperforms embedding-based and self-attribution baselines, improving both text-level and multimodal attribution accuracy (e.g., +15% F1 over embedding baselines). Caption-based attribution achieves competitive performance with raw-image attention while being more lightweight and practical. These findings highlight attention-guided attribution as a promising step toward interpretable and deployable clinical summarization systems.

Referring Expression Segmentation (RES) is a core vision-language segmentation task that enables pixel-level understanding of targets via free-form linguistic expressions, supporting critical applications such as human-robot interaction and augmented reality. Despite the progress of Multimodal Large Language Model (MLLM)-based approaches, existing RES methods still suffer from two key limitations: first, the coarse bounding boxes from MLLMs lead to redundant or non-discriminative point prompts; second, the prevalent reliance on textual coordinate reasoning is unreliable, as it fails to distinguish targets from visually similar distractors. To address these issues, we propose \textbf{\model}, a novel RES framework integrating \textbf{E}ntropy-\textbf{B}ased Point \textbf{D}iscovery (\textbf{EBD}) and \textbf{V}ision-\textbf{B}ased \textbf{R}easoning (\textbf{VBR}). Specifically, EBD identifies high-information candidate points by modeling spatial uncertainty within coarse bounding boxes, treating point selection as an information maximization process. VBR verifies point correctness through joint visual-semantic alignment, abandoning text-only coordinate inference for more robust validation. Built on these components, \model implements a coarse-to-fine workflow: bounding box initialization, entropy-guided point discovery, vision-based validation, and mask decoding. Extensive evaluations on four benchmark datasets (RefCOCO, RefCOCO+, RefCOCOg, and ReasonSeg) demonstrate that \model achieves new state-of-the-art performance across all four benchmarks, highlighting its effectiveness in generating accurate and semantically grounded segmentation masks with minimal prompts.

With the advent of agentic AI, Software Engineering is transforming to a new era dubbed Software Engineering 3.0. Software project management (SPM) must also evolve with such transformations to boost successful project completion, while keeping humans at the heart of it. Building on our preliminary ideas of "agentic SPM", and supporting literature, we present our vision of an "Agentic Project Manager (PM)" as a multi-agent system for SPM 3.0. They will work like a "junior project manager", or an "intern project manager" collaboratively with software teams. We introduce four working modes, with varying autonomy levels to choose from, based on the SPM task. This addresses concerns with ethics, accountability, and trust related to agentic PMs. We also share insights on human PM role evolution and new skill requirements as a "strategic leader" and a "coach" for humans and agents. While creating the foundation for agentic SPM research, we present a research agenda for the wider research community.

Large language models exhibit strong multilingual capabilities, yet significant performance gaps persist between dominant and non-dominant languages. Prior work attributes this gap to imbalances between shared and language-specific neurons in multilingual representations. We propose Cross-Lingual Activation Steering (CLAS), a training-free inference-time intervention that selectively modulates neuron activations. We evaluate CLAS on classification and generation benchmarks, achieving average improvements of 2.3% (Acc.) and 3.4% (F1) respectively, while maintaining high-resource language performance. We discover that effective transfer operates through functional divergence rather than strict alignment; performance gains correlate with increased language cluster separation. Our results demonstrate that targeted activation steering can unlock latent multilingual capacity in existing models without modification to model weights.

Multimodal language models (MLMs) perform well on semantic vision-language tasks but fail at spatial reasoning that requires adopting another agent's visual perspective. These errors reflect a persistent egocentric bias and raise questions about whether current models support allocentric reasoning. Inspired by human spatial cognition, we introduce perspective tokens, specialized embeddings that encode orientation through either (1) embodied body-keypoint cues or (2) abstract representations supporting mental rotation. Integrating these tokens into LLaVA-1.5-13B yields performance on level-2 visual perspective-taking tasks. Across synthetic and naturalistic benchmarks (Isle Bricks V2, COCO, 3DSRBench), perspective tokens improve accuracy, with rotation-based tokens generalizing to non-human reference agents. Representational analyses reveal that fine-tuning enhances latent orientation sensitivity already present in the base model, suggesting that MLMs contain precursors of allocentric reasoning but lack appropriate internal structure. Overall, embedding cognitively grounded spatial structure directly into token space provides a lightweight, model-agnostic mechanism for perspective-taking and more human-like spatial reasoning.

On-demand Polymer discovery is essential for various industries, ranging from biomedical to reinforcement materials. Experiments with polymers have a long trial-and-error process, leading to long procedures and extensive resources. For these processes, machine learning has accelerated scientific discovery at the property prediction and latent space search fronts. However, laboratory researchers cannot readily access codes and these models to extract individual structures and properties due to infrastructure limitations. We present a closed-loop polymer structure-property predictor integrated in a terminal for early-stage polymer discovery. The framework is powered by LLM reasoning to provide users with property prediction, property-guided polymer structure generation, and structure modification capabilities. The SMILES sequences are guided by the synthetic accessibility score and the synthetic complexity score (SC Score) to ensure that polymer generation is as close as possible to synthetically accessible monomer-level structures. This framework addresses the challenge of generating novel polymer structures for laboratory researchers, thereby providing computational insights into polymer research.

Community detection (CD) on signed networks is crucial for understanding how positive and negative relations jointly shape network structure. However, existing CD methods often yield inconsistent communities due to noisy or conflicting edge signs. In this paper, we propose ReCon, a model-agnostic post-processing framework that progressively refines community structures through four iterative steps: (1) structural refinement, (2) boundary refinement, (3) contrastive learning, and (4) clustering. Extensive experiments on eighteen synthetic and four real-world networks using four CD methods demonstrate that ReCon consistently enhances community detection accuracy, serving as an effective and easily integrable solution for reliable CD across diverse network properties.

This paper provides a fresh view of the neural network (NN) data flow problem, i.e., identifying the NN connections that are most important for the performance of the full model, through the lens of graph theory. Understanding the NN data flow provides a tool for symbolic NN analysis, e.g.,~robustness analysis or model repair. Unlike the standard approach to NN data flow analysis, which is based on information theory, we employ the notion of graph curvature, specifically Ollivier-Ricci curvature (ORC). The ORC has been successfully used to identify important graph edges in various domains such as road traffic analysis, biological and social networks. In particular, edges with negative ORC are considered bottlenecks and as such are critical to the graph's overall connectivity, whereas positive-ORC edges are not essential. We use this intuition for the case of NNs as well: we 1)~construct a graph induced by the NN structure and introduce the notion of neural curvature (NC) based on the ORC; 2)~calculate curvatures based on activation patterns for a set of input examples; 3)~aim to demonstrate that NC can indeed be used to rank edges according to their importance for the overall NN functionality. We evaluate our method through pruning experiments and show that removing negative-ORC edges quickly degrades the overall NN performance, whereas positive-ORC edges have little impact. The proposed method is evaluated on a variety of models trained on three image datasets, namely MNIST, CIFAR-10 and CIFAR-100. The results indicate that our method can identify a larger number of unimportant edges as compared to state-of-the-art pruning methods.

The software engineering research community is productive, yet it faces a constellation of challenges: swamped review processes, metric-driven incentives, distorted publication practices, and increasing pressures from AI, scale, and outright scams. These issues are often treated in isolation, yet they arise from deep structural dynamics within the research ecosystem itself and distract us from the larger role of research in society. Meaningful progress requires a holistic system-level view. We sketch such a framework drawing on ideas from complex systems, ecosystems, and theory of change. Reframing SE's challenges through this lens reveals non-linear feedback loops that sustain current dysfunctions, and it helps to identify leverage points for reform. These are less a matter of isolated fixes and more a matter of exploring coordinated sets of fixes that operate across the SE ecosystem

A desirable property of any deployed artificial intelligence is generalization across domains, i.e. data generation distribution under a specific acquisition condition. In medical imagining applications the most coveted property for effective deployment is Single Domain Generalization (SDG), which addresses the challenge of training a model on a single domain to ensure it generalizes well to unseen target domains. In multi-center studies, differences in scanners and imaging protocols introduce domain shifts that exacerbate variability in rare class characteristics. This paper presents our experience on SDG in real life deployment for two exemplary medical imaging case studies on seizure onset zone detection using fMRI data, and stress electrocardiogram based coronary artery detection. Utilizing the commonly used application of diabetic retinopathy, we first demonstrate that state-of-the-art SDG techniques fail to achieve generalized performance across data domains. We then develop a generic expert knowledge integrated deep learning technique DL+EKE and instantiate it for the DR application and show that DL+EKE outperforms SOTA SDG methods on DR. We then deploy instances of DL+EKE technique on the two real world examples of stress ECG and resting state (rs)-fMRI and discuss issues faced with SDG techniques.

Humans act via a nuanced process that depends both on rational deliberation and also on identity and contextual factors. In this work, we study how large language models (LLMs) can simulate human action in the context of social dilemma games. While prior work has focused on "steering" (weak binding) of chat models to simulate personas, we analyze here how deep binding of base models with extended backstories leads to more faithful replication of identity-based behaviors. Our study has these findings: simulation fidelity vs human studies is improved by conditioning base LMs with rich context of narrative identities and checking consistency using instruction-tuned models. We show that LLMs can also model contextual factors such as time (year that a study was performed), question framing, and participant pool effects. LLMs, therefore, allow us to explore the details that affect human studies but which are often omitted from experiment descriptions, and which hamper accurate replication.

Although boosting software development performance, large language model (LLM)-powered code generation introduces intellectual property and data security risks rooted in the fact that a service provider (cloud) observes a client's prompts and generated code, which can be proprietary in commercial systems. To mitigate this problem, we propose NOIR, the first framework to protect the client's prompts and generated code from the cloud. NOIR uses an encoder and a decoder at the client to encode and send the prompts' embeddings to the cloud to get enriched embeddings from the LLM, which are then decoded to generate the code locally at the client. Since the cloud can use the embeddings to infer the prompt and the generated code, NOIR introduces a new mechanism to achieve indistinguishability, a local differential privacy protection at the token embedding level, in the vocabulary used in the prompts and code, and a data-independent and randomized tokenizer on the client side. These components effectively defend against reconstruction and frequency analysis attacks by an honest-but-curious cloud. Extensive analysis and results using open-source LLMs show that NOIR significantly outperforms existing baselines on benchmarks, including the Evalplus (MBPP and HumanEval, Pass@1 of 76.7 and 77.4), and BigCodeBench (Pass@1 of 38.7, only a 1.77% drop from the original LLM) under strong privacy against attacks.

This study investigates regional bias in large language models (LLMs), an emerging concern in AI fairness and global representation. We evaluate ten prominent LLMs: GPT-3.5, GPT-4o, Gemini 1.5 Flash, Gemini 1.0 Pro, Claude 3 Opus, Claude 3.5 Sonnet, Llama 3, Gemma 7B, Mistral 7B, and Vicuna-13B using a dataset of 100 carefully designed prompts that probe forced-choice decisions between regions under contextually neutral scenarios. We introduce FAZE, a prompt-based evaluation framework that measures regional bias on a 10-point scale, where higher scores indicate a stronger tendency to favor specific regions. Experimental results reveal substantial variation in bias levels across models, with GPT-3.5 exhibiting the highest bias score (9.5) and Claude 3.5 Sonnet scoring the lowest (2.5). These findings indicate that regional bias can meaningfully undermine the reliability, fairness, and inclusivity of LLM outputs in real-world, cross-cultural applications. This work contributes to AI fairness research by highlighting the importance of inclusive evaluation frameworks and systematic approaches for identifying and mitigating geographic biases in language models.

Weather conditions can drastically alter the state of crops and rangelands, and in turn, impact the incomes and food security of individuals worldwide. Satellite-based remote sensing offers an effective way to monitor vegetation and climate variables on regional and global scales. The annual peak Normalized Difference Vegetation Index (NDVI), derived from satellite observations, is closely associated with crop development, rangeland biomass, and vegetation growth. Although various machine learning methods have been developed to forecast NDVI over short time ranges, such as one-month-ahead predictions, long-term forecasting approaches, such as one-year-ahead predictions of vegetation conditions, are not yet available. To fill this gap, we develop a two-phase machine learning model to forecast the one-year-ahead peak NDVI over high-resolution grids, using the Four Corners region of the Southwestern United States as a testbed. In phase one, we identify informative climate attributes, including precipitation and maximum vapor pressure deficit, and develop the generalized parallel Gaussian process that captures the relationship between climate attributes and NDVI. In phase two, we forecast these climate attributes using historical data at least one year before the NDVI prediction month, which then serve as inputs to forecast the peak NDVI at each spatial grid. We developed open-source tools that outperform alternative methods for both gross NDVI and grid-based NDVI one-year forecasts, providing information that can help farmers and ranchers make actionable plans a year in advance.

Data science agents promise to accelerate discovery and insight-generation by turning data into executable analyses and findings. Yet existing data science benchmarks fall short due to fragmented evaluation interfaces that make cross-benchmark comparison difficult, narrow task coverage and a lack of rigorous data grounding. In particular, we show that a substantial portion of tasks in current benchmarks can be solved without using the actual data. To address these limitations, we introduce DSGym, a standardized framework for evaluating and training data science agents in self-contained execution environments. Unlike static benchmarks, DSGym provides a modular architecture that makes it easy to add tasks, agent scaffolds, and tools, positioning it as a live, extensible testbed. We curate DSGym-Tasks, a holistic task suite that standardizes and refines existing benchmarks via quality and shortcut solvability filtering. We further expand coverage with (1) DSBio: expert-derived bioinformatics tasks grounded in literature and (2) DSPredict: challenging prediction tasks spanning domains such as computer vision, molecular prediction, and single-cell perturbation. Beyond evaluation, DSGym enables agent training via execution-verified data synthesis pipeline. As a case study, we build a 2,000-example training set and trained a 4B model in DSGym that outperforms GPT-4o on standardized analysis benchmarks. Overall, DSGym enables rigorous end-to-end measurement of whether agents can plan, implement, and validate data analyses in realistic scientific context.

With the growing ubiquity of multi-core architectures, concurrent systems have become essential but increasingly prone to complex issues such as data races and deadlocks. While modern issue-tracking systems facilitate the reporting of such problems, labeling concurrency-related bug reports remains a labor-intensive and error-prone task. This paper presents a linguistic-pattern-based framework for automatically identifying concurrency bug reports. We derive 58 distinct linguistic patterns from 730 manually labeled concurrency bug reports, organized across four levels: word-level (keywords), phrase-level (n-grams), sentence-level (semantic), and bug report-level (contextual). To assess their effectiveness, we evaluate four complementary approaches-matching, learning, prompt-based, and fine-tuning-spanning traditional machine learning, large language models (LLMs), and pre-trained language models (PLMs). Our comprehensive evaluation on 12 large-scale open-source projects (10,920 issue reports from GitHub and Jira) demonstrates that fine-tuning PLMs with linguistic-pattern-enriched inputs achieves the best performance, reaching a precision of 91% on GitHub and 93% on Jira, and maintaining strong precision on post cut-off data (91%). The contributions of this work include: (1) a comprehensive taxonomy of linguistic patterns for concurrency bugs, (2) a novel fine-tuning strategy that integrates domain-specific linguistic knowledge into PLMs, and (3) a curated, labeled dataset to support reproducible research. Together, these advances provide a foundation for improving the automation, precision, and interpretability of concurrency bug classification.

Simulating and validating coordination among multiple autonomous vehicles (AVs) is a challenging task as most existing simulation architectures are limited to single-vehicle operation or rely on centralized control. This paper presents a Distributed Multi-AV Architecture (DMAVA) that enables synchronized, real-time autonomous driving simulation across multiple physical hosts. Each vehicle runs its own complete AV stack and operates independently from other AVs. The vehicles in the simulation maintain synchronized coordination through a low-latency data-centric communication layer. The proposed system integrates ROS 2 Humble, Autoware Universe, AWSIM Labs, and Zenoh to support concurrent execution of multiple Autoware stacks within a shared Unity-based environment. Experiments conducted on multiple-host configurations demonstrate stable localization, reliable inter-host communication, and fully synchronized closed-loop control. The DMAVA also serves as a foundation for Multi-Vehicle Autonomous Valet Parking, demonstrating its extensibility toward higher-level cooperative autonomy. Demo videos and source code are available at: https://github.com/zubxxr/distributed-multi-autonomous-vehicle-architecture.

Foundation models are used for many real-world applications involving language generation from temporally-ordered multimodal events. In this work, we study the ability of models to identify the most important sub-events in a video, which is a fundamental prerequisite for narrating or summarizing multimodal events. Specifically, we focus on football games and evaluate models on their ability to distinguish between important and non-important sub-events in a game. To this end, we construct a new dataset by leveraging human preferences for importance implicit in football game highlight reels, without any additional annotation costs. Using our dataset, which we will publicly release to the community, we compare several state-of-the-art multimodal models and show that they are not far from chance level performance. Analyses of models beyond standard evaluation metrics reveal their tendency to rely on a single dominant modality and their ineffectiveness in synthesizing necessary information from multiple sources. Our findings underline the importance of modular architectures that can handle sample-level heterogeneity in multimodal data and the need for complementary training procedures that can maximize cross-modal synergy.

We propose a novel training objective for GPs constructed using lower-dimensional linear projections of the data, referred to as \emph{projected likelihood} (PL). We provide a closed-form expression for the information loss related to the PL and empirically show that it can be reduced with random projections on the unit sphere. We show the superiority of the PL, in terms of accuracy and computational efficiency, over the exact GP training and the variational free energy approach to sparse GPs over different optimisers, kernels and datasets of moderately large sizes.

This paper presents the DMV-AVP System, a distributed simulation of Multi-Vehicle Autonomous Valet Parking (AVP). The system was implemented as an application of the Distributed Multi-Vehicle Architecture (DMAVA) for synchronized multi-host execution. Most existing simulation approaches rely on centralized or non-distributed designs that constrain scalability and limit fully autonomous control. This work introduces two modules built on top of the DMAVA: 1) a Multi-Vehicle AVP Node that performs state-based coordination, queuing, and reservation management across multiple vehicles, and 2) a Unity-Integrated YOLOv5 Parking Spot Detection Module that provides real-time, vision-based perception within AWSIM Labs. Both modules integrate seamlessly with the DMAVA and extend it specifically for multi-vehicle AVP operation, supported by a Zenoh-based communication layer that ensures low-latency topic synchronization and coordinated behavior across hosts. Experiments conducted on two- and three-host configurations demonstrate deterministic coordination, conflict-free parking behavior, and scalable performance across distributed Autoware instances. The results confirm that the proposed Distributed Multi-Vehicle AVP System supports cooperative AVP simulation and establishes a foundation for future real-world and hardware-in-the-loop validation. Demo videos and source code are available at https://github.com/zubxxr/multi-vehicle-avp

College students experience many stressors, resulting in high levels of anxiety and depression. Wearable technology provides unobtrusive sensor data that can be used for the early detection of mental illness. However, current research is limited concerning the variety of psychological instruments administered, physiological modalities, and time series parameters. In this research, we collect the Student Mental and Environmental Health (StudentMEH) Fitbit dataset from students at our institution during the pandemic. We provide a comprehensive assessment of the ability of predictive machine learning models to screen for depression, anxiety, and stress using different Fitbit modalities. Our findings indicate potential in physiological modalities such as heart rate and sleep to screen for mental illness with the F1 scores as high as 0.79 for anxiety, the former modality reaching 0.77 for stress screening, and the latter modality achieving 0.78 for depression. This research highlights the potential of wearable devices to support continuous mental health monitoring, the importance of identifying best data aggregation levels and appropriate modalities for screening for different mental ailments.

We introduce an efficient few-shot keyword spotting model for edge devices, EdgeSpot, that pairs an optimized version of a BC-ResNet-based acoustic backbone with a trainable Per-Channel Energy Normalization frontend and lightweight temporal self-attention. Knowledge distillation is utilized during training by employing a self-supervised teacher model, optimized with Sub-center ArcFace loss. This study demonstrates that the EdgeSpot model consistently provides better accuracy at a fixed false-alarm rate (FAR) than strong BC-ResNet baselines. The largest variant, EdgeSpot-4, improves the 10-shot accuracy at 1% FAR from 73.7% to 82.0%, which requires only 29.4M MACs with 128k parameters.

Large language models (LLMs) enable rapid and consistent automated evaluation of open-ended exam responses, including dimensions of content and argumentation that have traditionally required human judgment. This is particularly important in cases where a large amount of exams need to be graded in a limited time frame, such as nation-wide graduation exams in various countries. Here, we examine the applicability of automated scoring on two large datasets of trial exam essays of two full national cohorts from Estonia. We operationalize the official curriculum-based rubric and compare LLM and statistical natural language processing (NLP) based assessments with human panel scores. The results show that automated scoring can achieve performance comparable to that of human raters and tends to fall within the human scoring range. We also evaluate bias, prompt injection risks, and LLMs as essay writers. These findings demonstrate that a principled, rubric-driven, human-in-the-loop scoring pipeline is viable for high-stakes writing assessment, particularly relevant for digitally advanced societies like Estonia, which is about to adapt a fully electronic examination system. Furthermore, the system produces fine-grained subscore profiles that can be used to generate systematic, personalized feedback for instruction and exam preparation. The study provides evidence that LLM-assisted assessment can be implemented at a national scale, even in a small-language context, while maintaining human oversight and compliance with emerging educational and regulatory standards.

Research in AI4Science has shown promise in many science applications, including polymer design. However, current LLMs prove ineffective on this problem space because: (i) most models lack polymer-specific knowledge (ii) existing aligned models lack coverage of knowledge and capabilities relevant to polymer design. Addressing this, we introduce PolyBench, a large scale training and test benchmark dataset of more than 125K polymer design related tasks, leveraging a knowledge base of 13M+ data points obtained from experimental and synthetic sources to ensure broad coverage of polymers and their properties. For effective alignment using PolyBench, we introduce a knowledge-augmented reasoning distillation method that augments this dataset with structured CoT. Furthermore, tasks in PolyBench are organized from simple to complex analytical reasoning problems, enabling generalization tests and diagnostic probes across the problem space. Experiments show that small language models (SLMs), of 7B to 14B parameters, trained on PolyBench data outperform similar sized models, and even closed source frontier LLMs on PolyBench test dataset while demonstrating gains on other polymer benchmarks as well.

We introduce DNIPRO, a novel longitudinal corpus of 246K news articles documenting the Russo-Ukrainian war from Feb 2022 to Aug 2024, spanning eleven media outlets across five nation states (Russia, Ukraine, U.S., U.K., and China) and three languages (English, Russian, and Mandarin Chinese). This multilingual resource features consistent and comprehensive metadata, and multiple types of annotation with rigorous human evaluations for downstream tasks relevant to systematic transnational analyses of contentious wartime discourse. DNIPRO's distinctive value lies in its inclusion of competing geopolitical perspectives, making it uniquely suited for studying narrative divergence, media framing, and information warfare. To demonstrate its utility, we include use case experiments using stance detection, sentiment analysis, topical framing, and contradiction analysis of major conflict events within the larger war. Our explorations reveal how outlets construct competing realities, with coverage exhibiting polarized interpretations that reflect geopolitical interests. Beyond supporting computational journalism research, DNIPRO provides a foundational resource for understanding how conflicting narratives emerge and evolve across global information ecosystems.

Recent foundational video-to-video diffusion models have achieved impressive results in editing user provided videos by modifying appearance, motion, or camera movement. However, real-world video editing is often an iterative process, where users refine results across multiple rounds of interaction. In this multi-turn setting, current video editors struggle to maintain cross-consistency across sequential edits. In this work, we tackle, for the first time, the problem of cross-consistency in multi-turn video editing and introduce Memory-V2V, a simple, yet effective framework that augments existing video-to-video models with explicit memory. Given an external cache of previously edited videos, Memory-V2V employs accurate retrieval and dynamic tokenization strategies to condition the current editing step on prior results. To further mitigate redundancy and computational overhead, we propose a learnable token compressor within the DiT backbone that compresses redundant conditioning tokens while preserving essential visual cues, achieving an overall speedup of 30%. We validate Memory-V2V on challenging tasks including video novel view synthesis and text-conditioned long video editing. Extensive experiments show that Memory-V2V produces videos that are significantly more cross-consistent with minimal computational overhead, while maintaining or even improving task-specific performance over state-of-the-art baselines. Project page: https://dohunlee1.github.io/MemoryV2V

General Matrix Multiplication (GEMM) is the cornerstone of Deep Learning and HPC workloads; accordingly, academia and industry have heavily optimized this kernel. Modern platforms with matrix multiplication accelerators exhibit high FLOP/Byte machine balance, which makes implementing optimal matrix multiplication challenging. On modern CPU platforms with matrix engines, state-of-the-art vendor libraries tune input tensor layouts, parallelization schemes, and cache blocking to minimize data movement across the memory hierarchy and maximize throughput. However, the best settings for these parameters depend strongly on the target platform (number of cores, memory hierarchy, cache sizes) and on the shapes of the matrices, making exhaustive tuning infeasible; in practice this leads to performance "glass jaws". In this work we revisit space filling curves (SFC) to alleviate the problem of this cumbersome tuning. SFC convert multi-dimensional coordinates (e.g. 2D) into a single dimension (1D), keeping nearby points in the high-dimensional space close in the 1D order. We partition the Matrix Multiplication computation space using recent advancements in generalized SFC (Generalized Hilbert Curves), and we obtain platform-oblivious and shape-oblivious matrix-multiplication schemes that exhibit inherently high degree of data locality. Furthermore, we extend the SFC-based work partitioning to implement Communication-Avoiding (CA) algorithms that replicate the input tensors and provably minimize communication/data-movement on the critical path. The integration of CA-algorithms is seamless and yields compact code (~30 LOC), yet it achieves state-of-the-art results on multiple CPU platforms, outperforming vendor libraries by up to 2x(geometric-mean speedup) for a range of GEMM shapes.

Agent-based modeling (ABM) has emerged as an indispensable methodology for studying complex adaptive systems across the natural and social sciences. However, Python-based ABM frameworks face a fundamental tension between the accessibility that has made Python dominant in scientific computing and the performance requirements of large-scale simulations. This paper introduces AMBER, a framework that resolves this tension through a novel architectural approach: replacing the conventional object-per-agent representation with columnar state management using the Polars DataFrame library. We analyze the computational characteristics of both paradigms, present the architectural design of AMBER including its core abstractions, spatial environments, experiment management, and optimization capabilities. Empirical evaluation on three canonical benchmarks demonstrates that AMBER achieves speedups of 1.2x to 93x depending on workload characteristics, with the greatest advantages for models dominated by population-wide attribute operations. Memory profiling reveals 30-50% reduction in peak usage compared to object-oriented frameworks. Our results establish columnar state management as a viable architectural foundation for high-performance ABM in interpreted languages.

Active learning for photonic crystals explores the integration of analytic approximate Bayesian last layer neural networks (LL-BNNs) with uncertainty-driven sample selection to accelerate photonic band gap prediction. We employ an analytic LL-BNN formulation, corresponding to the infinite Monte Carlo sample limit, to obtain uncertainty estimates that are strongly correlated with the true predictive error on unlabeled candidate structures. These uncertainty scores drive an active learning strategy that prioritizes the most informative simulations during training. Applied to the task of predicting band gap sizes in two-dimensional, two-tone photonic crystals, our approach achieves up to a 2.6x reduction in required training data compared to a random sampling baseline while maintaining predictive accuracy. The efficiency gains arise from concentrating computational resources on high uncertainty regions of the design space rather than sampling uniformly. Given the substantial cost of full band structure simulations, especially in three dimensions, this data efficiency enables rapid and scalable surrogate modeling. Our results suggest that analytic LL-BNN based active learning can substantially accelerate topological optimization and inverse design workflows for photonic crystals, and more broadly, offers a general framework for data efficient regression across scientific machine learning domains.

Agentic AI pipelines suffer from a hidden inefficiency: they frequently reconstruct identical intermediate logic, such as metric normalization or chart scaffolding, even when the user's natural language phrasing is entirely novel. Conventional boundary caching fails to capture this inefficiency because it treats inference as a monolithic black box. We introduce SemanticALLI, a pipeline-aware architecture within Alli (PMG's marketing intelligence platform), designed to operationalize redundant reasoning. By decomposing generation into Analytic Intent Resolution (AIR) and Visualization Synthesis (VS), SemanticALLI elevates structured intermediate representations (IRs) to first-class, cacheable artifacts. The impact of caching within the agentic loop is substantial. In our evaluation, baseline monolithic caching caps at a 38.7% hit rate due to linguistic variance. In contrast, our structured approach allows for an additional stage, the Visualization Synthesis stage, to achieve an 83.10% hit rate, bypassing 4,023 LLM calls with a median latency of just 2.66 ms. This internal reuse reduces total token consumption, offering a practical lesson for AI system design: even when users rarely repeat themselves, the pipeline often does, at stable, structured checkpoints where caching is most reliable.

