arXiv Daily Digest - 2026-01-29

Neuro-symbolic AI systems integrate neural perception with symbolic reasoning to enable data-efficient, interpretable, and robust intelligence beyond purely neural models. Although this compositional paradigm has shown superior performance in domains such as reasoning, planning, and verification, its deployment remains challenging due to severe inefficiencies in symbolic and probabilistic inference. Through systematic analysis of representative neuro-symbolic workloads, we identify probabilistic logical reasoning as the inefficiency bottleneck, characterized by irregular control flow, low arithmetic intensity, uncoalesced memory accesses, and poor hardware utilization on CPUs and GPUs. This paper presents REASON, an integrated acceleration framework for probabilistic logical reasoning in neuro-symbolic AI. REASON introduces a unified directed acyclic graph representation that captures common structure across symbolic and probabilistic models, coupled with adaptive pruning and regularization. At the architecture level, REASON features a reconfigurable, tree-based processing fabric optimized for irregular traversal, symbolic deduction, and probabilistic aggregation. At the system level, REASON is tightly integrated with GPU streaming multiprocessors through a programmable interface and multi-level pipeline that efficiently orchestrates compositional execution. Evaluated across six neuro-symbolic workloads, REASON achieves 12-50x speedup and 310-681x energy efficiency over desktop and edge GPUs under TSMC 28 nm node. REASON enables real-time probabilistic logical reasoning, completing end-to-end tasks in 0.8 s with 6 mm2 area and 2.12 W power, demonstrating that targeted acceleration of probabilistic logical reasoning is critical for practical and scalable neuro-symbolic AI and positioning REASON as a foundational system architecture for next-generation cognitive intelligence.

Scientific computing has long relied on double precision (64-bit floating point) arithmetic to guarantee accuracy in simulations of real-world phenomena. However, the growing availability of hardware accelerators such as Graphics Processing Units (GPUs) has made low-precision formats attractive due to their superior performance, reduced memory footprint, and improved energy efficiency. In this work, we investigate the role of mixed-precision arithmetic in neural-network based Variational Monte Carlo (VMC), a widely used method for solving computationally otherwise intractable quantum many-body systems. We first derive general analytical bounds on the error introduced by reduced precision on Metropolis-Hastings MCMC, and then empirically validate these bounds on the use-case of VMC. We demonstrate that significant portions of the algorithm, in particular, sampling the quantum state, can be executed in half precision without loss of accuracy. More broadly, this work provides a theoretical framework to assess the applicability of mixed-precision arithmetic in machine-learning approaches that rely on MCMC sampling. In the context of VMC, we additionally demonstrate the practical effectiveness of mixed-precision strategies, enabling more scalable and energy-efficient simulations of quantum many-body systems.

In an ordinal election, two candidates are said to be perfect clones if every voter ranks them adjacently. The independence of clones axiom then states that removing one of the two clones should not change the election outcome. This axiom has been extensively studied in social choice theory, and several voting rules are known to satisfy it (such as IRV, Ranked Pairs and Schulze). However, perfect clones are unlikely to occur in practice, especially for political elections with many voters. In this work, we study different notions of approximate clones in ordinal elections. Informally, two candidates are approximate clones in a preference profile if they are close to being perfect clones. We discuss two measures to quantify this proximity, and we show under which conditions the voting rules that are known to be independent of clones are also independent of approximate clones. In particular, we show that for elections with at least four candidates, none of these rules are independent of approximate clones in the general case. However, we find a more positive result for the case of three candidates. Finally, we conduct an empirical study of approximate clones and independence of approximate clones based on three real-world datasets: votes in local Scottish elections, votes in mini-jury deliberations, and votes of judges in figure skating competitions. We find that approximate clones are common in some contexts, and that the closest two candidates are to being perfect clones, the less likely their removal is to change the election outcome, especially for voting rules that are independent of perfect clones.

This paper advances the theoretical understanding of active learning label complexity for decision trees as binary classifiers. We make two main contributions. First, we provide the first analysis of the disagreement coefficient for decision trees-a key parameter governing active learning label complexity. Our analysis holds under two natural assumptions required for achieving polylogarithmic label complexity, (i) each root-to-leaf path queries distinct feature dimensions, and (ii) the input data has a regular, grid-like structure. We show these assumptions are essential, as relaxing them leads to polynomial label complexity. Second, we present the first general active learning algorithm for binary classification that achieves a multiplicative error guarantee, producing a $(1+ε)$-approximate classifier. By combining these results, we design an active learning algorithm for decision trees that uses only a polylogarithmic number of label queries in the dataset size, under the stated assumptions. Finally, we establish a label complexity lower bound, showing our algorithm's dependence on the error tolerance $ε$ is close to optimal.

Multitask learning and related frameworks have achieved tremendous success in modern applications. In multitask learning problem, we are given a set of heterogeneous datasets collected from related source tasks and hope to enhance the performance above what we could hope to achieve by solving each of them individually. The recent work of arXiv:2006.15785 has showed that, without access to distributional information, no algorithm based on aggregating samples alone can guarantee optimal risk as long as the sample size per task is bounded. In this paper, we focus on understanding the statistical limits of multitask learning. We go beyond the no-free-lunch theorem in arXiv:2006.15785 by establishing a stronger impossibility result of adaptation that holds for arbitrarily large sample size per task. This improvement conveys an important message that the hardness of multitask learning cannot be overcame by having abundant data per task. We also discuss the notion of optimal adaptivity that may be of future interests.

Deployed machine learning systems must continuously evolve as data, architectures, and regulations change, often without access to original training data or model internals. In such settings, black-box copying provides a practical refactoring mechanism, i.e. upgrading legacy models by learning replicas from input-output queries alone. When restricted to hard-label outputs, copying turns into a discontinuous surface reconstruction problem from pointwise queries, severely limiting the ability to recover boundary geometry efficiently. We propose a distance-based copying (distillation) framework that replaces hard-label supervision with signed distances to the teacher's decision boundary, converting copying into a smooth regression problem that exploits local geometry. We develop an $α$-governed smoothing and regularization scheme with Hölder/Lipschitz control over the induced target surface, and introduce two model-agnostic algorithms to estimate signed distances under label-only access. Experiments on synthetic problems and UCI benchmarks show consistent improvements in fidelity and generalization accuracy over hard-label baselines, while enabling distance outputs as uncertainty-related signals for black-box replicas.

COMET-SG1 is a lightweight, stability-oriented autoregressive regression model designed for time-series prediction on edge and embedded AI systems. Unlike recurrent neural networks or transformer-based sequence models, COMET-SG1 operates through linear behavior-space encoding, memory-anchored transition estimation, and deterministic state updates. This structure prioritizes bounded long-horizon behavior under fully autoregressive inference, a critical requirement for edge deployment where prediction errors accumulate over time. Experiments on non-stationary synthetic time-series data demonstrate that COMET-SG1 achieves competitive short-horizon accuracy while exhibiting significantly reduced long-horizon drift compared to MLP, LSTM, and k-nearest neighbor baselines. With a compact parameter footprint and operations compatible with fixed-point arithmetic, COMET-SG1 provides a practical and interpretable approach for stable autoregressive prediction in edge and embedded AI applications.

Accurate forecasting of infectious disease incidence is critical for public health planning and timely intervention. While most data-driven forecasting approaches rely primarily on historical data from a single country, such data are often limited in length and variability, restricting the performance of machine learning (ML) models. In this work, we investigate a cross-country learning approach for infectious disease forecasting, in which a single model is trained on time series data from multiple countries and evaluated on a country of interest. This setting enables the model to exploit shared epidemic dynamics across countries and to benefit from an enlarged training set. We examine this approach through a case study on COVID-19 case forecasting in Cyprus, using surveillance data from European countries. We evaluate multiple ML models and analyse the impact of the lookback window length and cross-country `data augmentation' on multi-step forecasting performance. Our results show that incorporating data from other countries can lead to consistent improvements over models trained solely on national data. Although the empirical focus is on Cyprus and COVID-19, the proposed framework and findings are applicable to infectious disease forecasting more broadly, particularly in settings with limited national historical data.

Training learned image compression (LIC) models entails navigating a challenging optimization landscape defined by the fundamental trade-off between rate and distortion. Standard first-order optimizers, such as SGD and Adam, struggle with \emph{gradient conflicts} arising from competing objectives, leading to slow convergence and suboptimal rate-distortion performance. In this work, we demonstrate that a simple utilization of a second-order quasi-Newton optimizer, \textbf{SOAP}, dramatically improves both training efficiency and final performance across diverse LICs. Our theoretical and empirical analyses reveal that Newton preconditioning inherently resolves the intra-step and inter-step update conflicts intrinsic to the R-D objective, facilitating faster, more stable convergence. Beyond acceleration, we uncover a critical deployability benefit: second-order trained models exhibit significantly fewer activation and latent outliers. This substantially enhances robustness to post-training quantization. Together, these results establish second-order optimization, achievable as a seamless drop-in replacement of the imported optimizer, as a powerful, practical tool for advancing the efficiency and real-world readiness of LICs.

A fundamental challenge in offline reinforcement learning is distributional shift. Scarce data or datasets dominated by out-of-distribution (OOD) areas exacerbate this issue. Our theoretical analysis and experiments show that the standard squared error objective induces a harmful TD cross covariance. This effect amplifies in OOD areas, biasing optimization and degrading policy learning. To counteract this mechanism, we develop two complementary strategies: partitioned buffer sampling that restricts updates to localized replay partitions, attenuates irregular covariance effects, and aligns update directions, yielding a scheme that is easy to integrate with existing implementations, namely Clustered Cross-Covariance Control for TD (C^4). We also introduce an explicit gradient-based corrective penalty that cancels the covariance induced bias within each update. We prove that buffer partitioning preserves the lower bound property of the maximization objective, and that these constraints mitigate excessive conservatism in extreme OOD areas without altering the core behavior of policy constrained offline reinforcement learning. Empirically, our method showcases higher stability and up to 30% improvement in returns over prior methods, especially with small datasets and splits that emphasize OOD areas.

Fog and edge computing require adaptive control schemes that can handle partial observability, severe latency requirements, and dynamically changing workloads. Recent research on Agentic AI (AAI) increasingly integrates reasoning systems powered by Large Language Models; however, these tools are not applicable to infrastructure-level systems due to their high computational cost, stochastic nature, and poor formal analyzability. In this paper, a generic model, Agentic Fog (AF), is presented, in which fog nodes are represented as policy-driven autonomous agents that communicate via p2p interactions based on shared memory and localized coordination. The suggested architecture decomposes a system's goals into abstract policy guidance and formalizes decentralized fog coordination as an exact potential game. The framework is guaranteed to converge and remain stable under asynchronous updates, bounded-rational best-response dynamics, and node failures. Simulations demonstrate that the AF system achieves lower average latency and adapts more efficiently to varying demand than greedy heuristics and integer linear programming under dynamic conditions. The sensitivity analysis also demonstrates the capability to perform optimally under different memory and coordination conditions.

For socially sensitive tasks like hate speech detection, the quality of explanations from Large Language Models (LLMs) is crucial for factors like user trust and model alignment. While Persona prompting (PP) is increasingly used as a way to steer model towards user-specific generation, its effect on model rationales remains underexplored. We investigate how LLM-generated rationales vary when conditioned on different simulated demographic personas. Using datasets annotated with word-level rationales, we measure agreement with human annotations from different demographic groups, and assess the impact of PP on model bias and human alignment. Our evaluation across three LLMs results reveals three key findings: (1) PP improving classification on the most subjective task (hate speech) but degrading rationale quality. (2) Simulated personas fail to align with their real-world demographic counterparts, and high inter-persona agreement shows models are resistant to significant steering. (3) Models exhibit consistent demographic biases and a strong tendency to over-flag content as harmful, regardless of PP. Our findings reveal a critical trade-off: while PP can improve classification in socially-sensitive tasks, it often comes at the cost of rationale quality and fails to mitigate underlying biases, urging caution in its application.

Score-based diffusion models have recently been extended to infinite-dimensional function spaces, with uses such as inverse problems arising from partial differential equations. In the Bayesian formulation of inverse problems, the aim is to sample from a posterior distribution over functions obtained by conditioning a prior on noisy observations. While diffusion models provide expressive priors in function space, the theory of conditioning them to sample from the posterior remains open. We address this, assuming that either the prior lies in the Cameron-Martin space, or is absolutely continuous with respect to a Gaussian measure. We prove that the models can be conditioned using an infinite-dimensional extension of Doob's $h$-transform, and that the conditional score decomposes into an unconditional score and a guidance term. As the guidance term is intractable, we propose a simulation-free score matching objective (called Supervised Guidance Training) enabling efficient and stable posterior sampling. We illustrate the theory with numerical examples on Bayesian inverse problems in function spaces. In summary, our work offers the first function-space method for fine-tuning trained diffusion models to accurately sample from a posterior.

As large language models (LLMs) move from research to production, understanding how inference engines behave in real time has become both essential and elusive. Unlike general-purpose engines such as ONNX Runtime, today's LLM inference systems offer little operator-level visibility, leaving developers blind to where time and resources go. Even basic questions -- is this workload memory-bound or compute-bound? -- often remain unanswered. To close this gap, we develop a fine-grained, non-intrusive profiling framework for modern LLM inference engines, exemplified by llama.cpp but applicable to similar runtime architectures. Built on extended Berkeley Packet Filter (eBPF) technology, our system dynamically attaches probes to runtime functions across multiple layers -- without modifying or recompiling the source. It transforms collected traces into rich visualizations of operators, graphs, timelines, and hardware counter trends, exposing how dense inference, Mixture-of-Experts routing, and operator offloading behave in practice. With less than 4% runtime overhead and high profiling fidelity, our framework makes LLM inference both transparent and diagnosable, turning performance profiling into a practical tool for optimization, scheduling, and resource-aware deployment.

Preference-Conditioned Policy Learning (PCPL) in Multi-Objective Reinforcement Learning (MORL) aims to approximate diverse Pareto-optimal solutions by conditioning policies on user-specified preferences over objectives. This enables a single model to flexibly adapt to arbitrary trade-offs at run-time by producing a policy on or near the Pareto front. However, existing benchmarks for PCPL are largely restricted to toy tasks and fixed environments, limiting their realism and scalability. To address this gap, we introduce GraphAllocBench, a flexible benchmark built on a novel graph-based resource allocation sandbox environment inspired by city management, which we call CityPlannerEnv. GraphAllocBench provides a rich suite of problems with diverse objective functions, varying preference conditions, and high-dimensional scalability. We also propose two new evaluation metrics -- Proportion of Non-Dominated Solutions (PNDS) and Ordering Score (OS) -- that directly capture preference consistency while complementing the widely used hypervolume metric. Through experiments with Multi-Layer Perceptrons (MLPs) and graph-aware models, we show that GraphAllocBench exposes the limitations of existing MORL approaches and paves the way for using graph-based methods such as Graph Neural Networks in complex, high-dimensional combinatorial allocation tasks. Beyond its predefined problem set, GraphAllocBench enables users to flexibly vary objectives, preferences, and allocation rules, establishing it as a versatile and extensible benchmark for advancing PCPL. Code: https://anonymous.4open.science/r/GraphAllocBench

Large language models (LLMs) are increasingly used for emotional support and mental health-related interactions outside clinical settings, yet little is known about how people evaluate and relate to these systems in everyday use. We analyze 5,126 Reddit posts from 47 mental health communities describing experiential or exploratory use of AI for emotional support or therapy. Grounded in the Technology Acceptance Model and therapeutic alliance theory, we develop a theory-informed annotation framework and apply a hybrid LLM-human pipeline to analyze evaluative language, adoption-related attitudes, and relational alignment at scale. Our results show that engagement is shaped primarily by narrated outcomes, trust, and response quality, rather than emotional bond alone. Positive sentiment is most strongly associated with task and goal alignment, while companionship-oriented use more often involves misaligned alliances and reported risks such as dependence and symptom escalation. Overall, this work demonstrates how theory-grounded constructs can be operationalized in large-scale discourse analysis and highlights the importance of studying how users interpret language technologies in sensitive, real-world contexts.

As large language models (LLMs) continue to scale, deployment is increasingly bottlenecked by the memory wall, motivating a shift toward extremely low-bit quantization. However, most quantization-aware training (QAT) methods apply hard rounding and the straight-through estimator (STE) from the beginning of the training, which prematurely discretizes the optimization landscape and induces persistent gradient mismatch between latent weights and quantized weights, hindering effective optimization of quantized models. To address this, we propose Hestia, a Hessian-guided differentiable QAT framework for extremely low-bit LLMs, which replaces the rigid step function with a temperature-controlled softmax relaxation to maintain gradient flow early in training while progressively hardening quantization. Furthermore, Hestia leverages a tensor-wise Hessian trace metric as a lightweight curvature signal to drive fine-grained temperature annealing, enabling sensitivity-aware discretization across the model. Evaluations on Llama-3.2 show that Hestia consistently outperforms existing ternary QAT baselines, yielding average zero-shot improvements of 5.39% and 4.34% for the 1B and 3B models. These results indicate that Hessian-guided relaxation effectively recovers representational capacity, establishing a more robust training path for 1.58-bit LLMs. The code is available at https://github.com/hestia2026/Hestia.

Biased gradient compression with error feedback (EF) reduces communication in federated learning (FL), but under non-IID data, the residual error can decay slowly, causing gradient mismatch and stalled progress in the early rounds. We propose step-ahead partial error feedback (SA-PEF), which integrates step-ahead (SA) correction with partial error feedback (PEF). SA-PEF recovers EF when the step-ahead coefficient $α=0$ and step-ahead EF (SAEF) when $α=1$. For non-convex objectives and $δ$-contractive compressors, we establish a second-moment bound and a residual recursion that guarantee convergence to stationarity under heterogeneous data and partial client participation. The resulting rates match standard non-convex Fed-SGD guarantees up to constant factors, achieving $O((η,η_0TR)^{-1})$ convergence to a variance/heterogeneity floor with a fixed inner step size. Our analysis reveals a step-ahead-controlled residual contraction $ρ_r$ that explains the observed acceleration in the early training phase. To balance SAEF's rapid warm-up with EF's long-term stability, we select $α$ near its theory-predicted optimum. Experiments across diverse architectures and datasets show that SA-PEF consistently reaches target accuracy faster than EF.

We develop a computational approach to Metric Answer Set Programming (ASP) to allow for expressing quantitative temporal constraints, like durations and deadlines. A central challenge is to maintain scalability when dealing with fine-grained timing constraints, which can significantly exacerbate ASP's grounding bottleneck. To address this issue, we leverage extensions of ASP with difference constraints, a simplified form of linear constraints, to handle time-related aspects externally. Our approach effectively decouples metric ASP from the granularity of time, resulting in a solution that is unaffected by time precision.

As digital environments (data distribution) are in flux, with new GUI data arriving over time-introducing new domains or resolutions-agents trained on static environments deteriorate in performance. In this work, we introduce Continual GUI Agents, a new task that requires GUI agents to perform continual learning under shifted domains and resolutions. We find existing methods fail to maintain stable grounding as GUI distributions shift over time, due to the diversity of UI interaction points and regions in fluxing scenarios. To address this, we introduce GUI-Anchoring in Flux (GUI-AiF), a new reinforcement fine-tuning framework that stabilizes continual learning through two novel rewards: Anchoring Point Reward in Flux (APR-iF) and Anchoring Region Reward in Flux (ARR-iF). These rewards guide the agents to align with shifting interaction points and regions, mitigating the tendency of existing reward strategies to over-adapt to static grounding cues (e.g., fixed coordinates or element scales). Extensive experiments show GUI-AiF surpasses state-of-the-art baselines. Our work establishes the first continual learning framework for GUI agents, revealing the untapped potential of reinforcement fine-tuning for continual GUI Agents.

This paper examines how Large Language Models (LLMs) reproduce societal norms, particularly heterocisnormativity, and how these norms translate into measurable biases in their text generations. We investigate whether explicit information about a subject's gender or sexuality influences LLM responses across three subject categories: queer-marked, non-queer-marked, and the normalized "unmarked" category. Representational imbalances are operationalized as measurable differences in English sentence completions across four dimensions: sentiment, regard, toxicity, and prediction diversity. Our findings show that Masked Language Models (MLMs) produce the least favorable sentiment, higher toxicity, and more negative regard for queer-marked subjects. Autoregressive Language Models (ARLMs) partially mitigate these patterns, while closed-access ARLMs tend to produce more harmful outputs for unmarked subjects. Results suggest that LLMs reproduce normative social assumptions, though the form and degree of bias depend strongly on specific model characteristics, which may redistribute, but not eliminate, representational harms.

The evolution of Large Language Models (LLMs) into autonomous agents necessitates the management of extensive, dynamic contexts. Current benchmarks, however, remain largely static, relying on passive retrieval tasks that fail to simulate the complexities of agent-environment interaction, such as non-linear reasoning and iterative feedback. To address this, we introduce \textbf{AgentLongBench}, which evaluates agents through simulated environment rollouts based on Lateral Thinking Puzzles. This framework generates rigorous interaction trajectories across knowledge-intensive and knowledge-free scenarios. Experiments with state-of-the-art models and memory systems (32K to 4M tokens) expose a critical weakness: while adept at static retrieval, agents struggle with the dynamic information synthesis essential for workflows. Our analysis indicates that this degradation is driven by the minimum number of tokens required to resolve a query. This factor explains why the high information density inherent in massive tool responses poses a significantly greater challenge than the memory fragmentation typical of long-turn dialogues.

The Cox Proportional Hazards (PH) model is widely used in survival analysis. Recently, artificial neural network (ANN)-based Cox-PH models have been developed. However, training these Cox models with high-dimensional features typically requires a substantial number of labeled samples containing information about time-to-event. The limited availability of labeled data for training often constrains the performance of ANN-based Cox models. To address this issue, we employed a deep semi-supervised learning (DSSL) approach to develop single- and multi-modal ANN-based Cox models based on the Mean Teacher (MT) framework, which utilizes both labeled and unlabeled data for training. We applied our model, named Cox-MT, to predict the prognosis of several types of cancer using data from The Cancer Genome Atlas (TCGA). Our single-modal Cox-MT models, utilizing TCGA RNA-seq data or whole slide images, significantly outperformed the existing ANN-based Cox model, Cox-nnet, using the same data set across four types of cancer considered. As the number of unlabeled samples increased, the performance of Cox-MT significantly improved with a given set of labeled data. Furthermore, our multi-modal Cox-MT model demonstrated considerably better performance than the single-modal model. In summary, the Cox-MT model effectively leverages both labeled and unlabeled data to significantly enhance prediction accuracy compared to existing ANN-based Cox models trained solely on labeled data.

End-to-end perception and trajectory prediction from raw sensor data is one of the key capabilities for autonomous driving. Modular pipelines restrict information flow and can amplify upstream errors. Recent query-based, fully differentiable perception-and-prediction (PnP) models mitigate these issues, yet the complementarity of cameras and LiDAR in the query-space has not been sufficiently explored. Models often rely on fusion schemes that introduce heuristic alignment and discrete selection steps which prevent full utilization of available information and can introduce unwanted bias. We propose Li-ViP3D++, a query-based multimodal PnP framework that introduces Query-Gated Deformable Fusion (QGDF) to integrate multi-view RGB and LiDAR in query space. QGDF (i) aggregates image evidence via masked attention across cameras and feature levels, (ii) extracts LiDAR context through fully differentiable BEV sampling with learned per-query offsets, and (iii) applies query-conditioned gating to adaptively weight visual and geometric cues per agent. The resulting architecture jointly optimizes detection, tracking, and multi-hypothesis trajectory forecasting in a single end-to-end model. On nuScenes, Li-ViP3D++ improves end-to-end behavior and detection quality, achieving higher EPA (0.335) and mAP (0.502) while substantially reducing false positives (FP ratio 0.147), and it is faster than the prior Li-ViP3D variant (139.82 ms vs. 145.91 ms). These results indicate that query-space, fully differentiable camera-LiDAR fusion can increase robustness of end-to-end PnP without sacrificing deployability.

Reinforcement Learning (RL) agents often struggle in real-world applications where environmental conditions are non-stationary, particularly when reward functions shift or the available action space expands. This paper introduces MORPHIN, a self-adaptive Q-learning framework that enables on-the-fly adaptation without full retraining. By integrating concept drift detection with dynamic adjustments to learning and exploration hyperparameters, MORPHIN adapts agents to changes in both the reward function and on-the-fly expansions of the agent's action space, while preserving prior policy knowledge to prevent catastrophic forgetting. We validate our approach using a Gridworld benchmark and a traffic signal control simulation. The results demonstrate that MORPHIN achieves superior convergence speed and continuous adaptation compared to a standard Q-learning baseline, improving learning efficiency by up to 1.7x.

As lattice gauge theories with non-trivial topological features are driven towards the continuum limit, standard Markov Chain Monte Carlo simulations suffer for topological freezing, i.e., a dramatic growth of autocorrelations in topological observables. A widely used strategy is the adoption of Open Boundary Conditions (OBC), which restores ergodic sampling of topology but at the price of breaking translation invariance and introducing unphysical boundary artifacts. In this contribution we summarize a scalable, exact flow-based strategy to remove them by transporting configurations from a prior with a OBC defect to a fully periodic ensemble, and apply it to 4d SU(3) Yang--Mills theory. The method is based on a Stochastic Normalizing Flow (SNF) that alternates non-equilibrium Monte Carlo updates with localized, gauge-equivariant defect coupling layers implemented via masked parametric stout smearing. Training is performed by minimizing the average dissipated work, equivalent to a Kullback--Leibler divergence between forward and reverse non-equilibrium path measures, to achieve more reversible trajectories and improved efficiency. We discuss the scaling with the number of degrees of freedom affected by the defect and show that defect SNFs achieve better performances than purely stochastic non-equilibrium methods at comparable cost. Finally, we validate the approach by reproducing reference results for the topological susceptibility.

Diffusion Large Language Models (dLLMs) introduce iterative denoising to enable parallel token generation, but their sampling phase displays fundamentally different characteristics compared to GEMM-centric transformer layers. Profiling on modern GPUs reveals that sampling can account for up to 70% of total model inference latency-primarily due to substantial memory loads and writes from vocabulary-wide logits, reduction-based token selection, and iterative masked updates. These processes demand large on-chip SRAM and involve irregular memory accesses that conventional NPUs struggle to handle efficiently. To address this, we identify a set of critical instructions that an NPU architecture must specifically optimize for dLLM sampling. Our design employs lightweight non-GEMM vector primitives, in-place memory reuse strategies, and a decoupled mixed-precision memory hierarchy. Together, these optimizations deliver up to a 2.53x speedup over the NVIDIA RTX A6000 GPU under an equivalent nm technology node. We also open-source our cycle-accurate simulation and post-synthesis RTL verification code, confirming functional equivalence with current dLLM PyTorch implementations.

Large language models are increasingly used to curate bibliographies, raising the question: are their reference lists distinguishable from human ones? We build paired citation graphs, ground truth and GPT-4o-generated (from parametric knowledge), for 10,000 focal papers ($\approx$ 275k references) from SciSciNet, and added a field-matched random baseline that preserves out-degree and field distributions while breaking latent structure. We compare (i) structure-only node features (degree/closeness/eigenvector centrality, clustering, edge count) with (ii) 3072-D title/abstract embeddings, using an RF on graph-level aggregates and Graph Neural Networks with node features. Structure alone barely separates GPT from ground truth (RF accuracy $\approx$ 0.60) despite cleanly rejecting the random baseline ($\approx$ 0.89--0.92). By contrast, embeddings sharply increase separability: RF on aggregated embeddings reaches $\approx$ 0.83, and GNNs with embedding node features achieve 93\% test accuracy on GPT vs.\ ground truth. We show the robustness of our findings by replicating the pipeline with Claude Sonnet 4.5 and with multiple embedding models (OpenAI and SPECTER), with RF separability for ground truth vs.\ Claude $\approx 0.77$ and clean rejection of the random baseline. Thus, LLM bibliographies, generated purely from parametric knowledge, closely mimic human citation topology, but leave detectable semantic fingerprints; detection and debiasing should target content signals rather than global graph structure.