Contemporary automated scientific discovery has focused on agents for generating scientific experiments, while systems that perform higher-level scientific activities such as theory building remain underexplored. In this work, we formulate the problem of synthesizing theories consisting of qualitative and quantitative laws from large corpora of scientific literature. We study theory generation at scale, using 13.7k source papers to synthesize 2.9k theories, examining how generation using literature-grounding versus parametric knowledge, and accuracy-focused versus novelty-focused generation objectives change theory properties. Our experiments show that, compared to using parametric LLM memory for generation, our literature-supported method creates theories that are significantly better at both matching existing evidence and at predicting future results from 4.6k subsequently-written papers

Multi-agent systems powered by large language models (LLMs) are transforming enterprise automation, yet systematic evaluation methodologies for assessing tool-use reliability remain underdeveloped. We introduce a comprehensive diagnostic framework that leverages big data analytics to evaluate procedural reliability in intelligent agent systems, addressing critical needs for SME-centric deployment in privacy-sensitive environments. Our approach features a 12-category error taxonomy capturing failure modes across tool initialization, parameter handling, execution, and result interpretation. Through systematic evaluation of 1,980 deterministic test instances spanning both open-weight models (Qwen2.5 series, Functionary) and proprietary alternatives (GPT-4, Claude 3.5/3.7) across diverse edge hardware configurations, we identify actionable reliability thresholds for production deployment. Our analysis reveals that procedural reliability, particularly tool initialization failures, constitutes the primary bottleneck for smaller models, while qwen2.5:32b achieves flawless performance matching GPT-4.1. The framework demonstrates that mid-sized models (qwen2.5:14b) offer practical accuracy-efficiency trade-offs on commodity hardware (96.6\% success rate, 7.3 s latency), enabling cost-effective intelligent agent deployment for resource-constrained organizations. This work establishes foundational infrastructure for systematic reliability evaluation of tool-augmented multi-agent AI systems.

Large Language Models (LLMs) have demonstrated remarkable multilingual capabilities, making them promising tools in both high- and low-resource languages. One particularly valuable use case is generating synthetic samples that can be used to train smaller models in low-resource scenarios where human-labelled data is scarce. In this work, we investigate whether these synthetic data generation capabilities can serve as a form of distillation, producing smaller models that perform on par with or even better than massive LLMs across languages and tasks. To this end, we use a state-of-the-art multilingual LLM to generate synthetic datasets covering 11 languages and 4 classification tasks. These datasets are then used to train smaller models via fine-tuning or instruction tuning, or as synthetic in-context examples for compact LLMs. Our experiments show that even small amounts of synthetic data enable smaller models to outperform the large generator itself, particularly in low-resource languages. Overall, the results suggest that LLMs are best utilised as generators (teachers) rather than classifiers, producing data that empowers smaller and more efficient multilingual models.

The Edelstein effect, which enables charge-to-spin conversion and is therefore highly promising for future spintronic devices, can be realized and non-volatilely manipulated in ferroelectric materials owing to their broken inversion symmetry and switchable polarization states. To date, most ferroelectric systems reported to exhibit the Edelstein effect are semiconductors, requiring extrinsic doping for functionality. In contrast, the Edelstein effect has rarely been reported in metallic ferroelectric systems, where doping is unnecessary. Using first-principles calculations, we predict that a pronounced Edelstein effect can be realized in the recently proposed intrinsic two-dimensional ferroelectric metal PtBi$_{2}$ monolayer, where the sign of the Edelstein coefficient is coupled to the direction of ferroelectric polarization through the polarization-switching-induced reversal of spin textures, thereby enabling non-volatile control of charge-spin conversion. The Edelstein effect reaches a magnitude of $10^{11}~\hbar/(\textup{A} \cdot \textup{cm})$, which is sizable compared to previously reported ferroelectric systems. Microscopically, the Edelstein effect in a PtBi$_2$ monolayer originates from competing contributions of inner Rashba-like electron pockets and outer hole pockets with opposite signs; an upward shift of the Fermi level alters their balance and can reverse the sign of the Edelstein effect. Upon applying biaxial strain, the Fermi-surface electronic structure is strongly modified, resulting in a pronounced change of the Edelstein effect: a 2 \% compressive strain suppresses the Edelstein effect by about 50 \%. Our results not only identify a promising material platform for tunable charge-spin conversion but also provide new insights into the functional potential of metallic ferroelectric systems.

Symplectic quantization is a functional approach to quantum field theory that allows sampling of quantum fluctuations directly in Minkowski space-time by means of a generalized microcanonical ensemble similar to the one of the standard microcanonical approach to lattice field theory. In a previous paper we showed that, for an interacting scalar field theory in 1+1-dimensions, this formalism allows to capture numerically some crucial real-time features inaccessible to any Euclidean approach to lattice field theory. Yet, the new approach was plagued by two main limitations: an ill-defined non-interacting limit and the absence of a direct formal correspondence between its correlation functions and those generated by the Feynman path integral approach. In this paper, we introduce the new \emph{"constrained symplectic quantization"} approach, for which the perfect equivalence with the Feynman path integral is proved and which is perfectly well defined for the free theory. This new approach is characterized by the analytical continuation of all fields and of the action from $\mathbb{R}$ to $\mathbb{C}$ and the presence of some constraints which guarantee the stability of the generalized Hamiltonian dynamics and the convergence of the corresponding generalized microcanonical partition function, hence the name of the theory. We show the application of this formalism to the quantum harmonic oscillator on a Minkowskian-time lattice, finding perfect agreement between one- and two-point numerical correlators and the exact quantum-mechanical results. We observe genuine real-time features such as the oscillatory propagator and the discrete excited-state energy spectrum. Our results provide strong numerical evidence that constrained symplectic quantization can sample real-time quantum-mechanical observables, offering a concrete route to overcome the limitations of Euclidean-time importance sampling.

Understanding the coupling between lattice vibrations and magnetic order is crucial for controlling properties of two-dimensional magnetic materials. Here, we investigate the vibrational properties of bulk and thick-flake CrCl$_\textrm{3}$ using polarization-resolved Raman spectroscopy, complemented by photoluminescence, photoluminescence excitation, and optical absorption measurements. Symmetry analysis, supported by first-principles phonon calculations, enables the unambiguous assignment of all eight Raman-active modes, four $\textrm{A}_\textrm{g}$ and four $\textrm{E}_\textrm{g}$, previously predicted only theoretically. Excitation-energy-dependent measurements reveal that the strong enhancement of selected phonon modes originates primarily from interference effects rather than resonant Raman scattering. Temperature-dependent Raman spectroscopy further reveals pronounced signatures of spin-phonon coupling across the transition from a fully antiferromagnetic phase, through an intermediate regime with local, domain-like ferromagnetic order, to the paramagnetic phase, accompanied by a clear rhombohedral-to-monoclinic structural transition. Together, these results demonstrate how lattice, electronic, and magnetic degrees of freedom collectively govern the Raman response of CrCl$_\textrm{3}$.

We study the fluctuation statistics of integrated currents in noisy quantum diffusive systems, focusing on the Quantum Symmetric Simple Exclusion and Inclusion Processes (QSSEP/QSSIP). These one-dimensional fermionic (QSSEP) and bosonic (QSSIP) models feature stochastic nearest-neighbor hopping driven by Brownian noise, together with boundary injection and removal processes. They provide solvable microscopic settings in which quantum coherence coexists with diffusion. Upon noise averaging, their dynamics reduce to those of the classical SSEP/SSIP. We show that the cumulant generating function of the integrated current, at large scales, obeys a large deviation principle. To leading order in system size and for each noise realization, it converges to that of the corresponding classical process, establishing a classical typicality of current fluctuations in these noisy quantum systems. We further demonstrate a direct connection with Macroscopic Fluctuation Theory (MFT), showing that the large-scale equations satisfied by biased quantum densities coincide with the steady-state Hamilton equations of MFT, thereby providing a microscopic quantum justification of the MFT framework in these models. Finally, we identify the leading finite-size corrections to the current statistics. We show the existence of subleading contributions of purely quantum origin, which are absent in the corresponding classical setting, and provide their explicit expressions for the second and third current cumulants. These quantum corrections are amenable to direct experimental or numerical verification, provided sufficient control over the noise realizations can be achieved. Their presence points toward the necessity of a quantum extension of Macroscopic Fluctuation Theory.

While electron-electron (e-e) interactions are known to influence resistivity in non-Galilean invariant two-dimensional (2D) systems, their effect on magnetotransport is not fully understood. Conventional models for simple bands often predict a vanishing magnetoresistivity from e-e interactions alone. In this work, we investigate magnetotransport in a gapless 6.3 nm HgTe quantum well, a hybrid 2D band system that hosts coexisting holes with both linear (Dirac-like) and parabolic energy bands. Focusing on the high temperature regime where particle-particle collisions dominate scattering, we observe significant corrections to both the magnetoresistivity and the Hall effect. The high temperature transport coefficients are in good agreement with the theoretical model describing transport in massive-massless fermion mixtures governed by a frictional mechanism and intervalley scattering. Our findings provide strong experimental validation for this theoretical framework, demonstrating that collisions between particles with different dispersions are a key mechanism governing magnetotransport in hybrid band semimetals.

Magnetotactic bacteria synthesize linear chains of magnetite nanoparticles within their bodies, which allow the bacteria to navigate the Earth's magnetic field in search of the best habitat. Biogenic magnetite particles, called magnetosomes, are very promising for use in biomedicine. Magnetosome chains have also been found in ancient fossils and sediments. The study of magnetofossils provides valuable information about the Earth's biological past. The presence of biogenic magnetite in ancient rock samples can be detected by measuring ferromagnetic resonance spectra, first-order magnetization reversal curves, or quasi-static hysteresis loops. Theoretical analyses of these experiments generally assume that magnetosomes are spherical nanoparticles, although the shape of some types of magnetosomes is close to spheroidal one. In this work, simple formulas for describing the magneto-dipole interaction of oriented spheroids are obtained and quasi-static hysteresis loops of randomly oriented magnetosome chain assembly consisting of elongated spheroids are calculated.

We derive and validate a partition function for low-dimensional systems interacting with a heat bath, addressing the general issue of thermodynamic modeling of nanoscale systems. In contrast to bulk systems in the canonical (NVT) ensemble where the partition function is solely determined by the Hamiltonian of the system and the temperature of the heat bath, our formulation demonstrates that accounting for the interactions with the heat bath is essential for describing the statistical mechanics of low-dimensional materials. To validate our theoretical findings, we develop a molecular dynamics (MD) algorithm for directly modeling the heat bath as a gas reservoir. We first validate our approach using a 1D harmonic oscillator, calculating its length distribution through explicit numerical integration and confirming these results with MD simulations. We then extend our method to investigate the out-of-plane fluctuations of a 2D graphene monolayer immersed in a gas at finite temperature and pressure. Comparisons with conventional NVT ensemble simulations controlled by a thermostat reveal that environmental interactions significantly influence the properties of the 2D material system.

We present a comprehensive first-principles study of twisted bilayer graphene (tBLG) for a wide range of twist angles, with a focus on structural and electronic properties. By employing density functional theory (DFT) with an optimized local basis set, we simulate tBLG, obtaining fully relaxed commensurate structures for twist angles down to 0.987°. For all angles the lattice relaxation agrees well with continuum elastic models. For angles accessible to plane-wave DFT (VASP), we provide a detailed comparison with our local basis DFT (SIESTA) calculations, demonstrating excellent agreement in both the atomic and electronic structure. The dependence of the Fermi velocity and band width on the twist angle shows qualitative agreement with results from an `exact' $\mathbf{k \cdot p}$ continuum model, but reveals a small twist angle offset. Additionally, we provide details of the low-energy wavefunction character, band inversion and symmetries. Our results provide an ab initio reference point for the microscopic structure and electronic properties of tBLG which will serve as the foundation for future studies incorporating many-body effects.

Fractional quantum Hall (FQH) states host fractionally charged anyons with exotic exchange statistics. Of particular interest are FQH phases supporting non-Abelian anyons, which can encode topologically protected quantum information. In this work, we construct quasihole bases for a broad family of non-Abelian FQH states using parton wave functions, which reproduces the fusion-space dimensionality expected from their underlying conformal field theory, consistent with level-rank duality across the parton family. As an application, we numerically compute braiding matrices for representative parton states for large systems, providing a general framework for diagnosing non-Abelian characteristics in candidate FQH states.

Flat-band bound states in the continuum (BICs) are topological states with suppressed group velocity and robustness against radiation loss, offering a powerful platform for the exploration of non-Hermitian, nonlinear, topological phenomena and device applications. Van der Waals (vdW) metasurfaces have recently emerged as promising candidates for sustaining BICs and hybridizing with material transitions. However, the realization of flat-band BICs remains elusive. Here, we experimentally demonstrate polaritonic high-order BICs on a wide-angle flat band utilizing a subwavelength metasurface made of a vdW magnet CrSBr. The large oscillator strength of direct excitons in CrSBr enables near ultrastrong coupling with BICs, leading to strongly suppressed polaritonic angular dispersions. Remarkably, second-order polaritonic BICs become flat-band across a wide angular range, with corresponding Q factors exceeding 1500. Additionally, we find that these polaritonic BICs vanish in the transverse magnetic configuration, while leading to fascinating surface hyperbolic exciton-polaritons within the Reststrahlen band. Our findings underscore CrSBr as an exceptional platform for exploring flat-band photonics and polaritonics, paving the new avenue for advances in next-generation optical and quantum technologies.

It is well known that a solid (e.g. wood or rubber) can be put under tensile stress by pulling on it. Once a critical stress is overcome, the solid breaks, leaving an empty space. Similarly, due to internal cohesion, a liquid can withstand tension (i.e. negative pressure), up to a critical point where a large bubble spontaneously forms, releasing the tension and leaving a void (the bubble). This process is known as cavitation. While water at negative pressure is metastable, such a state can be long-lived. In fact, water under tension is found routinely in the plant kingdom, as a direct effect of dehydration, e.g. by evaporation. In this chapter, we provide a brief overview of occurrences of water stress and cavitation in plants, then use a simple thermodynamic and fluid mechanical framework to describe the basic physics of water stress and cavitation. We focus specifically on situations close to those in plants, that is water at negative pressure nested within a structure that is solid, but porous and potentially deformable. We also discuss insights from these simple models as well as from experiments with artificial structures mimicking some essential aspects of the structures found within plants.

The intricate wobbling motion of flagellated bacteria, characterized by the periodic precession of the cell body, is a determinant factor in their motility and navigation within complex fluid environments. While well-studied in quiescent fluids, bacterial wobbling under ubiquitous flow conditions remains unexplored. In this work, we investigate the wobbling dynamics of \textit{Escherichia coli} swimming near surfaces under steady shear flow. Our experiments reveal that the wobbling amplitude intensifies with flow strength before reaching a plateau, with this amplification exhibiting a strong dependence on the swimming orientation relative to the flow direction. It turns out that the enhanced wobbling remains governed by the misalignment between the cell body and the flagellar bundle. Furthermore, we observe that the wobbling frequency increases monotonically with flow strength, and that shorter bacteria exhibit more pronounced variations in both amplitude and frequency. By linking the wobbling motion to the intrinsic body-flagella misalignment, we attribute the flow-enhanced precession to a combination of shear- and chirality-induced torques acting on the flexible flagellar hook. This mechanical coupling ultimately suppresses the net migration velocity as the flow rate increases. These findings elucidate the elastohydrodynamic mechanisms by which shear flow modifies bacterial locomotion near surfaces, with implications for microbial transport in physiological and ecological environments.

Lecture notes on elements of nonequilibrium statistical mechanics: (1) a characterization of the nonequilibrium condition, largely by contrast to equilibrium; (2) a retelling of some of the great performances of the more distant past, including the perspectives of Boltzmann and Onsager; and (3) more recent methods and concepts, from local detailed balance and the identification of entropy fluxes to dynamical fluctuation theory, and the importance of dynamical activity.

Dispersionless electronic bands lead to an extremely high density of states and suppressed kinetic energy, thereby increasing electronic correlations and instabilities that can shape emergent ordered states, such as excitonic, ferromagnetic, and superconducting phases. A flat band that extends over the entire momentum space and is well isolated from other dispersive bands is, therefore, particularly interesting. Here, the band structure of the van der Waals crystal NbOCl$_2$ is revealed by utilizing photoelectron momentum microscopy. We directly map out an electronic band that is flat throughout the entire Brillouin zone and features a width of only $\sim$100 meV. This band is well isolated from both the conduction and remote valence bands. Moreover, the quasiparticle band gap shows a high tunability upon the deposition of caesium atoms on the surface. By combining the single-particle band structure with the optical transmission spectrum, the optical gap is identified. The fully isolated flat band in a van der Waals crystal provides a qualitatively new testbed for exploring flat-band physics.

Carbon dioxide hydrates are ice-like nonstoichiometric inclusion solid compounds with importance to global climate change, and gas transportation and storage. The thermodynamic and kinetic mechanisms that control carbon dioxide nucleation critically depend on hydrate-water interfacial free energy. Interfacial energies show large uncertainties due to the conditions at which experiments are performed. Under these circumstances, we hypothesize that accurate molecular models for water and carbon dioxide combined with computer simulation tools can offer an alternative but complementary way to estimate interfacial energies at coexistence conditions from a molecular perspective. We have evaluated the interfacial free energy of carbon dioxide hydrates at coexistence conditions (three-phase equilibrium or dissociation line) implementing advanced computational methodologies, including the novel Mold Integration methodology. Our calculations are based on the definition of the interfacial free energy, standard statistical thermodynamic techniques, and the use of the most reliable and used molecular models for water (TIP4P/Ice) and carbon dioxide (TraPPE) available in the literature. We find that simulations provide an interfacial energy value, at coexistence conditions, consistent with the experiments from its thermodynamic definition. Our calculations are reliable since are based on the use of two molecular models that accurately predict: (1) The ice-water interfacial free energy; and (2) the dissociation line of carbon dioxide hydrates. Computer simulation predictions provide alternative but reliable estimates of the carbon dioxide interfacial energy. Our pioneering work demonstrates that is possible to predict interfacial energies of hydrates from a truly computational molecular perspective and opens a new door to the determination of free energies of hydrates.

We employ inhomogeneous integral equation theory to investigate the equilibrium properties of hard disks confined to a channel of width $L$ by hard parallel walls. If the channel width is narrowed below two disk diameters, then the system enters a quasi one-dimensional regime for which the particles cannot move past each other. In the limit when $L$ is equal to one particle diameter the system reduces to the one-dimensional bulk along the center of the channel. We study first the dimensional crossover properties of the inhomogeneous Percus-Yevick (PY) integral equation as $L$ is reduced and then investigate the behaviour of a quasi one-dimensional system as the packing of the particles is increased for a fixed value of $L$. We find that the inhomogeneous PY equation is highly accurate for situations of quasi one-dimensional confinement and that it predicts the onset of a structural transition to a zigzag state at higher packing. The excellent performance of this integral equation method and the ease with which it handles confinement-induced dimensional crossover is a consequence of the improved resolution which comes from treating explicitly the inhomogeneous two-body correlation functions.

Exploiting current-induced spin-orbit torques (SOTs) to manipulate the magnetic state of dipolar-coupled nanomagnet systems with in-plane magnetic anisotropy, such as artificial spin ices, provides a route to local, electrically-programmable control of the magnetization, with relevance for applications including neuromorphic computing. Here, we demonstrate how the orientation of a nanomagnet relative to the direction of an applied electrical current impacts the threshold current density needed for all-electrical magnetization switching, and how dipolar coupling between the nanomagnets influences the switching of interacting pairs and ensembles of nanomagnets. Using a material system designed to generate SOTs in response to electrical currents, we find that the current required to switch the magnetization of isolated nanomagnets varies non-monotonically as the angle between the nanomagnet long axis and the current increases. In small artificial spin ice systems, we observe similar angular dependence of the switching current, which can be used to control the magnetization orientation of specific subsets of nanomagnets. These experimental results are supported by micromagnetic modeling, which illustrates how the various current induced torques can be exploited to control magnetization switching in nanomagnetic systems. These results establish SOT switching as a practical method for programmable manipulation of dipolar nanomagnetic systems.

We investigate integrable boundary states in the anisotropic Heisenberg chain under periodic or twisted boundary conditions, for both even and odd system lengths. Our work demonstrates that the concept of integrable boundary states can be readily generalized. For the XXZ spin chain, we present a set of factorized integrable boundary states using the KT-relation, and these states are also applicable to the XYZ chain. It is shown that a specific set of eigenstates of the transfer matrix can be selected by each boundary state, resulting in an explicit selection rule for the Bethe roots.

Precise control and nanoscale confinement of terahertz (THz) fields are essential requirements for emerging applications in photonics, quantum technologies, wireless communications, and sensing. Here, we demonstrate a polaritonic cavity enhanced THz photoresponse in an antenna coupled device based on chemical vapor deposited (CVD) monolayer graphene. The dipole antenna lobes simultaneously serve as two gate electrodes, concentrate the impinging THz field, and efficiently launch acoustic graphene plasmons (AGPs), which drive a strong photo-thermoelectric (PTE) signal. Between 6 and 90 K, the photovoltage exhibits pronounced peaks, modulating the PTE response by up to 40\%, that we attribute to AGPs forming a Fabry Pérot THz cavity in the full or half graphene channel. Combined full wave and transport thermal simulations accurately reproduce the gate controlled plasmon wavelength, spatial absorption profile, and the resulting nonuniform electron heating responsible for the PTE response. The lateral and vertical maximum confinement factors of the AGP wavelength relative to the incident wavelength are 165 and 4000, respectively, for frequencies from 1.83 to 2.52 THz. These results demonstrate that wafer scalable CVD graphene, without hBN encapsulation, can host coherent AGP resonances and exhibit an efficient polaritonic enhanced photoresponse under appropriate gating, antenna coupling, and AGP cavity design, opening a route to scalable, polarization and frequency selective, liquid nitrogen cooled, and low power consumption THz detection platforms based on plasmon thermoelectric transduction.

Magnetic skyrmions and related topological spin textures have emerged as a central topic in condensed-matter physics, combining fundamental significance with potential for transformative applications in spintronics, magnonics, and beyond. Over the past decade, advances in material platforms, imaging techniques, theoretical modeling, and device concepts have established skyrmionics as a rapidly expanding field. At the same time, challenges remain in stabilizing, controlling, and integrating such textures into functional architectures, while novel phenomena such as antiskyrmions, higher-order skyrmions, hopfions, and antiferromagnetic textures arise. The 2026 Skyrmionics Roadmap represents a collective effort of many authors, providing a comprehensive perspective on the current state-of-the-art and the outlook for the coming years. In 33 focused sections, each co-authored by two researchers, we chart progress in theory and modeling, material systems, skyrmion dynamics, and skyrmion technologies. By offering a consolidated vision, this Roadmap aims to guide both fundamental research and application-driven efforts, accelerating the transition of skyrmionics from conceptual breakthroughs toward practical technologies.

Phase-separating active systems can display phenomenology that is impossible in equilibrium. The binodal densities are not solely determined by a bulk (effective) free energy, but also affected by gradient terms, while capillary waves and Ostwald processes are determined by three distinct interfacial tensions. These and related phenomena were so far explained at continuum level using a top-down minimal theory (Active Model B+). This theory, by Taylor-expanding in the scalar order parameter (or density), effectively assumes that phase separation is weak, which is not true across most of the phase diagram. Here we develop a quantitative account of active phase separation, by introducing an active counterpart of Cahn-Hilliard theory, constructing the density current from all possible terms with up to four spatial derivatives without Taylor-expanding in the density. From this O(grad^4) theory, we show how to compute binodals and interfacial tensions for arbitrary choices of the five density-dependent 'coefficient functions' that specify the theory (replacing the four constant coefficients of Active Model B+). We further consider a particle model composed of thermal quorum-sensing active particles (tQSAPs) yielding a fully specified example of the O(grad^4) theory upon coarse-graining. We find that to coarse-grain consistently at O(grad^4) requires a novel procedure, based on multiple-scale analysis, to systematically eliminate fast-evolving orientational moments. Using this, we calculate from microscopic physics all five coefficient functions of the active Cahn-Hilliard theory for tQSAPs. We identify contributions that were missed in previous continuum theories, and show how neglecting them becomes justified only in the limit of large quorum-sensing range parameter. Comparison with particle simulations of tQSAPs shows that our O(grad^4) theory improves on previous continuum models [...]

Quantum technologies aim to assemble devices whose operation is controlled by the quantum state of individual atoms. Achieving this level of control in a practical, scalable design remains, however, a major obstacle to mass societal adoption. By working at the level of interatomic bonding, molecular engineering has enabled exquisite control over the electronic properties of individual atoms and their interactions with neighboring atoms. This positions molecular electronics as a potentially disruptive quantum technology, but serious technological challenges have prevented it from being included in technical road maps. The main obstacle is that conventional, mass scalable nanodevice technologies utilize resists and solvents that can degrade molecules. Some approaches involve exposing junction interfaces to contaminants (e.g. air, resist etc...), which can be particularly problematic for spintronics. In this technical paper, we present our decade-long work into building a nanotechnological chain that can process entire metal/molecule heterostructures into vertical nanopillars electronic devices. We discuss the advantages and pitfalls of the various iterations of this process that were implemented. We also discuss outlooks for this unique technology.

Metal-organic framework (MOF) derived materials formed through high temperature processes show great potential as catalysts. However, understanding of structure-property relationships between the initial MOF and the resulting MOF-derived catalyst is limited because the amorphous nature of the catalyst challenges standard structural characterization methods. Neural network approaches that learn interatomic potentials from density functional theory offer a promising solution. We simulated the pyrolysis of UiO-66, UiO-67 and MIP-206 using both foundational and fine-tuned machine learned interatomic potentials (MLIPs). To mimic experimental conditions, an atmosphere of CO2 and H2 was introduced and the structures were doped with 20 wt% copper to probe the effect of copper on the structural evolution of MOFs. These simulations provide atomistic insights into gas evolution, metal nanoparticle formation, and linker decomposition that were compared to available experimental data. Overall, this work demonstrates the potential of MLIPs to accurately model high temperature MOF dynamics under experimentally relevant conditions and guide the design of new catalytic materials.

The efficient generation of random quantum states is a long-standing challenge, motivated by their diverse applications in quantum information processing tasks. In this work, we identify entanglement as the key resource that enables local random unitaries to generate global random states by effectively gluing randomness across the system. Specifically, we demonstrate that approximate random states can be produced from an entangled state $|ψ\rangle$ through the application of local random unitaries. We show that the resulting ensemble forms an approximate state design with an error saturating as $Θ(e^{-\mathcal{N}_2(ψ)})$, where $\mathcal{N}_2(ψ)$ is the second Rényi entanglement entropy of $|ψ\rangle$. Furthermore, we prove that this tight bound also applies to the second Rényi entropy of coherence when the ensemble is constructed using coherence-free operations. These results imply that, when restricted to resource-free gates, the quality of the generated random states is determined entirely by the resource content of the initial state. Notably, we find that among all $α$-Rényi entropeis, the second Rényi entropy yields the tightest bounds. Consequently, these second Rényi entropies can be interpreted as the maximal capacities for generating randomness using resource-free operations. Finally, moving beyond approximate state designs, we utilize this entanglement-assisted gluing mechanism to present a novel method for generating pseudorandom states in multipartite systems from a locally entangled state via pseudorandom unitaries in each of parties.

Energy and charge transfer between molecules and metal surfaces underpin heterogeneous catalysis, surface-enhanced spectroscopies and plasmon-driven chemistry, yet the microscopic origins of vibrational excitation at metal interfaces remain unresolved. Here we use temperature-dependent surface-enhanced Raman scattering (SERS) to directly distinguish plasmon-vibration optomechanical coupling from hot-electron-driven excitation.By probing thionine adsorbed on gold nanostructures at 295 K and 3.5 K, we show that pronounced anti-Stokes scattering at cryogenic temperature arises from optical pumping of vibrational populations, whereas room-temperature spectra are governed by thermal population. Bromide co-adsorbates play a decisive role by guiding molecular alignment, inducing surface atom displacements, and enabling transient adsorption geometries that activate otherwise Raman-inactive vibrational modes. In the absence of bromide, distinct excitation pathways emerge, reflecting competition between optomechanical coupling and charge-transfer processes associated with molecular polarization along the optical field or orientation relative to the metal surface. These results establish molecular optomechanics as a sensitive probe of surface-molecule interactions and demonstrate how anion-mediated surface dynamics regulate energy flow at plasmonic interfaces.

We present a theory for the interaction between active particles and a passive flexible membrane. By explicitly solving for the pressure exerted by the active particles, we show that they reduce the membrane tension and bending modulus and introduce novel non-local contributions to the membrane mechanics. This theory predicts activity-induced instabilities and their morphology are in agreement with recent experimental and simulation data.