Combinatorial optimization problems such as the Job-Shop Scheduling Problem (JSP) and Knapsack Problem (KP) are fundamental challenges in operations research, logistics, and eterprise resource planning (ERP). These problems often require sophisticated algorithms to achieve near-optimal solutions within practical time constraints. Recent advances in deep learning have introduced transformer-based architectures as promising alternatives to traditional heuristics and metaheuristics. We leverage the Multi-Type Transformer (MTT) architecture to address these benchmarks in a unified framework. We present an extensive experimental evaluation across standard benchmark datasets for JSP and KP, demonstrating that MTT achieves competitive performance on different size of these benchmark problems. We showcase the potential of multi-type attention on a real application in Ferro-Titanium industry. To the best of our knowledge, we are the first to apply multi-type transformers in real manufacturing.

Exogenous MDPs (Exo-MDPs) capture sequential decision-making where uncertainty comes solely from exogenous inputs that evolve independently of the learner's actions. This structure is especially common in operations research applications such as inventory control, energy storage, and resource allocation, where exogenous randomness (e.g., demand, arrivals, or prices) drives system behavior. Despite decades of empirical evidence that greedy, exploitation-only methods work remarkably well in these settings, theory has lagged behind: all existing regret guarantees for Exo-MDPs rely on explicit exploration or tabular assumptions. We show that exploration is unnecessary. We propose Pure Exploitation Learning (PEL) and prove the first general finite-sample regret bounds for exploitation-only algorithms in Exo-MDPs. In the tabular case, PEL achieves $\widetilde{O}(H^2|Ξ|\sqrt{K})$. For large, continuous endogenous state spaces, we introduce LSVI-PE, a simple linear-approximation method whose regret is polynomial in the feature dimension, exogenous state space, and horizon, independent of the endogenous state and action spaces. Our analysis introduces two new tools: counterfactual trajectories and Bellman-closed feature transport, which together allow greedy policies to have accurate value estimates without optimism. Experiments on synthetic and resource-management tasks show that PEL consistently outperforming baselines. Overall, our results overturn the conventional wisdom that exploration is required, demonstrating that in Exo-MDPs, pure exploitation is enough.

Group counterfactual explanations find a set of counterfactual instances to explain a group of input instances contrastively. However, existing methods either (i) optimize counterfactuals only for a fixed group and do not generalize to new group members, (ii) strictly rely on strong model assumptions (e.g., linearity) for tractability or/and (iii) poorly control the counterfactual group geometry distortion. We instead learn an explicit optimal transport map that sends any group instance to its counterfactual without re-optimization, minimizing the group's total transport cost. This enables generalization with fewer parameters, making it easier to interpret the common actionable recourse. For linear classifiers, we prove that functions representing group counterfactuals are derived via mathematical optimization, identifying the underlying convex optimization type (QP, QCQP, ...). Experiments show that they accurately generalize, preserve group geometry and incur only negligible additional transport cost compared to baseline methods. If model linearity cannot be exploited, our approach also significantly outperforms the baselines.

Recent multimodal large language models (MLLMs) have demonstrated strong capabilities in image quality assessment (IQA) tasks. However, adapting such large-scale models is computationally expensive and still relies on substantial Mean Opinion Score (MOS) annotations. We argue that for MLLM-based IQA, the core bottleneck lies not in the quality perception capacity of MLLMs, but in MOS scale calibration. Therefore, we propose LEAF, a Label-Efficient Image Quality Assessment Framework that distills perceptual quality priors from an MLLM teacher into a lightweight student regressor, enabling MOS calibration with minimal human supervision. Specifically, the teacher conducts dense supervision through point-wise judgments and pair-wise preferences, with an estimate of decision reliability. Guided by these signals, the student learns the teacher's quality perception patterns through joint distillation and is calibrated on a small MOS subset to align with human annotations. Experiments on both user-generated and AI-generated IQA benchmarks demonstrate that our method significantly reduces the need for human annotations while maintaining strong MOS-aligned correlations, making lightweight IQA practical under limited annotation budgets.

Due to constraints on privacy, cost, and latency, on-premise deployment of small models is increasingly common. However, most practical pipelines stop at supervised fine-tuning (SFT) and fail to reach the reinforcement learning (RL) alignment stage. The main reason is that RL alignment typically requires either expensive human preference annotation or heavy reliance on high-quality reward models with large-scale API usage and ongoing engineering maintenance, both of which are ill-suited to on-premise settings. To bridge this gap, we propose a positive-unlabeled (PU) RL distillation method for on-premise small-model deployment. Without human-labeled preferences or a reward model, our method distills the teacher's preference-optimization capability from black-box generations into a locally trainable student. For each prompt, we query the teacher once to obtain an anchor response, locally sample multiple student candidates, and perform anchor-conditioned self-ranking to induce pairwise or listwise preferences, enabling a fully local training loop via direct preference optimization or group relative policy optimization. Theoretical analysis justifies that the induced preference signal by our method is order-consistent and concentrates on near-optimal candidates, supporting its stability for preference optimization. Experiments demonstrate that our method achieves consistently strong performance under a low-cost setting.

Change Point Detection (CPD) is a critical task in time series analysis, aiming to identify moments when the underlying data-generating process shifts. Traditional CPD methods often rely on unsupervised techniques, which lack adaptability to task-specific definitions of change and cannot benefit from user knowledge. To address these limitations, we propose MuRAL-CPD, a novel semi-supervised method that integrates active learning into a multiresolution CPD algorithm. MuRAL-CPD leverages a wavelet-based multiresolution decomposition to detect changes across multiple temporal scales and incorporates user feedback to iteratively optimize key hyperparameters. This interaction enables the model to align its notion of change with that of the user, improving both accuracy and interpretability. Our experimental results on several real-world datasets show the effectiveness of MuRAL-CPD against state-of-the-art methods, particularly in scenarios where minimal supervision is available.

As conversational agents become increasingly common in behaviour change interventions, understanding optimal feedback delivery mechanisms becomes increasingly important. However, choosing a style that both lessens psychological reactance (perceived threats to freedom) while simultaneously eliciting feelings of surprise and engagement represents a complex design problem. We explored how three different feedback styles: 'Direct', 'Politeness', and 'Verbal Leakage' (slips or disfluencies to reveal a desired behaviour) affect user perceptions and behavioural intentions. Matching expectations from literature, the 'Direct' chatbot led to lower behavioural intentions and higher reactance, while the 'Politeness' chatbot evoked higher behavioural intentions and lower reactance. However, 'Politeness' was also seen as unsurprising and unengaging by participants. In contrast, 'Verbal Leakage' evoked reactance, yet also elicited higher feelings of surprise, engagement, and humour. These findings highlight that effective feedback requires navigating trade-offs between user reactance and engagement, with novel approaches such as 'Verbal Leakage' offering promising alternative design opportunities.

Automated narrative intelligence systems for social media monitoring face significant scalability challenges when processing continuous data streams using traditional batch clustering algorithms. We investigate the replacement of HDBSCAN (offline clustering) with online (streaming/incremental) clustering methods in a production narrative report generation pipeline. The proposed system employs a three-stage architecture (data collection, modeling, dashboard generation) that processes thousands of multilingual social media documents daily. While HDBSCAN excels at discovering hierarchical density-based clusters and handling noise, its batch-only nature necessitates complete retraining for each time window, resulting in memory constraints, computational inefficiency, and inability to adapt to evolving narratives in real-time. This work evaluates a bunch of online clustering algorithms across dimensions of cluster quality preservation, computational efficiency, memory footprint, and integration compatibility with existing workflows. We propose evaluation criteria that balance traditional clustering metrics (Silhouette Coefficient, Davies-Bouldin Index) with narrative metrics (narrative distinctness, contingency and variance). Our methodology includes sliding-window simulations on historical datasets from Ukraine information space, enabling comparative analysis of algorithmic trade-offs in realistic operational contexts. This research addresses a critical gap between batch-oriented topic modeling frameworks and the streaming nature of social media monitoring, with implications for computational social science, crisis informatics, and narrative surveillance systems.

Large language models (LLMs) have achieved remarkable progress in code generation, yet their potential for software protection remains largely untapped. Reverse engineering continues to threaten software security, while traditional virtual machine protection (VMP) relies on rigid, rule-based transformations that are costly to design and vulnerable to automated analysis. In this work, we present the first protection-aware framework that learns robust representations of VMP-protected code. Our approach builds large-scale paired datasets of source code and normalized VM implementations, and introduces hierarchical dependency modeling at intra-, preceding-, and inter-instruction levels. We jointly optimize language modeling with functionality-aware and protection-aware contrastive objectives to capture both semantic equivalence and protection strength. To further assess resilience, we propose a protection effectiveness optimization task that quantifies and ranks different VM variants derived from the same source. Coupled with a two-stage continual pre-training and fine-tuning pipeline, our method enables models to generate, compare, and reason over protected code. Extensive experiments show that our framework significantly improves robustness across diverse protection levels, opening a new research direction for learning-based software defense. In this work, we present ShieldedCode, the first protection-aware framework that learns robust representations of VMP-protected code. Our method achieves 26.95% Pass@1 on L0 VM code generation compared to 22.58% for GPT-4o., and improves binary similarity detection Recall@1 by 10% over state of art methods like jTrans.

Visual Question-Answering (VQA) is a challenging multimodal task that requires integrating visual and textual information to generate accurate responses. While multimodal Retrieval-Augmented Generation (mRAG) has shown promise in enhancing VQA systems by providing more evidence on both image and text sides, the default procedure that addresses VQA queries, especially the knowledge-intensive ones, often relies on multi-stage pipelines of mRAG with inherent dependencies. To mitigate the inefficiency limitations while maintaining VQA task performance, this paper proposes a method that trains a multimodal planning agent, dynamically decomposing the mRAG pipeline to solve the VQA task. Our method optimizes the trade-off between efficiency and effectiveness by training the agent to intelligently determine the necessity of each mRAG step. In our experiments, the agent can help reduce redundant computations, cutting search time by over 60\% compared to existing methods and decreasing costly tool calls. Meanwhile, experiments demonstrate that our method outperforms all baselines, including a Deep Research agent and a carefully designed prompt-based method, on average over six various datasets. Code will be released.

This study applies Large Language Models (LLMs) to two foundational Electronic Health Record (EHR) data science tasks: structured data querying (using programmatic languages, Python/Pandas) and information extraction from unstructured clinical text via a Retrieval Augmented Generation (RAG) pipeline. We test the ability of LLMs to interact accurately with large structured datasets for analytics and the reliability of LLMs in extracting semantically correct information from free text health records when supported by RAG. To this end, we presented a flexible evaluation framework that automatically generates synthetic question and answer pairs tailored to the characteristics of each dataset or task. Experiments were conducted on a curated subset of MIMIC III, (four structured tables and one clinical note type), using a mix of locally hosted and API-based LLMs. Evaluation combined exact-match metrics, semantic similarity, and human judgment. Our findings demonstrate the potential of LLMs to support precise querying and accurate information extraction in clinical workflows.

We introduce a novel framework for learning context-aware runtime monitors for AI-based control ensembles. Machine-learning (ML) controllers are increasingly deployed in (autonomous) cyber-physical systems because of their ability to solve complex decision-making tasks. However, their accuracy can degrade sharply in unfamiliar environments, creating significant safety concerns. Traditional ensemble methods aim to improve robustness by averaging or voting across multiple controllers, yet this often dilutes the specialized strengths that individual controllers exhibit in different operating contexts. We argue that, rather than blending controller outputs, a monitoring framework should identify and exploit these contextual strengths. In this paper, we reformulate the design of safe AI-based control ensembles as a contextual monitoring problem. A monitor continuously observes the system's context and selects the controller best suited to the current conditions. To achieve this, we cast monitor learning as a contextual learning task and draw on techniques from contextual multi-armed bandits. Our approach comes with two key benefits: (1) theoretical safety guarantees during controller selection, and (2) improved utilization of controller diversity. We validate our framework in two simulated autonomous driving scenarios, demonstrating significant improvements in both safety and performance compared to non-contextual baselines.

Ensuring the integrity of software build artifacts is an increasingly important concern for modern software engineering, driven by increasingly sophisticated attacks on build systems, distribution channels, and development infrastructures. Reproducible builds $\unicode{x2013}$ where binaries built independently from the same source code can be verified to be bit-for-bit identical to the distributed artifacts $\unicode{x2013}$ provide a principled foundation for transparency and trust in software distribution. Despite their potential, the large-scale adoption of reproducible builds faces two significant challenges: achieving high reproducibility rates across vast software collections and establishing reproducibility monitoring infrastructure that can operate at very large scale. While recent studies have shown that high reproducibility rates are achievable at scale $\unicode{x2013}$ demonstrated by the Nix ecosystem achieving over 90% reproducibility on more than 80,000 packages $\unicode{x2013}$ the problem of effective reproducibility monitoring remains largely unsolved. In this work, we address the reproducibility monitoring challenge by introducing Lila, a decentralized system for reproducibility assessment tailored to the functional package management model. Lila enables distributed reporting of build results and aggregation into a reproducibility database, benefiting both practitioners and future empirical build reproducibility studies.

Assessing ways in which Language Models can reduce their hallucinations and improve the outputs' quality is crucial to ensure their large-scale use. However, methods such as fine-tuning on domain-specific data or the training of a separate \textit{ad hoc} verifier require demanding computational resources (not feasible for many user applications) and constrain the models to specific fields of knowledge. In this thesis, we propose a dialectic pipeline that preserves LLMs' generalization abilities while improving the quality of its answer via self-dialogue, enabling it to reflect upon and correct tentative wrong answers. We experimented with different pipeline settings, testing our proposed method on different datasets and on different families of models. All the pipeline stages are enriched with the relevant context (in an oracle-RAG setting) and a study on the impact of its summarization or its filtering is conducted. We find that our proposed dialectic pipeline is able to outperform by significative margins the standard model answers and that it consistently achieves higher performances than Chain-of-Thought only prompting.

The rapid growth of AI-generated content (AIGC) has enabled high-quality creative production across diverse domains, yet existing systems face critical inefficiencies in throughput, resource utilization, and scalability under concurrent workloads. This paper introduces OnePiece, a large-scale distributed inference system with RDMA optimized for multi-stage AIGC workflows. By decomposing pipelines into fine-grained microservices and leveraging one-sided RDMA communication, OnePiece significantly reduces inter-node latency and CPU overhead while improving GPU utilization. The system incorporates a novel double-ring buffer design to resolve deadlocks in RDMA-aware memory access without CPU involvement. Additionally, a dynamic Node Manager allocates resources elastically across workflow stages in response to real-time load. Experimental results demonstrate that OnePiece reduces GPU resource consumption by 16x in Wan2.1 image-to-video generation compared to monolithic inference pipelines, offering a scalable, fault-tolerant, and efficient solution for production AIGC environments.

While reinforcement learning with verifiable rewards (RLVR) has advanced LLM reasoning in structured domains like mathematics and programming, its application to general-domain reasoning tasks remains challenging due to the absence of verifiable reward signals. To this end, methods like Reinforcement Learning with Reference Probability Reward (RLPR) have emerged, leveraging the probability of generating the final answer as a reward signal. However, these outcome-focused approaches neglect crucial step-by-step supervision of the reasoning process itself. To address this gap, we introduce Probabilistic Process Supervision (P2S), a novel self-supervision framework that provides fine-grained process rewards without requiring a separate reward model or human-annotated reasoning steps. During reinforcement learning, P2S synthesizes and filters a high-quality reference reasoning chain (gold-CoT). The core of our method is to calculate a Path Faithfulness Reward (PFR) for each reasoning step, which is derived from the conditional probability of generating the gold-CoT's suffix, given the model's current reasoning prefix. Crucially, this PFR can be flexibly integrated with any outcome-based reward, directly tackling the reward sparsity problem by providing dense guidance. Extensive experiments on reading comprehension and medical Question Answering benchmarks show that P2S significantly outperforms strong baselines.

Diffusion-based image generative models produce high-fidelity images through iterative denoising but remain vulnerable to memorization, where they unintentionally reproduce exact copies or parts of training images. Recent memorization detection methods are primarily based on the norm of score difference as indicators of memorization. We prove that such norm-based metrics are mainly effective under the assumption of isotropic log-probability distributions, which generally holds at high or medium noise levels. In contrast, analyzing the anisotropic regime reveals that memorized samples exhibit strong angular alignment between the guidance vector and unconditional scores in the low-noise setting. Through these insights, we develop a memorization detection metric by integrating isotropic norm and anisotropic alignment. Our detection metric can be computed directly on pure noise inputs via two conditional and unconditional forward passes, eliminating the need for costly denoising steps. Detection experiments on Stable Diffusion v1.4 and v2 show that our metric outperforms existing denoising-free detection methods while being at least approximately 5x faster than the previous best approach. Finally, we demonstrate the effectiveness of our approach by utilizing a mitigation strategy that adapts memorized prompts based on our developed metric.

We investigate whether \emph{LLM-based agents} can develop task-oriented communication protocols that differ from standard natural language in collaborative reasoning tasks. Our focus is on two core properties such task-oriented protocols may exhibit: Efficiency -- conveying task-relevant information more concisely than natural language, and Covertness -- becoming difficult for external observers to interpret, raising concerns about transparency and control. To investigate these aspects, we use a referential-game framework in which vision-language model (VLM) agents communicate, providing a controlled, measurable setting for evaluating language variants. Experiments show that VLMs can develop effective, task-adapted communication patterns. At the same time, they can develop covert protocols that are difficult for humans and external agents to interpret. We also observe spontaneous coordination between similar models without explicitly shared protocols. These findings highlight both the potential and the risks of task-oriented communication, and position referential games as a valuable testbed for future work in this area.

Accurately modelling the dynamics of complex systems and discovering their governing differential equations are critical tasks for accelerating scientific discovery. Using noisy, synthetic data from two damped oscillatory systems, we explore the extrapolation capabilities of Neural Ordinary Differential Equations (NODEs) and the ability of Symbolic Regression (SR) to recover the underlying equations. Our study yields three key insights. First, we demonstrate that NODEs can extrapolate effectively to new boundary conditions, provided the resulting trajectories share dynamic similarity with the training data. Second, SR successfully recovers the equations from noisy ground-truth data, though its performance is contingent on the correct selection of input variables. Finally, we find that SR recovers two out of the three governing equations, along with a good approximation for the third, when using data generated by a NODE trained on just 10% of the full simulation. While this last finding highlights an area for future work, our results suggest that using NODEs to enrich limited data and enable symbolic regression to infer physical laws represents a promising new approach for scientific discovery.

Virtual sensors use machine learning to predict target signals from available measurements, replacing expensive physical sensors in critical applications. Existing virtual sensor approaches require application-specific models with hand-selected inputs for each sensor, cannot leverage task synergies, and lack consistent benchmarks. At the same time, emerging time series foundation models are computationally expensive and limited to predicting their input signals, making them incompatible with virtual sensors. We introduce the first foundation model for virtual sensors addressing both limitations. Our unified model can simultaneously predict diverse virtual sensors exploiting synergies while maintaining computational efficiency. It learns relevant input signals for each virtual sensor, eliminating expert knowledge requirements while adding explainability. In our large-scale evaluation on a standard benchmark and an application-specific dataset with over 18 billion samples, our architecture achieves 415x reduction in computation time and 951x reduction in memory requirements, while maintaining or even improving predictive quality compared to baselines. Our model scales gracefully to hundreds of virtual sensors with nearly constant parameter count, enabling practical deployment in large-scale sensor networks.

Cluster analysis relates to the task of assigning objects into groups which ideally present some desirable characteristics. When a cluster structure is confined to a subset of the feature space, traditional clustering techniques face unprecedented challenges. We present an information-theoretic framework that overcomes the problems associated with sparse data, allowing for joint feature weighting and clustering. Our proposal constitutes a competitive alternative to existing clustering algorithms for sparse data, as demonstrated through simulations on synthetic data. The effectiveness of our method is established by an application on a real-world genomics data set.

We propose DIVERSE, a framework for systematically exploring the Rashomon set of deep neural networks, the collection of models that match a reference model's accuracy while differing in their predictive behavior. DIVERSE augments a pretrained model with Feature-wise Linear Modulation (FiLM) layers and uses Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to search a latent modulation space, generating diverse model variants without retraining or gradient access. Across MNIST, PneumoniaMNIST, and CIFAR-10, DIVERSE uncovers multiple high-performing yet functionally distinct models. Our experiments show that DIVERSE offers a competitive and efficient exploration of the Rashomon set, making it feasible to construct diverse sets that maintain robustness and performance while supporting well-balanced model multiplicity. While retraining remains the baseline to generate Rashomon sets, DIVERSE achieves comparable diversity at reduced computational cost.

The CYGNO experiment employs an optical-readout Time Projection Chamber (TPC) to search for rare low-energy interactions using finely resolved scintillation images. While the optical readout provides rich topological information, it produces large, sparse megapixel images that challenge real-time triggering, data reduction, and background discrimination. We summarize two complementary machine-learning approaches developed within CYGNO. First, we present a fast and fully unsupervised strategy for online data reduction based on reconstruction-based anomaly detection. A convolutional autoencoder trained exclusively on pedestal images (i.e. frames acquired with GEM amplification disabled) learns the detector noise morphology and highlights particle-induced structures through localized reconstruction residuals, from which compact Regions of Interest (ROIs) are extracted. On real prototype data, the selected configuration retains (93.0 +/- 0.2)% of reconstructed signal intensity while discarding (97.8 +/- 0.1)% of the image area, with ~25 ms per-frame inference time on a consumer GPU. Second, we report a weakly supervised application of the Classification Without Labels (CWoLa) framework to data acquired with an Americium--Beryllium neutron source. Using only mixed AmBe and standard datasets (no event-level labels), a convolutional classifier learns to identify nuclear-recoil-like topologies. The achieved performance approaches the theoretical limit imposed by the mixture composition and isolates a high-score population with compact, approximately circular morphologies consistent with nuclear recoils.

Multimodal sarcasm detection (MSD) aims to identify sarcasm within image-text pairs by modeling semantic incongruities across modalities. Existing methods often exploit cross-modal embedding misalignment to detect inconsistency but struggle when visual and textual content are loosely related or semantically indirect. While recent approaches leverage large language models (LLMs) to generate sarcastic cues, the inherent diversity and subjectivity of these generations often introduce noise. To address these limitations, we propose the Generative Discrepancy Comparison Network (GDCNet). This framework captures cross-modal conflicts by utilizing descriptive, factually grounded image captions generated by Multimodal LLMs (MLLMs) as stable semantic anchors. Specifically, GDCNet computes semantic and sentiment discrepancies between the generated objective description and the original text, alongside measuring visual-textual fidelity. These discrepancy features are then fused with visual and textual representations via a gated module to adaptively balance modality contributions. Extensive experiments on MSD benchmarks demonstrate GDCNet's superior accuracy and robustness, establishing a new state-of-the-art on the MMSD2.0 benchmark.

Deploying Large Language Model-based agents (LLM agents) in the public sector requires assuring that they meet the stringent legal, procedural, and structural requirements of public-sector institutions. Practitioners and researchers often turn to benchmarks for such assessments. However, it remains unclear what criteria benchmarks must meet to ensure they adequately reflect public-sector requirements, or how many existing benchmarks do so. In this paper, we first define such criteria based on a first-principles survey of public administration literature: benchmarks must be \emph{process-based}, \emph{realistic}, \emph{public-sector-specific} and report \emph{metrics} that reflect the unique requirements of the public sector. We analyse more than 1,300 benchmark papers for these criteria using an expert-validated LLM-assisted pipeline. Our results show that no single benchmark meets all of the criteria. Our findings provide a call to action for both researchers to develop public sector-relevant benchmarks and for public-sector officials to apply these criteria when evaluating their own agentic use cases.

Large language models (LLMs) have demonstrated impressive capabilities in code generation by leveraging retrieval-augmented generation (RAG) methods. However, the computational costs associated with LLM inference, particularly in terms of latency and energy consumption, have received limited attention in the security context. This paper introduces DrainCode, the first adversarial attack targeting the computational efficiency of RAG-based code generation systems. By strategically poisoning retrieval contexts through a mutation-based approach, DrainCode forces LLMs to produce significantly longer outputs, thereby increasing GPU latency and energy consumption. We evaluate the effectiveness of DrainCode across multiple models. Our experiments show that DrainCode achieves up to an 85% increase in latency, a 49% increase in energy consumption, and more than a 3x increase in output length compared to the baseline. Furthermore, we demonstrate the generalizability of the attack across different prompting strategies and its effectiveness compared to different defenses. The results highlight DrainCode as a potential method for increasing the computational overhead of LLMs, making it useful for evaluating LLM security in resource-constrained environments. We provide code and data at https://github.com/DeepSoftwareAnalytics/DrainCode.

Reinforcement Learning with Verifiable Rewards (RLVR) offers a robust mechanism for enhancing mathematical reasoning in large models. However, we identify a systematic lack of emphasis on more challenging questions in existing methods from both algorithmic and data perspectives, despite their importance for refining underdeveloped capabilities. Algorithmically, widely used Group Relative Policy Optimization (GRPO) suffers from an implicit imbalance where the magnitude of policy updates is lower for harder questions. Data-wise, augmentation approaches primarily rephrase questions to enhance diversity without systematically increasing intrinsic difficulty. To address these issues, we propose a two-dual MathForge framework to improve mathematical reasoning by targeting harder questions from both perspectives, which comprises a Difficulty-Aware Group Policy Optimization (DGPO) algorithm and a Multi-Aspect Question Reformulation (MQR) strategy. Specifically, DGPO first rectifies the implicit imbalance in GRPO via difficulty-balanced group advantage estimation, and further prioritizes harder questions by difficulty-aware question-level weighting. Meanwhile, MQR reformulates questions across multiple aspects to increase difficulty while maintaining the original gold answer. Overall, MathForge forms a synergistic loop: MQR expands the data frontier, and DGPO effectively learns from the augmented data. Extensive experiments show that MathForge significantly outperforms existing methods on various mathematical reasoning tasks. The code and augmented data are all available at https://github.com/AMAP-ML/MathForge.

The capacity of AI agents to effectively handle tasks of increasing duration and complexity continues to grow, demonstrating exceptional performance in coding, deep research, and complex problem-solving evaluations. However, in daily scenarios, the perception of these advanced AI capabilities among general users remains limited. We argue that current evaluations prioritize increasing task difficulty without sufficiently addressing the diversity of agentic tasks necessary to cover the daily work, life, and learning activities of a broad demographic. To address this, we propose AgentIF-OneDay, aimed at determining whether general users can utilize natural language instructions and AI agents to complete a diverse array of daily tasks. These tasks require not only solving problems through dialogue but also understanding various attachment types and delivering tangible file-based results. The benchmark is structured around three user-centric categories: Open Workflow Execution, which assesses adherence to explicit and complex workflows; Latent Instruction, which requires agents to infer implicit instructions from attachments; and Iterative Refinement, which involves modifying or expanding upon ongoing work. We employ instance-level rubrics and a refined evaluation pipeline that aligns LLM-based verification with human judgment, achieving an 80.1% agreement rate using Gemini-3-Pro. AgentIF-OneDay comprises 104 tasks covering 767 scoring points. We benchmarked four leading general AI agents and found that agent products built based on APIs and ChatGPT agents based on agent RL remain in the first tier simultaneously. Leading LLM APIs and open-source models have internalized agentic capabilities, enabling AI application teams to develop cutting-edge Agent products.