We study current flow around a circular density depletion in a two-dimensional electron gas in the presence of a strong magnetic field. The depletion is parametrized by a power-law tail with an exponent $β> 2$. We show that current and electrochemical potential are exponentially suppressed inside a surrounding area much larger than the geometric size of the depletion region. The corresponding ``no-go'' radius grows as a certain power of the magnetic field. Residual current and potential exhibit spiraling patterns inside the no-go region. Outside of it, they acquire corrections inversely proportional to the distance, which is known as the Landauer resistivity dipole. The Landauer dipole is rotated by the angle $π(1 - 1 / β)$ with respect to the direction of the average electric field. We also consider the effect of electron viscosity and show that the variation of the no-go radius with magnetic field becomes more rapid if viscosity is large enough. In that regime the size of the Landauer dipole is set by the Gurzhi length, which is much larger than the no-go radius, which is in turn much larger than the geometric size of the depletion. Our results may be useful for interpreting nanoimaging of current distribution in graphene and other two-dimensional systems.

We establish a unified fluctuation-response relation for Langevin dynamics. By exploiting the common mathematical structures underlying fluctuations and responses of empirical density and current, we derive a unified identity that generalizes the fluctuation-dissipation theorem from equilibrium to nonequilibrium settings. This relation connects global fluctuations of observables with their local responses to perturbations in force, mobility, and temperature. We further derive finite-time fluctuation-response inequalities, leading to response uncertainty relations that complement the identity by providing more practical bounds. These derivations establish a unified theoretical framework linking the fluctuation-dissipation theorem and thermodynamic uncertainty relations. Using the $F_1$-ATPase molecular motor model, we illustrate how these response-based bounds constrain the long-time diffusion coefficient.

Monolayer transition metal dichalcogenides (TMDs) are characterized by relatively large carrier effective masses and suppressed screening of the Coulomb interaction, which substantially enhances the correlation effects in these structures. The direct band gap allows to effectively optically probe these correlations. Here, we present an experimental observation of Wigner crystallization in monolayer $\mathrm{WSe}_2$ probed by the measurement of the exciton diffraction on the Wigner crystal (WC) periodic potential. We observe the formation of the WC phase in the absence of external magnetic fields at temperature range $T<26~\mathrm{K}$ and carrier concentrations $n$ $<2\times10^{11}~\mathrm{cm}^{-2}$. The direct observation of the exciton diffraction is enabled by the strong exciton longitudinal-transverse splitting induced by the long-range intervalley exchange interaction, leading to the large detuning between main exciton peak and first diffraction peak. Our findings highlight that the valley degree of freedom of charge carriers in TMDs facilitates optical probing of correlated electron phases in these structures.

We derive a modified Lindblad equation for the state of quantum dot tunnel coupled to a Markovian lead when the spin state of the dot is driven by an oscillating magnetic field. We show that the equation is a completely positive, trace-preserving map and find the jump operators. This is a driven-dissipative regime in which coherent driving is relevant to the tunneling and cannot be treated as simply a rotation modifying the system with a bath derived under a static magnetic field. This work was motivated by an experimental desire to determine the Zeeman splitting of an electron spin on a quantum dot (a spin qubit), and in a related work we show that this splitting energy can be found by measuring the charge occupancy of the dot while sweeping the frequency of the driving field \ arXiv:2503.17481. Here we cover the full derivation of the equation and give the jump operators. These jump operators are potentially useful for describing the stochastic behavior of more complex systems with coherent driving of a spin capable of tunneling on or off of a device, such as in electron spin resonance scanning tunneling microscopy. The jump operators have the interesting feature of combining jumps of electrons onto and off of the device.

Using the example of Ar, Kr, and Xe atoms, it is shown that graphene nanobubbles on flat substrates are multistable systems. A nanobubble can have many stable stationary states, each characterized by the number of layers, $l$, within the cluster of internal atoms. The layers are circular in shape, concentrically stacked on top of each other, forming an $l$-stepped pyramid with a flat top. The covering of this pyramid with a graphene sheet is achieved through its local stretching. The valence bonds of the sheet stretch only over the group of internal atoms; outside the coverage zone, the sheet remains undeformed and lies flush against the substrate. The maximum possible number of layers, $l_m$, increases monotonically with the number of atoms $N$ ($l_m=6$ for $N=4000$). The graphene sheet, interacting with the substrate, compresses the internal atom cluster against it, generating an internal pressure of $P\sim 1$ GPa. Numerical simulations of thermal vibrations reveal that among all $l$-layer configurations of a nanobubble, there is always one "ground"\ state. Upon heating, this ground state smoothly transitions into a layerless liquid state. All other stationary states transform into this ground state once a certain temperature is reached (for $N=4000$, the ground state corresponds to state with $l=4$). The coexistence of several stable states with different numbers of layers at low temperatures leads to the absence of a universal shape for the nanobubbles. In this scenario, the height-to-radius ratio, $H/R$, can vary from 0 to 0.24, depending on the number of layers.

We present a theoretical framework for non-collinear antiferromagnetic spin torque oscillators (NC-AFM STO) by unifying spin dynamics under the Poisson Bracket formalism. Shifting from traditional torque-based descriptions to an operational symmetry perspective, we develop two complementary viewpoints: a vector perspective identifying infinite degenerate Rigid Body Precession (RBP) states where exchange energy depends solely on the total magnetic momentum, and a particle perspective decomposing dynamics into Center-of-Mass (CM) translation and Relative Motion (RM) oscillation. Using time-dependent rotational and translational transformation techniques, we analytically resolve the rapid (~10 ps) transient evolution into a stable RBP state driven by SOT and damping. We demonstrate that the out-of-plane anisotropy (OPA) lifts the exchange degeneracy, triggering a long-term (~1 ns) oscillatory decay toward a steady state characterized by uniform spin z-components and a 120-degree inter-spin locking angle. This state is accurately governed by our Terminal Velocity Motion (TVM) model [arXiv:2305.14013], where exchange coupling transforms into kinetic energy with a light effective mass. The model precisely predicts SOT-driven transients, hysteretic excitation, and the dynamic phase diagram. Finally, we account for the sub-critical current regime mismatch by identifying a 'Rigid-Body Breaking' effect: a surge in effective friction caused by the self-resonance of RM variables induced by CM translation, mediated by the in-plane anisotropy (IPA).

When parameters of a dynamical system change sufficiently fast, critical transitions can take place even in the absence of bifurcations. This phenomenon is known as rate-induced tipping and has been reported in a variety of systems, from simple ordinary differential equations and maps to mathematical models in climate sciences and ecology. In most examples, the transition happens at a critical rate of parameter change, a rate-induced tipping point, and is associated with a simple unstable orbit (edge state). In this work, we show how this simple picture changes when non-attracting fractal sets exist in the autonomous system, a ubiquitous situation in non-linear dynamics. We show that these fractals in phase space induce fractals in parameter space, which control the rates and parameter changes that result in tipping. We explain how such rate-induced fractals appear and how the fractal dimensions of the different sets are related to each other. We illustrate our general theory in three paradigmatic systems: a piecewise linear one-dimensional map, the two-dimensional Hénon map, and a forced pendulum.

We present a rapid, large-field bimodal imaging platform that integrates conventional brightfield microscopy with a lensless IR imaging scanner, enabling whole-slide IR image stack acquisition in minutes. Using a dedicated deep learning model, we implement an optical HE staining strategy based on subcellular morpho-spectral fingerprinting.

The growing body of atmospheric observations of exoplanets from space and ground-based facilities showcases how the great diversity of the planetary population is not limited to their physical properties but extends to their compositions. The ESA space mission Ariel will observe and characterise hundreds of exoplanetary atmospheres to explore and understand the roots of this compositional diversity. To lay the foundations for the Ariel mission, the OPAL Key Science Project is tasked with creating an unprecedented library of realistic synthetic atmospheres spanning tens of elements and hundreds of molecules on which the Ariel consortium will test and validate its codes and pipelines ahead of launch. In this work we describe the aims and the pipeline of codes of the OPAL project, as well as the process through which we trace the genetic link connecting planets to their native protoplanetary disks and host stars. We present the early results of this complex and unprecedented endeavour and discuss how they highlight the great diversity of outcomes that emerge from the large degeneracy in the parameter space of possible initial conditions to the planet formation process. This, in turn, illustrates the growing importance of interdisciplinary modelling studies supported by high-performance computing methods and infrastructures to properly investigate this class of high-dimensionality problems.

Accurate and fast calculations of localized surface plasmon resonances (LSPR) in metallic nanoparticles is essential for applications in sensing, nano-optics, and energy harvesting. Although full-wave numerical techniques such as the boundary element method (BEM) or the discrete dipole approximation (DDA) provide high accuracy, their computational cost often hinders rapid parametric studies. Here it is presented an ultrafast method that avoids solving large eigenproblems. Instead, only the dipolar component of the induced surface charge density \((σ_{dipolar})\) is retained through a expansion into Cartesion dipole basis, yielding a compact $3\times3$ geometric formulation that avoids full boundary-integral solves. The spectral response is obtained in a similar way, by projecting the Neumann--Poincaré surface operator onto the dipole subspace and evaluating a Rayleigh quotient, giving geometry-only eigenvalues again without an $N\times N$ eigenproblem. A major advantage of this method is that all geometry-dependent quantities are computed once per nanoparticle, while material dispersion and environmental changes enter only through simple algebraic expressions for the polarizability, enabling rapid evaluation across wavelengths. Retardation effects are incorporated through the modified long-wavelength approximation (MLWA), extending accuracy into the weakly retarded regime. The resulting framework provides a valuable tool for fast modelling and optimization of plasmonic nanoparticles at a significant lesser computational cost than BEM, DDA, and other standard tools.

Forming planetary systems are populated by large numbers of gravitationally interacting planetary bodies, spanning from massive giant planets to small planetesimals akin to present-day asteroids and comets. All these planetary bodies are embedded in the gaseous embrace of their native protoplanetary disks, and their interactions with the disk gas play a central role in shaping their dynamical evolution and the outcomes of planet formation. These factors make realistic planet formation simulations extremely computationally demanding, which in turn means that accurately modeling the formation of planetary systems requires the use of high-performance methods. The planet formation code Mercury-Ar$χ$es was developed to address these challenges and, since its first implementation, has been used in multiple exoplanetary and Solar System studies. Mercury-Ar$χ$es is a parallel n-body code that builds on the widely used Mercury code and is capable of modeling the growth and migration of forming planets, the interactions between planetary bodies and the disk gas, as well as the evolving impact flux of planetesimals on forming planets across the different stages of their formation process. In this work we provide the up-to-date overview of its physical modeling capabilities and the first detailed description of its high-performance implementation based on the OpenMP directive-based parallelism for shared memory environments, to harness the multi-thread and vectorization features of modern processor architectures.

Heat generation and temperature reading at the nanoscale have attracted increasing attention due to their direct relevance in thermal therapeutic approaches. Consequently, huge progress has been made toward the design of dual-function nanoplatforms that integrate heating and thermometry capabilities at the nanoscale. However, in most cases, dual nanoheater nanothermometer platforms rely either on specifically engineered materials or on complex readout schemes, which limits translational potential due to complex implementation procedures. To overcome these challenges, we present a methodology for directly extracting temperature information based on dynamical magnetization measurements of cobalt ferrite magnetic nanoflowers. We demonstrate that these nanocrystals monitor temperature changes through variations in their magnetization cycles measured under alternating magnetic fields. Importantly, this thermometric functionality is preserved after surface functionalization and under chemical variations in the nanoparticle environment. Interestingly, we show that we can simultaneously generate heat and report temperature changes within the same agent. This is thanks to the photothermal conversion of cobalt ferrite nanoparticles subjected to near infrared irradiation and the tight reported relationship between magnetization dynamics and Brownian relaxation. Together, these results establish cobalt ferrite magnetic nanoparticles as a label-free platform for simultaneous heat generation and intrinsic temperature readout, enabling real-time nanoscale thermal control.

Cooperation underlies many natural and artificial systems. While voluntary participation can sustain cooperation without informational assumptions, real interactions are rarely anonymous, leaving the joint effects of participation and reputation insufficiently understood. We propose a reputation-based voluntary Prisoner's Dilemma in which agents incur a monitoring cost to inspect opponents and decide whether to exit an interaction for a fixed incentive to avoid exploitation or to default to cooperation or defection. We show that reputation-conditioned exit generates multiple coexistence pathways that sustain cooperation across population structures. In well-mixed populations, cooperation persists through stable mixed coexistence, whereas in structured populations, exit-incentive-dependent regimes emerge, including local cyclic dominance and persistent oscillations. Together, these results extend voluntary participation frameworks and underscore the role of exit-incentive design in cooperative multi-agent systems.

The deep neural network multigrid solver (DNN-MG) combines a coarse-grid finite element simulation with a deep neural network that corrects the solution on finer grid levels, thereby improving the computational efficiency. In this work, we discuss various design choices for the DNN-MG method and demonstrate significant improvements in accuracy and generalizability when applied to the solution of the instationary Navier-Stokes equations. We investigate the stability of the hybrid simulation and show how the neural networks can be made more robust with the help of replay buffers. By retraining on data derived from the hybrid simulation, the error caused by the neural network over multiple time-steps can be minimized without the need for a differentiable numerical solver. Furthermore, we compare multiple neural network architectures, including recurrent neural networks and Transformers, and study their ability to utilize more information from an increased temporal and spatial receptive field. Transformers allow us to make use of information from cells outside the predicted patch even with unstructured meshes while maintaining the locality of our approach. This can further improve the accuracy of DNN-MG without a significant impact on performance.

Multiphase gas -- ranging from cold molecular clouds ($\lesssim 100\,$K) to hot, diffuse plasma ($\gtrsim 10^6\,$K) is a defining feature of the interstellar, circumgalactic, intracluster, and intergalactic media. Accurately simulating its dynamics is critical to improving our understanding of galaxy formation and evolution, however, due to their multi-scale and multi-physics nature, multiphase systems are highly challenging to model. In this review, we provide a comprehensive overview of numerical simulations of multiphase gas in and around galaxies. We begin by outlining the environments where multiphase gas arises and the physical and computational challenges associated with its modeling. Key quantities that characterize multiphase gas dynamics are discussed, followed by an in-depth look at idealized setups such as turbulent mixing layers, cloud-wind interactions, thermal instability, and turbulent boxes. The review then transitions to less idealized and/or larger-scale simulations, covering radiative supernovae bubbles, tall box simulations, isolated galaxy models including dwarf and Milky Way-mass systems, and cosmological zoom-in simulations, with a particular focus on simulations that enhance resolution in the halo. Throughout, we emphasize the importance of connecting scales, extracting robust diagnostics, and comparing simulations to observations. We conclude by outlining persistent challenges and promising directions for future work in simulating the multiphase Universe.

We investigate the thermodynamics and phase stability of the AlTiVCr lightweight high-entropy alloy using a combination of ab initio electronic structure calculations, a concentration wave analysis, and atomistic Monte Carlo simulations. In alignment both with experimental data and with results obtained using other computational approaches, we predict a $\textrm{B2}$ (CsCl) chemical ordering emerging in this alloy at comparatively high temperatures, which is driven by Al and Ti moving to separate sublattices, while V and Cr express weaker site preferences. The impact of this $\textrm{B2}$ chemical ordering on the electronic transport properties of the alloy is investigated within a Kubo-Greenwood linear response framework and it is found that, counter-intuitively, the alloy's residual resistivity increases as the material transitions from the $\textrm{A2}$ (disordered bcc) phase to our predicted $\textrm{B2}$ (partially) ordered structure. This is understood to result primarily from a reduction in the density of electronic states at the Fermi level induced by the chemical ordering. At low temperatures, our atomistic Monte Carlo simulations then reveal subsequent sublattice orderings, with the ground-state configuration predicted to be a fully-ordered, single-phase structure with vanishing associated residual resistivity. These results give fresh, insight into the atomic-scale structure and consequent physical properties of this well-studied, technologically relevant material.

Nonlinear stochastic motion presents significant challenges for Bayesian particle tracking. To address this challenge, this paper proposes a framework to construct an invertible transformation that maps the nonlinear state-space model (SSM) into a higher-dimensional linear Gaussian SSM. This approach allows the application of standard linear-Gaussian inference techniques while maintaining a connection to the dynamics of the original system. The paper derives the necessary conditions for such transformations using Ito's lemma and variational calculus, and illustrates the method on a bistable cubic motion model, radial Brownian process model, and a logistic model with multiplicative noise. Simulations confirm that the transformed linear systems, when projected back, accurately reconstruct the nonlinear dynamics and, in distinct regimes of stiffness and singularity, yield tracking accuracy competitive with conventional filters, while avoiding their structural instabilities.

Social media platforms facilitate echo chambers through feedback loops between user preferences and recommendation algorithms. While algorithmic homogeneity is well-documented, the distinct evolutionary pathways driven by content-based versus link-based recommendations remain unclear. Using an extended dynamic Bounded Confidence Model (BCM), we show that content-based algorithms--unlike their link-based counterparts--steer social networks toward a segregation-before-polarization (SbP) pathway. Along this trajectory, structural segregation precedes opinion divergence, accelerating individual isolation while delaying but ultimately intensifying collective polarization. Furthermore, we reveal a paradox in information sharing: Reposting increases the number of connections in the network, yet it simultaneously reinforces echo chambers because it amplifies small, latent opinion differences that would otherwise remain inconsequential. These findings suggest that mitigating polarization requires stage-dependent algorithmic interventions, shifting from content-centric to structure-centric strategies as networks evolve.

Jet tagging, identifying the origin of jets produced in particle collisions, is a critical classification task in high-energy physics. Despite the revolutionary impact of deep learning on jet tagging over the past decade, the paradigm has remained unchanged. In particular, jets are classified independently, one at a time. This single-jet approach ignores correlations, overlaps, and wider event context between jets. We introduce PanopTag, a new paradigm for jet tagging that departs from traditional single-jet tagging approaches. Rather than classifying jets independently, PanopTag simultaneously tags all jets by employing an encoder-decoder architecture that uses jet kinematics as queries to cross-attend to particle flow object embeddings. We evaluate PanopTag on heavy-flavor $(b/c)$-tagging and demonstrate remarkable performance improvements over state-of-the-art single-jet baselines that are only accessible by exploiting event-level features and correlations between jets.

Learning the full family of solutions to parameterized partial differential equations (PDEs) is a central challenge to our ability to model the behavior of heterogeneous systems, with a variety of fundamental and application-oriented implications in fields such as hydrogeology where system properties exhibit significant (and often uncertain) spatial heterogeneity. We address this by reformulating a Physics-Informed Neural Network (PINN) as a differentiable solver that learns the continuous solution manifold for steady-state Darcy flow. Our framework requires only a single training run, circumventing the need for costly re-training for each new parameter instance. Its versatility is demonstrated through two representations of spatially heterogeneous hydraulic conductivity fields: a direct analytical form and a novel data-driven formulation resting on an autoencoder to create a low-dimensional latent encoding. A key innovation is the integration of the differentiable decoder into the physics-informed loss function, enabling on-the-fly reconstruction of complex conductivity fields via automatic differentiation. The approach yields accurate, mass-conserving flow solutions and supports efficient uncertainty quantification, providing a general methodology for physics-constrained data-driven modeling of heterogeneous systems.

As quantum technologies (QT) move from foundational research toward industrial and societal deployment, national strategies have become critical instruments for shaping the future of this emerging field. In this study, we conduct the first large-scale, data-driven analysis of 62 national quantum strategic documents (QSDs) from 20 countries. Using AI-based natural language processing (topic modeling), we identify 12 topics present in the text, ranging from technical development areas to transversal aspects such as workforce development and governance. Temporal analysis reveals a distinct shift in policy discourse toward applications of QT and commercialisation, and relatively away from basic science. Our findings highlight the increasing diversification of the QT field, and contribute to the growing area of quantum policy studies. We advocate for more AI and data-driven analyses of the quantum ecosystem, to work toward a scalable framework for understanding the technological and societal challenges of the second quantum revolution.

Hybrid light-matter states, known as molecular polaritons, arise from electronic or vibrational strong coupling (ESC and VSC) with confined electromagnetic fields. While these have been widely studied, the influence of electron-nuclear dynamics in driven cavities remains largely unknown. Here, we report a previously unrecognized mechanism of vibrational activation that emerges under collective ESC in driven optical cavities. Using semiclassical simulations that self-consistently combine Maxwell's equations with quantum molecular dynamics, we show that collective electronic Rabi oscillations coherently drive nuclear motion. This effect is captured using both vibrational wave-packet dynamics in a minimal two-level model and atomistic simulations based on time-dependent density-functional tight-binding with Ehrenfest dynamics. Vibrational activation depends non-monotonically on the Rabi frequency and is maximized when the collective polaritonic splitting resonates with a molecular vibrational mode. The mechanism exhibits features consistent with a stimulated Raman-like relaxation mechanism. Our results establish a self-consistent framework for realistic cavity-electron-nuclear dynamics.

A general, compact way of achieving second-order in finite-volume numerical methods is to perform a MUSCL-like, piecewise linear reconstruction of flow properties at each cell interface. To avoid the surge of spurious oscillations in the discrete solution, a limiter function is commonly employed. This strategy, however, can add a series of drawbacks to the overall numerical scheme. The present paper investigates this behavior by considering three different limiter formulations in the context of a second-order, finite volume scheme for the simulation of steady, turbulent flows on unstructured meshes. Three limiter formulations are considered: the original Venkatakrishnan limiter, Wang's modification to the Venkatakrishnan limiter and Nishikawa's recently introduced R3 limiter. Three different configurations of the fully-developed, two-dimensional, transonic NACA 0012 airfoil are analyzed, configured with different angles of attack and similar freestream properties. The gas dynamics are modeled using the Reynolds-averaged Navier-Stokes (RANS) equations, where the negative Spalart-Allmaras turbulence model is used to solve the closure problem. All limiters are shown to yield similar results for all configurations of this case, although with different dissipative characteristics, provided their control constants are used within appropriate intervals. The presented numerical results are in good agreement with experimental data available in the literature.

We introduce a quantitative method to compare arbitrary pairs of graph centrality measures, based on the ordering of vertices induced by them. The proposed method is conceptually simple, mathematically elegant, and allows for a quantitative restatement of many conjectures that were previously cumbersome to formalize. Moreover, it produces an approximation scheme useful for network scientists. We explore some of these uses and formulate new conjectures that are of independent interest.

Contagion processes in social systems often involve interactions that go beyond pairwise contacts. Higher-order networks, represented as hypergraphs, have been widely used to model multi-body interactions, and their presence can drastically alter contagion dynamics compared to traditional network models. However, existing analytical approaches typically assume independence between pairwise and higher-order degrees, and thus study their roles in isolation. In this paper, we develop an effective hyperdegree model (EHDM) to describe Susceptible-Infected-Susceptible (SIS) dynamics on hypergraphs that explicitly captures correlations between the distribution of groups with different sizes. Our effective hyperdegree model shows excellent agreement with stochastic simulations across different types of higher-order networks, including those with heterogeneous degree distributions. We explore the critical role of cross-order degree correlations, specifically, whether nodes that are hubs in pairwise interactions also serve as hubs in higher-order interactions. We show that positive correlation decreases the epidemic threshold and anti-correlation temporally desynchronizes infection pathways (pairwise and group interactions). Finally, we demonstrate that, depending on the level of correlation, the optimal control strategy shifts -- from one that is purely pairwise- or higher-order-focused to one in which a mixed strategy becomes optimal.

Following limb amputation and targeted muscle reinnervation (TMR), nerves supplying agonist and antagonist muscles are rerouted into separate targeted muscles, disrupting natural neuromechanical coupling between muscle groups. Using high-density intramuscular microelectrode arrays in reinnervated muscles, we show that neural signals for agonist and antagonist tasks remain functionally coupled: motor units active during agonist tasks were also recruited during corresponding antagonist tasks, despite no visual feedback on coactivation being provided.

Long-term evolutionary processes can strongly influence common-pool resource conservation by generating new traits or behaviours that modify the feedback between population strategies and the resource state. Here we develop an eco-evolutionary framework in which individuals repeatedly interact with the same opponent and follow direct reciprocity through reactive strategies. The strategic dynamics is coupled to a renewable common resource and analyzed using adaptive dynamics. After our exhaustive non-linear dynamical analysis of $2\times2$ strategic games, we focus on comparative and combined usefulness of institutional incentives in the form of rewards and punishments in preventing the Tragedy of the Commons even when defection dominates in the replete resource state. We also report possibility of robust stable oscillations -- emerging via Hopf bifurcation -- in resource state and population strategies.

Digital traces of daily activities, such as e-commerce (EC) purchase histories, provide scalable signals for public health surveillance, yet their epidemiological validity remains unclear. This study validates a behavioral proxy for disease onset, defined as transitions from regular to therapeutic diets, by comparing large-scale EC data (N=55,645) against independent insurance-derived clinical records. Using feline lower urinary tract disease (FLUTD) as a case study, the proxy showed strong agreement with clinical data for ingredient-level risk patterns (r=0.74) and seasonal dynamics (r=0.82). Furthermore, analysis using EC data alone reproduced the established protective association of wet food consumption. These results demonstrate that validated behavioral signals from EC data can serve as cost-effective complements to traditional surveillance, with potential applicability to monitoring lifestyle-related diseases in human populations.

Quantum combs are powerful conceptual tools for capturing multi-time processes in quantum information theory, constituting the most general quantum mechanical process. But, despite their causal nature, they lack a meaningful physical connection to time -- and are, by and large, arguably incompatible with it without extra structure. The subclass of quantum combs which assumes an underlying process is described by the so-called process tensor framework, which has been successfully used to study and characterise non-Markovian open quantum systems. But, although process tensors are motivated by an underlying dynamics, it is not a priori clear how to connect to a continuous process tensor object mathematically -- leaving an uncomfortable conceptual gap. In this work, we take a decisive step toward remedying this situation. We introduce a fully continuous process tensor framework by showing how the discrete multi-partite Choi state becomes a field-theoretic state in bosonic Fock space, which is intrinsically and rigorously defined in the continuum. With this equipped, we lay out the core structural elements of this framework and its properties. This translation allows for an information-theoretic treatment of multi-time correlations in the continuum via the analysis of their continuous matrix product state representatives. Our work closes a gap in the quantum information literature, and opens up the opportunity for the application of many-body physics insights to our understanding of quantum stochastic processes in the continuum.

Building on our previous work [1], where the Local Sum Rules (LSR) were established, we investigate the construction of compact objects in Randall-Sundrum braneworlds supported by matter fields that are dynamically consistent and localizable. We begin by revisiting the Chamblin et al. black string, highlighting its role as a foundational higher-dimensional solution. We then show that the Ellis-Bronnikov wormhole can be consistently embedded in this framework via a localized free scalar field, providing a simple yet nontrivial example of a braneworld compact object. Finally, we derive two novel black string solutions sourced by a localized nonlinear electrodynamics (NED) theory with Lagrangian $\mathcal{L}(\mathcal{F}) = -β\sqrt{\mathcal{F}}$, corresponding to purely magnetic and dyonic configurations. The purely magnetic solution reproduces the classical Letelier string cloud on the brane, while the dyonic solution generalizes it to include electric charge, closely paralleling the Letelier-Alencar construction. Both NED solutions reduce smoothly to the Chamblin et al. black string in the limit $β\to 0$, illustrating how localized higher-dimensional matter fields can consistently support braneworld compact objects and connect higher-dimensional physics with well-known four-dimensional solutions.

Quantum entanglement distributed via satellites enable global-scale quantum communication. However, onboard sources are susceptible to misalignment due to dynamical orbital conditions. Here, we present two recalibration techniques for efficient generation of high quality entanglement using a periodically poled lithium niobate (PPLN)-based spontaneous parametric down-conversion (SPDC) source with minimum intervention. The first is a heuristic algorithm (HA) which mimics the manual alignment process in a laboratory. The second is based on reinforcement learning (RL). Our simulation demonstrates superior performance of RL with AUC=0.9119 compared to HA's 0.7042 in the modified ROC analysis (60 min threshold). RL achieves perfect alignment in 10 min as opposed to HA's 30 min. Both the methods operate within feasible satellite constraints, offering scalable automation for complex quantum communication scenarios.

We introduce a method for engineering discrete local dynamics in globally-driven dual-species neutral atom experiments, allowing us to study emergent digital models through uniform analog controls. Leveraging the new opportunities offered by dual-species systems, such as species-alternated driving, our construction exploits simple Floquet protocols on static atom arrangements, and benefits of generalized blockade regimes (different inter- and intra-species interactions). We focus on discrete dynamical models that are special examples of Quantum Cellular Automata (QCA), and explicitly consider a number of relevant examples, including the kicked-Ising model, the Floquet Kitaev honeycomb model, and the digitization of generic translation-invariant nearest-neighbor Hamiltonians (e.g., for Trotterized evolution). As an application, we study chaotic features of discretized many-body dynamics that can be detected by leveraging only demonstrated capabilities of globally-driven experiments, and benchmark their ability to discriminate chaotic evolution.