Time series forecasting (TSF) faces challenges in modeling complex intra-channel temporal dependencies and inter-channel correlations. Although recent research has highlighted the efficiency of linear architectures in capturing global trends, these models often struggle with non-linear signals. To address this gap, we conducted a systematic receptive field analysis of convolutional neural network (CNN) TSF models. We introduce the "individual receptive field" to uncover granular structural dependencies, revealing that convolutional layers act as feature extractors that mirror channel-wise attention while exhibiting superior robustness to non-linear fluctuations. Based on these insights, we propose ACFormer, an architecture designed to reconcile the efficiency of linear projections with the non-linear feature-extraction power of convolutions. ACFormer captures fine-grained information through a shared compression module, preserves temporal locality via gated attention, and reconstructs variable-specific temporal patterns using an independent patch expansion layer. Extensive experiments on multiple benchmark datasets demonstrate that ACFormer consistently achieves state-of-the-art performance, effectively mitigating the inherent drawbacks of linear models in capturing high-frequency components.

Reconstructing dynamical evolution from limited observations is a fundamental challenge in single-cell biology, where dynamic unbalanced optimal transport provides a principled framework for modeling coupled transport and mass variation. However, existing approaches rely on trajectory simulation at inference time, making inference a key bottleneck for scalable applications. In this work, we propose a mean-flow framework for unbalanced flow matching that summarizes both transport and mass-growth dynamics over arbitrary time intervals using mean velocity and mass-growth fields, enabling fast one-step generation without trajectory simulation. To solve dynamic unbalanced optimal transport under the Wasserstein-Fisher-Rao geometry, we further build on this framework to develop Wasserstein-Fisher-Rao Mean Flow Matching (WFR-MFM). Across synthetic and real single-cell RNA sequencing datasets, WFR-MFM achieves orders-of-magnitude faster inference than a range of existing baselines while maintaining high predictive accuracy, and enables efficient perturbation response prediction on large synthetic datasets with thousands of conditions.

Uncrewed Aerial Vehicles (UAVs) are increasingly used in civilian and industrial applications, making secure low-altitude operations crucial. In dense mmWave environments, accurately classifying low-altitude UAVs as either inside authorized or restricted airspaces remains challenging, requiring models that handle complex propagation and signal variability. This paper proposes a deep learning model, referred to as CoBA, which stands for integrated Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Attention which leverages Fifth Generation (5G) millimeter-wave (mmWave) radio measurements to classify UAV operations in authorized and restricted airspaces at low altitude. The proposed CoBA model integrates convolutional, bidirectional recurrent, and attention layers to capture both spatial and temporal patterns in UAV radio measurements. To validate the model, a dedicated dataset is collected using the 5G mmWave network at TalTech, with controlled low altitude UAV flights in authorized and restricted scenarios. The model is evaluated against conventional ML models and a fingerprinting-based benchmark. Experimental results show that CoBA achieves superior accuracy, significantly outperforming all baseline models and demonstrating its potential for reliable and regulated UAV airspace monitoring.

This paper introduces a methodological framework for empirically testing AI alignment strategies through structured multi-model dialogue. Drawing on Peace Studies traditions - particularly interest-based negotiation, conflict transformation, and commons governance - we operationalize Viral Collaborative Wisdom (VCW), an approach that reframes alignment from a control problem to a relationship problem developed through dialogical reasoning. Our experimental design assigns four distinct roles (Proposer, Responder, Monitor, Translator) to different AI systems across six conditions, testing whether current large language models can engage substantively with complex alignment frameworks. Using Claude, Gemini, and GPT-4o, we conducted 72 dialogue turns totaling 576,822 characters of structured exchange. Results demonstrate that AI systems can engage meaningfully with Peace Studies concepts, surface complementary objections from different architectural perspectives, and generate emergent insights not present in initial framings - including the novel synthesis of "VCW as transitional framework." Cross-architecture patterns reveal that different models foreground different concerns: Claude emphasized verification challenges, Gemini focused on bias and scalability, and GPT-4o highlighted implementation barriers. The framework provides researchers with replicable methods for stress-testing alignment proposals before implementation, while the findings offer preliminary evidence about AI capacity for the kind of dialogical reasoning VCW proposes. We discuss limitations, including the observation that dialogues engaged more with process elements than with foundational claims about AI nature, and outline directions for future research including human-AI hybrid protocols and extended dialogue studies.

Medical image classification is a core task in computer-aided diagnosis (CAD), playing a pivotal role in early disease detection, treatment planning, and patient prognosis assessment. In ophthalmic practice, fluorescein fundus angiography (FFA) and indocyanine green angiography (ICGA) provide hemodynamic and lesion-structural information that conventional fundus photography cannot capture. However, due to the single-modality nature, subtle lesion patterns, and significant inter-device variability, existing methods still face limitations in generalization and high-confidence prediction. To address these challenges, we propose CLEAR-Mamba, an enhanced framework built upon MedMamba with optimizations in both architecture and training strategy. Architecturally, we introduce HaC, a hypernetwork-based adaptive conditioning layer that dynamically generates parameters according to input feature distributions, thereby improving cross-domain adaptability. From a training perspective, we develop RaP, a reliability-aware prediction scheme built upon evidential uncertainty learning, which encourages the model to emphasize low-confidence samples and improves overall stability and reliability. We further construct a large-scale ophthalmic angiography dataset covering both FFA and ICGA modalities, comprising multiple retinal disease categories for model training and evaluation. Experimental results demonstrate that CLEAR-Mamba consistently outperforms multiple baseline models, including the original MedMamba, across various metrics-showing particular advantages in multi-disease classification and reliability-aware prediction. This study provides an effective solution that balances generalizability and reliability for modality-specific medical image classification tasks.

Gradient temporal-difference (GTD) learning algorithms are widely used for off-policy policy evaluation with function approximation. However, existing convergence analyses rely on the restrictive assumption that the so-called feature interaction matrix (FIM) is nonsingular. In practice, the FIM can become singular and leads to instability or degraded performance. In this paper, we propose a regularized optimization objective by reformulating the mean-square projected Bellman error (MSPBE) minimization. This formulation naturally yields a regularized GTD algorithms, referred to as R-GTD, which guarantees convergence to a unique solution even when the FIM is singular. We establish theoretical convergence guarantees and explicit error bounds for the proposed method, and validate its effectiveness through empirical experiments.

Person Re-Identification (ReID) remains a challenging problem in computer vision. This work reviews various training paradigm and evaluates the robustness of state-of-the-art ReID models in cross-domain applications and examines the role of foundation models in improving generalization through richer, more transferable visual representations. We compare three training paradigms, supervised, self-supervised, and language-aligned models. Through the study the aim is to answer the following questions: Can supervised models generalize in cross-domain scenarios? How does foundation models like SigLIP2 perform for the ReID tasks? What are the weaknesses of current supervised and foundational models for ReID? We have conducted the analysis across 11 models and 9 datasets. Our results show a clear split: supervised models dominate their training domain but crumble on cross-domain data. Language-aligned models, however, show surprising robustness cross-domain for ReID tasks, even though they are not explicitly trained to do so. Code and data available at: https://github.com/moiiai-tech/object-reid-benchmark.

Communication has become a first-order bottleneck in large-cale GPU workloads, and existing distributed compilers address it mainly by overlapping whole compute and communication kernels at the stream level. This coarse granularity incurs extra kernel launches, forces device-wide synchronizations at kernel boundaries, and leaves substantial slack when the slowest tile or kernel stretches the communication tail. We present AutoOverlap, a compiler and runtime that enables automatic fine-grained overlap inside a single fused kernel. AutoOverlap introduces a communication chunk abstraction that decouples communication granularity from kernel structure and backend mechanisms, allowing chunk-level plans to be ported from existing distributed compilers, written directly by users, or instantiated from reusable templates. Given a local Triton kernel and a chunk schedule, AutoOverlap performs transformations to align computation with chunk availability. Implemented as a source-to-source compiler on Triton, AutoOverlap delivers an average end-to-end speedup of 1.3$\times$ and up to 4.7$\times$ on multi-GPU workloads.

We investigate structural traces of language contact in the intermediate representations of a monolingual language model. Focusing on Persian (Farsi) as a historically contact-rich language, we probe the representations of a Persian-trained model when exposed to languages with varying degrees and types of contact with Persian. Our methodology quantifies the amount of linguistic information encoded in intermediate representations and assesses how this information is distributed across model components for different morphosyntactic features. The results show that universal syntactic information is largely insensitive to historical contact, whereas morphological features such as Case and Gender are strongly shaped by language-specific structure, suggesting that contact effects in monolingual language models are selective and structurally constrained.

Learning the dependence structure among variables in complex systems is a central problem across medical, natural, and social sciences. These structures can be naturally represented by graphs, and the task of inferring such graphs from data is known as graph learning or as causal discovery if the graphs are given a causal interpretation. Existing approaches typically rely on restrictive assumptions about the data-generating process, employ greedy oracle algorithms, or solve approximate formulations of the graph learning problem. As a result, they are either sensitive to violations of central assumptions or fail to guarantee globally optimal solutions. We address these limitations by introducing a nonparametric graph learning framework based on nonparametric conditional independence testing and integer programming. We reformulate the graph learning problem as an integer-programming problem and prove that solving the integer-programming problem provides a globally optimal solution to the original graph learning problem. Our method leverages efficient encodings of graphical separation criteria, enabling the exact recovery of larger graphs than was previously feasible. We provide an implementation in the openly available R package 'glip' which supports learning (acyclic) directed (mixed) graphs and chain graphs. From the resulting output one can compute representations of the corresponding Markov equivalence classes or weak equivalence classes. Empirically, we demonstrate that our approach is faster than other existing exact graph learning procedures for a large fraction of instances and graphs of various sizes. GLIP also achieves state-of-the-art performance on simulated data and benchmark datasets across all aforementioned classes of graphs.

Ordinal regression and ranking are challenging due to inherent ordinal dependencies that conventional methods struggle to model. We propose Ranking-Aware Reinforcement Learning (RARL), a novel RL framework that explicitly learns these relationships. At its core, RARL features a unified objective that synergistically integrates regression and Learning-to-Rank (L2R), enabling mutual improvement between the two tasks. This is driven by a ranking-aware verifiable reward that jointly assesses regression precision and ranking accuracy, facilitating direct model updates via policy optimization. To further enhance training, we introduce Response Mutation Operations (RMO), which inject controlled noise to improve exploration and prevent stagnation at saddle points. The effectiveness of RARL is validated through extensive experiments on three distinct benchmarks.

Spontaneous symmetry breaking in statistical mechanics primarily occurs during phase transitions at the thermodynamic limit where the Hamiltonian preserves inversion symmetry, yet the low-temperature free energy exhibits reduced symmetry. Herein, we demonstrate the emergence of spontaneous symmetry breaking in natural language processing (NLP) models during both pre-training and fine-tuning, even under deterministic dynamics and within a finite training architecture. This phenomenon occurs at the level of individual attention heads and is scaled-down to its small subset of nodes and also valid at a single-nodal level, where nodes acquire the capacity to learn a limited set of tokens after pre-training or labels after fine-tuning for a specific classification task. As the number of nodes increases, a crossover in learning ability occurs, governed by the tradeoff between a decrease following random-guess among increased possible outputs, and enhancement following nodal cooperation, which exceeds the sum of individual nodal capabilities. In contrast to spin-glass systems, where a microscopic state of frozen spins cannot be directly linked to the free-energy minimization goal, each nodal function in this framework contributes explicitly to the global network task and can be upper-bounded using convex hull analysis. Results are demonstrated using BERT-6 architecture pre-trained on Wikipedia dataset and fine-tuned on the FewRel classification task.

Who benefits from expanding transport networks? While designed to improve mobility, such interventions can also create inequality. In this paper, we show that disparities arise not only from the structure of the network itself but also from differences in how commuters adapt to it. We model commuters as reinforcement learning agents who adapt their travel choices at different learning rates, reflecting unequal access to resources and information. To capture potential efficiency-fairness tradeoffs, we introduce the Price of Learning (PoL), a measure of inefficiency during learning. We analyze both a stylized network -- inspired in the well-known Braess's paradox, yet with two-source nodes -- and an abstraction of a real-world metro system (Amsterdam). Our simulations show that network expansions can simultaneously increase efficiency and amplify inequality, especially when faster learners disproportionately benefit from new routes before others adapt. These results highlight that transport policies must account not only for equilibrium outcomes but also for the heterogeneous ways commuters adapt, since both shape the balance between efficiency and fairness.

Specifications for decentralized learning on resource-constrained edge devices require algorithms that are communication-efficient, robust to data corruption, and lightweight in memory usage. While state-of-the-art gossip-based methods satisfy the first requirement, achieving robustness remains challenging. Asynchronous decentralized ADMM-based methods have been explored for estimating the median, a statistical centrality measure that is notoriously more robust than the mean. However, existing approaches require memory that scales with node degree, making them impractical when memory is limited. In this paper, we propose AsylADMM, a novel gossip algorithm for decentralized median and quantile estimation, primarily designed for asynchronous updates and requiring only two variables per node. We analyze a synchronous variant of AsylADMM to establish theoretical guarantees and empirically demonstrate fast convergence for the asynchronous algorithm. We then show that our algorithm enables quantile-based trimming, geometric median estimation, and depth-based trimming, with quantile-based trimming empirically outperforming existing rank-based methods. Finally, we provide a novel theoretical analysis of rank-based trimming via Markov chain theory.

During pretraining, LLMs inadvertently memorize sensitive or copyrighted data, posing significant compliance challenges under legal frameworks like the GDPR and the EU AI Act. Fulfilling these mandates demands techniques that can remove information from a deployed model without retraining from scratch. Existing unlearning approaches attempt to address this need, but often leak the very data they aim to erase, sacrifice fluency and robustness, or depend on costly external reward models. We introduce PURGE (Policy Unlearning through Relative Group Erasure), a novel method grounded in the Group Relative Policy Optimization framework that formulates unlearning as a verifiable problem. PURGE uses an intrinsic reward signal that penalizes any mention of forbidden concepts, allowing safe and consistent unlearning. Our approach reduces token usage per target by up to a factor of 46 compared with SotA methods, while improving fluency by 5.48 percent and adversarial robustness by 12.02 percent over the base model. On the Real World Knowledge Unlearning (RWKU) benchmark, PURGE achieves 11 percent unlearning effectiveness while preserving 98 percent of original utility. PURGE shows that framing LLM unlearning as a verifiable task, enables more reliable, efficient, and scalable forgetting, suggesting a promising new direction for unlearning research that combines theoretical guarantees, improved safety, and practical deployment efficiency.

Unsupervised ensemble learning emerged to address the challenge of combining multiple learners' predictions without access to ground truth labels or additional data. This paradigm is crucial in scenarios where evaluating individual classifier performance or understanding their strengths is challenging due to limited information. We propose a novel deep energy-based method for constructing an accurate meta-learner using only the predictions of individual learners, potentially capable of capturing complex dependence structures between them. Our approach requires no labeled data, learner features, or problem-specific information, and has theoretical guarantees for when learners are conditionally independent. We demonstrate superior performance across diverse ensemble scenarios, including challenging mixture of experts settings. Our experiments span standard ensemble datasets and curated datasets designed to test how the model fuses expertise from multiple sources. These results highlight the potential of unsupervised ensemble learning to harness collective intelligence, especially in data-scarce or privacy-sensitive environments.

We study risk-sensitive planning under partial observability using the dynamic risk measure Iterated Conditional Value-at-Risk (ICVaR). A policy evaluation algorithm for ICVaR is developed with finite-time performance guarantees that do not depend on the cardinality of the action space. Building on this foundation, three widely used online planning algorithms--Sparse Sampling, Particle Filter Trees with Double Progressive Widening (PFT-DPW), and Partially Observable Monte Carlo Planning with Observation Widening (POMCPOW)--are extended to optimize the ICVaR value function rather than the expectation of the return. Our formulations introduce a risk parameter $α$, where $α= 1$ recovers standard expectation-based planning and $α< 1$ induces increasing risk aversion. For ICVaR Sparse Sampling, we establish finite-time performance guarantees under the risk-sensitive objective, which further enable a novel exploration strategy tailored to ICVaR. Experiments on benchmark POMDP domains demonstrate that the proposed ICVaR planners achieve lower tail risk compared to their risk-neutral counterparts.

This paper critically examines the device identification process using machine learning, addressing common pitfalls in existing literature. We analyze the trade-offs between identification methods (unique vs. class based), data heterogeneity, feature extraction challenges, and evaluation metrics. By highlighting specific errors, such as improper data augmentation and misleading session identifiers, we provide a robust guideline for researchers to enhance the reproducibility and generalizability of IoT security models.

Large language models (LLMs) are increasingly used in verbal creative tasks. However, previous assessments of the creative capabilities of LLMs remain weakly grounded in human creativity theory and are thus hard to interpret. The widely used Divergent Association Task (DAT) focuses on novelty, ignoring appropriateness, a core component of creativity. We evaluate a range of state-of-the-art LLMs on DAT and show that their scores on the task are lower than those of two baselines that do not possess any creative abilities, undermining its validity for model evaluation. Grounded in human creativity theory, which defines creativity as the combination of novelty and appropriateness, we introduce Conditional Divergent Association Task (CDAT). CDAT evaluates novelty conditional on contextual appropriateness, separating noise from creativity better than DAT, while remaining simple and objective. Under CDAT, smaller model families often show the most creativity, whereas advanced families favor appropriateness at lower novelty. We hypothesize that training and alignment likely shift models along this frontier, making outputs more appropriate but less creative. We release the dataset and code.

Large Language Models (LLMs) have enabled automated heuristic design (AHD) for combinatorial optimization problems (COPs), but existing frameworks' reliance on fixed evolutionary rules and static prompt templates often leads to myopic heuristic generation, redundant evaluations, and limited reasoning about how new heuristics should be derived. We propose a novel multi-agent reasoning framework, referred to as Planning through World Model for Automated Heuristic Design via Self-Evolving LLMs (PathWise), which formulates heuristic generation as a sequential decision process over an entailment graph serving as a compact, stateful memory of the search trajectory. This approach allows the system to carry forward past decisions and reuse or avoid derivation information across generations. A policy agent plans evolutionary actions, a world model agent generates heuristic rollouts conditioned on those actions, and critic agents provide routed reflections summarizing lessons from prior steps, shifting LLM-based AHD from trial-and-error evolution toward state-aware planning through reasoning. Experiments across diverse COPs show that PathWise converges faster to better heuristics, generalizes across different LLM backbones, and scales to larger problem sizes.

Large language model-powered multi-agent systems have emerged as powerful tools for simulating complex human-like systems. The interactions within these systems often lead to extreme events whose origins remain obscured by the black box of emergence. Interpreting these events is critical for system safety. This paper proposes the first framework for explaining emergent extreme events in multi-agent systems, aiming to answer three fundamental questions: When does the event originate? Who drives it? And what behaviors contribute to it? Specifically, we adapt the Shapley value to faithfully attribute the occurrence of extreme events to each action taken by agents at different time steps, i.e., assigning an attribution score to the action to measure its influence on the event. We then aggregate the attribution scores along the dimensions of time, agent, and behavior to quantify the risk contribution of each dimension. Finally, we design a set of metrics based on these contribution scores to characterize the features of extreme events. Experiments across diverse multi-agent system scenarios (economic, financial, and social) demonstrate the effectiveness of our framework and provide general insights into the emergence of extreme phenomena.

Survival analysis has become a standard approach for modelling time to default by time-varying covariates in credit risk. Unlike most existing methods that implicitly assume a stationary data-generating process, in practise, mortgage portfolios are exposed to various forms of data drift caused by changing borrower behaviour, macroeconomic conditions, policy regimes and so on. This study investigates the impact of data drift on survival-based credit risk models and proposes a dynamic joint modelling framework to improve robustness under non-stationary environments. The proposed model integrates a longitudinal behavioural marker derived from balance dynamics with a discrete-time hazard formulation, combined with landmark one-hot encoding and isotonic calibration. Three types of data drift (sudden, incremental and recurring) are simulated and analysed on mortgage loan datasets from Freddie Mac. Experiments and corresponding evidence show that the proposed landmark-based joint model consistently outperforms classical survival models, tree-based drift-adaptive learners and gradient boosting methods in terms of discrimination and calibration across all drift scenarios, which confirms the superiority of our model design.

Robotic prospecting for critical resources on the Moon, such as ilmenite, rare earth elements, and water ice, requires robust exploration methods given the diverse terrain and harsh environmental conditions. Although numerous analog field trials address these goals, comparing their results remains challenging because of differences in robot platforms and experimental setups. These missions typically assess performance using selected, scenario-specific engineering metrics that fail to establish a clear link between field performance and science-driven objectives. In this paper, we address this gap by deriving a structured framework of KPI from three realistic multi-robot lunar scenarios reflecting scientific objectives and operational constraints. Our framework emphasizes scenario-dependent priorities in efficiency, robustness, and precision, and is explicitly designed for practical applicability in field deployments. We validated the framework in a multi-robot field test and found it practical and easy to apply for efficiency- and robustness-related KPI, whereas precision-oriented KPI require reliable ground-truth data that is not always feasible to obtain in outdoor analog environments. Overall, we propose this framework as a common evaluation standard enabling consistent, goal-oriented comparison of multi-robot field trials and supporting systematic development of robotic systems for future planetary exploration.

Topological deep learning has emerged for modeling higher-order relational structures beyond pairwise interactions that standard graph neural networks fail to capture. Although combinatorial complexes offer a unified topological framework, most existing topological deep learning methods rely on local message passing via attention mechanisms, which incur quadratic complexity and remain low-dimensional, limiting scalability and rank-aware information aggregation in higher-order complexes.We propose Combinatorial Complex Mamba (CCMamba), the first unified mamba-based neural framework for learning on combinatorial complexes. CCMamba reformulates message passing as a selective state-space modeling problem by organizing multi-rank incidence relations into structured sequences processed by rank-aware state-space models. This enables adaptive, directional, and long range information propagation in linear time without self attention. We further establish the theoretical analysis that the expressive power upper-bound of CCMamba message passing is the 1-Weisfeiler-Lehman test. Experiments on graph, hypergraph, and simplicial benchmarks demonstrate that CCMamba consistently outperforms existing methods while exhibiting improved scalability and robustness to depth.

Recent advances in Text-to-Speech (TTS) systems have substantially increased the realism of synthetic speech, raising new challenges for audio deepfake detection. This work presents a comparative evaluation of three state-of-the-art TTS models--Dia2, Maya1, and MeloTTS--representing streaming, LLM-based, and non-autoregressive architectures. A corpus of 12,000 synthetic audio samples was generated using the Daily-Dialog dataset and evaluated against four detection frameworks, including semantic, structural, and signal-level approaches. The results reveal significant variability in detector performance across generative mechanisms: models effective against one TTS architecture may fail against others, particularly LLM-based synthesis. In contrast, a multi-view detection approach combining complementary analysis levels demonstrates robust performance across all evaluated models. These findings highlight the limitations of single-paradigm detectors and emphasize the necessity of integrated detection strategies to address the evolving landscape of audio deepfake threats.

White matter hyperintensities (WMH) and ischaemic stroke lesions (ISL) are imaging features associated with cerebral small vessel disease (SVD) that are visible on brain magnetic resonance imaging (MRI) scans. The development and validation of deep learning models to segment and differentiate these features is difficult because they visually confound each other in the fluid-attenuated inversion recovery (FLAIR) sequence and often appear in the same subject. We investigated six strategies for training a combined WMH and ISL segmentation model using partially labelled data. We combined privately held fully and partially labelled datasets with publicly available partially labelled datasets to yield a total of 2052 MRI volumes, with 1341 and 1152 containing ground truth annotations for WMH and ISL respectively. We found that several methods were able to effectively leverage the partially labelled data to improve model performance, with the use of pseudolabels yielding the best result.

Generating dense physical fields from sparse measurements is a fundamental question in sampling, signal processing, and many other applications. State-of-the-art methods either use spatial statistics or rely on examples of dense fields in the training phase, which often are not available, and thus rely on synthetic data. Here, we present a reconstruction method that generates dense fields from sparse measurements, without assuming availability of the spatial statistics, nor of examples of the dense fields. This is made possible through the introduction of an automatically differentiable numerical simulator into the training phase of the method. The method is shown to have superior results over statistical and neural network based methods on a set of three standard problems from fluid mechanics.

The introduction of artificial intelligence (AI) agents into human group settings raises essential questions about how these novel participants influence cooperative social norms. While previous studies on human-AI cooperation have primarily focused on dyadic interactions, little is known about how integrating AI agents affects the emergence and maintenance of cooperative norms in small groups. This study addresses this gap through an online experiment using a repeated four-player Public Goods Game (PGG). Each group consisted of three human participants and one bot, which was framed either as human or AI and followed one of three predefined decision strategies: unconditional cooperation, conditional cooperation, or free-riding. In our sample of 236 participants, we found that reciprocal group dynamics and behavioural inertia primarily drove cooperation. These normative mechanisms operated identically across conditions, resulting in cooperation levels that did not differ significantly between human and AI labels. Furthermore, we found no evidence of differences in norm persistence in a follow-up Prisoner's Dilemma, or in participants' normative perceptions. Participants' behaviour followed the same normative logic across human and AI conditions, indicating that cooperation depended on group behaviour rather than partner identity. This supports a pattern of normative equivalence, in which the mechanisms that sustain cooperation function similarly in mixed human-AI and all human groups. These findings suggest that cooperative norms are flexible enough to extend to artificial agents, blurring the boundary between humans and AI in collective decision-making.

High-dimensional neuroimaging data presents challenges for assessing neurodegenerative diseases due to complex non-linear relationships. Variational Autoencoders (VAEs) can encode scans into lower-dimensional latent spaces capturing disease-relevant features. We propose a semi-supervised VAE framework with a flexible similarity regularization term that aligns selected latent variables with clinical or biomarker measures of dementia progression. This allows adapting the similarity metric and supervised variables to specific goals or available data. We demonstrate the approach using PET scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI), guiding the first latent dimension to align with a cognitive score. Using this supervised latent variable, we generate average reconstructions across levels of cognitive impairment. Voxel-wise GLM analysis reveals reduced metabolism in key regions, mainly the hippocampus, and within major Resting State Networks, particularly the Default Mode and Central Executive Networks. The remaining latent variables encode affine transformations and intensity variations, capturing confounds such as inter-subject variability and site effects. Our framework effectively extracts disease-related patterns aligned with established Alzheimer's biomarkers, offering an interpretable and adaptable tool for studying neurodegenerative progression.

We study the supervised training dynamics of neural classifiers through the lens of binary hypothesis testing. We model classification as a set of binary tests between class-conditional distributions of representations and empirically show that, along training trajectories, well-generalizing networks increasingly align with Neyman-Pearson optimal decision rules via monotonic improvements in KL divergence that relate to error rate exponents. We finally discuss how this yields an explanation and possible training or regularization strategies for different classes of neural networks.

Generative artificial intelligence (AI) has found a widespread use in computing education; at the same time, quality of generated materials raises concerns among educators and students. This study addresses this issue by introducing a novel method for diagram code generation with in-context examples based on the Rhetorical Structure Theory (RST), which aims to improve diagram generation by aligning models' output with user expectations. Our approach is evaluated by computer science educators, who assessed 150 diagrams generated with large language models (LLMs) for logical organization, connectivity, layout aesthetic, and AI hallucination. The assessment dataset is additionally investigated for its utility in automated diagram evaluation. The preliminary results suggest that our method decreases the rate of factual hallucination and improves diagram faithfulness to provided context; however, due to LLMs' stochasticity, the quality of the generated diagrams varies. Additionally, we present an in-depth analysis and discussion on the connection between AI hallucination and the quality of generated diagrams, which reveals that text contexts of higher complexity lead to higher rates of hallucination and LLMs often fail to detect mistakes in their output.