We explore dark matter like fluids in a spherically symmetric Lemaitre Tolman Bondi (LTB) minisuperspace, tracking how symmetry properties of the Hamiltonian constraint control the emergence of effective dark sources in General Relativity (GR) and Horava Lifshitz (HL) gravity. We first deform the GR Hamiltonian by adding an extra weight $+1$ density to the potential. We show that potential deformations of this type leave the (reduced) Dirac algebra unchanged and the modification is naturally reinterpreted as an effective anisotropic stress energy contribution. While the fluid reproduces an isothermal-like mass scaling, its pressure anisotropy prevents it from giving flat rotation curves. We then turn to HL gravity, where the deformed Dirac algebra induces a controlled nonconservation law for an emergent dust component. Generalizing earlier results, we identify a restricted class of LTB backgrounds for which the HL source term yields a positive scaling dark matter density, consistent with ghost-freedom, and recovery of GR in the infrared. The analysis is conditional on a prescribed background: obtaining a fully backreacted areal radius solution consistent with the HL field equations is left as a natural direction for future work.

In this work, we experimentally investigate the classical-light emulation of different notions of nonclassicality in the simplest scenario. We implement this prepare-and-measure scenario involving four preparations and two binary-outcome measurements using two distinct experimental setups that exploit different degrees of freedom of light: polarization and first-order Hermite-Gaussian transverse modes. We additionally model experimental noise through an all-optical setup that reproduces the operational effect of a depolarizing channel. Our experimental results are consistent with the findings of Khoshbin et al. [Phys. Rev. A 109, 032212 (2024)]: under the assumption that the two measurements performed form a tomographically complete set, the observed statistics violate their noise-robust inequalities, indicating inconsistencies with preparation noncontextuality and bounded ontological distinctness for preparations. Although our implementation uses classical light, it reproduces the statistics predicted for the simplest scenario. Since the states and measurements of this scenario underpin computational advantages in tasks such as two-bit quantum random access codes -- among the simplest communication primitives enabling semi-device-independent certification of nonclassicality -- our implementation is directly relevant for such applications.

We investigate the boundary critical phenomena of the one-dimensional quantum Ashkin-Teller model using boundary conformal field theory and density matrix renormalization group (DMRG) simulations. Based on the $\mathbb{Z}_2$-orbifold of the $c=1$ compactified boson boundary conformal field theory, we construct microscopic lattice boundary terms that renormalize to the stable conformal boundary conditions,, utilizing simple current extensions and the underlying $\mathrm{SU}(2)$ symmetry to explicitly characterize the four-state Potts point. We validate these theoretical identifications via finite-size spectroscopy of the lattice energy spectra, confirming their consistency with $D_4$ symmetry and Kramers-Wannier duality. Finally, we discuss the boundary renormalization group flows among these identified fixed points to propose a global phase diagram for the boundary criticality.

We theoretically compare the quantum Fisher information (QFI) for three configurations of absorption spectroscopy with undetected idler photons: an SU(1,1) interferometer with inter-source idler loss, an induced-coherence (IC) setup in which the idler partially seeds a second squeezer together with a vacuum ancilla, and a distributed-loss (DL) scheme with in-medium attenuation. We calculate the QFI as a function of parametric gain for both full and signal-only detection access. For losses below 99% and low to moderate gain, the SU(1,1) configuration provides the largest QFI. At high gain and intermediate loss, the IC scheme performs best, while under extreme attenuation (transmission $<$ 1%) the DL model becomes optimal. These results delineate the measurement regimes in which each architecture is optimal in terms of information theory.

We present analytical results toward the Wigner entropy conjecture, which posits that among all physical Wigner non-negative states the Wigner entropy is minimized by pure Gaussian states for which it attains the value $1+\lnπ$.Working under a minimal set of constraints on the Wigner function, namely, non-negativity, normalization, and the pointwise bound $πW\le 1$, we construct an explicit hierarchy of lower bounds $B_n$ on $S[W]$ by combining a truncated series lower bound for $-\ln x$ with moment identities of the Wigner function.This yields closed-form purity-based sufficient conditions ensuring $S[W]\ge 1+\lnπ$.In particular, we first prove that all Wigner non-negative states with $μ\le 4-2\sqrt3$ satisfy the Wigner entropy conjecture. We further obtain a systematic purity-only relaxation of the hierarchy, yielding the simple sufficient condition $μ\le 2/e$. On top of aforesaid results, our analysis clarifies why additional physicality constraints are necessary for purity-based approaches that aim to approach the extremal case $μ\leq1$.

Many applications require or benefit from being able to securely localize remote parties. In classical physics, adversaries can in principle have complete knowledge of such a party's devices, and secure localization is fundamentally impossible. This limitation can be overcome with quantum technologies, but proposals to date require trusting vulnerable hardware. Here we develop and experimentally demonstrate a protocol for device-independent quantum position verification that guarantees security with only observed correlations from a loophole-free Bell test across a quantum network. The protocol certifies the position of a remote party against adversaries who, before each instance of the test, are weakly entangled, but otherwise have unlimited quantum computation and communication capabilities. Our demonstration achieves a one-dimensional localization that is 2.47(2) times smaller than the best, necessarily non-remote, classical localization protocol. Compared to such a classical protocol having identical latencies, the localization is 4.53(5) times smaller. This work anchors digital security in the physical world.

Supergravity theories in de Sitter spacetime are known to be very constrained, and rather unnatural within String/M Theory. We revisit the seminal paper by Pilch, van Nieuwenhuizen and Sohnius, where the possible existence of a real Lagrangian for ${\cal N}=2$ pure supergravity in four-dimensional de Sitter spacetime was pointed out. We clarify several issues related to the non-unitarity of the theory and explicitly construct the unique, complete theory searched for long ago by the aforementioned authors. We argue that the lack of unitarity of the Lorentzian theory may be revisited in the Euclidean approach to de Sitter quantum gravity, where alternative definitions of unitarity can be introduced.

We derive general relativistic Gaussian equations for osculating elements for orbits under the influence of a perturbing force without any restrictions in an underlying Schwarzschild space-time. Such a formulation provides a way to describe the evolution of orbital parameters in strong gravity relativistic settings. As examples of external forces we considered Kerr and $q$-metric space-times generated forces, for which we solve equations for osculating elements in linear approximation. For the Kerr space-time in the post-Newtonian limit, our result reproduces the well-known Lense--Thirring precession of the longitude of the ascending node.

Control over the internal states of trapped ions makes them the ideal system to generate single and two-photon states. Coupling a single ion to an optical cavity enables efficient emission of single photons into a single spatial mode and grants control over their temporal shape, phase and frequency. Using the long coherence time of the ion's internal states and employing a scheme to protect the coherence of the ion-cavity interaction, we demonstrate the generation of a two-photon entangled state with full control over the phase. Initially, ion-photon entanglement is generated. A second photon is subsequently generated, mapping the ion's state onto the second photon. By adjusting the drive field the phase of the entangled state can be fully controlled. We implement this scheme in the most resource efficient way by utilizing a single $^{40}$Ca$^+$ ion coupled to an optical cavity and demonstrate the generation of a two-photon entangled stated with full phase control with a fidelity of up to 82\%.

The cosmological tensions present in the $Λ$ cold dark matter model that have emerged and strengthened over recent years motivate model independent approaches to analysing data. Cosmography is useful for interpreting data in cosmology without imposing assumptions about the field equations of gravity or the matter content in the Universe. Some cosmography methods, denoted covariant cosmography, go even further and stay agnostic to the underlying space-time metric. Due to their high level of generality, covariant cosmography methods can incorporate the anisotropies and inhomogeneities in the observer's vicinity, and may in turn inform about the associated curvature of the relevant structures in our cosmic neighbourhood. Thus, covariant cosmography is a powerful model-independent tool for analysing cosmological data while also enabling the mapping of our local cosmic neighbourhood. In order to be able to explore the covariant cosmography framework to its fullest, it must be tested in tractable models and simulations. In this paper we derive the cosmography of luminosity distance to fourth order in redshift and investigate it in the special case of axially symmetric Szekeres models. We compare the numerical results for the distance-redshift relations of synthetic observers placed within the Szekeres structures with the predictions from the cosmography, and comment on the found level of approximation of the cosmography in relation to other results in the litterature.

Sharing genuine multipartite entanglement by considering collective use of copies of biseparable states, which are entangled across all bipartitions but lack genuine multipartite entanglement at the single-copy level, plays a central role in several quantum information processing protocols, and has been referred as genuine multipartite entanglement activation. We present a protocol for three-qutrit systems showing that two copies of rank-two biseparable states, entangled across every bipartition, are sufficient to generate a genuinely multipartite entangled state with nonzero probability. This contrasts with the three-qubit scenario where many copies of biseparable states might be required for sharing genuine multipartite entanglement. We subsequently generalize our protocols to the case of an arbitrary number of parties. Our protocol does not rely on the implementation of joint measurements on the copies of states. Interestingly, the proposed construction naturally leads to the activation of genuinely nonlocal correlations, yielding a result that is stronger than genuine multipartite entanglement activation alone.

Developing high-performance materials is critical for diverse energy applications to increase efficiency, improve sustainability and reduce costs. Classical computational methods have enabled important breakthroughs in energy materials development, but they face scaling and time-complexity limitations, particularly for high-dimensional or strongly correlated material systems. Quantum computing (QC) promises to offer a paradigm shift by exploiting quantum bits with their superposition and entanglement to address challenging problems intractable for classical approaches. This perspective discusses the opportunities in leveraging QC to advance energy materials research and the challenges QC faces in solving complex and high-dimensional problems. We present cases on how QC, when combined with classical computing methods, can be used for the design and simulation of practical energy materials. We also outline the outlook for error-corrected, fault-tolerant QC capable of achieving predictive accuracy and quantum advantage for complex material systems.

Wegner duality is essential for Z2 lattice gauge theory, yet the duality on non-trivial topologies has remained implicit. We extend Wegner duality to arbitrary topology and dimension, obtaining a new class of Ising models, in which topology is encoded in non-local domain-wall patterns. Without the overhead of gauge constraints, simulating this model on an L*L torus requires only L*L qubits with two-body couplings, halving the conventional four-body coupled 2L*L qubits, enabling full experimental realization of Z2 lattice gauge theory on near-term devices.

Variational Quantum Algorithms (VQAs) are a class of hybrid quantum-classical algorithms that leverage on classical optimization tools to find the optimal parameters for a parameterized quantum circuit. One relevant application of VQAs is the Variational Quantum Eigensolver (VQE), which aims at steering the output of the quantum circuit to the ground state of a certain Hamiltonian. Recent works have provided global convergence guarantees for VQEs under suitable local surjectivity and smoothness hypotheses, but little has been done in characterizing convergence of these algorithms when the underlying quantum circuit is affected by noise. In this work, we characterize the effect of different coherent and incoherent noise processes on the optimal parameters and the optimal cost of the VQE, and we study their influence on the convergence guarantees of the algorithm. Our work provides novel theoretical insight into the behavior of parameterized quantum circuits. Furthermore, we accompany our results with numerical simulations implemented via Pennylane.

Extending the high-harmonic cutoff with experimentally accessible fields is essential for advancing tabletop coherent extreme ultraviolet (EUV) and soft X-ray sources. Although terahertz (THz) assistance offers a promising route, cutoff extension at weak, laboratory-accessible THz strengths remain poorly understood. In this report, we comprehensively investigate THz-assisted high-order harmonic generation (HHG) using time-dependent Schrödinger equation simulations supported by classical trajectory analysis and Bohmian-based quantum dynamics. By mapping the plateau evolution versus THz strength, we show that even weak THz fields can extend the cutoff, producing a pronounced ``fish-fin'' structure whose prominent rays saturate near $I_p + 8 U_p$. We trace this extension to long electron excursions spanning several optical cycles before recombination, and provide a fully consistent explanation using both classical analysis and Bohmian trajectories flow. Our findings reveal that this cutoff-extension mechanism is remarkably robust, persisting across different atomic species and remaining insensitive to variations in the driving parameters. These results demonstrate that cutoff control is achievable with laboratory-scale THz fields, offering practical guidelines for engineering coherent high-energy HHG, and providing a robust pathway for tracking ultrafast electron motion in real time.

We introduce SeeMPS, a Python library dedicated to implementing tensor network algorithms based on the well-known Matrix Product States (MPS) and Quantized Tensor Train (QTT) formalisms. SeeMPS is implemented as a complete finite precision linear algebra package where exponentially large vector spaces are compressed using the MPS/TT formalism. It enables both low-level operations, such as vector addition, linear transformations, and Hadamard products, as well as high-level algorithms, including the approximation of linear equations, eigenvalue computations, and exponentially efficient Fourier transforms. This library can be used for traditional quantum many-body physics applications and also for quantum-inspired numerical analysis problems, such as solving PDEs, interpolating and integrating multidimensional functions, sampling multivariate probability distributions, etc.

We investigate the cosmic censorship conjecture in a spherically symmetric collapse with shear and bulk viscosity, heat flux, and pressure anisotropy, imposing physically reasonable energy conditions. Using the semi-tetrad covariant formalism, we derive the dynamics of the collapsing fluid, including a master equation for the evolution of the Weyl curvature, to examine the role of viscosity. The analysis of null geodesic geometry uncovers a novel curvature-balance mechanism between Ricci (matter) and Weyl (free gravitational field) curvature on the apparent horizon; this balance determines the causal nature of the horizon and thereby governs the visibility of the singularity. We then derive necessary and sufficient covariant conditions for the central singularity to be locally naked. Our findings support a weaker form of cosmic censorship and extend the covariant censorship analysis to realistic dissipative, viscous collapse.

We develop a general canonical quantization scheme for $k$-essence cosmology in scalar-tensor theory. Utilizing the Dirac-Bergmann algorithm, we construct the Hamiltonian associated with the cosmological field equations and identify the first- and second-class constraints. The introduction of appropriate canonically conjugate variables with respect to Dirac brackets, allows for the canonical quantization of the model. In these new variables, the Hamiltonian constraint reduces to a quadratic function with no potential term. Its quantum realization leads to a Wheeler-DeWitt equation reminiscent of the massless Klein-Gordon case. As an illustrative example, we consider the action of a tachyonic field and investigate the conditions under which a phantom crossing can occur as a quantum tunneling effect. For the simplified constant potential case, we investigate the consequences of different boundary conditions on the singularity avoidance and to the mean expansion rate.

So far, entanglement harvesting has been extensively studied in free space setups. Here, we provide a detailed analytical and numerical analysis of entanglement harvesting in cavities. Specifically, we adiabatically couple the quantized electromagnetic field to two identical Gaussian detectors located on the symmetry axis of a cylindrical cavity. Our numerical investigations reveal a strong dependence on the cavity length, while showing invariance under changes in the cavity radius in regimes of maximal entanglement. Moreover, we identify different scalings of the detector system parameters for entanglement inside and outside the light cone. Finally, we uncover a strong dependence of the harvested correlations on the cavity induced parity of the electromagnetic field.

Quantum linear system (QLS) solvers are a fundamental class of quantum algorithms used in many potential quantum computing applications, including machine learning and solving differential equations. The performance of quantum algorithms is often measured by their query complexity, which quantifies the number of oracle calls required to access the input. The main parameters determining the complexity of QLS solvers are the condition number $κ$ and sparsity $s$ of the linear system, and the target error $ε$. To date, the best known query-complexity lower bound is $Ω(κ\log(1/ε))$, which establishes the optimality of the most recent QLS solvers. The original proof of this lower bound is attributed to Harrow and Kothari, but their result is unpublished. Furthermore, when discussing a more general lower bound including the sparsity $s$ of the linear system, it has become folklore that it should read as $Ω( κ\sqrt{s}\log(1/ε))$. In this work, we establish the rigorous lower bound capturing the sparsity dependence of QLS. We prove the lower bound of $Ω(κ\sqrt{s})$ for any quantum algorithm that solves QLS with constant error. While the dependence on all parameters $κ,s,ε$ remains an open problem, our result provides a crucial stepping stone toward the complete characterization of QLS complexity.

While quantum computers are a very promising tool for the far future, in their current state of the art they remain limited both in size and quality. This has given rise to hybrid quantum-classical algorithms, where the quantum device performs only a small but vital part of the overall computation. Among these, variational quantum algorithms (VQAs), which combine a classical optimization procedure with quantum evaluation of a cost function, have emerged as particularly promising. However, barren plateaus and ill-conditioned optimization landscapes remain among the primary obstacles faced by VQAs, often leading to unstable convergence and high sensitivity to initialization. Motivated by this challenge, we investigate whether a purely classical remedy, standard L2 squared-norm regularization, can systematically stabilize hybrid quantum-classical optimization. Specifically, we augment the Variational Quantum Eigensolver (VQE) objective with a quadratic penalty proportional to the squared norm of the parameters, without modifying the quantum circuit or measurement process. Across all tested Hamiltonians, H2, LiH, and the Random Field Ising Model (RFIM), we observe improved performance over a broad window of the regularization strength. Our large-scale numerical results demonstrate that classical regularization provides a robust, system-independent mechanism for mitigating VQE instability, enhancing the reliability and reproducibility of variational quantum optimization without altering the underlying quantum circuit.

We study a minimal model for charging a quantum battery consisting of a two-level system (TLS) acting as a charger, coupled to a harmonic oscillator that serves as the quantum battery. A single-photon quantum pulse of light excites the TLS, which subsequently transfers its excitation to the isolated battery. The TLS may also decay into the electromagnetic environment. We obtain analytical solutions for the dynamics of the battery and determine the optimal pulse shape that maximizes the stored energy. The optimal pulse saturates a universal bound for the stored energy, determined by the TLS decay rates into the pulse and the environment. Furthermore, we derive the minimum charging time and establish a quantum speed limit at the exceptional point, where a critical transition occurs in the system's dynamics. We also present analytical expressions for the charging power and investigate the pulse duration that maximizes it.

Quantum neural networks (QNNs) provide expressive probabilistic models by leveraging quantum superposition and entanglement, yet their practical training remains challenging due to highly oscillatory loss landscapes and noise inherent to near-term quantum devices. Existing training approaches largely rely on gradient-based procedural optimization, which often suffers from slow convergence, sensitivity to hyperparameters, and instability near sharp minima. In this work, we propose an alternative inverse-probability algebraic learning framework for QNNs. Instead of updating parameters through incremental gradient descent, our method treats learning as a local inverse problem in probability space, directly mapping discrepancies between predicted and target Born-rule probabilities to parameter corrections via a pseudo-inverse of the Jacobian. This algebraic update is covariant, does not require learning-rate tuning, and enables rapid movement toward the vicinity of a loss minimum in a single step. We systematically compare the proposed method with gradient descent and Adam optimization in both regression and classification tasks using a teacher-student QNN benchmark. Our results show that algebraic learning converges significantly faster, escapes loss plateaus, and achieves lower final errors. Under finite-shot sampling, the method exhibits near-optimal error scaling, while remaining robust against intrinsic hardware noise such as dephasing. These findings suggest that inverse-probability algebraic learning offers a principled and practical alternative to procedural optimization for QNN training, particularly in resource-constrained near-term quantum devices.

We develop and numerically analyze an algebraic-geometric framework for spin-adapted coupled-cluster (CC) theory. Since the electronic Hamiltonian is SU(2)-invariant, physically relevant quantum states lie in the spin singlet sector. We give an explicit description of the SU(2)-invariant (spin singlet) many-body space by identifying it with an Artinian commutative ring, called the excitation ring, whose dimension is governed by a Narayana number. We define spin-adapted truncation varieties via embeddings of graded subspaces of this ring, and we identify the CCS truncation variety with the Veronese square of the Grassmannian. Compared to the spin-generalized formulation, this approach yields a substantial reduction in dimension and degree, with direct computational consequences. In particular, the CC degree of the truncation variety -- governing the number of homotopy paths required to compute all CC solutions -- is reduced by orders of magnitude. We present scaling studies demonstrating asymptotic improvements and we exploit this reduction to compute the full solution landscape of spin-adapted CC equations for water and lithium hydride.

We investigate the gravitational lensing effects of a static spherically symmetric black hole (BH) within the framework of the revised Deser-Woodard (D-W) nonlocal gravity. By analyzing the deflection angle in both the weak and strong field limits, we derive several distinguishing features of the model. In the weak field limit, we report a leading-order correction to the deflection angle directly attributed to the non-local nature of the theory. In the strong field limit, we find that the lensing corrections are almost linearly dependent on the coupling parameter $ζ$ while being exponentially suppressed by the exponent parameter $n$. Furthermore, the gravitational lensing effect in the revised D-W model at a given time shares similar scale-invariant behavior to General Relativity and conformal gravity, offering a potential pathway to distinguish it from other alternatives using astronomical observations.

The accurate characterization of quantum systems is essential for the advancement of quantum technologies. In particular, certifying convex functions of quantum states plays a central role in many applications. We present a certification method for experimentally prepared quantum states that accounts for both shot noise and measurement imperfections in the data-acquisition stage. Building upon previous work, our method extends confidence regions to accommodate imperfect control over measurements. The values of the functions can then be bounded using convex optimization techniques. We provide explicit prescriptions for quantifying the noise contribution from finite statistics and for estimating the effect of measurement imperfections. By jointly incorporating statistical and systematic errors, the method yields a robust certification framework for quantum experiments.

Gravitational waves (GWs) in the $10^{-3}-0.1$ Hz band encode unique signatures of the early universe and merging compact objects, but they are beyond the reach of existing observatories. Theoretical models suggest that the Moon could act as a resonant detector, but the unknown influence of its rugged surface and heterogeneous interior has cast doubt on this prospect. Here, we resolve this long-standing uncertainty by constructing the first high-resolution, structurally realistic model of the lunar GW response. We achieve this by combining high-fidelity spectral-element simulations with the analytical power of normal-mode perturbation theory, thereby resolving topographical effects down to $3.7$ km grid spacing while maintaining the capacity to discern global free-oscillation patterns. This dual-methodology approach not only recovers the expected predominant quadrupole ($l=2$) oscillation mode, but also exposes a systematic signal amplification of $(10-20)\%$ in thick-crust regions. This enhancement is traced by our normal-mode analysis to a mode-coupling process, in which the original quadrupolar oscillation induced by the passing GWs distributes energy into a series of higher-order modes, the hybridized eigenmodes of the laterally heterogeneous Moon. Near certain eigen-frequencies and at specific locations, we observe up to tenfold amplification, highlighting the power of numerical simulations in resolving these structurally fine-tuned features. Our work establishes the Moon as an accurately calibrated resonant GW detector, and the resulting amplification maps provide quantitative guide for the optimal landing site selection.

Neutral atoms for quantum computing applications show promise in terms of scalability and connectivity. We demonstrate the realization of a versatile apparatus capable of stochastically loading a 5x5 array of optical tweezers with single $^{88}$Sr atoms featuring flexible magnetic field control and excellent optical access. A custom-designed oven, spin-flip Zeeman slower, and deflection stage produce a controlled flux of Sr directed to the science chamber. In the science chamber, featuring a vacuum pressure of $3 \times 10^{-11}$ mbar, the Sr is cooled using two laser cooling stages, resulting in $\sim 3 \times 10^5$ atoms at a temperature of 5(1) $μ$K. The optical tweezers feature a $1/e^2$ waist of 0.81(2) $μ$m, and loaded atoms can be imaged with a fidelity of $\sim 0.997$ and a survival probability of $0.99^{+0.01}_{-0.02}$. The atomic array presented here forms the core of a full-stack quantum computing processor targeted for quantum chemistry computational problems.

We demonstrate how resonances in a quantum graph consisting of a compact core and semi-infinite leads can be identified from the eigenvalue behavior of the cut-off system.

Towards the scalable realization of a quantum computer, a quantum charge-coupled device (QCCD) based on ion shuttling has been considered a promising approach. However, the processes of detaching an ion from an array, reintegrating it, and driving non-uniform motion introduce severe heating, requiring significant time and laser power for re-cooling and stabilization. To mitigate these challenges, we propose a novel entangling scheme between a stationary ion qubit and a continuously transported mobile ion, which remains in uniform motion and minimizes motional heating. We theoretically demonstrate a gate error on the order of 0.01%, within reach of current technology. This approach enables resource-efficient quantum operations and facilitates long-distance entanglement distribution, where stationary trapped-ion arrays serve as memory units and mobile ions act as communication qubits passing beside them. Our results pave the way for an alternative trapped-ion architecture beyond the QCCD paradigm.

Quantum sensing based on two-photon interferometry manifests quantum superiority beyond the classical precision limit. However, this superiority is usually diminished inevitably by the noise. Here, we analyze the sensitivity of two typical two-photon interferometries to the noise, that is, Hong-Ou-Mandel (HOM) and N00N state interferometry. It is found that HOM (N00N state) interference, which depends on the biphoton frequency difference (sum), is insensitive (sensitive) to the phase noise in both the manners of spectrally non-resolved and resolved detections in practice, suggesting their potential applications of sensing for different noise scenarios. Furthermore, spectrally resolved detection outperforms spectrally non-resolved one for the two interferometries, especially for the scope that exceeds the coherence time of biphotons. The findings provide an optimal strategy for the practical applications of two-photon interferometric sensing in diverse noise environments.

Indefinite causal order (ICO) has been studied via higher-order quantum processes (e.g.\ the quantum switch), process matrices, and quantum-gravity proposals involving superposed causal structure, yet the meaning of ``indefiniteness'' and its relation to definite-order explanations often remain opaque. Part~I develops a category-theoretic formulation of definite-order explainability as a gluing problem: each definite causal ordering (a partial order/DAG type) is treated as a context, and causal separability amounts to a consistent global section (possibly after convex mixing), whereas causal nonseparability is a failure-to-glue. We also introduce a compact seven-valued contextual classifier -- an intuitionistic elaboration -- that separates variation across contexts from genuine indeterminacy. Part~II applies this framework to a quantum-gravity motivated setting where the fundamental time is a parametric ordering variable $τ$, distinct from geometric (spacetime) time. Adopting a stochastic-quantization perspective on spin-network dynamics (Hilbert space not assumed fundamental) and reading the Wheeler--DeWitt condition as an equilibrium/stationarity constraint, we interpret ICO as indeterminacy of the parametric order of coarse-grained relational interventions, even when the microscopic update process is globally ordered by $τ$. Together, the two parts provide a common language for comparing ICO criteria and for stating precisely what ``no hidden definite order'' means.

Quantum technologies have surpassed classical systems by leveraging the unique properties of superposition and entanglement in photons and matter. Recent advancements in integrated quantum photonics, especially in silicon-based and lithium niobate platforms, are pushing the technology toward greater scalability and functionality. Silicon circuits have progressed from centimeter-scale, dual-photon systems to millimeter-scale, high-density devices that integrate thousands of components, enabling sophisticated programmable manipulation of multi-photon states. Meanwhile, lithium niobate, thanks to its wide optical transmission window, outstanding nonlinear and electro-optic coefficients, and chemical stability, has emerged as an optimal substrate for fully integrated photonic quantum chips. Devices made from this material exhibit high efficiency in in generating, manipulating, converting, storing, and detecting photon states, thereby establishing a basis for deterministic multi-photon generation and single-photon quantum interactions, as well as comprehensive frequency-state control. This review explores the development of integrated photonic quantum technologies based on both silicon and lithium niobate, highlighting invaluable insights gained from silicon-based systems that can assist the scaling of lithium niobate technologies. It examines the functional integration mechanisms of lithium niobate in electro-optic tuning and nonlinear energy conversion, showcasing its transformative impact throughout the photonic quantum computing process. Looking ahead, we speculate on the developmental pathways for lithium niobate platforms and their potential to revolutionize areas such as quantum communication, complex system simulation, quantum sampling, and optical quantum computing paradigms.

Quantum phase estimation is an important routine in many quantum algorithms, particularly for estimating the ground state energy in quantum chemistry simulations. This estimation involves applying powers of a unitary to the ground state, controlled by an auxiliary state prepared on a control register. In many applications the goal is to provide a confidence interval for the phase estimate, and optimal performance is provided by a discrete prolate spheroidal sequence. We show how to prepare the corresponding state in a far more efficient way than prior work. We find that a matrix product state representation with a bond dimension of 4 is sufficient to give a highly accurate approximation for all dimensions tested, up to $2^{24}$. This matrix product state can be efficiently prepared using a sequence of simple three-qubit operations. When the dimension is a power of 2, the phase estimation can be performed with only three qubits for the control register, making it suitable for early-generation fault-tolerant quantum computers with a limited number of logical qubits.

This note introduces a family of circulant quantum channels -- a subclass of the mixed-permutation channels -- and investigates its key structural and operational properties. We show that the image of the circulant quantum channel is precisely the set of circulant matrices. This characterization facilitates the analysis of arbitrary $n$-th order Bargmann invariants. Furthermore, we prove that the channel is entanglement-breaking, implying a substantially reduced resource cost for erasing quantum correlations compared to a general mixed-permutation channel. Applications of this channel are also discussed, including the derivation of tighter lower bounds for $\ell_p$-norm coherence and a characterization of its action in bipartite systems.

Neuronal electrical activity underlies human cognition, yet its direct, noninvasive measurement in the living human brain remains a fundamental challenge. Existing neuroimaging techniques, including EEG, MEG, and fMRI, are limited by trade-offs in sensitivity and spatial or temporal resolution. Here we propose quantum sensing MRI (qsMRI), a noninvasive approach that enables direct detection of neuronal firing-induced magnetic fields using a clinical MRI system. qsMRI exploits endogenous proton (1H) nuclear spins in water molecules as intrinsic quantum sensors and decodes time-resolved phase information from free induction decay (FID) signals to infer neuronal magnetic fields. We validate qsMRI through simulations, phantom experiments, and human studies at rest and during motor tasks, and provide open experimental procedures to facilitate independent validation. We further present a case study demonstrating potential applications to neurological disorders. qsMRI represents a first-in-human application of quantum sensing on a clinical MRI platform, establishes a non-BOLD functional imaging modality, and enables interrogation of neuronal firing dynamics in both cortical and deep brain regions.

Airbridges are essential for creating high-performance, low-parasitic interconnects in integrated circuits and quantum devices. Conventional multi-step fabrication methods hinder miniaturization and introduce process-related defects. We report a simplified process for fabricating nanoscale airbridges using only a single electron-beam lithography step. By optimizing a multilayer resist stack with a triple-exposure-dose scheme and a thermal reflow step, we achieve smooth, suspended metallic bridges with sub-200-nm features that exhibit robust mechanical stability. Fabricated within a gradiometric SQUID design for superconducting transmon qubits, these airbridges introduce no measurable additional loss in the relaxation time $T_1$, while enabling a 2.5-fold enhancement of the dephasing time $T_2^*$. This efficient method offers a practical route toward integrating high-performance three-dimensional interconnects in advanced quantum and nano-electronic devices.

Efficient spectrum sharing in the Citizens Broadband Radio Service (CBRS) band is essential for maximizing 5G network capacity, particularly when high-traffic base stations require simultaneous access to multiple channels. Standard formulations of the Quantum Approximate Optimization Algorithm (QAOA) impose such multi-channel constraints using penalty terms, so most of the explored Hilbert space corresponds to invalid assignments. We propose a subspace-confined QAOA tailored to CBRS multi-channel allocation, in which each node-wise channel register is initialized in a Generalized Dicke state and evolved under an intra-register XY mixer. This ansatz confines the dynamics to a tensor product of Johnson graphs that exactly encode per-node Hamming-weight constraints. For an 8-node CBRS interference graph with 24 qubits, the effective search space is reduced from the full Hilbert space of size $2^{24}$ to 2916 feasible configurations. Within this subspace, the algorithm converges rapidly to low-conflict assignments without large penalty coefficients. Simulations on instances with up to eight nodes show that the proposed ansatz achieves near-optimal conflict levels and consistently outperforms standard penalty-based QAOA and a greedy classical heuristic in terms of feasibility. Noise simulations with depolarizing channels further indicate that the constraint-preserving structure maintains a high feasibility ratio in NISQ-relevant error regimes.

We demonstrate a substantial reduction in two-level system loss in tantalum coplanar waveguide resonators fabricated on high-resistivity silicon substrates through the use of an ultrathin titanium sacrificial layer. A 0.2nm titanium film, deposited atop pre-sputtered α-tantalum, acts as a solid-state oxygen getter that chemically modifies the native Ta oxide at the metal-air interface. After device fabrication, the titanium layer is removed using buffered oxide etchant, leaving behind a chemically reduced Ta oxide surface. Subsequent high-vacuum annealing further suppresses two-level system loss. Resonators treated with this process exhibit internal quality factors Qi exceeding an average of 1.5 million in the single-photon regime across ten devices, over three times higher than otherwise identical devices lacking the titanium layer. These results highlight the critical role of interfacial oxide chemistry in superconducting loss and reinforce atomic-scale surface engineering as an effective approach to improving coherence in tantalum-based quantum circuits. The method is compatible with existing fabrication workflows applicable to tantalum films, offering a practical route to further extending T1 lifetimes in superconducting qubits.

Intrinsic randomness is generated when a quantum state is measured in any basis in which it is not diagonal. In an adversarial scenario, we quantify this randomness by the probability that a correlated eavesdropper could correctly guess the measurement outcomes. What if the eavesdropper is never wrong, but can sometimes return an inconclusive outcome? Inspired by analogous concepts in quantum state discrimination, we introduce the unambiguous randomness of a quantum state and measurement, and, relaxing the assumption of perfect accuracy, randomness with a fixed rate of inconclusive outcomes. We solve these problems for any state and projective measurement in dimension two, as well as for an isotropically noisy state measured in an unbiased basis of any dimension. In the latter case, we find that, given a fixed amount of total noise, an eavesdropper correlated only to the noisy state is always outperformed by an eavesdropper with joint correlations to both a noisy state and a noisy measurement. In fact, we identify a critical error parameter beyond which the joint eavesdropper achieves perfect guessing probability, ruling out any possibility of private randomness.

Neutral atoms have become a competitive platform for quantum metrology, simulation, sensing, and computing. Current magic trapping techniques are insufficient to engineer magic trapping conditions for qudits encoded in hyperfine states with $J \neq 0$, compromising qudit coherence. In this paper we propose a scheme to engineer magic trapping conditions for qudits via bichromatic tweezers. We show it is possible to suppress differential light shifts across all magnetic sublevels of the $5s5p$ $\mathrm{^{3}P_2}$ state by using two carefully chosen wavelengths (with comparable tensor light shift magnitude and opposite sign) at an appropriate intensity ratio, thus suppressing light-shift induced dephasing, enabling scalar magic conditions between the ground state and $5s5p$ $\mathrm{^{3}P_2}$, and tensor magic conditions for qudits encoded within it. Furthermore, this technique enables robust operation at the tensor magic angle 54.7$^\circ$ with linear trap polarization via reduced sensitivity to uncertainty in experimental parameters. We expect this technique to enable new loading protocols, enhance cooling efficiency, and enhance nuclear spins' coherence times, thus facilitating qudit-based quantum computing in ${}^{87}$Sr in the $5s5p$ $\mathrm{^{3}P_2}$ manifold.

Does gravity care about electric charge? Precision tests of the weak equivalence principle achieve remarkable sensitivity but deliberately minimize electric charge on test masses, leaving this fundamental question experimentally open. We present a minimalist framework coupling electromagnetism to linearized gravity through conservation of a complex charge-mass current, predicting charge-dependent violations $Δa/g = κ(q/m)$. Remarkably, this prediction occupies unexplored experimental territory precisely because precision gravity tests avoid charge variation. We identify this as a significant gap and propose a modified torsion balance experiment where $q/m$ is treated as a controlled variable. Such an experiment could test whether gravitational acceleration depends on electric charge, probing physics in genuinely new parameter space. This work exemplifies how theoretical minimalism can reveal overlooked opportunities in fundamental physics.

Noise and errors are unavoidable in any realistic quantum process, including processes designed to reduce noise and errors in the first place. In particular, quantum thermodynamical protocols for cooling can be significantly affected, potentially altering both their performance and efficiency. Analytically characterizing the impact of such errors becomes increasingly challenging as the system size grows, particularly in deep quantum circuits where noise can accumulate in complex ways. To address this, we introduce a general framework for approximating the cumulative effect of gate-dependent noise using a global depolarizing channel. We specify the regime in which this approximation provides a reliable description of the noisy dynamics. Applying our framework to the thermodynamical two-sort algorithmic cooling (TSAC) protocol, we analytically derive its asymptotic cooling limit in the presence of noise. Using the cooling limit, the optimal cooling performance is achieved by a finite number of qubits--distinguished from the conventional noiseless TSAC protocol by an infinite number of qubits--and fundamental bounds on the achievable ground-state population are derived. This approach opens new avenues for exploring noisy quantum thermodynamical processes.

We develop a geometry-first model that maps measured thin-disk water megamaser observables--sky angles, frequency shifts, their secular drifts and the angular redshift rate--to the black hole parameters in a generic static, spherically symmetric (SSS) spacetime written in the Schwarzschild gauge. The core of the approach is local: dot-product relations in the equatorial curved geometry relate the conserved light-deflection parameter to the observed detector angle at finite distance, providing a connection between sky positions and photon constants of motion. These local identities feed a closed model for the frequency shift of photons traveling between a maser clump circularly orbiting a black hole and a finite-distance detector, making explicit the dependence on the metric at emission and detection radii. We also apply the Gauss-Bonnet theorem to this construction on the equatorial two-manifold as an intrinsic cross-check. This theorem provides a global consistency relation between the local emission and detection angles, helping to validate sign conventions and angle branch choices in the local setup. In this sense, the local and global perspectives on the megamaser system support each other. To supplement the instantaneous information contained in frequency shifts, we incorporate the time-domain general relativistic invariant, the redshift rapidity. We further introduce a prospective angular-domain observable, the angular redshift rate, and give its analytic expression in the SSS framework. The results are formulated for generic SSS backgrounds, providing closed relations suited for likelihood-based inference from VLBI positions and spectral monitoring. In particular, for a Schwarzschild background, the black hole mass, its distance to Earth and megamaser orbital radius are fully constrained in the language of astrophysical observables.

In this work, we present a theoretical analysis of null geodesics, critical photon orbits, and shadow formation associated with a wormhole generated by a geometric defect. The propagation of light in this spacetime is examined through the deflection angle in both weak- and strong-field regimes. Analytical expansions are derived in each regime and employed to characterize gravitational lensing observables. By varying the global monopole charge, we evaluate its impact on these observables and determine parameter ranges that may be accessible to current or future observational probes. Finally, we calculate the quasinormal modes as well as the time-domain solution for scalar perturbations as well.

The (pion) decays controlled by axial anomaly imply the specific entanglement between photons having also the counterparts for classical electromagnetic waves. This is also a specific case of Eisnstein-Podolsky-Rosen-Bohm-Aharonov effect. The absence of causality and non-locality in (angular) momentum conservation is manifested, being especially clear for the generalization to the case of time rather than space separation corresponds to the polarization of dileptons described by time-like pion transition formfactors which may be studied experimentally. The similar decays in external magnetic field manifest the interplay with vacuum conductivity in external magnetic field and longitudinal polarization of vector mesons observed in heavy-ion collisions.

We establish anisotropic uncertainty principles (UPs) for general metaplectic operators acting on $L^2(\mathbb{R}^d)$, including degenerate cases associated with symplectic matrices whose $B$-block has nontrivial kernel. In this setting, uncertainty phenomena are shown to be intrinsically directional and confined to an effective phase-space dimension given by $\mathrm{rank}(B)$. First, we prove sharp Heisenberg-Pauli-Weyl type inequalities involving only the directions corresponding to $\ker(B)^\perp$, with explicit lower bounds expressed in terms of geometric quantities associated with the underlying symplectic transformation. We also provide a complete characterization of all extremizers, which turn out to be partially Gaussian functions with free behavior along the null directions of $B$. Building on this framework, we extend the Beurling-Hörmander theorem to the metaplectic setting, obtaining a precise polynomial-Gaussian structure for functions satisfying suitable exponential integrability conditions involving both $f$ and its metaplectic transform. Finally, we prove a Morgan-type (or Gel'fand--Shilov type) uncertainty principle for metaplectic operators, identifying a sharp threshold separating triviality from density of admissible functions and showing that this threshold is invariant under metaplectic transformations. Our results recover the classical Fourier case and free metaplectic transformations as special instances, and reveal the geometric and anisotropic nature of uncertainty principles in the presence of symplectic degeneracies.

Conformal field theories (CFTs) feature prominently in high-energy physics, statistical mechanics, and condensed matter. For example, CFTs govern emergent universal properties of systems tuned to quantum phase transitions, including their entanglement, correlations, and low-energy excitation spectra. Much of the rich structure predicted by CFTs nevertheless remains unobserved in experiment. Here we directly observe the energy excitation spectra of emergent CFTs at quantum phase transitions -- recovering universal energy ratios characteristic of the underlying field theories. Specifically, we develop and implement a modulation technique to resolve a Rydberg chain's finite-size spectra, variably tuned to quantum phase transitions described by either Ising or tricritical Ising CFTs. We also employ local control to distinguish parities of excitations under reflection and, in the tricritical Ising chain, to induce transitions between distinct CFT spectra associated with changing boundary conditions. By utilizing a variant of the modulation technique, we furthermore study the dynamical structure factor of the critical system, which is closely related to the correlation of an underlying Ising conformal field. Our work not only probes the emergence of CFT features in a quantum simulator, but also provides a technique for diagnosing a priori unknown universality classes in future experiments.

We combined experimental and theoretical investigations of an effective two-level atomic system operating in the near-infrared telecom wavelength regime, realized using hot rubidium vapor confined within a sub-micron-thick cell. In this strongly confined geometry, atomic coherence is profoundly influenced by wall-induced relaxation arising from frequent atom-surface collisions. By analyzing both absorption and fluorescence spectra, we demonstrate that the optical response is dominated by a closed cycling transition, which effectively isolates the atomic dynamics to a two-level configuration despite the presence of multiple hyperfine states. This confinement-induced selection suppresses optical pumping into uncoupled states and enables robust, controllable light-matter interaction at telecom wavelengths within a miniature atomic platform. Our results establish a practical route to realizing near-infrared atomic two-level systems in compact vapor-cell devices, opening new opportunities for integrated quantum photonic technologies, including on-chip quantum memories, telecom-band frequency references, and scalable quantum information processing.

Informationally overcomplete POVMs are known to outperform minimally complete measurements in many tomography and estimation tasks, and they also leave a purely classical freedom in shadow tomography: the same observable admits infinitely many unbiased linear reconstructions from identical measurement data. We formulate the choice of reconstruction coefficients as a convex minimax problem and give an algorithm with guaranteed convergence that returns the tightest state-independent variance bound achievable by post-processing for a fixed POVM and observable. Numerical examples show that the resulting estimators can dramatically reduce sampling complexity relative to standard (canonical) reconstructions, and can even improve the qualitative scaling with system size for structured noncommuting targets.

Non-adiabatic dynamics at molecule-metal interfaces govern diverse and technologically important phenomena, from heterogeneous catalysis to dye-sensitized solar energy conversion and charge transport across molecular junctions. Realistic modeling of such dynamics necessitates taking into account various charge and energy transfer channels involving the coupling of nuclear motion with a very large number of electronic states, leading to prohibitive cost using classical computational methods. In this work we introduce a generalization of the Anderson-Newns Hamiltonian and develop a highly optimized quantum algorithm for simulating the non-adiabatic dynamics of realistic molecule-metal interfaces. Using the PennyLane software platform, we perform resource estimations of our algorithm, showing its remarkably low implementation cost for model systems representative of various scientifically and industrially relevant molecule-metal systems. Specifically, we find that time evolution for models including $100$ metal orbitals, $8$ molecular orbitals, and $20$ nuclear degrees of freedom, requires only $271$ qubits and $7.9 \times 10^7$ Toffoli gates for $1000$ Trotter steps, suggesting non-adiabatic molecule-metal dynamics as a fruitful application of first-generation fault-tolerant quantum computers.

The Superconducting Quasiparticle-Amplifying Transmon (SQUAT) is a new sensor architecture for THz (meV) detection based on a weakly charge-sensitive transmon directly coupled to a transmission line. In such devices, energy depositions break Cooper pairs in the qubit capacitor islands, generating quasiparticles. Quasiparticles that tunnel across the Josephson junction change the transmon qubit parity, generating a measurable signal. In this paper, we present the design of first-generation SQUATs and demonstrate an architecture validation. We summarize initial characterization measurements made with prototype devices, comment on background sources that influence the observed parity-switching rate, and present experimental results showing simultaneous detection of charge and quasiparticle signals using aluminum-based SQUATs.

Multipartite entanglement is a natural generalization of bipartite entanglement, but is relatively poorly understood. In this paper, we develop tools to calculate a class of multipartite entanglement measures - known as multi-invariants - for stabilizer states. We give an efficient numerical algorithm that computes multi-invariants for stabilizer states. For tripartite stabilizer states, we also obtain an explicit formula for any multi-invariant using the GHZ-extraction theorem. We then present a counting argument that calculates any Coxeter multi-invariant of a q-partite stabilizer state. We conjecture a closed form expression for the same. We uncover hints of an interesting connection between multi-invariants, stabilizer states and topology. We show how our formulas are further simplified for a restricted class of stabilizer states that appear as ground states of interesting models like the toric code and the X-cube model.

As quantum devices scale to larger and larger sizes, a significant challenge emerges in scaling their coherent controls accordingly. Quantum cellular automata (QCAs) constitute a promising framework that bypasses this control problem: universal dynamics can be achieved using only a static qubit array and global control operations. We realize QCAs on a dual-species Rydberg array of rubidium and cesium atoms, leveraging independent global control of each species to perform a myriad of quantum protocols. With simple pulse sequences, we explore many-body dynamics and generate a variety of entangled states, including GHZ states, 96.7(1.7)%-fidelity Bell states, 17-qubit cluster states, and high-connectivity graph states. The versatility and scalability of QCAs offers compelling routes for scaling quantum information systems with global controls, as well as new perspectives on quantum many-body dynamics.

Using the worldline quantum field theory formalism, we compute the conservative scattering angle and impulse for classical black hole scattering at fifth post-Minkowskian (5PM) order by providing the second self-force (2SF) contributions. This four-loop calculation involves non-planar Feynman integrals and requires advanced integration-by-parts reduction, novel differential-equation strategies, and efficient boundary-integral algorithms to solve a system of hundreds of master integrals in four integral families on high-performance computing systems. The resulting function space includes multiple polylogarithms as well as iterated integrals with a K3 period, which generate a spurious velocity divergence at $v/c=\sqrt{8}/3$. This divergence is present in the potential region and must be cancelled by conservative memory contributions from radiative regions. We find that the standard use of Feynman propagators to access the conservative sector fails to ensure this cancellation. We propose a conservative propagator prescription which realises both cancellations leading to a physically sensible answer. All available low-velocity checks of our result against the post-Newtonian literature are satisfied.

In four-dimensional vacuum general relativity the only known static, exact and analytical black hole solution is given by the Schwarzschild spacetime. In this paper this renowned metric is generalised by adding another integrating constant, a hair that switches the metric from the Petrov type D to the type I. This new parameter represents the intensity of an external gravitational field, which can be considered the hyperbolic generalisation of the Witten's bubble of nothing. No curvature or conical singularities are present outside the event horizon. The no hair arguments are circumvented because the metric is not asymptotically flat, and neither the black hole is spherical. The gravitational hair continuously deforms the Schwarzschild geometry: the horizon becomes oblate, while its area is reduced. Conserved charges and thermodynamic properties of the black hole are studied.

The historical and conceptual foundations of General Relativity are revisited, putting the main focus on the physical meaning of the invariant ds, the Equivalence Principle, and the precise interpretation of spacetime geometry. It is argued that Albert Einstein initially sought a dynamical formulation in which ds encoded the gravitational effects, without invoking curvature as a physical entity. The now more familiar geometrical interpretation (identifying gravitation with spacetime curvature) gradually emerged through his collaboration with Marcel Grossmann and the adoption of the Ricci tensor in 1915. Anyhow, in his 1920 Leiden lecture, Einstein explicitly reinterpreted spacetime geometry as the state of a physical medium (an ether endowed with metrical properties but devoid of mechanical substance) thereby actually rejecting geometry as an independent ontological reality. Building upon this mature view, gravitation is reconstructed from the Weak Equivalence Principle, understood as the exact compensation between inertial and gravitational forces acting on a body under a uniform gravitational field. From this fundamental principle, together with an extension of Fermat Principle to massive objects, the invariant ds is obtained, first in the static case, where the gravitational potential modifies the flow of proper time. Then, by applying the Lorentz transformation to this static invariant, its general form is derived for the case of matter in motion. The resulting invariant reproduces the relativistic form of Newton second law in proper time and coincides with the weak field limit of General Relativity in the harmonic gauge.

Prime factorization on quantum processors is typically implemented either via circuit-based approaches such as Shor's algorithm or through Hamiltonian optimization methods based on adiabatic, annealing, or variational techniques. While Shor's algorithm demands high-fidelity quantum gates, Hamiltonian optimization schemes, with prime factors encoded as degenerate ground states of a problem Hamiltonian, generally require substantial classical post-processing to determine control parameters. We propose an all-quantum, measurement-based feedback approach that iteratively steers a quantum system toward the target ground state, eliminating the need for classical computation of drive parameters once the problem Hamiltonian is determined and realized. As a proof of principle, we experimentally factor the biprime 551 using a three-qubit NMR quantum register and numerically analyze the robustness of the method against control field-errors. We further demonstrate scalability by numerically implementing the FALQON factorization of larger biprimes, 9,167 and 2,106,287, using 5 and 9 qubits, respectively.

Fault-tolerant quantum computation in photonic architectures relies on the efficient preparation of high-fidelity logical magic states under realistic constraints imposed by finite squeezing and photon loss. In this work, we study logical T-gate magic-state preparation in GKP-encoded photonic qubits using a repeat-until-success injection protocol combined with outer surface-code protection. We develop an architecture-level modeling framework based on a lightweight density-matrix simulator implemented with standard numerical linear algebra. Finite squeezing is mapped to effective logical dephasing, depolarizing noise is included at the logical level, and photon loss is treated as a heralded erasure process. This approach avoids explicit continuous-variable wavefunction simulation, hardware-specific photonic models, and quantum software frameworks, enabling transparent and computationally efficient exploration of architectural trade-offs. We perform systematic parameter sweeps over squeezing values from 8 to 16 dB, baseline loss probabilities between 0.01 and 0.03, and surface-code distances d = 1, 3, 5, and 7. Across this regime, we evaluate repeat-until-success probability, average injection overhead, and logical magic-state fidelity. We find that success probabilities exceed 0.94 across all studied parameters, with an average overhead close to unity. After outer-code protection, logical fidelities reach approximately 0.77 to 0.80 and show weak sensitivity to moderate photon loss but a strong dependence on squeezing. Phase-boundary analysis identifies minimum squeezing requirements needed to simultaneously achieve high success probability and logical fidelity. These results provide quantitative design guidance for scalable photonic fault-tolerant quantum architectures.

It has been proposed that any coupling constant in a covariant action can be treated as a conserved charge by promoting the coupling constant to auxiliary fields, typically realized by a scalar field paired with a higher-form gauge field. However, the procedure may break local symmetries, which can be explicitly shown in a simpler setting such as Maxwell theory. The Hamiltonian analysis of Maxwell theory with the auxiliary fields reveals that some of the constraints are second-class. Applying the BFT formalism, we restore the broken local symmetries and obtain a fully symmetric action defined on an extended configuration space. Despite the restoration of the local symmetries, no additional conserved charges are associated with the recovered symmetries. Consequently, the original theory turns out to be the gauge-fixed version of the extended theory.

We derive exact general solutions (as opposed to attractor particular solutions) for the evolution of a scalar field $φ$ in a universe dominated by a background fluid with equation of state parameter $w_B = -1$, extending earlier work on exact solutions with $w_B > -1$. Straightfoward exact solutions exist when the evolution is described by a linear differential equation, corresponding to constant, linear, and quadratic potentials. In the nonlinear case, exact solutions are derived for $V = V_0\ln φ$, $V = V_0 φ^{1/2}$ and $V = V_0/φ$, and the logarithmic potential also yields an exact first integral. These complicated parametric solutions are considerably less useful than those derived previously for a universe dominated by a barotropic fluid such as matter or radiation with $w_B > -1$. However, we generalize the slow-roll approximation and show that it applies to all sufficiently flat potentials in the case of a vacuum-dominated expansion, while it never applies when the universe is dominated by a background fluid with $w_B > -1$.

Using $e^+e^-$ collision data corresponding to an integrated luminosity of 7.33~${\rm fb^{-1}}$ recorded by the BESIII detector at center-of-mass energies between 4.128 and 4.226~${\rm GeV}$, we present the first search for the semileptonic decays $D^+_s\to\ f_1(1420)e^+ν_e$ and $D^+_s\to\ f_1(1285)e^+ν_e$. The first evidence for the decay $D^+_s\to\ f_1(1420)e^+ν_e$ is found with a statistical significance of 3.4$σ$, and its product branching fraction $\mathcal{B}(D^+_s\to\ f_1(1420)e^+ν_e)\cdot\mathcal{B}(f_1(1420)\to\ K^+K^-π^0)$ is determined to be $\rm (4.5^{+2.0}_{-1.7}(stat) \pm0.4(syst)) \times 10^{-4} $, corresponding to an upper limit of $7.6 \times10^{-4}$ at the 90% confidence level. No significant signal of the decay $D^+_s\to\ f_1(1285)e^+ν_e$ is observed and the upper limit on the product branching fraction is set to be $\mathcal{B}(D^+_s\to\ f_1(1285)e^+ν_e)\cdot\mathcal{B}(f_1(1285)\to\ π^+π^-η) < 1.7\times10^{-4}$ at the 90% confidence level.

We report on the measurement of neutron energy spectra at the newly established Yemi Underground Laboratory (Yemilab) in the Republic of Korea, designed to host dark matter and rare-event search experiments. A high-sensitivity neutron spectrometer was employed, consisting of ten cylindrical {}^{3}He proportional counters, eight of which were embedded in cylindrical high-density polyethylene moderators of various sizes. To quantify and mitigate contributions from internal α-backgrounds, each detector underwent a dedicated background measurement using a cadmium-shielded box. These backgrounds, primarily originating from trace amounts of U and Th in the stainless-steel housings, were characterized and subtracted during data analysis. Neutron measurements were carried out at three locations within the Yemilab between March to October 2023. After waveform-based event selection and correction for \alphasym-backgrounds, neutron count rates were estimated and corresponding energy spectra were reconstructed using the unfolding method. The total neutron fluence rates were measured ranged from (3.24 $\pm$ 0.11) to (4.01 $\pm$ 0.10) $\times~10^{-5}~ {cm}^{-2}~{s}^{-1}$, with thermal and fast neutron components (1 - 10 MeV) ranging from (1.32 $\pm$ 0.05) to (1.51 $\pm$ 0.05) $\times 10^{-5}~{cm}^{-2}~{s}^{-1}$ and (0.27 $\pm$ 0.03) to (0.34 $\pm$ 0.10) $\times~10^{-5}~{cm}^{-2}~{s}^{-1}$, respectively.

The Low Gain Avalanche Diodes (LGADs) are promising particle detectors for timing resolution better than $50$ ps under a high radiation environment. This study investigates n-in-p LGAD architecture, focusing on ultra-thin sensors of thickness less than $50\ μ$m using the WeightField2 program. The capabilities of WeightField2 are demonstrated by comparing its results with irradiation measurements from an FBK LGAD wafer, showing good agreement across unirradiated and neutron-irradiated conditions. This paper presents device simulations in High Luminosity LHC conditions (lifetime integrated fluence $ \mathcal{O} (10^{14})\ \mathrm{n_{eq}~cm^{-2}}$, temperature $ \approx 243\ \mathrm{K} $), and taking into account radiation damage, gain reduction due to fluence, and lattice defects. It is shown that a 20 $μ$m thick sensor achieves the best timing performance. Among Silicon (Si), Diamond (C), and 4H-Silicon Carbide (4H-SiC), we found 4H-SiC to be the most promising: it provides the highest gain value for a fixed thickness and gain implant layer configuration, and best retains high charge collection value and timing capability under increasing fluence up to $50\times10^{14}\ \mathrm{n_{eq}~cm^{-2}}$. A time resolution less than 25 ps is reported with different gain implant concentrations for a $20 μ$m 4H-SiC sensor. This work presents the potential of SiC-based LGADs in high-radiation collider environments.

Higgs boson pair production in the massless quark-antiquark channel proceeds at leading order (LO) via electroweak boson loops. We calculate the next-to-leading order QCD corrections to this process. For the corresponding two-loop amplitudes, an analytic representation has been achieved. Even though the size of this contribution at the level of total cross sections is below 1% compared to the LO gluon channel, the effect on differential observables can be in the 10% range and therefore this contribution should be taken into account when comparing to LHC data.

The quest to understand three-body dynamics from first-principle QCD includes the study of non-resonant and resonant systems. The isospin $I=2$ system is of particular interest having no three-body resonance but featuring a resonance in a sub-channel, while also being a coupled-channel problem. In this study, we calculate the finite-volume spectrum from lattice QC at two different pion masses, map the amplitude to the infinite volume through a generalized FVU three-body quantization condition, investigate the limit of a narrow $ρ$, and compare with an effective Lagrangian prediction at leading order. Chiral extrapolations between different pion masses are performed.

Current multi-ton detectors put stringent constraints on the GeV-scale galactic dark matter, pushing the allowed cross-section almost towards the neutrino fog, yet remain mostly insensitive to the light dark matter. Cosmic rays can upscatter the non-relativistic halo dark matter particles, making a sub-population of them gain sufficient kinetic energy to be discernible in current direct search experiments. In this work, we explore this alternate strategy to probe sub-MeV electrophilic dark matter boosted by cosmic rays with the latest data of LZ 2025 (WS2024 run) and improve the constraint on the MeV scale dark matter by almost $\sim\mathcal{O}(1)$ compared to the previous XENONnT limit for energy-independent cross-section. Using realistic energy-dependent cross-sections, we also analyse such a scenario, where the associated mediator mass plays a crucial role in governing the event rate and hence the expected limits too. With energy-dependent cross-sections, our obtained limits also remain stronger than the existing constraints from current direct detection experiments. Even compared to the limits from the neutrino detectors with a larger target size, LZ 2025 can put stringent constraints in certain parameter space of the mediator, excluding the previously unexplored regions.