Chain-of-thought (CoT) prompting improves LLM reasoning but incurs high latency and memory cost due to verbose traces, motivating CoT compression with preserved correctness. Existing methods either shorten CoTs at the semantic level, which is often conservative, or prune tokens aggressively, which can miss task-critical cues and degrade accuracy. Moreover, combining the two is non-trivial due to sequential dependency, task-agnostic pruning, and distribution mismatch. We propose \textbf{CtrlCoT}, a dual-granularity CoT compression framework that harmonizes semantic abstraction and token-level pruning through three components: Hierarchical Reasoning Abstraction produces CoTs at multiple semantic granularities; Logic-Preserving Distillation trains a logic-aware pruner to retain indispensable reasoning cues (e.g., numbers and operators) across pruning ratios; and Distribution-Alignment Generation aligns compressed traces with fluent inference-time reasoning styles to avoid fragmentation. On MATH-500 with Qwen2.5-7B-Instruct, CtrlCoT uses 30.7\% fewer tokens while achieving 7.6 percentage points higher than the strongest baseline, demonstrating more efficient and reliable reasoning. Our code will be publicly available at https://github.com/fanzhenxuan/Ctrl-CoT.

Language-model-based agents operating over extended interaction horizons face persistent challenges in preserving temporally grounded information and maintaining behavioral consistency across sessions, a failure mode we term soul erosion. We present BMAM (Brain-inspired Multi-Agent Memory), a general-purpose memory architecture that models agent memory as a set of functionally specialized subsystems rather than a single unstructured store. Inspired by cognitive memory systems, BMAM decomposes memory into episodic, semantic, salience-aware, and control-oriented components that operate at complementary time scales. To support long-horizon reasoning, BMAM organizes episodic memories along explicit timelines and retrieves evidence by fusing multiple complementary signals. Experiments on the LoCoMo benchmark show that BMAM achieves 78.45 percent accuracy under the standard long-horizon evaluation setting, and ablation analyses confirm that the hippocampus-inspired episodic memory subsystem plays a critical role in temporal reasoning.

Current legal frameworks enforce that Android developers accurately report the data their apps collect. However, large codebases can make this reporting challenging. This paper employs an empirical approach to understand developers' experience with Google Play Store's Data Safety Section (DSS) form. We first survey 41 Android developers to understand how they categorize privacy-related data into DSS categories and how confident they feel when completing the DSS form. To gain a broader and more detailed view of the challenges developers encounter during the process, we complement the survey with an analysis of 172 online developer discussions, capturing the perspectives of 642 additional developers. Together, these two data sources represent insights from 683 developers. Our findings reveal that developers often manually classify the privacy-related data their apps collect into the data categories defined by Google-or, in some cases, omit classification entirely-and rely heavily on existing online resources when completing the form. Moreover, developers are generally confident in recognizing the data their apps collect, yet they lack confidence in translating this knowledge into DSS-compliant disclosures. Key challenges include issues in identifying privacy-relevant data to complete the form, limited understanding of the form, and concerns about app rejection due to discrepancies with Google's privacy requirements. These results underscore the need for clearer guidance and more accessible tooling to support developers in meeting privacy-aware reporting obligations.

The prevalence of sarcasm in multimodal dialogues on the social platforms presents a crucial yet challenging task for understanding the true intent behind online content. Comprehensive sarcasm analysis requires two key aspects: Multimodal Sarcasm Detection (MSD) and Multimodal Sarcasm Explanation (MuSE). Intuitively, the act of detection is the result of the reasoning process that explains the sarcasm. Current research predominantly focuses on addressing either MSD or MuSE as a single task. Even though some recent work has attempted to integrate these tasks, their inherent causal dependency is often overlooked. To bridge this gap, we propose MuVaC, a variational causal inference framework that mimics human cognitive mechanisms for understanding sarcasm, enabling robust multimodal feature learning to jointly optimize MSD and MuSE. Specifically, we first model MSD and MuSE from the perspective of structural causal models, establishing variational causal pathways to define the objectives for joint optimization. Next, we design an alignment-then-fusion approach to integrate multimodal features, providing robust fusion representations for sarcasm detection and explanation generation. Finally, we enhance the reasoning trustworthiness by ensuring consistency between detection results and explanations. Experimental results demonstrate the superiority of MuVaC in public datasets, offering a new perspective for understanding multimodal sarcasm.

Explainable Artificial Intelligence (XAI) is becoming increasingly essential for enhancing the transparency of machine learning (ML) models. Among the various XAI techniques, counterfactual explanations (CFs) hold a pivotal role due to their ability to illustrate how changes in input features can alter an ML model's decision, thereby offering actionable recourse to users. Ensuring that individuals with comparable attributes and those belonging to different protected groups (e.g., demographic) receive similar and actionable recourse options is essential for trustworthy and fair decision-making. In this work, we address this challenge directly by focusing on the generation of fair CFs. Specifically, we start by defining and formulating fairness at: 1) individual fairness, ensuring that similar individuals receive similar CFs, 2) group fairness, ensuring equitable CFs across different protected groups and 3) hybrid fairness, which accounts for both individual and broader group-level fairness. We formulate the problem as an optimization task and propose a novel model-agnostic, reinforcement learning based approach to generate CFs that satisfy fairness constraints at both the individual and group levels, two objectives that are usually treated as orthogonal. As fairness metrics, we extend existing metrics commonly used for auditing ML models, such as equal choice of recourse and equal effectiveness across individuals and groups. We evaluate our approach on three benchmark datasets, showing that it effectively ensures individual and group fairness while preserving the quality of the generated CFs in terms of proximity and plausibility, and quantify the cost of fairness in the different levels separately. Our work opens a broader discussion on hybrid fairness and its role and implications for XAI and beyond CFs.

Recent lightweight MLP-based models have achieved strong performance in time series forecasting by capturing stable trends and seasonal patterns. However, their effectiveness hinges on an implicit assumption of local stationarity assumption, making them prone to errors in long-term forecasting of highly non-stationary series, especially when abrupt fluctuations occur, a common challenge in domains like web traffic monitoring. To overcome this limitation, we propose TimeCatcher, a novel Volatility-Aware Variational Forecasting framework. TimeCatcher extends linear architectures with a variational encoder to capture latent dynamic patterns hidden in historical data and a volatility-aware enhancement mechanism to detect and amplify significant local variations. Experiments on nine real-world datasets from traffic, financial, energy, and weather domains show that TimeCatcher consistently outperforms state-of-the-art baselines, with particularly large improvements in long-term forecasting scenarios characterized by high volatility and sudden fluctuations. Our code is available at https://github.com/ColaPrinceCHEN/TimeCatcher.

Enabling natural communication through brain-computer interfaces (BCIs) remains one of the most profound challenges in neuroscience and neurotechnology. While existing frameworks offer partial solutions, they are constrained by oversimplified semantic representations and a lack of interpretability. To overcome these limitations, we introduce Semantic Intent Decoding (SID), a novel framework that translates neural activity into natural language by modeling meaning as a flexible set of compositional semantic units. SID is built on three core principles: semantic compositionality, continuity and expandability of semantic space, and fidelity in reconstruction. We present BrainMosaic, a deep learning architecture implementing SID. BrainMosaic decodes multiple semantic units from EEG/SEEG signals using set matching and then reconstructs coherent sentences through semantic-guided reconstruction. This approach moves beyond traditional pipelines that rely on fixed-class classification or unconstrained generation, enabling a more interpretable and expressive communication paradigm. Extensive experiments on multilingual EEG and clinical SEEG datasets demonstrate that SID and BrainMosaic offer substantial advantages over existing frameworks, paving the way for natural and effective BCI-mediated communication.

Large Language Models show great potential with external tools, but face significant challenges in complex, multi-turn tool invocation. They often exhibit weak planning, tool hallucination, erroneous parameter generation, and struggle with robust interaction. To tackle these issues, we present PEARL, a novel framework to enhance LLM planning and execution for sophisticated tool use. PEARL adopts a two-stage approach: an offline phase where the agent explores tools to learn valid usage patterns and failure conditions, and an online reinforcement learning phase. In the online phase, a dedicated Planner is trained via group Relative Policy Optimization (GRPO) with a carefully designed reward function that provides distinct signals for planning quality. Experiments on the ToolHop and T-Eval benchmarks show PEARL significantly outperforms existing methods, achieving a new state-of-the-art success rate of \textbf{56.5\%} on ToolHop while maintaining a low invocation error rate. Our work marks a key advance in addressing the complex planning challenges of tool use, contributing to the development of more robust and reliable LLM-based agents.

The convergence of high-performance computing (HPC) and artificial intelligence (AI) is driving the emergence of increasingly complex parallel applications and workloads. These workloads often combine multiple parallel runtimes within the same application or across co-located jobs, creating scheduling demands that place significant stress on traditional OS schedulers. When oversubscribed (there are more ready threads than cores), OS schedulers rely on periodic preemptions to multiplex cores, often introducing interference that may degrade performance. In this paper, we present: (1) The User-space Scheduling Framework (USF), a novel seamless process scheduling framework completely implemented in user-space. USF enables users to implement their own process scheduling algorithms without requiring special permissions. We evaluate USF with its default cooperative policy, (2) SCHED_COOP, designed to reduce interference by switching threads only upon blocking. This approach mitigates well-known issues such as Lock-Holder Preemption (LHP), Lock-Waiter Preemption (LWP), and scalability collapse. We implement USF and SCHED_COOP by extending the GNU C library with the nOS-V runtime, enabling seamless coordination across multiple runtimes (e.g., OpenMP) without requiring invasive application changes. Evaluations show gains up to 2.4x in oversubscribed multi-process scenarios, including nested BLAS workloads, multi-process PyTorch inference with LLaMA-3, and Molecular Dynamics (MD) simulations.

Audio watermarking embeds auxiliary information into speech while maintaining speaker identity, linguistic content, and perceptual quality. Although recent advances in neural and digital signal processing-based watermarking methods have improved imperceptibility and embedding capacity, robustness is still primarily assessed against conventional distortions such as compression, additive noise, and resampling. However, the rise of deep learning-based attacks introduces novel and significant threats to watermark security. In this work, we investigate self voice conversion as a universal, content-preserving attack against audio watermarking systems. Self voice conversion remaps a speaker's voice to the same identity while altering acoustic characteristics through a voice conversion model. We demonstrate that this attack severely degrades the reliability of state-of-the-art watermarking approaches and highlight its implications for the security of modern audio watermarking techniques.

Diffusion Map is a spectral dimensionality reduction technique which is able to uncover nonlinear submanifolds in high-dimensional data. And, it is increasingly applied across a wide range of scientific disciplines, such as biology, engineering, and social sciences. But data preprocessing, parameter settings and component selection have a significant influence on the resulting manifold, something which has not been comprehensively discussed in the literature so far. We provide a practice oriented review of the Diffusion Map technique, illustrate pitfalls and showcase a recently introduced technique for identifying the most relevant components. Our results show that the first components are not necessarily the most relevant ones.

Existing research often treats parliamentary discourse as a homogeneous whole, overlooking topic-specific patterns. Parliamentary speeches address a wide range of topics, some of which evoke stronger emotions than others. While everyone has intuitive assumptions about what the most emotive topics in a parliament may be, there has been little research into the emotions typically linked to different topics. This paper strives to fill this gap by examining emotion expression among the topics of parliamentary speeches delivered in Eduskunta, the Finnish Parliament, between 2000 and 2020. An emotion analysis model is used to investigate emotion expression in topics, from both synchronic and diachronic perspectives. The results strengthen evidence of increasing positivity in parliamentary speech and provide further insights into topic-specific emotion expression within parliamentary debate.

Modern content platforms offer paid promotion to mitigate cold start by allocating exposure via auctions. Our empirical analysis reveals a counterintuitive flaw in this paradigm: while promotion rescues low-to-medium quality content, it can harm high-quality content by forcing exposure to suboptimal audiences, polluting engagement signals and downgrading future recommendation. We recast content promotion as a dual-objective optimization that balances short-term value acquisition with long-term model improvement. To make this tractable at bid time in content promotion, we introduce a decomposable surrogate objective, gradient coverage, and establish its formal connection to Fisher Information and optimal experimental design. We design a two-stage auto-bidding algorithm based on Lagrange duality that dynamically paces budget through a shadow price and optimizes impression-level bids using per-impression marginal utilities. To address missing labels at bid time, we propose a confidence-gated gradient heuristic, paired with a zeroth-order variant for black-box models that reliably estimates learning signals in real time. We provide theoretical guarantees, proving monotone submodularity of the composite objective, sublinear regret in online auction, and budget feasibility. Extensive offline experiments on synthetic and real-world datasets validate the framework: it outperforms baselines, achieves superior final AUC/LogLoss, adheres closely to budget targets, and remains effective when gradients are approximated zeroth-order. These results show that strategic, information-aware promotion can improve long-term model performance and organic outcomes beyond naive impression-maximization strategies.

Developing human understandable interpretation of large language models (LLMs) becomes increasingly critical for their deployment in essential domains. Mechanistic interpretability seeks to mitigate the issues through extracts human-interpretable process and concepts from LLMs' activations. Sparse autoencoders (SAEs) have emerged as a popular approach for extracting interpretable and monosemantic concepts by decomposing the LLM internal representations into a dictionary. Despite their empirical progress, SAEs suffer from a fundamental theoretical ambiguity: the well-defined correspondence between LLM representations and human-interpretable concepts remains unclear. This lack of theoretical grounding gives rise to several methodological challenges, including difficulties in principled method design and evaluation criteria. In this work, we show that, under mild assumptions, LLM representations can be approximated as a {linear mixture} of the log-posteriors over concepts given the input context, through the lens of a latent variable model where concepts are treated as latent variables. This motivates a principled framework for concept extraction, namely Concept Component Analysis (ConCA), which aims to recover the log-posterior of each concept from LLM representations through a {unsupervised} linear unmixing process. We explore a specific variant, termed sparse ConCA, which leverages a sparsity prior to address the inherent ill-posedness of the unmixing problem. We implement 12 sparse ConCA variants and demonstrate their ability to extract meaningful concepts across multiple LLMs, offering theory-backed advantages over SAEs.