Motivated by recent results in spin chains we study dissipation and relaxation in a two-dimensional holographic interface conformal field theory (ICFT) in which degrees of freedom on one side of the interface are coupled to an external bath, while the other side remains isolated. In the bulk description this setup is realized by gluing a supersymmetric Janus geometry to a BTZ black hole region, with the coupling implemented through a double-trace deformation. We determine the quasinormal modes in the bulk by solving the double-trace matching conditions of the system and bath. The lowest imaginary part of the modes defines a Liouvillian gap, and following earlier work in spin chains we introduce the dimensionless ratio crelax as a measure of interface-induced suppression of relaxation. Numerically we find that, crelax is independent of coupling details to the bath. It is a strong candidate for a universal interface observable characterizing dissipation and relaxation across the interface.

We investigate whether the hidden zeros and associated factorisations found for massless colour-ordered amplitudes persist under massive deformations. Using the kinematic mesh construction, we show that hidden zeros survive only for symmetry controlled mass generation. For massive $\text{Tr} Φ^3$ with a uniform mass, the zeros and their factorisation patterns are inherited after a massive shift of planar variables, and an analogous statement holds for Kaluza-Klein reductions where the relevant non-planar variables are modified by conserved mode numbers. For the non-linear sigma model (NLSM), a naive pion mass term generically spoils hidden zeros, while a spurion induced potential restores them. This allows factorisation near zeros, including odd point channels described by an appropriately mass deformed NLSM + $φ^3$ theory, and leads to a hidden zero based on-shell recursion for massive NLSM amplitudes. For spin-one, a simple massive Yang-Mills theory fails to exhibit hidden zeros, while spontaneously broken gauge theories preserve them.

The NA64 experiment at CERN H4 beamline recently started a high-energy positron-beam program to search for light dark matter particles through a thick-target, missing-energy measurement. To fulfil the energy resolution requirement of the physics measurement $σ_E/E\simeq2.5\%/\sqrt{E\mathrm{[GeV}]} \oplus 0.5\%$ and cope with the constraints and performance requests of the NA64 setup, a new high-resolution homogeneous electromagnetic calorimeter PKR-CAL has been designed. The detector is based on PbWO$_4$ crystals, each read by multiple SiPM sensors to maximize the light collection. The PKR-CAL design has been optimized to mitigate and control unavoidable SiPM saturation effects at high light levels, as well as to minimize the gain fluctuations induced by instantaneous variations of the H4 beam intensity. The $R\&D$ program culminated in the construction of a small-scale prototype, POKERINO. In this work, we present the results from the experimental characterization campaign of the POKERINO aiming at demonstrating that the obtained performances are compatible with the application requirements.

The Algebra of Physical Space (APS) is used to explore the Constructive Standard Model (CSM) of particle physics. Namely, this paper connects the spinor formalism of the APS to massive amplitudes in the CSM. A novel equivalency between traditional CSM and APS-CSM formalisms is introduced, called the Scattering Algebra (SA), with example calculations confirming the consistency of results between both frameworks. Through this all, two significant insights are revealed: The identification of traditional CSM spin spinors with Lorentz rotors in the APS, and the connection of the CSM to various formalisms through ray spinor structure. The CSM's results are replicated in massive cases, showcasing the power of the index-free, matrix-free, coordinate-free, geometric approach and paving the way for future research into massless cases, amplitude-construction, and Wigner little group methods within the APS.

Two machine learning techniques for jet measurements at the LHCb experiment are presented: a regression-based method for jet-energy calibration and a deep neural network algorithm for jet flavour tagging, distinguishing between $b$-quark, $c$-quark, and light parton jets. These techniques are applied to a search for inclusive $H \to \bbbar$ and $H \to c\barcc$ decays using a LHCb dataset corresponding to an integrated luminosity of 1.6\invfb. The observed (expected) 95\% confidence level upper limits correspond to 6.6 (11.1) times the SM cross-section for the $H \to b\bar b$ process, and 1003 (1834) times the SM cross-section for the $H \to c\bar c$ process.

Measurements of CP asymmetries in hadron decays integrated over time provide access to direct CP violation, which arises from differences between the amplitudes of CP-conjugate decay processes. These proceedings present an overview of recent measurements from the LHCb and Belle II experiments, including progress on the determination of the CKM angle $γ$, the first observation of CP violation in baryon decays, and new studies of direct CP violation in $D$ mesons. The outlook for measurements with datasets whose collection is underway is also discussed.

The Alexandrov-Kontsevich-Schwarz-Zaboronsky (AKSZ) construction encodes all the data of a topological sigma-model in the finite-dimensional symplectic $Q$-manifold. Relaxing the nondegeneracy condition i.e. considering a presymplectic form instead, extends the construction to non-topological models. The gauge-invariant action functional of (presymplectic) AKSZ sigma model is written in terms of space-time differential forms and can be seen as a covariant multidimensional analogue of the usual 1st order Hamiltonian action. In this work, we show that the AKSZ construction has a natural generalisation where the target space $Q$-manifold is equipped with a form of arbitrary degree $Ω$ (possibly inhomogeneous) which is $(\mathrm{d}+L_Q)$-closed. This data defines a higher-derivative generalisation of the AKSZ action which is still invariant under the natural gauge transformations determined by $Q$ and which is efficiently formulated in terms of a version of Chern-Weil map introduced by Kotov and Strobl. It turns out that a variety of interesting gauge theories, including higher-dimensional Chern-Simons theory, MacDowell-Mansouri-Stelle-West action and self-dual gravity as well as its higher spin extension, can be concisely reformulated as such multisymplectic AKSZ models. We also present a version of the construction in the setup of PDE geometry and demonstrate that the counterpart of the multisymplectic AKSZ action is precisely the standard multisymplectic formulation, where the Chern-Weil map corresponds to the usual pullback map.

BESIII has recently accumulated a large data sample near the $ψ(3770)$ production threshold corresponding to an integrated luminosity of $20\text{ fb}^{-1}$. Neutral $D^0\bar{D}^0$ pairs produced at the $ψ(3770)$ are in a $C$-odd correlated state, providing a unique laboratory to measure the strong-phase differences between $D^0$ and $\bar{D}^0$ decays. These parameters are essential inputs to the study of CP violation in heavy-flavor physics, primarily in the determinations of the CKM angle gamma and charm-mixing parameters. These proceedings report new measurements of strong-phase differences in different neutral $D$ decays at BESIII, including new measurements of the strong phases in $D\to K^+K^-π^+π^-$ decays. Additionally reported is the first observation of correlated $DD$ pairs produced at $e^+e^-$ center-of-mass energies above the $ψ(3770)$ threshold, where $e^+e^-\to D^{*}\bar{D}$ and $e^+e^-\to D^{*}\bar{D}^{*}$ processes also occur. These processes produce both $C$-odd and $C$-even correlated $D^0\bar{D}^0$ pairs, which allow for new measurement techniques to determine strong-phases from previously unused datasets.

The thermodynamic geometry of the hadron resonance gas model with (without) excluded volume effects (EVE) of baryons is investigated. The case with imaginary mu, where mu is the baryon chemical potential, is investigated as well as the one with real mu. We calculate the scalar curvature R and use the R=0 criterion to investigate the phase structure in the mu^2-T plane where T is the temperature. The curve on which R=0 continues analytically from the imaginary mu region, where the lattice QCD is feasible, to the real mu one. In the presence of EVE, there are rich phase structures in the large real mu region as well as the Roberge-Weiss like region where mu is imaginary and a singularity appears, while there is no phase structure in the large real $μ$ region in the absence of EVE. The limitation temperature of the baryon gas is also obtained by using the baryon number fluctuation. The LQCD predicted critical point locates almost on the curve of the limitation temperature we determined. A simple sufficient condition, n_B>1/(2v_B)$, is obtained for the quark deconfinement in the large real mu region, where n_B and v_B are the net baryon number density and the volume of a baryon, respectively.

The BESIII collaboration has collected the world's largest datasets at the energy thresholds for producing a variety of open-charm hadrons. Most recently, the BESIII collaboration has collected datasets that significantly increase the number of $D^0$, $D^+$, and $Λ_c^+$ hadrons for analysis, further improving on previous datasets collected by BESIII. These proceedings highlight recent results that leverage these datasets to study the pure leptonic and semileptonic decays of these hadrons, as well as study the polarization of baryon pairs produced in electron-positron collisions.

Fundamental scale invariance implies the scale invariant standard model. Both the Fermi scale and the Planck mass are given by fields, and their ratio is dictated by a dimensionless cosmon-Higgs coupling. For an ultraviolet fixed point of quantum gravity this coupling is an irrelevant parameter of the renormalization flow and becomes predictable. An analytic scaling solution for quantum gravity admits no free parameter for the mass term of the Higgs boson. If the largest intrinsic mass scale generated by the renormalisation flow away from the fixed point is sufficiently below the Fermi scale, the couplings of the scale invariant standard model are determined by the scaling solution. For a given short distance model remaining valid to infinitely small distances the ratio Fermi scale over Planck mass can be predicted. With reasonable assumptions for an ultraviolet fixed point a numerical solution finds a tiny value for the ratio between the Fermi and Planck scales, very close to a second order quantum electroweak phase transition. This could explain the observed gauge hierarchy.

We study the evolution of primordial black holes (PBHs) formed in the early universe in the presence of a surrounding thermal bath. By incorporating the effects of thermal absorption, we show that PBHs can undergo significant mass growth, leading to extended lifetimes and substantial deviations from the standard Hawking evaporation scenario. We find a critical collapse efficiency, $γ_{\rm c} \simeq 0.395$, above which the PBH mass grows without bound. This correction has profound implications for both PBH-induced reheating and dark matter (DM) production. Specifically, we find that the reheating temperature can be suppressed, and the DM parameter space for the PBH reheating scenario can undergo $\mathcal{O}(10)$-$\mathcal{O}(10^4)$ corrections, depending on the PBH formation mass and collapse efficiency. Moreover, our results significantly shift the parameter space in which PBHs can account for the entirety of the DM. To the best of our knowledge, this is the first comprehensive phenomenological study to incorporate thermal absorption into PBH evolution and quantify its impact on cosmological observables.

We question the compatibility of recent ${\cal S}=-2$ hypernuclear assignments of J-PARC E07 $Ξ^-$-capture emulsion events with assignments deduced from other experiments.

In this HYP2025 talk I pay tribute to Tullio Bressani (1940-2024), Bogdan Povh (1932-2024) and Toshimitsu Yamazaki (1934-2025), all of whom made lasting contributions to shaping up Strangeness Nuclear Physics. Yoshinori Akaishi's (1941-2025) record is also noted.

We study flavour-changing neutral current decays of the top quark in the democratic Three Higgs Doublet Model featuring a $Z_3$-symmetric scalar potential and Natural Flavour Conservation. In this framework, while such processes are absent at tree-level, the extended scalar sector induces new one-loop contributions to rare top decays. We compute the branching ratios for processes of the form $t \to q X$ (with $q = u, c$ and $X$ denoting a boson of the model), and explore the viable regions of the parameter space under theoretical consistency conditions and current experimental constraints. Several alignment-limit scenarios corresponding to different hierarchies among the CP-even Higgs states are analysed, and we find that the predicted branching ratios can significantly exceed their Standard Model expectations while remaining consistent with existing limits. In particular, we identify scenarios with light non-standard scalars that can lead to rates within the projected sensitivity of the High-Luminosity LHC. Our results therefore highlight rare top decays as a promising probe of the extended scalar sector of the Three Higgs Doublet Model.

The BABAR experiment participates to the global endeavor for a precise prediction of the anomalous magnetic moment of the muon by evaluating the contribution from hadronic vacuum polarization, in particular through cross section measurements of hadronic final states from $e^+e^-$ collisions. After a first measurement in 2009 of the largest input that comes from the $e^+e^- \rightarrow π^+π^-(γ)$ cross section, we present preliminary results from a new study on 460 ${\rm fb}^{-1}$ of BABAR data, involving a blind and independent procedure. The results of the two analyses are shown to be consistent.

High-energy proton-proton collisions at the LHC offer a stringent test of Quantum Chromodynamics (QCD) in the small-$x$, gluon-dominated regime. This study focus on a minimal, gluon-driven framework to describe the charged-particle multiplicities and their pseudorapidity densities in high energy collisions. The two-component model taken here includes the hard gluon-gluon fusion process and the soft quark recombination process, which directly relates to both integrated and unintegrated parton distributions. We begin by evolving Parton Distribution Functions (PDFs) using the Modified Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (MD-DGLAP) equations. These PDFs are then converted into unintegrated PDFs (UPDFs) via the Kimber-Martin-Ryskin (KMR) scheme. The resulting PDFs and UPDFs are incorporated into the two-component model to predict the charged-particle pseudorapidity density $\left(1 / N_{\mathrm{ev}}\right) d N_{\mathrm{ch}} / d η$ in $pp$ collisions at LHC energies. Our predictions are compared to the data from the ATLAS experiment, revealing that the model effectively captures the features of the observed pseudorapidity distributions, despite its simplicity. Within this framework, the gluon-gluon fusion processes are found to dominate particle production for $\sqrt { s } \ge 9 0 0 \ \mathrm { GeV }$.These findings provide phenomenological support for MD-DGLAP-based PDFs and the associated small-$x$ gluon dynamics. Furthermore,a comparative analysis of results from alternative PDF sets--including CTEQ, MSHT, NNPDF, HERAPDF, and GRV--is performed, with particular focus on examining their consistency with the relative shapes of experiment data in the small-$x$ region.

Dilatons, the CP-even pseudo-Nambu-Goldstone bosons arising from spontaneous scale symmetry breaking, offer a compelling alternative to axion-like particles (ALPs) yet lack a comprehensive low-energy framework. We address this by constructing a systematic effective field theory (EFT) for the dilaton based on a manifestly scale-invariant regularization scheme. This approach derives universal linear couplings to the trace anomaly while preserving consistent renormalization group evolution. We establish a hierarchical EFT tower connecting the ultraviolet conformal sector to the infrared, encompassing the dilaton-extended SMEFT, low-energy EFT up to dimension-7, and a chiral Lagrangian describing meson and baryon interactions. We perform a comprehensive phenomenological analysis across two distinct mass regimes, where dilaton manifests as either conventional particle or wave-like particle. For MeV-scale dilatons behaving as conventional particles, we obtain constraints from LHC production, semi-invisible $B$- and $K$-meson decays, and supernova cooling. For ultralight dilatons acting as dark matter, we project sensitivities for atomic clocks and atom interferometers. This unified EFT framework would pave the way for extended phenomenological studies across the full mass spectrum of the light dilaton.

We reexamine the energy-momentum tensor in classical electrodynamics from the perspective of spacetime-dependent translations, i.e., diffeomorphism invariance in flat spacetime. When energy-momentum is identified through local translations rather than constant ones, a unique, symmetric, and gauge-invariant energy-momentum tensor emerges that satisfies a genuine off shell Noether identity without invoking the equations of motion. For the free electromagnetic field, this tensor coincides with the familiar Belinfante-Rosenfeld and Bessel-Hagen expressions, but arises here directly from spacetime-dependent translation symmetry rather than from improvement procedures or compensating gauge transformations. In interacting classical electrodynamics, comprising a point charge coupled to the electromagnetic field, diffeomorphism invariance yields well-defined energy-momentum tensors for the field and the particle, while the interaction term itself generates no independent local energy-momentum tensor. Its role is instead entirely encoded in the coupled equations of motion governing energy-momentum exchange, thereby resolving ambiguities in energy-momentum localization present in canonical and improvement-based approaches.

Using 7.33 fb$^{-1}$ of $e^+e^-$ collision data samples collected with the BESIII detector at center-of-mass energies between 4.128 and 4.226 GeV, we perform a simultaneous fit to search for the radiative decay $D_s^+\toγK^*(892)^+$ via $K^*(892)^+\to K^+π^0$ and $K^*(892)^+\to K_S^0π^+$ for the first time. No significant signals are observed. The upper limit on the branching fraction of $D^+_s\toγK^*(892)^+$ is set to be $2.3\times10^{-4}$ at the $90\%$ confidence level.

We develop a systematic approach to the computation of gauge invariant quartic interactions between reducible massive and massless higher spin fields. Extending the BRST formulation of existing cubic results, we obtain a single constraint for each off-shell quartic vertex that ensures both the gauge invariance of the Lagrangian and associativity of the gauge transformations at quartic order. A solution to these equations is presented. The general equation is then reduced to an on-shell version to reduce complexity. We find example solutions for the off-shell and on-shell quartic vertices in low spin examples relevant to the problem of black hole scattering.

We study lepton-number-violating interactions at dimension seven in the Standard Model effective field theory that contribute to the meson decays $B \to K νν$ and $K \to πνν$. Such interactions could washout the baryon asymmetry of the Universe and also contribute to the neutrinoless double beta decay, even though the interactions involve a change in down-type quark flavors. We clarify conditions under which excesses in meson decay rates over the Standard Model predictions can be successfully observed. We also show that, although these interactions contribute to neutrino masses at the two-loop level, the Weinberg operator can be introduced consistently without spoiling the scenario.

We present experimental results on the angular distributions of Drell-Yan muons produced by a 120 GeV/$c$ proton beam interacting with liquid hydrogen and deuterium targets. The dimuon angular distributions in both polar ($θ$) and azimuthal ($φ$) angles in the Collins-Soper frame are measured within the kinematic range of $4.5 < m_{μμ} < 10\ \mathrm{GeV}/c^2$, $0.19 < p_T < 2.24\ \mathrm{GeV}/c$, and $0 < x_F < 0.95$. Unlike the results of a previous proton-induced Drell-Yan experiment at a higher energy, the data reveal a pronounced $\cos 2φ$ modulation in the angular distributions. Comparison with perturbative QCD (pQCD) predictions shows statistically significant deviations, with p-values of 3.5\% for the $p+p$ and 1.5\% for the $p+d$ Drell-Yan processes. These results suggest the presence of nonperturbative QCD contributions.

The universality of the color dipole amplitude is a cornerstone of high-energy Quantum Chromodynamics (QCD). However, standard phenomenological approaches typically rely on rigid parametric ansatzes and often require ad-hoc geometric adjustments to reconcile inclusive and diffractive measurements. To resolve this tension, we introduce Physics-Informed Neural Networks (PINNs) employing a ``Teacher--Student'' strategy. The rigorous momentum-space Balitsky-Kovchegov evolution dynamics act as the ``Teacher,'' constraining the solution manifold, while the network ``Student'' is refined against inclusive HERA $F_2$ data. This approach extracts a model-independent dipole amplitude without assuming initial states. Strikingly, we demonstrate that this amplitude -- without parameter retuning or geometric rescaling -- successfully predicts exclusive $J/ψ$ photoproduction cross-sections. This zero-parameter prediction rigorously confirms the universality of the gluon saturation scale and establishes PINNs as a transformative paradigm for uncovering non-perturbative QCD structures.

We present a detailed characterization of a new generation of athermal-phonon single-charge sensitive Si HVeV detectors, the best of which achieved 612 meV $\pm$ 4 meV baseline resolution. Our sub-eV energy resolution enables precise measurements of single-photon events and reveal consistent energy losses of 0.81 eV $\pm$ 0.03 eV per charge excitation across two facilities. We demonstrate that the noise for these detectors is well described using a standard Transition Edge Sensor noise model. We also place upper bounds on the nominal phonon collection efficiency of 45\%, establishing these detectors as the most efficient athermal phonon detectors to date, limited only by intrinsic limitations of quasiparticle generation.

Non-observation of proton decays as well as the smallness of the neutrino masses can naturally be explained by the accidental baryon and lepton number symmetry in the Standard Model, where the approximate symmetries are a consequence of the absence of the baryon or lepton number violating operators at the renormalizable level. The neutrino masses at sub-eV scales can be explained by the presence of the dimension-five, $\ell\ell HH/Λ$, term in the Lagrangian, suggesting that a more fundamental theory takes over beyond the energy scale $Λ$. We consider the possibility that the theory above the scale $Λ$ generates general higher dimensional operators with the flavor structure implied by the Yukawa interactions in the Standard Model. Such a set-up can be realized, for example, in the composite Higgs scenario with partial compositeness of fermions. The fermion masses and the neutrino masses are explained for $Λ\sim 10^{11}$GeV. The lifetime of proton in this scenario is, interestingly, consistent with the observed event of the $p \to π^0 μ^+$ decay at the Super-Kamiokande experiment. The Hyper-Kamiokande experiments should see a large number of events soon after the data taking.

Accurate knowledge of material reflectance in the vacuum ultraviolet (VUV) range is crucial for optimizing photon detection in noble gas detectors such as DUNE. Despite its importance, reflectance values for detector materials in the VUV region remain poorly characterized, with literature values showing significant variation depending on surface termination and finish. We present an angular-resolved reflectance measurement system developed at IFIC that operates in a gaseous argon atmosphere, enabling realistic measurements of detector materials under controlled conditions. The setup couples a deuterium lamp to a monochromator and employs a motorized PMT rotating around the sample to measure reflected light distributions across a wide angular range. We have characterized two key DUNE materials -- aluminum field cage profiles and stainless steel cryostat membranes -- in both the UV-VIS (300-500 nm) and VUV (128-200 nm) ranges. In the UV-VIS region, we confirm literature values of approximately 60% reflectance for aluminum and 40% for stainless steel. Preliminary VUV measurements at 45° angle of incidence yield reflectance values of 10-15% for both materials, significantly lower than their UV-VIS counterparts. The reflected light distributions exhibit a mixed character between specular and diffuse reflection. These results have direct implications for detector simulations and light yield predictions in next-generation experiments.

Gapped two-dimensional gauge theories with massless fermions generically have rich vacuum structures consisting of many degenerate vacua related by the action of topological line operators. The algebra of such operators has been used to calculate ratios of vacuum expectation values of local operators and to predict nontrivial particle-soliton degeneracies. In this paper, we use recently-developed tensor network methods to study several examples of such theories via their Hamiltonian lattice descriptions. Our lattice results agree with all previously-made predictions. Furthermore, we identify the lattice strong-coupling states that can be adiabatically continued to the degenerate vacua in the continuum limit. We conjecture a procedure, referred to as a lattice decay rule, for how this identification works in general. This rule allows us to compute the continuum vacuum degeneracy by studying the lattice Hamiltonian in the strong-coupling limit.

The density of ultralight dark matter can be modified in the vicinity of macroscopic bodies when the dark matter possesses quadratic couplings to the Standard Model. If these couplings are sufficiently strong, Earth's atmosphere acts to shield the dark matter, thereby limiting the effectiveness of laboratory-based experiments. Experiments performed at altitudes exceeding the dark matter de Broglie wavelength experience the same orbit-averaged field amplitude as in the absence of scattering. Quantum clocks are capable of detecting variations in fundamental parameters due to the dark matter background. If based on the International Space Station, they are therefore well-suited to probe dark matter masses $m_{\rm DM}\gtrsim 10^{-9} \text{\, eV}$. Moreover, when the dark matter de Broglie wavelength is smaller than Earth's radius ($m_{\rm DM} \gtrsim 10^{-10}$ eV), the dark matter profile around Earth exhibits a dipole feature. In Low Earth Orbits this dipole temporally modulates potential dark matter signals. This provides a powerful cross-check of the orbit-averaged effect and can enhance the sensitivity of these experiments. We find optical clocks could give rise to world-leading constraints in some cases. Orbiting nuclear clocks could probe even more of the parameter space inaccessible to ground-based experiments.

The time-dependent $CP$ asymmetry in $B^{0}\!\to{J\mskip-3mu/\mskip-2muψ}ρ(770)^0$ decays is measured using proton-proton collision data corresponding to an integrated luminosity of $6\,\text{fb}^{-1}$, collected with the LHCb detector at a center-of-mass energy of $13\,\text{TeV}$ during the years 2015-2018. The $CP$-violation parameters for this process are determined to be $2β^{\rm eff}_{c\bar{c}d} = 0.710 \pm 0.084 \pm 0.028\,\text{rad}$ and $|λ| = 1.019 \pm 0.034 \pm 0.009$, where the first uncertainty is statistical and the second systematic. This constitutes the first observation of time-dependent $CP$ violation in $B$ meson decays to charmonium final states mediated by a $b\!\to{c\bar{c}d}$ transition. These results are consistent with, and two times more precise than, the previous LHCb measurement based on a data sample collected at 7 and $8\,\text{TeV}$ corresponding to an integrated luminosity of $3\,\text{fb}^{-1}$. Assuming approximate SU(3) flavor symmetry, these two measurements are combined to set the most stringent constraint on the enguin contribution, $Δφ_{s}$, to the $CP$-violating phase $φ_{s}$ in $B^{0}_{s}\!\to{J\mskip-3mu/\mskip-2muψ}φ(1020)$ decays, yielding $Δφ_{s} = 5.0 \pm 4.2\,\text{mrad}$.

We examine three exclusive nuclear ground state shell models implemented in the NEUT neutrino event generator and benchmark them against the recent JSNS$^2$ measurement of missing energy using a monoenergetic neutrino source. The nature of the measurement allows a detailed investigation of nuclear ground-state modeling using a neutrino source, and gives access to a direct measurement of the neutron spectral function in a $^{12}$C nucleus. The NEUT intranuclear cascade and nuclear deexcitation \textsc{NucDeEx} are used to simulate inelastic final-state interactions and nuclear deexcitations respectively. We find that the spectral function (SF) models perform better than relativistic mean field models in modeling both the ground state and the tail of the missing energy distribution when the NEUT cascade and nuclear excitation channels are turned on. We also find that taking into account the missing energy threshold for single nucleon knockout interactions results in all nuclear models being accepted based on the obtained $p$-values.

Through a holographic model of QCD, we present a phenomenological approach to study the running of the strong coupling constant α_s in both non-perturbative and perturbative regimes. The renormalization of the metric tensor, driven by the Ricci Flow, and the breaking of conformal and chiral symmetries -- thanks to introducing a double dilaton model and large-$N_c$ corrections -- allow us to relate the existence of an infrared fixed point in the coupling constant with a smooth matching to pQCD well above 2 GeV. This is done through a model with two fit parameters and one matching point. The proposed dilaton model yields linear Regge trajectories and decay constants for scalar, vector, and tensor meson families similar to their experimental counterparts. We finally study neutral and charged pion form factors to show an application of the running coupling constant obtained.

The environs of other stellar systems may be directly probed by analyzing the cometary activity of interstellar objects. The recently discovered interstellar object 3I/ATLAS was the subject of an intensive worldwide follow-up campaign in its pre-perihelion approach. Now, 3I/ATLAS has begun its post-perihelion departure from the Solar System. In this letter, we report the first post-perihelion blue-sensitive integral-field unit spectroscopy of 3I/ATLAS using the Keck Cosmic Web Imager on November 16, 2025. We confirm previously reported CN, Fe, and Ni outgassing along with detections of carbon chain molecules $\mathrm{C}_2$ and $\mathrm{C}_3$. We calculate production rates for each species. We find Fe and Ni production rates of $\mathrm{Q_{Fe}} = (9.55\pm3.96)\times10^{25}$ atoms s$^{-1}$, and $\mathrm{Q_{Ni}} = (6.61\pm2.74)\times10^{25}$ atoms s$^{-1}$, resulting in a ratio of $\log(\mathrm{Q_{Ni}} / \mathrm{Q_{Fe}}) = -0.16\pm0.03$, which matches Solar System comets well and continues the pre-perihelion trend of declining $\log(\mathrm{Q_{Ni}} / \mathrm{Q_{Fe}})$ with $r_h$. We investigate the radial distributions of these elemental species and find characteristic $e$-folding radii of 3880$\pm$39 km for Ni, 6053$\pm$68 km for CN, 4194$\pm$45 km for $\mathrm{C}_2$, and 3833$\pm$45 km for $\mathrm{C}_3$. Compared to pre-perihelion measurements, these radii have increased by a factor of $\sim$6.5--7. Our post-perihelion observations reveal that 3I/ATLAS continues to exhibit cometary behavior broadly consistent with Solar System comets.

We present near- and mid-infrared spectra of eight Low-Luminosity Active Galactic Nuclei (LLAGN), spanning nearly four orders of magnitude in black hole mass and Eddington ratio, obtained with JWST/NIRSpec and MIRI as part of the ReveaLLAGN program along with identical archival data of Cen A. The high spatial resolution of JWST cleanly separates AGN emission from host-galaxy contamination, enabling detections of high-ionization potential lines more than an order of magnitude fainter than previously measured. Emission-line diagnostics reveal a transition at log($L_{bol}/L_{Edd}$) ~ -3.5, where the spectral energy distribution becomes increasingly deficient in ultraviolet photons. We find that rotational H$_2$ excitation temperatures are elevated (~500 K higher) compared to both higher-luminosity AGN and star-forming galaxies, while the H$_2$(0-0)S(3)/PAH$_{11.3 μm}$ ratios are consistent with those observed in the AGN population. We discuss the possible roles of outflows, jets, and X-ray dominated regions in shaping the interstellar medium surrounding LLAGN. Silicate emission at ~10 $μ$m, localized to the nuclear region, is detected in most ReveaLLAGN targets. This dataset offers the first comprehensive JWST-based characterization of infrared emission lines in the nuclear regions of LLAGN.

We explore the feasibility of using Lyman-$α$ (Ly$α$) forests to calibrate the ensemble redshift distribution of the high-redshift tail ($2<z<3$) of photometric galaxies. We use \texttt{CoLoRe} simulations to create mock DESI 5-year Ly$α$ forests and Rubin Observatory LSST 10-year photometric galaxies up to $z=3$, and measure the galaxy redshift distribution via their angular cross-correlations. Due to large redshift-space distortions in the Ly$α$ forest, the conventional $n(z)$ estimator for clustering redshifts does not apply, and we develope a theoretical framework to model the angular cross-correlation directly. Using the simulations, we explore effects of instrumental noise, continuum fitting, and contamination in the Ly$α$ forest, cross-correlation angular scales ($θ$), and redshift bin size ($Δz$) on the signal-to-noise (SNR) of the measurements. We find that continuum fitting methods strongly impact the SNR of the measurements. With our baseline continuum fitting method, \texttt{LyCAN}, at angular scales $θ\sim10$ arcmin and $Δz=0.1$, we measure the cross-correlation signal at $24σ$. If the shape of the redshift distribution and galaxy bias evolution are known well for $z<2$, the cross-correlation can constrain the mean redshift of the galaxy sample to $σ_z/(1+\bar{z}) = 0.006$ at a mean redshift of $\bar{z}=2$. This demonstrates that Ly$α$ cross-correlation is a reliable and promising method to calibrate the high-redshift tails of photometric Stage IV galaxy surveys.

Angular redshift fluctuations (ARF) are a new cosmological observable, recently proposed in the literature. It measures the 2D angular deviations of the average redshift of a given matter tracer under an input redshift shell. Since it depends on the galaxy bias, it can be used to constrain primordial non-Gaussianity through the scale-dependent bias effect. We analyze a sample of quasars built upon the Gaia satellite and unWISE data, Quaia, to measure the local non-Gaussianity parameter $f_{\rm NL}$. This sample is particularly suitable for measuring $f_{\rm NL}$ due to its large volume coverage. We measure the ARF power spectra from the Quaia catalog and combine their information with the 2D (projected) galaxy density and their cross-correlation with the $Planck$ PR4 CMB lensing maps lensing to jointly constrain $f_{\rm NL}$. Assuming the universality relation, we measure $f_{\rm NL} = -3 \pm 14$ at 68% confidence level by combining Quaia quasar angular density and ARF with the CMB lensing. This result is the second tightest constraint on $f_{\rm NL}$ using LSS two-point statistics to date and the best measurement achieved using two-point projected summary statistics, improving by $\sim$25% the previous measurement from Quaia. Our results motivate the inclusion of ARF as an additional cosmological observable in future 2D analysis of upcoming datasets from large surveys.

The Stage~IV \textit{Euclid} mission will deliver spectroscopic galaxy redshifts together with photometric positions and shapes, enabling cosmological analyses through spectroscopic galaxy clustering (GCsp), photometric galaxy clustering (GCph), weak-lensing cosmic shear (WL), and their cross-correlation (XC). In this work we forecast the constraining power of a Euclid-like survey on the Generalised Dark Matter (GDM) parameters \(w_{\rm gdm}\) and \(c^{2}_{s,{\rm gdm}}\). Our analysis extends previous forecasting pipeline used for standard cold dark matter. For GCsp, we adopt a semi-analytic nonlinear RSD model, with free terms for each bin. For the photometric probes, we compute the nonlinear GDM matter power spectrum using dedicated simulations, and we modify the lensing and clustering window and the intrinsic-alignment prescription. We consider several survey configurations and explore three fiducial values of \(σ_8\) motivated by current CMB and low-redshift measurements. In an optimistic setting, for fiducial values \(σ_8 \simeq 0.81\) and \(σ_8 \simeq 0.77\), we find relative errors of \(4.01\%\) (GCsp), \(5.01\%\) (GCph+WL+XC), and \(1.96\%\) (all probes) on \(c^{2}_{s,{\rm gdm}}\), and \(3.26\%\) (GCph+WL+XC) and \(1.85\%\) (all probes) on \(w_{\rm gdm}\). For a lower fiducial value \(σ_8 \simeq 0.67\), that could strongly disfavor $Λ$GDM, we find constraints of \(5\%\) (GCsp), \(5\%\) (GCph+WL+XC), and \(2.45\%\) (all probes) on \(c^{2}_{s,{\rm gdm}}\), and \(3.43\%\) (GCph+WL+XC) and \(2.04\%\) (all probes) on \(w_{\rm gdm}\). We also found that, combining all probes, whether in the pessimistic or optimistic settings, a Euclid-like survey will be able to disentangle between the three scenarios. These results show that the survey will be able to constrain the GDM parameters and distinguish between normalisations of the matter fluctuations.(Abridged)

We present a panoramic view of several scaling relations (ScRs) of galaxies of different morphology. The ScRs are obtained from the data of two large surveys (WINGS and MANGA). We analyze the distribution (parameterized by the percent over the total) of galaxies in each region of the diagnostic planes that are set up by means of suitable physical quantities. In addition to this, we discuss the origin of the differences observed in the ScRs between the two samples. Finally, we compare the observational data with the theoretical ones taken from two subsets of the Illustris large scale simulations (TNG50 and TNG100) and we discuss how the comparison should be performed for a correct statistical answer.

The growth index $γ$ is a powerful trigger for detecting deviations from $Λ$CDM. However, its value is often determined by considering an asymptotic constant value that works for all redshift, or else following a chosen parameterisation. Here we formulate the growth index as function of three quantities that could be directly related to observables in redshift bins, $fσ_8(z_i)$, $f(z_i)$ and $H(z_i)$. We determine its value and its derivative at observed nodal center of redshift bins and use the shape function method, after showing insightful connection with its underlying governing virtual-work conservation principle, to construct a redshift dependence of the $γ$ without assuming a specific parameterization. We then use the resulting shape function to test if we can disentangle between different scenarios where there are discrepancies between its three constituent measured components. We also tested whether it can be used to rule out models of modified gravity, or extended parametric models of the growth index that capture more general behaviors with an additional parameter as function of the scale factor or dark energy. Adopting forecasted measurements from next generation surveys on the three quantities used to construct $γ$, we find that reported discrepancies between them could be detected with our method, but at the bins where the errors and lost of precision from our addition of degrees of freedom is small with respect to the deviation of $γ$. The same could be concluded for first order extensions to $γ$ or common modified gravity models, and to a lesser degree for dynamical dark energy models after supposing the latest DESI values. We conclude that this method is a strong tool to investigate cosmology in a model-independent way especially with forthcoming data delivered by further stage-IV surveys with more stringent uncertainties.(Abridged)

Understanding how cosmological parameters influence the cosmic microwave background (CMB) power spectra is a central component of modern cosmology education, but interactive exploration is often limited by computational cost or technical complexity. We present CosmoSlider, a lightweight visualization tool that enables real-time exploration of CMB power spectra as multiple cosmological parameters are varied simultaneously. The tool employs a neural-network emulator implemented using TensorFlow Lite, allowing rapid evaluation of spectra without relying on large grids of precomputed models or on-demand execution of Einstein--Boltzmann solvers. CosmoSlider is available both as an iOS application and as a web-based tool, making it accessible across platforms and suitable for use in classrooms, lectures, and self-guided study. By providing immediate visual feedback, CosmoSlider supports the development of intuition for the physical processes underlying CMB anisotropies and serves as a complementary resource to traditional theoretical instruction.

Electron-capture supernovae (ECSNe) are commonly thought to result in a collapse to a neutron star. Recent work has shown that a thermonuclear explosion is also a possible outcome. The division between the two regimes has not yet been mapped out. In this study, we investigate the conditions under which the transition from thermonuclear explosion to collapse occurs, and what physical mechanisms drive each outcome. We conducted a parameter study of 56 3D hydrodynamic simulations of ECSN in ONe white dwarfs using a level set based flame model implemented in the Leafs code. We varied both the ignition location and the central density at ignition to determine the conditions of the transition regime. Additionally, we explored two different laminar flame parameterizations and how they impact the simulation outcome. From our parameter study, we find a transition density in the range of $\logρ_c^{ini}=10.0$ and $10.15$ g cm$^{-3}$, depending on the ignition location and utilized laminar flame speed parameterization. Importantly, we find that for sufficiently high central densities, the burned ashes can sink into the core and trap large amounts of neutron-rich material in the bound remnant. In the transition regime between explosion and collapse, we find that the laminar flame speed plays a critical role by suppressing the formation of instabilities and thereby reducing the nuclear energy generation needed to overcome the collapse. We find that a thermonuclear explosion is possible for a wide range of parameters, whereby a more off-center ignition allows for higher central densities to still result in an explosion. Both the conditions at ignition and the flame physics are critical in determining the outcome. Detailed 3D hydrodynamic simulations of the preceding stellar evolution and the ignition process of the thermonuclear flame are necessary to accurately predict the outcome of ECSNe.

As the central galaxy in the nearest cluster, M87 provides the best spatial resolution for disentangling the complex interactions between AGN jets and the surrounding environment. We investigate the velocity structure of the multitemperature X-ray gas in M87, particularly in the eastern and southwestern arms associated with past AGN outbursts, using high-resolution spectroscopy from XRISM/Resolve. We analyze a mosaic of XRISM/Resolve observations covering the core of M87, fitting single- and multi-temperature models to spectra extracted from different regions and energy bands. We assess the line-of-sight velocities and velocity dispersions of the hotter ambient and cooler uplifted gas phases, and evaluate systematic uncertainties related to instrumental gain calibration. The hotter ICM phase, traced by Fe He-$α$ emission, shows velocity dispersions below $\sim100$ km/s, and no significant velocity shifts between the arms and a relaxed offset region, suggesting limited dynamical impact from older AGN lobes. In contrast, the cooler gas phase appears to exhibit larger line of sight velocity gradients up to several hundred km/s as well as a higher velocity dispersion than the ambient hot phase, although these conclusions remain tentative pending improvements in the robustness of the gain calibration at lower energies. The first microcalorimeter-resolved map of gas dynamics in M87 supports the uplift scenario for the X-ray arms, with the cooler gas in the east and southwest seemingly moving in opposite directions along the line of sight. The kinetic energy is a small fraction of the gravitational potential energy associated with the gas uplift, and XRISM further suggests that AGN-driven motions may be short-lived in the hot ambient ICM. These constraints provide important input towards shaping future models of AGN feedback.

We forecast constraints on an effective dark fluid equation of state parameter $w_{\rm eff}$ that encapsulates modified gravity theories that modifies both the Universe background expansion as well as its large scale structures growth. This is achieved through relating Friedmann equations' dark fluid pressure and density content, thus $w_{\rm eff}$, to modified gravity parameterized models by mean of the Newtonian potential equation parameter $μ_0$, the gravitational slip parameter $η_0$ and a redshift dependent Hubble parameter $H_{0,{\rm bck}}$. We adopt next stage SKA survey specifications, alone or in combination with concurrently expected DR3 Euclid survey release, paying attention to the modeling and recipe of the implementation of the galaxy clustering and lensing probes obtained from the two surveys. We consider two data mock models: one with deviation of the intermediate parameters at the level of 10 \% (yielding however $w_{\rm eff}=-1.03$) and another sub-percently close to $Λ$CDM. We found that the three parameters deviation from $Λ$CDM could only be detected at 1 $σ$ from SKA alone, while this improves to $\sim$ 2 $σ$ when we combine with Euclid. An improvement of the order of 30\% on the bounds is reached after projecting the three parameters into a single $w_{\rm eff}$ parameter. However, this affects both cases and thus it does not change much, though it improves the level of detection with respect to $Λ$CDM values. We conclude that synergy from both surveys benefits to tighten our constraints, but also that our highly generalized parameterization, although impacting at both the background and the perturbation level, will be hard to disentangle from $Λ$CDM at the level at which our forecast is performed and it still needs, to the least, data from more advanced stages of the adopted surveys to hope reach this target.

The nuclear equation of state, which determines the structure and properties of neutron stars, remains subject to substantial theoretical uncertainties, leading to model dependence in predicted observables. Universal relations have emerged as a powerful tool to mitigate this dependence by linking neutron star observables in a framework-independent manner. In this work, we introduce a new universal relation that \emph{bridges} finite nuclei and neutron stars through the dimensionless quantity $ζ= β_{1.4}\tilde{L}^{-1}$, which couples the compactness of a $1.4~M_{\odot}$ neutron star to the slope of the nuclear symmetry energy at saturation. The relation is examined under a broad set of relativistic energy density functionals with point-coupling and meson-exchange interactions, as well as non-relativistic Skyrme functionals. We demonstrate that $ζ$ exhibits a strong exponential correlation with the electric dipole polarizability $α_D$ in finite nuclei across all considered equations of state. By exploiting experimental $α_D$ data for selected neutron-rich nuclei, we constrain $ζ$ and translate these constraints into equation-of-state-independent bounds on the neutron star radius $R_{1.4}$ and the symmetry-energy slope $L$, providing insights into the properties of neutron star matter.

The discrepancy of the Hubble parameter H0 as measured from the cosmic microwave background versus that found from traditional distance ladder measurements has produced considerable discussion about the need for another force in cosmology. However the significance of the discrepancy depends on understanding the systematic associated with crowding, metallicity effects, and extinction of the stellar tracers. Thus additional precision distance indicators in the local universe are desperately needed for investigating the H0 tension. The analysis of MUSE archival data makes the case that the Planetary Nebula Luminosity Function (PNLF) has become such an indicator, as the method can reach distances comparable to HST distances of Cepheid at a fraction of a cost, in terms of telescope time and ground-based. With new wide-field spectroscopic facilities it becomes possible to measure distances to early-type galaxies (ETGs) using the PNLF out to 100 Mpc distance, achieving a precise estimate for the H0 value which is independent of the Type Ia supernova calibration, with only single-epoch measurements.

It has been suggested since recent time that the magnitude of the interaction between galaxies could be measured from the level of kinematic disturbance of their outer regions with respect to the innermost ones. Here, I proved that the outer north-eastern region of the Small Magellanic Cloud (SMC), a relatively recent stellar structure with a tidal origin from the interaction with the Large Magellanic Cloud, is imprinted by a residual velocity pattern. I obtained from GEMINI GMOS spectra mean radial velocities of star clusters formed in situ, which added to derived mean proper motions and heliocentric distances, allowed to compute their 3D space velocity components. These space velocities differentiate from those that the clusters would have if they instead orderly rotated with the galaxy, i.e., their residual velocities are larger than the upper limit for an object pertaining to the SMC main body rotation disk. The level of kinematic disturbance depends on the SMC rotation disk adopted; galaxy rotation disks traced using relatively old objects are discouraged.The resulting kinematic disturbance arises in younger and older stellar populations, so that the epoch of close interaction between both Magellanic Clouds cannot be uncovered on the basis of the kinematics behavior of stellar populations populating the outer SMC

Context. Among known young stellar objects (YSOs), those exhibiting the most dramatic increases in brightness due to sudden increase in mass accretion rate are eruptive young stars. Gaia20dsk is one of the Gaia-alerted young star candidates that has displayed a double, nonperiodic brightening resembling that of other young eruptive stars. Aims. The goal of this work is to determine the physical and accretion properties of Gaia20dsk to confirm its classification as an eruptive young star. Methods. We combined publicly available optical and near-infrared (NIR) photometry with our X-shooter optical/NIR spectrum. In our analysis, we examined the optical and IR light curves from the bursts, reviewing the color-magnitude diagrams across different bands, reporting the detection of emission lines, and providing estimates of the star's accretion rates during the burst. Results. The optical light curve shows two major and one brief brightening events with a maximum amplitude of ~1.8 mag in the last five years. A classification based on spectral index indicates that Gaia20dsk is a flat-spectrum star. The X-shooter spectrum exhibit emission lines characteristic of accreting low-to-intermediate-mass young stars, displaying features typical of MNor-type objects. The mass accretion rate is between (0.5-1.8)*10^{-6} M_sun/yr. Conclusions. Gaia20dsk is an eruptive YSO that exhibits photometric features similar to those of MNors, including its characteristic brightening amplitude and burst duration, along with similar spectroscopic features and accretion rates.

Strongly lensed quasars in cusp configurations provide a uniquely sensitive probe of small-scale dark matter structure. Using the largest microlensing-free flux ratios for 17 quadruply imaged cusps, we combine these with extensive Monte Carlo simulations of mock lens realizations under cold dark matter (CDM), self-interacting dark matter (SIDM), and fuzzy dark matter (FDM) scenarios. Building on this, we propose a region (minor-axis and narrow major-axis cusp lenses) where flux-ratio anomalies persist even under globally parameterized models ("macromodels") with multipole freedom (capturing disk, asymmetric, or merger-driven structures). Within this region, J1042+1641 is $>3σ$ incompatible with both CDM and SIDM. Our results yield a Bayes factor exceeding $100$, providing very strong evidence for FDM over even the most optimistic CDM and SIDM scenarios. As only 11 cusp lenses lie within this region, extending to larger samples will be essential for assessing its statistical generality and for decisively confirming these findings with future microlensing-free flux ratio data.

During the star formation process, the interplay between gravity, turbulence, and B-fields is significant, with B-fields apparently serving a regulatory function. However, the extent to which B-fields are decisive relative to turbulence and gravity remains uncertain. This study aims to ascertain the role of B-fields in the fragmentation of molecular clouds. We examine the B-field observed with ALMA at core scales towards the infrared dark cloud G14.225-0.506, focusing on 3 regions with shared physical conditions, and juxtapose it with prior observations at the Hub-filament system scale. Our findings indicate a similar B-field strength and fragmentation level between the 2 hubs. However, distinct B-field morphologies are identified across the 3 regions where polarized emission is detected. In the region N, the large-scale B-field, which is perpendicular to the filamentary structure, persists at smaller scales in the southern half but becomes distorted near the more massive condensations in the northern half. Notably, these condensations exhibit signs of impending collapse, as evidenced by supercritical mass-to-flux values. In the region S, the B-field is considerably inhomogeneous among the detected condensations, and we do not observe a direct correlation between the field morphology and the condensation density. Lastly, in an isolated dust clump located within a southern filament of the northern hub, the B-field aligns parallel to the elongated emission, suggesting a transition in the field geometry. The B-field shows a clear evolution with spatial scales. We propose that the most massive condensations detected in the northern Hub are undergoing gravitational collapse, as revealed by the relative significance of the magnetic field and gravitational potential and mass-to-flux ratio. The distortion of the B-field could be a response to the flow of material due to the collapse.

[HP99] 159 is remarkable as the first supersoft X-ray source (SSS) identified with an evolved helium star donor. With a likely orbital period of 1.164 d or 2.327 d, the origin of the SSS component is controversial, with the two current models being either steady He-burning on the white dwarf surface, or that it is a helium nova in the decaying phase. To help resolve this issue we present extensive new long-term spectroscopy (with SALT) and photometry (at SAAO and with OGLE) of [HP99] 159 which (a) supports 2.327 d as the orbital period, and (b) finds only a small He II radial velocity modulation. The latter is surprising as it implies a very low inclination system, whereas our light curve modelling suggests $i{\sim}50^\circ$, and hence that the He II must be produced in outflowing material further above, or beyond, the disc. We find that the decaying nova model cannot fit our OGLE light curve and the observed SSS flux level. [HP99] 159 has been essentially constant as an SSS over several decades, implying a sustained high level of mass-transfer from its He star donor, making it the only confirmed single-degenerate scenario SN Ia progenitor. We have updated the known SSS binary parameters and find a clear $\sim$1.5 mag difference in their $M_{\rm V}$ when compared to the $M_{\rm V} - Σ$ properties of LMXBs, likely due to the larger irradiated areas and more luminous donors.

From James Webb Space Telescope (JWST) surveys, 31 galaxies with average redshift 7.3 are selected containing large Balmer break, Lyman-$α$ break (V-shaped SED versus $λ$). Apart from Hubble Space Telescope (HST) and JWST-NIRCam (Near-infrared camera) photometry for these galaxies, there are JWST-NIRSpec (Near-infrared spectrograph) spectra for 13 galaxies and mid-infrared photometry (mostly JWST-MIRI) for 15 of them. Spectroscopical analyses included Balmer emission lines, Balmer + 4000 angstroms breaks or CaII lines. Spectral energy distribution (SED) fitting with photometry include old and young stellar populations, emission lines associated to HII regions, AGN, interstellar dust extinction and intergalactic extinction from neutral hydrogen. By adopting realistic extinction curves and taking into account the V-shaped SED and low emission at near infrared at rest, the analyses show that AGN contribution in these galaxies ('little red dots' most of them) should be small on average in the reddest wavelengths, though important for few of the 31 galaxies. Average age of the 31 galaxies: $0.61\pm 0.31$(95% CL) Gyr, while the average age of the $Λ$CDM universe is 0.70 Gyr. This corresponds to a formation epoch $z_{ form.}>11.2$(97.5% CL). Reddest galaxies present largest ages. One of these very red galaxies gets an age incompatible to be younger than the age of the Universe within $>4.7σ$. TP-AGB effect cannot explain this tension. None the less, there may be other uncertainties in the models, so this tension is a provisional result and further research is needed to confirm it.

Understanding the Universe's origins and evolution remains one of the most fundamental challenges in modern cosmology. This white paper explores three key science priorities in this field: unravelling the physics of cosmic inflation, investigating the accelerating expansion of the Universe, and precisely measuring the sum of the neutrino masses. Achieving these goals requires a dedicated survey to map the large-scale structure at high redshift in unprecedented detail. We describe how this can be achieved through a mission concept called SIRMOS, providing a high-throughput, highly multiplexed spectroscopic capability to obtain accurate redshifts for over 100 million galaxies over a wide sky area. Such a survey would leverage the deepest existing wide-area photometric catalogues for targeting, with spectra offering continuous 1.25-2.5~$μ$m wavelength coverage at moderate resolution, allowing precise redshift measurements in the $1<z<4$ range with minimal bias. We outline the scientific opportunities this presents. Recent years have seen significant advances in instrumentation, including digital micromirror devices, complex telescope mirrors, large detector arrays, and data processing pipelines. While these technologies have been demonstrated in terrestrial applications, such a survey is a unique opportunity to apply these proven capabilities in space to address fundamental questions in cosmology. Participation in such a mission will simultaneously deliver a compelling science case, help align UK Space Agency and STFC strategies, demonstrate the UK's growing capability in end-to-end space missions, and strengthen the national space economy through high-value industrial participation.

A search for pulsars was carried out using a Large Phased Array (LPA) radio telescope at a frequency of 110.4 MHz with a time resolution of 3.072 ms and a frequency resolution of 19.5 kHz with a 2.5 MHz bandwidth used. The survey was conducted in a site with declinations of $+53^\circ < δ< +87^\circ$. The viewing area is approximately 4100 sq.deg. The search was carried out using Fourier power spectra. To increase sensitivity, multiple observations were made in each direction in the sky, and the resulting power spectra were summarized. This made it possible to increase sensitivity by about 5-10 times, depending on the direction in the sky. A blind search opened 35 known pulsars. Estimates of the flux density for 33 pulsars have been obtained.

We present the detection and analysis of H2 absorption at z = 4.24 towards the bright quasar J0007-5705, observed with the Very Large Telescope as part of the ESPRESSO QUasar Absorption Line Survey (EQUALS). The high resolving power, R~120000, enables the identification of extremely weak H2 lines in several rotational levels at a total column density of N(H2)~2x10^14 cm^-2, among the lowest ever measured in quasar absorption systems. Remarkably, this constitutes the highest-redshift H2 detection to date. Two velocity components are resolved, separated by only 3 km/s: a narrow (b~1.7 km/s) and a broader (b~6.2 km/s) component. Modelling the rotational population of H2 yields density of log nH/cm^-3 ~ 2.8 with temperature of ~40K (typical of the cold neutral medium) for the narrow component and log nH/cm^-3 ~ 1.4 , T~600K for the warmer, more turbulent component under a moderate ultraviolet (UV) field, suggesting at least several Mpc distance from the quasar. This system reveals the existence of tiny (down to ~0.01 pc), cold overdensities in the neutral medium. Their detection among only 7 damped Lyman-alpha systems in EQUALS suggests that they may be widespread yet usually remain undetected. H2 provides an exceptionally sensitive probe of these structures: even a minute molecular fraction produces measurable Lyman-Werner absorption lines along the extremely narrow optical beam -- the size of the quasar's accretion disc -- when observed at sufficiently high spectral resolution. High-resolution spectroscopy on extremely large telescopes may routinely detect and resolve such structures in the distant Universe, when 21-cm absorption will trace the collective contribution of many cold cloudlets toward larger radio background sources.

Twin stars are predicted to exist in nature if the hadron-to-quark phase transition is strong enough to form a new branch of hybrid stars, separated from the branch of neutron stars. We adopt an agnostic approach, using transition energy density, transition pressure, the discontinuity strength, and a constant speed of sound for quark matter as our parameter space to construct a large possibility of hybrid equations of state, and thereby encapsulating a comprehensive picture of the twin star scenario. First, we report the complete conditions on our parameter space imposed by the general relativistic hydrostatic equilibrium solutions. For a fixed transition energy density and speed of sound for quark matter, we define distinct ranges of transition pressures based on the allowed strengths of discontinuity. Below a maximum transition pressure, a range of discontinuity exists that increases as the transition pressure decreases. Thereby, we identify the loci of the limits on discontinuities as the `witch-hat' curves. Based on the causality limit, the witch-hat curves can be punctured or incomplete. Strong constraints on this picture are drawn from the inferences from GW170817 and the NICER measurements. We computed the maximum mass for twin stars to be $2.05~M_\odot$, the allowed strongest discontinuity in rest-mass density to be $7.76ρ_\mathrm{sat}$, and the upper bound on transition rest-mass density to be $4.03ρ_\mathrm{sat}$. Subsequently, we compute the implications of the stiffness of the quark matter equation of state on this picture. Different confidence levels for observational inferences are considered to assess the extent of inclusion (and rejection) of hybrid equations of state and, consequently, their effects on the limits of the maximum mass of twin stars and phase transition properties.

Since viable $f(R)$ gravity models must reconcile early-universe inflation with late-time acceleration, we specifically study the dynamical behavior of such a theory during the matter-dominated to dark-energy-dominated transition epoch. By using $y_{H}(z)$ versus $z$ and the Hubble parameter, we solved the field equations. After appropriately choosing appropriate parameter values , we plotted a series of images. We mentioned that their current values are similar to latest observations data and $Λ$CDM-model values. Furthermore, we plotted the fitting of the distance modulus about this model using SN Ia observation data. Therefore we find that the $f(R)$ gravity model is consistent with the SN Ia data, meanwhile, explains the late-stage acceleration of the Universe. Finally, we used various diagnostic tools including $( r, s)$, $( r, q)$, $w_{D}-w'_{D}$ plane, growth rate analysis, statefinder hierarchy and $Om(z)$-diagnostic to evaluate the observational viability of our model, we perform a systematic comparison with the standard $Λ$CDM. We found that evolutionary images can be clearly distinguished this model from the $Λ$CDM.

Results from GRBAlpha, VZLUSAT-2 and GRBBeta CubeSats and their on-board gamma-ray detectors for monitoring transients are summarised in this article. GRBAlpha was a 1U CubeSat launched in March 2021 to a 550 km altitude polar orbit carrying a CsI(Tl) scintillator gamma-ray burst (GRB) detector with a sensitive range of approximately 30-900 keV. It successfully operated for over four years until June 2025 when it de-orbited. VZLUSAT-2 was a 3U CubeSat launched in January 2022 to a 535 km altitude polar orbit and de-orbited in November 2025 after almost four years of smooth operation. It carried on board two GRB detectors very similar to the one used on GRBAlpha. Both missions have detected about 360 gamma-ray transients, including over 170 long and short gamma-ray bursts (GRBs), and including the most intense GRB ever recorded GRB 221009A and the second brightest GRB 230307A. The new family member, GRBBeta 2U CubeSat, integrated at Masaryk University, was launched in July 2024 to a 580 km altitude, 62 degree inclination orbit. It has been detecting GRBs since its launch without any trouble. Gamma-ray detectors on these nanosatellites are based on CsI(Tl) scintillator readout by silicon photomultipliers (SiPMs). These missions also provide a unique opportunity to study the radiation damage of SiPMs in the low Earth orbit environment and monitor the radiation belts. We have demonstrated that CubeSats can be used in missions lasting beyond three years and routinely detect GRBs.

Detection of stellar flares at hard X-ray is still rare at the current stage. A transient was recently detected by the hard X-ray camera, ECLAIRs onboard the SVOM mission at 11:39:01.2UT on 2025, January 09. Simultaneous monitor in the optical band on the ground by SVOM/GWAC and follow-up spectroscopy enable us to confirm that the transient is caused by a superflare on HD~22468, a RS CVn-type star. The bolometric energy released in the flare is estimated to be $\sim7.2\times10^{37}-1.7\times10^{38}\ \mathrm{erg}$. The hard X-ray spectra of the event at the peak can be reproduced by the ``apec'' model of a hot plasma with a temperature of $106^{+27}_{-22}$~MK. In the optical range, the H$α$ emission-line profile obtained at $\sim1.7$ hrs after the trigger shows a bulk blueshift of $-96\pm20\ \mathrm{km\ s^{-1}}$, which can be explained by either a chromospheric evaporation or a prominence eruption. The ejected mass is estimated to be $3.9\times10^{20}$ g for the evaporating plasma, and to be $3.2\times10^{21}\ \mathrm{g}<M_{\mathrm{p}}<8.8\times10^{21}\ \mathrm{g}$ for the erupted prominence.

An effective operator is exactly equivalent to the long-wavelength form of the $M1$ operator in transition matrix elements. It allows us to analytically and numerically analyze the $M1$ contribution to the $α(d,γ)^6$Li reaction. Isoscalar $M1$ transitions from an initial $S$ wave are shown to be forbidden in radiative capture reactions when distortion is neglected in the initial state. A calculation in a three-body model with proton, neutron, and a structureless $α$ interacting through effective forces leads to a negligible $M1$ $S$-factor at small energies. The dominant $M1$ contribution comes from transitions from an initial $S$ wave to isospin 1 components of the $^6$Li ground state. It is suggested that using this effective $M1$ operator in other models should clarify the origin of large discrepancies between $M1$ $S$-factors appearing in the literature.

We introduce a new photometric catalog of RR Lyrae variables (RRLs, $\sim$300,000) mainly based on data available in public datasets. We also present the largest and most homogeneous spectroscopic dataset of RRLs and Blue Horizontal Branch [BHB] stars ever collected. This includes radial velocity measurements ($\sim$16,000) and iron abundances ($Δ$S method for 8,140 RRLs, plus 547 from literature). Elemental abundances based on high-resolution spectra are provided for 487 RRLs and 64 BHB stars. We identified candidate RRLs associated to the main Galactic components and their iron distribution function (IDF) becomes more metal-rich when moving from the Halo ([Fe/H]=-1.56) to the Thick (TCD; [Fe/H]=-1.47) and Thin (TND; [Fe/H]=-0.73) disk. Furthermore, Halo RRLs and RRLs in retrograde orbits are $α$-enhanced ([$α$/Fe]=0.27, $σ$=0.18), while TCD RRLs are either $α$-enhanced ([Fe/H]$\le$-1.0) or $α$-poor ([Fe/H]$>$-1.0), and TND RRLs are mainly $α$-poor ([$α$/Fe]=-0.01, $σ$=0.20). We also identified RRLs associated to the main stellar streams (Gaia-Sausage-Enceladus [GSE]; Sequoia, Helmi, Sagittarius) and we found that their IDFs are quite similar to Halo RRLs. However, GSE RRLs lack the metal-poor/metal-rich tails and their $α$-element distribution is quite compact. The iron radial gradient in Galactocentric distance for TND, TCD and Halo RRLs is negative and it decreases from -0.026, to -0.010, and to -0.002 dex/kpc. The iron radial gradient based on dry Halo (Halo without substructures) RRLs is, within the errors, equal to the global Halo. We also found a strong similarity between iron and [$α$/Fe] radial gradients of Milky Way RRLs and M31 globular clusters throughout the full range of galactocentric distances covered by the two samples.

The circumgalactic medium (CGM) is a multi-phase, dynamic interface between galaxy and the intergalactic medium, providing crucial diagnostics of galaxy evolution. However, direct evidence for a hot (million-Kelvin) CGM around present-day L* galaxies remains elusive. Here, we present the first systematic search of the hot CGM around nearby (< 50 Mpc) L* galaxies, by stacking their X-ray images and spectra from the SRG/eROSITA all-sky survey. Significant diffuse X-ray emission is detected out to ~ 50 kpc, with spectral signatures consistent with a hot gas but arguing against a predominantly non-thermal origin. The radial distribution and total amount of the hot gas are in agreement with prediction by IllustrisTNG simulations. The constraints on the hot CGM derived in this study hold promise for calibrating key physical processes in next-generation cosmological simulations.

Galaxies undergo perturbations, either gravitational or hydrodynamic in origin, which can generate extragalactic structures such as rings and tails, where in situ star formation may take place. We selected a sample consisting of JO201 and JW100, undergoing ram-pressure stripping, and NGC 5291 and NGC 7252, formed through gravitational interactions, to investigate how different perturbation mechanisms influence dust content and star formation in extragalactic features. In both cases, star formation can be observed outside the main disks of the galaxies. We present new results of dust attenuation for JO201 and JW100, while for NGC 5291 and NGC 7252 we use results from our previous study, based on high-resolution observations obtained with the Ultraviolet Imaging Telescope onboard AstroSat. Dust attenuation is determined from the ultraviolet continuum slope ($β$) calculated using the FUV-NUV colour, and the star formation rates of the star-forming knots are corrected accordingly. It is seen that dust attenuation and dust-corrected SFR densities of the knots in the ram-pressure stripped tails of JO201 and JW100 are comparable to those in the collisional ring of the NGC 5291 system and the tidal tails of the NGC 7252 system. We conclude that, though the formation scenarios of the tails of JO201 and JW100, the NGC 5291 ring, and the NGC 7252 tails are different, their dust content and star formation activity are notably similar.

Filaments are crucial components of the cosmic web, representing the extensive and aligned distributions of galaxies and gas. Using the Five-hundred-meter Aperture Spherical radio Telescope (FAST), we report the detection of a filament in the Ursa Major supergroup using atomic-hydrogen (HI) observations. This filament consists of sixteen various types of galaxies and five starless gas clumps, spanning a length of approximately 0.9 Mpc. Notably, it is extremely thin, with a thickness comparable to the diameter of a galaxy. We observed a galaxy-filament spin alignment and a velocity gradient within the filament. These findings strongly suggest a cold accretion flow along the filament, potentially contributing to the formation and growth of the galaxies. The thin filament, as a small group, is likely to be merged into the Ursa Major supergroup in the context of hierarchical structure formation.

We compute the density and velocity profiles along the tail induced by a body of mass $M$, embedded in the midplane of a vertically-stratified media with scaleheight $H$, adopting a one-dimensional model as in the Bondi-Hoyle-Lyttleton problem. In analogy to what occurs in the case of a homogeneous medium, there exist a family of solutions that satisfy the boundary conditions. A shooting method is employed to isolate those solutions that fulfill a specific set of physical and mathematical constraints. The tail is found to be both densest and slowest when the scaleheight $H$ is equal to the gravitational radius $ξ_{0}\equiv GM/v_{0}^{2}$, where $v_{0}$ its relative velocity with respect to the medium. The location of the stagnation point is evaluated as a function of $H$ and $ξ_{0}$, and an empirical fitting formula is provided. While the distance to the stagnation point is maximized when $H\simeq ξ_{0}$, the mass accretion rate attains its maximum value for $H \ll ξ_{0}$ at fixed surface density. When instead the midplane density is held constant and $H$ is varied, the accretion rate hardly changes once $H$ exceeds about $2ξ_{0}$. Additionally, we investigate how both the drag force resulting from mass accretion and the gravitational drag arising from its tail depend on $H/ξ_{0}$. We highlight how the effect of varying the degree of mixing in the tail influences the resulting drag force. Finally, for the particular case of an infinitely thin layer, we provide a simple analytical solution, which may serve as a useful pedagogical reference.

We present results from XMM-Newton observations of ten high-redshift ($0.81 < z < 1.17$) galaxy clusters selected from the CAMIRA catalog based on high richness ($N > 40$). These massive clusters, identified in the Hyper Suprime-Cam Subaru Strategic Program field, provide an ideal sample for probing the dynamical state of the intracluster medium (ICM) in the early Universe. We performed uniform X-ray imaging and spectral analyses to measure the ICM temperature and bolometric luminosity, and investigated cluster morphology through offsets between the brightest cluster galaxy (BCG) and the X-ray peak. Extended X-ray emission was detected from all targets, but only one system was classified as dynamically relaxed, indicating a low relaxed fraction ($\sim 10\%$) at high redshift. By combining this high-$z$ sample with a lower-redshift CAMIRA cluster sample, we derived scaling relations among richness, temperature, luminosity, and mass. The results are broadly consistent with predictions from both the self-similar model and the baseline model incorporating the mass--concentration relation. We find no significant redshift evolution, strengthening the view that cluster scaling relations are largely established by $z \sim 1$. We also examined the AGN fraction among member galaxies and found significantly higher AGN activity in high-redshift clusters, particularly in the outskirts, suggesting enhanced AGN triggering during early cluster assembly and a possible connection to the thermodynamic state of dynamically young clusters. These findings provide new insights into the formation and evolution of massive clusters and the thermodynamic history of the ICM, and complement large-area X-ray surveys such as eROSITA.

Tilted accretion disks in the magnetically arrested (MAD) state may be present in X-ray binaries and active galactic nuclei such as Sgr A* and M87. We have carried out 3D global GRMHD simulations to study the evolution of these accretion flows as a function of black hole spin and misalignment angle. Prograde MADs align with the spin through a two-stage process: an initial rapid alignment phase that operates on the magnetic flux saturation timescale, followed by a slower, spin-independent phase. In contrast, retrograde MADs remain persistently misaligned regardless of the black hole spin, displaying solid-body precession at rates four times higher than weakly magnetized flows at the same spin magnitude. By deriving torque equations in ideal GRMHD and evaluating them in a frame aligned with instantaneous disk orientation, we demonstrate that electromagnetic (EM) torques always act to align the disk with the BH spin, but are countered by opposing hydrodynamic fluxes in retrograde flows. We further develop a preliminary empirical model to explain the cause of two-stage prograde alignment and discuss the possibility of alignment in the retrograde MAD. Strongly magnetized, retrograde, misaligned accretion disks provide a candidate scenario for the low-frequency quasi-periodic oscillations in black hole X-ray binaries.

Galaxies and galaxy clusters trace the same cosmic density field, but their statistics have been modeled separately in cosmological analyses. We present a unified, simulation-based framework to model them using the galaxy-halo connection. Our analysis includes cluster lensing, galaxy clustering, and galaxy-cluster cross-correlation. We validate our method on the FLAMINGO hydrodynamic simulation. Relative to the cluster-only approach, combining these probes improves the $σ_8-Ω_m$ figure of merit by a factor of 15. Our framework enables stringent tests of cosmological models and exploits small-scale information.

Modern surveys such as Euclid report a decline in the fraction of barred galaxies from the local Universe to $z \sim 1$, whereas the TNG50 simulation predicts higher bar fractions, in tension with observations. This discrepancy may be due to observational biases in bar detectability when comparing simulations with observations. We present a proof-of-concept study quantifying how Euclid-like observational conditions affect bar detectability in TNG50. We analysed the entire galaxy sample at $z = 0.5$ and highlight one borderline case with a bar length of 2.1 kpc and bar strength $A_2 = 0.4$. Synthetic images were produced with Monte Carlo radiative transfer and realistic post-processing, and analysed with ellipse fitting and Fourier decomposition, as well as the recently constructed Zoobot analysis. Results were compared to idealised, noise-free stellar mass maps. In the illustrative case the bar is clearly detected in the mass map and remains visible in the Euclid VIS $I_{\rm E}$ filter, where Zoobot also classifies it as barred, but becomes undetectable in $Y_{\rm E}$ and in the VIS-NISP RGB composite, with all methods failing outside VIS. Extending to the full $z = 0.5$ sample, Zoobot recovers only 31/141 galaxies, while $A_2$ and ellipse fitting perform better (80/141 and 67/141) but still miss many short or weak bars. When non-detections are counted as unbarred, the bar fraction of 44 percent falls to $12\!-\!33$ percent depending on the method. These results demonstrate the strong impact of observational effects on bar detectability and motivate bar-fraction estimates which incorporate realistic instrumental conditions across redshift in cosmological simulations.

Giant radio haloes are diffuse synchrotron sources typically found in merging galaxy clusters, while smaller mini-haloes occur in cool-core clusters. Both trace cosmic-ray electrons in the intracluster medium, though recent observations suggest their distinction is not always clear. We present new 903-1655 MHz MeerKAT observations of Abell 1775 and Abell 1795, both hosting cool cores and cold fronts. Combined with reprocessed 120-168 MHz LOFAR Two-metre Sky Survey data, we perform imaging and spectral analyses of their radio emission. In both clusters, we detect radio haloes with distinct inner and outer components. In Abell 1775, the halo appears diffuse at 1.3 GHz, while LOFAR images reveal steep-spectrum filaments. In Abell 1795, the inner component corresponds to a previously reported mini-halo candidate, but the full structure extends to $\sim$1 Mpc with a spectral index of $α=-1.08\pm0.06$. The presence of such a large, flat-spectrum halo in a dynamically relaxed cluster makes Abell 1795 an outlier relative to typical merging systems. This suggests that some relaxed clusters may still retain sufficient turbulence to sustain particle re-acceleration, or that hadronic interactions producing secondary electrons play a significant role. Together with other recent discoveries in cool-core systems, our results indicate that some large radio haloes may have been overlooked in past studies due to limited dynamic range near bright central AGN. Finally, we detect steep-spectrum emission south of Abell 1795's central AGN, tracing a 45 kpc X-ray and optical filament that terminates in an X-ray cavity, likely linked to a past AGN outburst.

Star formation timescales are key to understanding fundamental physics like feedback mechanisms, as well as the abundance of bright galaxies at $z>10$. We investigate galaxy star formation histories (SFHs) and their evolution across $z\sim3$--9 by measuring the line-to-UV ratio (\rline) and line equivalent width (EW) of \hanii\ and \oiiihb\ directly from UNCOVER/MegaScience spectro-photometry without relying on a specific SFH or nebular line modeling. Our photometric measurements recover \rline\ and EW to $<10\%$ systematic accuracy compared to spectroscopy. This allows us to construct a large mass- (and flux-) complete sample and quantitatively examine how \rline\ evolves with redshift and stellar mass. We find that the intrinsic scatter in \rline\ does not significantly evolve with redshift across $3<z<7$, though it may increase at $z\gtrsim8$. We build population-level toy models using \texttt{fsps} to help interpret our observations, and find that scatter in \rline\ primarily reflects the amplitude of SFH fluctuations; this implies that our observed lack of evolution in the scatter of \rline\ is due to similar star formation burstiness from $z\sim3$ to $z\sim7$. Our observations are best reproduced by a set of SFHs with rising, long-duration, and large-amplitude bursts. Finally, we demonstrate that the toy model that best describes our $z\sim6$ data can boost UV brightness by up to $ΔM_{\rm UV}\sim-2.0\,{\rm mag}$ compared with a 200\,Myr constant SFH, and naturally produces a large number of galaxies at $z>10$. This suggests that no significant evolution in star formation burstiness is required to explain the abundance of UV-bright galaxies at high redshift.

We present the first sample of 222 high-redshift (z>0.5) star clusters, detected with JWST/NIRCam in 78 magnified galaxies from different galaxy cluster fields. The majority of the systems (~60%) is observed in the very deep NIRCam observations of the cluster AbellS1063 (GLIMPSE program), showing the power that deep observations, combined with lensing, has to reveal these primordial stellar structures. We perform simultaneous size-flux estimates in all available NIRCam filters and spectral energy distribution (SED) fitting analysis to recover star cluster physical properties. All star cluster candidates have very high magnification. Star clusters and clumps show similar ages and redshift distributions, although noticeable differences are seen in their masses, sizes and stellar surface densities inherent to the lack of resolution in the latter group. We reconstruct the formation redshift of star clusters and find that the large majority of the observed star clusters show young ages (<100 Myr) and seems to form at cosmic noon (CN,1<z<4). A small sample of CN star clusters is about 1 Gyr old, these potential globular clusters have formed well within cosmic reionization. Star clusters have stellar densities in the range 10^2 to 10^6 M/pc^2, with median values around 10^4 pc2. Their sizes and densities better overlap with those of nuclear star clusters in the local Universe. These intrinsic properties make high-z star clusters a viable channel to grow intermediate mass black holes. We use Bayesian inference to make first direct measurement of the star cluster mass function at z>1, based on a subsample of 60 star clusters younger than 100 Myr and with masses above 2e6 Msun. The star cluster mass function is well described by a power-law with slope beta = -1.89 suggesting that a power-law -2 function might already be in place in the distant Universe.

We present constraints on the sum of neutrino masses $\sum m_ν$ from a dataset incorporating the full SPT-3G 2018 TT/TE/EE+lensing spectra together with Planck PR4 lensing and low-$\ell$ parts of the Planck PR3 spectra. Using it as a baseline for the DESI DR2 BAO measurements, we arrive at a $95\%$ upper limit of $\sum m_ν< 0.11$ eV, relaxing the tension between $\rm Λ$CDM and lower bounds on $\sum m_ν$ based on neutrino oscillation experiments. When including DES Y1 weak lensing information and the Pantheon+ SNIa catalog, the limit is further loosened to $\sum m_ν<0.138$ eV with a slight preference for $\sum m_ν>0$. On contrast, replacing SPT-3G 2018 primary CMB and lensing spectra with ones from the SPT-3G 2019-2020 (D1) release tightens the overall constraint to $<0.082$ eV and pushes the $\sum m_ν$ posterior mode value to zero, indicating a preference for quasi-negative neutrino masses in line with the D1 analysis. This is a curious shift within SPT-3G measurements of the same field taken in 2018 and in 2019-2020 and processed with different analysis pipelines.

proto-Lightspeed is a new instrument that has been commissioned on the Nasmyth East port of the Magellan Clay Telescope at Las Campanas Observatory to deliver high-speed optical imaging with deep sub-electron read noise. Making use of commercial re-imaging lenses and the ORCA-Quest 2 camera from Hamamatsu, proto-Lightspeed images a field $1'$ in diameter at up to $200$ Hz or windowed fields at higher rates, up to 6600 Hz for a $1.6''\times 1'$ field of view. proto-Lightspeed delivers seeing-limited image quality in the $g'$, $r'$, and $i'$ bands and adjustable magnification for pixel scales between $0.017''-0.050''$. proto-Lightspeed is well suited to studying compact binary systems, exoplanet transits, rapid flaring associated with accretion, periodic optical emission from pulsars, occultations of background stars by small trans-Neptunian Objects, and any other rapidly variable source. proto-Lightspeed will be a P.I. instrument beginning in 2026B, available for use by members of the Magellan Consortium. In this paper, we discuss the design and performance of the instrument, results from its two commissioning runs, and plans for a facility instrument, Lightspeed, to support simultaneous multicolor imaging across a $7'\times4'$ field.

The Event Horizon Telescope (EHT) Collaboration produced the first image of the apparent shadow of the central black hole of Sagittarius\,A$^*$ (\sgra). \sgra source structure varies significantly on timescales shorter than the duration of an observation, preventing improved data coverage through Earth rotation aperture synthesis. This rapid variability provides the opportunity to quantify intrinsic variability and separate time-variable emission features from stable signatures of strong gravity and the accretion environment. To infer the properties \sgra and its surrounding accretion flow, we perform Bayesian inference on a series of EHT data segments (``snapshots''). We directly fit parameters of a semi-analytic emission model jointly with complex station gains to snapshot visibilities, then extract estimates of the time-averaged, persistent source structure and temporal variability by stacking snapshots in a Bayesian hierarchical model. This approach successfully reproduces parameters of General Relativistic Magnetohydrodynamics simulations using synthetic EHT observations. Even with physically motivated assumptions about the \sgra environment, black hole spin and magnetic field parameters are poorly constrained by 2017 EHT observations. Our inference constrains other parameters, favoring a nearly face-on observer inclination ($θ_{\rm o} = 9.2\degree \pm 3.6 \degree \pm_{\rm v} 11.6\degree$), an emission peak near the horizon ($R_{\rm peak} = 4.9 \pm 0.1 \pm_{\rm v} 0.5\,GM/c^2$), near-vertical projected spin position angle ($p.a. = 7.3\degree \pm 7.08 \degree \pm_{\rm v} 43.5\degree$ counterclockwise from vertical), and dominant emission $43.4\degree \pm 2.0\degree \pm_{\rm v} 5.9\degree$ above the equatorial plane, where we separate average structure uncertainty ($\pm$) from the impacts of temporal variability and model misspecification ($\pm_{\rm v}$).

We present a binary-based reassessment of the age of the intermediate-age open cluster NGC 7789, together with well-constrained stellar parameters for twelve components in six SB2 systems, including two eclipsing binaries. Our analysis employs a unified modelling framework that combines radial-velocity orbits, TESS light curves, and blue-to-IR spectral energy distributions (SEDs), providing a robust alternative to traditional isochrone-based age determinations. By adopting common cluster-wide parameters (age, distance, and line-of-sight extinction) when solving for the stellar parameters of the binary components, we obtain a coherent set of masses, radii, effective temperatures, and luminosities for all twelve stars. The combined SED, eclipsing-binary, and radial-velocity analysis yields a well-constrained cluster age of $1.26 \pm 0.09$ Gyr and an extinction of $A_V = 0.90 \pm 0.05$ mag, while remaining consistent with the Gaia DR3 distance of $d \simeq 2.06$ kpc used as an external prior. An independent Gaia DR3 astrometric analysis gives a distance of $2082 \pm 142$ pc and confirms the membership of all six systems. The twelve binary components occupy the turnoff and subgiant regions of the cluster, enabling stringent evolutionary tests: in the radius--mass, radius--temperature, and temperature--mass diagrams, they show excellent agreement with modern stellar evolution models for the derived cluster parameters. NGC 7789 thus serves as a valuable benchmark for multi-observable, binary-based age determinations in open cluster studies.

Tomographic reconstruction of reionization is a long-sought goal. It would move the field beyond global summary statistics, such as the volume-averaged ionised fraction, to direct, field-level constraints on the ionization topology. With this in mind, we present TORRCH (TOmographic Reconstruction of the Reionization of Cosmic Hydrogen), a deep-learning framework that reconstructs the neutral-hydrogen fraction field during the epoch of reionization from the spatial distributions of Ly$α$ emitters (LAEs) and non-Ly$α$-selected galaxies (NLSGs) at luminosity limits comparable to current surveys. Using hydrodynamical simulations post-processed with radiative transfer, we train a deterministic 3D U-Net on mock surveys spanning diverse reionization scenarios and predict the neutral-fraction field. We find that TORRCH recovers the large-scale ionization morphology from synthetic data comparable to current surveys with high fidelity, and reproduces both the one-point distribution and the 2D power spectrum of projected neutral fractions. The predicted galaxy-IGM cross-correlation is also captured well, including the expected small-scale anti-correlation and its decline towards zero at large separations. Reconstruction quality depends on tracer completeness, with deep joint LAE+NLSG samples yielding the most accurate morphology, while LAE-only selections retain bubble-scale topology but with reduced fidelity. Robustness tests show that the method is stable to variations in ionization conditions between training and test data, and to realistic redshift uncertainties. Our results suggest that galaxy-based tomography can potentially deliver reliable reionization maps across realistic survey redshift windows.

Obtaining a census of dark matter structures at low mass ($\leq 10^9 M_\odot$) can provide strong constraints on the nature of dark matter, though identifying such structures remains difficult. In this work, we study the galaxy cluster RX J0437.1+0043, taking advantage of its powerful "exotic" Hyperbolic-Umbilic (HU) lensing configuration to search for substructure candidates. Using a combination of high resolution imaging, IFU spectroscopy, and gravitational lensing modelling, we report on a tentative detection of a dark matter subhalo ($m_{\rm halo} = 2.25 \pm 0.94 \times 10^9 M_\odot$) near the vicinity of one of the largest HU images. We stress that this result is still preliminary and that deeper data and more advanced modelling techniques are needed to ultimately confirm this detection. Nevertheless, this work outlines the first steps towards understanding subhalo properties in dense cluster environments, developing HU cluster lenses as a potential new tool for investigating dark matter.

We report on the operation of a 13 g PbWO$_4$ crystal, grown from archaeological Pb and operated as a cryogenic calorimeter in an underground environment. Read out with a Ge thermistor, the detector achieves a low energy threshold and, for the first time, enables the derivation of a dark matter exclusion limit using PbWO$_4$ as target material, for both spin-dependent interactions on neutrons and spin-independent interactions. Although limited in mass and not representative of the final RES-NOVA detector design, this prototype demonstrates effective control of mechanical vibrations and low-energy noise in a cryogenic system, which is a key requirement for rare-event searches. The experiment therefore provides a proof of principle for the RES-NOVA detection concept, validating the use of archaeological Pb-based PbWO$_4$ crystals, low-background operation, and robust data-analysis procedures. These results establish a solid technological and methodological foundation for future RES-NOVA detectors employing larger target masses and advanced thermal readout technologies.

In the Gaia era, a comprehensive analysis of the binary open clusters NGC 869 (h Persei) and NGC 884 (chi Persei) system has been conducted to investigate its structural, astrophysical, kinematic, and Galactic orbital properties, along with its dynamical evolution. By applying the UPMASK algorithm to Gaia astrometric data for the estimation of cluster membership probabilities, it has been determined that 808 stars in NGC 869 and 707 stars in NGC 884 exhibit the highest statistical likelihood of being cluster members. The fundamental astrophysical parameters of the clusters were inferred within a Bayesian framework using Gaia data and PARSEC stellar evolutionary isochrones, through the application of the Markov Chain Monte Carlo (MCMC) technique. The estimated parameters are: colour excess E(B-V) = 0.516 +0.17/-0.24 mag and 0.516 +0.22/-0.33 mag, distances 2376 +301/-278 and 2273 +230/-290 pc, ages log(t/yr) = 7.31 +0.17/-0.32 and log(t/yr) = 7.30 +0.13/-0.29, and metallicities [Fe/H] = -0.24 +/- 0.12 and [Fe/H] = -0.25 +/- 0.12 dex for NGC 869 and NGC 884, respectively. Since spectroscopic observations are not available for the clusters, SED analysis was employed for the member stars, yielding results consistent with those obtained using the MCMC method. Kinematic and Galactic orbital analyses suggest that the open clusters originated in nearby regions of the Galaxy. This interpretation is supported by their similar space velocities and Galactic orbital parameters. Furthermore, orbital integration over 1 Gyr indicates a potential interaction between the clusters within the next 11 Myr. This study provides strong evidence of a common origin and a possible future dynamical interaction, contributing valuable insights into the formation and evolution of binary open clusters in the Milky Way.

FarView is a proposed low frequency radio interferometer for deployment on the lunar far side, enabled by the Moon's radio quiet environment. Operating over 1-50 MHz inaccessible from Earth, FarView will open a new observational window and promote discovery class science in cosmology, heliophysics, Galactic and exoplanet astrophysics. The primary science is measurement of the redshifted 21 cm signal from the Cosmic Dark Ages (z=30-100), identified by the Astro2020 Decadal Survey as a priority cosmology discovery area. FarView will deliver 3D tomographic measurements and precision power spectra of neutral hydrogen in a largely linear regime, enabling tests of inflationary initial conditions, primordial non Gaussianity, dark matter properties, neutrino masses, and early dark energy. The reference design consists of 100000 crossed dipole antennas in a dense core-halo configuration spanning 200 sq km. A compact 4 km core with 83000 dipoles maximizes sensitivity to large scale cosmological modes, while 20000 halo elements extending to 14 km provide angular resolution and calibration for foreground characterization. Sensitivity forecasts indicate a 10-sigma detection of the Dark Ages 21 cm power spectrum at z=30 over five years of half duty cycle lunar night observations. An FFT-based EPIC beamformer is identified as an efficient signal processing architecture. Beyond cosmology, FarView will enable interferometric imaging of low frequency solar radio bursts, advancing space weather studies. Additional capabilities include stellar space weather observations, Galactic cosmic ray tomography via free-free absorption, and searches for auroral radio emission from exoplanet magnetospheres, a probe of exoplanet habitability. FarView represents a flagship class opportunity to establish the Moon as a platform for foundational astrophysics while delivering unique observational capabilities.

Post-common envelope binaries (PCEBs) consisting of a white dwarf (WD) plus a main-sequence (MS) star can constrain current prescriptions of common envelope evolution (CEE) and calibrate theoretical models of binary formation and evolution. Most PCEBs studied to date have typical orbital periods of hours to a few days and can be well explained by assuming inefficient CEE to expel the envelope. However, there are currently several systems with relatively wide orbital periods ($>$18 days). To explain these wide PCEBs, additional sources of energy have been suggested to be taken into account. Here, we present the discovery and observational characterization of a compact object ($M\,\geq\,0.58\,\rm M_{\odot}$) with a K-type star companion in the solar neighborhood ($d\sim 112$ pc) and an orbital period of $P_{\rm orb}\sim 14$ days. The compact object binary is likely to be a system consisting of a WD and a barium dwarf. Such a system with an orbital period within the gap between tight and wide binaries provides a test of whether additional energy sources are required to explain its formation. Using binary evolution models, we investigate the evolutionary history of this wide PCEB system and find that the observed properties of this source can be explained without invoking any extra energy source.