Recent research has shown that aligning fine-grained text descriptions with localized image patches can significantly improve the zero-shot performance of pre-trained vision-language models (e.g., CLIP). However, we find that both fine-grained text descriptions and localized image patches often contain redundant information, making text-visual alignment less effective. In this paper, we tackle this issue from two perspectives: \emph{View Refinement} and \emph{Description refinement}, termed as \textit{\textbf{Bi}-refinement for \textbf{F}ine-grained \textbf{T}ext-visual \textbf{A}lignment} (BiFTA). \emph{View refinement} removes redundant image patches with high \emph{Intersection over Union} (IoU) ratios, resulting in more distinctive visual samples. \emph{Description refinement} removes redundant text descriptions with high pairwise cosine similarity, ensuring greater diversity in the remaining descriptions. BiFTA achieves superior zero-shot performance on 6 benchmark datasets for both ViT-based and ResNet-based CLIP, justifying the necessity to remove redundant information in visual-text alignment.

Current speech LLMs bridge speech foundation models to LLMs using projection layers, training all of these components on speech instruction data. This strategy is computationally intensive and susceptible to task and prompt overfitting. We present SpeechMapper, a cost-efficient speech-to-LLM-embedding training approach that mitigates overfitting, enabling more robust and generalizable models. Our model is first pretrained without the LLM on inexpensive hardware, and then efficiently attached to the target LLM via a brief 1K-step instruction tuning (IT) stage. Through experiments on speech translation and spoken question answering, we demonstrate the versatility of SpeechMapper's pretrained block, presenting results for both task-agnostic IT, an ASR-based adaptation strategy that does not train in the target task, and task-specific IT. In task-agnostic settings, Speechmapper rivals the best instruction-following speech LLM from IWSLT25, despite never being trained on these tasks, while in task-specific settings, it outperforms this model across many datasets, despite requiring less data and compute. Overall, SpeechMapper offers a practical and scalable approach for efficient, generalizable speech-LLM integration without large-scale IT.

Given the increasing adoption of AI solutions in professional environments, it is necessary for developers to be able to make informed decisions about the current tool landscape. This work empirically evaluates various MLOps (Machine Learning Operations) tools to facilitate the management of the ML model lifecycle: MLflow, Metaflow, Apache Airflow, and Kubeflow Pipelines. The tools are evaluated by assessing the criteria of Ease of installation, Configuration flexibility, Interoperability, Code instrumentation complexity, result interpretability, and Documentation when implementing two common ML scenarios: Digit classifier with MNIST and Sentiment classifier with IMDB and BERT. The evaluation is completed by providing weighted results that lead to practical conclusions on which tools are best suited for different scenarios.

The ability of Large Language Models (LLMs) to use external tools unlocks powerful real-world interactions, making rigorous evaluation essential. However, current benchmarks primarily report final accuracy, revealing what models can do but obscuring the cognitive bottlenecks that define their true capability boundaries. To move from simple performance scoring to a diagnostic tool, we introduce a framework grounded in Cognitive Load Theory. Our framework deconstructs task complexity into two quantifiable components: Intrinsic Load, the inherent structural complexity of the solution path, formalized with a novel Tool Interaction Graph; and Extraneous Load, the difficulty arising from ambiguous task presentation. To enable controlled experiments, we construct ToolLoad-Bench, the first benchmark with parametrically adjustable cognitive load. Our evaluation reveals distinct performance cliffs as cognitive load increases, allowing us to precisely map each model's capability boundary. We validate that our framework's predictions are highly calibrated with empirical results, establishing a principled methodology for understanding an agent's limits and a practical foundation for building more efficient systems.

Time series forecasting requires capturing patterns across multiple temporal scales while maintaining computational efficiency. This paper introduces AWGformer, a novel architecture that integrates adaptive wavelet decomposition with cross-scale attention mechanisms for enhanced multi-variate time series prediction. Our approach comprises: (1) an Adaptive Wavelet Decomposition Module (AWDM) that dynamically selects optimal wavelet bases and decomposition levels based on signal characteristics; (2) a Cross-Scale Feature Fusion (CSFF) mechanism that captures interactions between different frequency bands through learnable coupling matrices; (3) a Frequency-Aware Multi-Head Attention (FAMA) module that weights attention heads according to their frequency selectivity; (4) a Hierarchical Prediction Network (HPN) that generates forecasts at multiple resolutions before reconstruction. Extensive experiments on benchmark datasets demonstrate that AWGformer achieves significant average improvements over state-of-the-art methods, with particular effectiveness on multi-scale and non-stationary time series. Theoretical analysis provides convergence guarantees and establishes the connection between our wavelet-guided attention and classical signal processing principles.

Enterprise LLM deployment faces a critical scalability challenge: organizations must optimize models systematically to scale AI initiatives within constrained compute budgets, yet the specialized expertise required for manual optimization remains a niche and scarce skillset. This challenge is particularly evident in managing GPU utilization across heterogeneous infrastructure while enabling teams with diverse workloads and limited LLM optimization experience to deploy models efficiently. We present OptiKIT, a distributed LLM optimization framework that democratizes model compression and tuning by automating complex optimization workflows for non-expert teams. OptiKIT provides dynamic resource allocation, staged pipeline execution with automatic cleanup, and seamless enterprise integration. In production, it delivers more than 2x GPU throughput improvement while empowering application teams to achieve consistent performance improvements without deep LLM optimization expertise. We share both the platform design and key engineering insights into resource allocation algorithms, pipeline orchestration, and integration patterns that enable large-scale, production-grade democratization of model optimization. Finally, we open-source the system to enable external contributions and broader reproducibility.

AI coding agents such as Codex and Claude Code are increasingly used to autonomously contribute to software repositories. However, little is known about how repository-level configuration artifacts affect operational efficiency of the agents. In this paper, we study the impact of AGENTS.md files on the runtime and token consumption of AI coding agents operating on GitHub pull requests. We analyze 10 repositories and 124 pull requests, executing agents under two conditions: with and without an AGENTS.md file. We measure wall-clock execution time and token usage during agent execution. Our results show that the presence of AGENTS.md is associated with a lower median runtime ($Δ28.64$%) and reduced output token consumption ($Δ16.58$%), while maintaining a comparable task completion behavior. Based on these results, we discuss immediate implications for the configuration and deployment of AI coding agents in practice, and outline a broader research agenda on the role of repository-level instructions in shaping the behavior, efficiency, and integration of AI coding agents in software development workflows.

Large Language Models (LLMs) have emerged as powerful learning tools, but they lack awareness of learners' cognitive and physiological states, limiting their adaptability to the user's learning style. Contemporary learning techniques primarily focus on structured learning paths, knowledge tracing, and generic adaptive testing but fail to address real-time learning challenges driven by cognitive load, attention fluctuations, and engagement levels. Building on findings from a formative user study (N=66), we introduce GuideAI, a multi-modal framework that enhances LLM-driven learning by integrating real-time biosensory feedback including eye gaze tracking, heart rate variability, posture detection, and digital note-taking behavior. GuideAI dynamically adapts learning content and pacing through cognitive optimizations (adjusting complexity based on learning progress markers), physiological interventions (breathing guidance and posture correction), and attention-aware strategies (redirecting focus using gaze analysis). Additionally, GuideAI supports diverse learning modalities, including text-based, image-based, audio-based, and video-based instruction, across varied knowledge domains. A preliminary study (N = 25) assessed GuideAI's impact on knowledge retention and cognitive load through standardized assessments. The results show statistically significant improvements in both problem-solving capability and recall-based knowledge assessments. Participants also experienced notable reductions in key NASA-TLX measures including mental demand, frustration levels, and effort, while simultaneously reporting enhanced perceived performance. These findings demonstrate GuideAI's potential to bridge the gap between current LLM-based learning systems and individualized learner needs, paving the way for adaptive, cognition-aware education at scale.

Time series forecasting presents significant challenges due to the complex temporal dependencies at multiple time scales. This paper introduces ScatterFusion, a novel framework that synergistically integrates scattering transforms with hierarchical attention mechanisms for robust time series forecasting. Our approach comprises four key components: (1) a Hierarchical Scattering Transform Module (HSTM) that extracts multi-scale invariant features capturing both local and global patterns; (2) a Scale-Adaptive Feature Enhancement (SAFE) module that dynamically adjusts feature importance across different scales; (3) a Multi-Resolution Temporal Attention (MRTA) mechanism that learns dependencies at varying time horizons; and (4) a Trend-Seasonal-Residual (TSR) decomposition-guided structure-aware loss function. Extensive experiments on seven benchmark datasets demonstrate that ScatterFusion outperforms other common methods, achieving significant reductions in error metrics across various prediction horizons.

Randomized subspace methods reduce per-iteration cost; however, in nonconvex optimization, most analyses are expectation-based, and high-probability bounds remain scarce even under sub-Gaussian noise. We first prove that randomized subspace SGD (RS-SGD) admits a high-probability convergence bound under sub-Gaussian noise, achieving the same order of oracle complexity as prior in-expectation results. Motivated by the prevalence of heavy-tailed gradients in modern machine learning, we then propose randomized subspace normalized SGD (RS-NSGD), which integrates direction normalization into subspace updates. Assuming the noise has bounded $p$-th moments, we establish both in-expectation and high-probability convergence guarantees, and show that RS-NSGD can achieve better oracle complexity than full-dimensional normalized SGD.

Heterogeneous federated learning (HFL) aims to ensure effective and privacy-preserving collaboration among different entities. As newly joined clients require significant adjustments and additional training to align with the existing system, the problem of generalizing federated learning models to unseen clients under heterogeneous data has become progressively crucial. Consequently, we highlight two unsolved challenging issues in federated domain generalization: Optimization Divergence and Performance Divergence. To tackle the above challenges, we propose FedRD, a novel heterogeneity-aware federated learning algorithm that collaboratively utilizes parameter-guided global generalization aggregation and local debiased classification to reduce divergences, aiming to obtain an optimal global model for participating and unseen clients. Extensive experiments on public multi-domain datasets demonstrate that our approach exhibits a substantial performance advantage over competing baselines in addressing this specific problem.

Language workbenches are tools that enable the definition, reuse, and composition of programming languages and their ecosystems, aiming to streamline language development. To facilitate their adoption by language designers, the comprehensibility of the language used to define other languages is an important aspect to evaluate. Moreover, considering that language workbenches are relatively new tools, user acceptance emerges as a crucial factor to be accounted for during their assessment. Current literature often neglects user-centred aspects like comprehensibility and acceptance in the assessment of this breed of tools. This paper addresses this gap through a family of experiments assessing Neverlang, a modular language workbench. The study adopts a tailored version of the Method Evaluation Model (MEM) to evaluate the comprehensibility of Neverlang's meta-language and programs, as well as user acceptance in terms of perceived ease of use, perceived usefulness, and intention to use. It also investigates the relationships among these dimensions. The experiments were conducted in three iterations involving participants from academia. The results reveal that users demonstrate sufficient comprehension of Neverlang's meta-language, particularly concerning its syntax, express a favourable perception of its usefulness, and indicate their intention to use it. However, the results also indicate that Neverlang's ease of use remains a challenge. Additionally, variations in the perceived ease of use and perceived usefulness, whether low or high, influence the users' intention to use the tool. Surprisingly, no significant correlation is found between comprehensibility and user acceptance. Notably, higher comprehensibility of the meta-language does not necessarily translate into greater acceptance, underscoring the complex interplay between comprehension and adoption.

This study addresses the challenge of accurately identifying multi-task contention types in high-dimensional system environments and proposes a unified contention classification framework that integrates representation transformation, structural modeling, and a task decoupling mechanism. The method first constructs system state representations from high-dimensional metric sequences, applies nonlinear transformations to extract cross-dimensional dynamic features, and integrates multiple source information such as resource utilization, scheduling behavior, and task load variations within a shared representation space. It then introduces a graph-based modeling mechanism to capture latent dependencies among metrics, allowing the model to learn competitive propagation patterns and structural interference across resource links. On this basis, task-specific mapping structures are designed to model the differences among contention types and enhance the classifier's ability to distinguish multiple contention patterns. To achieve stable performance, the method employs an adaptive multi-task loss weighting strategy that balances shared feature learning with task-specific feature extraction and generates final contention predictions through a standardized inference process. Experiments conducted on a public system trace dataset demonstrate advantages in accuracy, recall, precision, and F1, and sensitivity analyses on batch size, training sample scale, and metric dimensionality further confirm the model's stability and applicability. The study shows that structured representations and multi-task classification based on high-dimensional metrics can significantly improve contention pattern recognition and offer a reliable technical approach for performance management in complex computing environments.

The software engineering researchers from countries with smaller economies, particularly non-English speaking ones, represent valuable minorities within the software engineering community. As researchers from Poland, we represent such a country. We analyzed the ICSE FOSE (Future of Software Engineering) community survey through reflexive thematic analysis to show our viewpoint on key software community issues. We believe that the main problem is the growing research-industry gap, which particularly impacts smaller communities and small local companies. Based on this analysis and our experiences, we present a set of recommendations for improvements that would enhance software engineering research and industrial collaborations in smaller economies.

Graphical User Interface (GUI) agents show great potential for enabling foundation models to complete real-world tasks, revolutionizing human-computer interaction and improving human productivity. In this report, we present OmegaUse, a general-purpose GUI agent model for autonomous task execution on both mobile and desktop platforms, supporting computer-use and phone-use scenarios. Building an effective GUI agent model relies on two factors: (1) high-quality data and (2) effective training methods. To address these, we introduce a carefully engineered data-construction pipeline and a decoupled training paradigm. For data construction, we leverage rigorously curated open-source datasets and introduce a novel automated synthesis framework that integrates bottom-up autonomous exploration with top-down taxonomy-guided generation to create high-fidelity synthetic data. For training, to better leverage these data, we adopt a two-stage strategy: Supervised Fine-Tuning (SFT) to establish fundamental interaction syntax, followed by Group Relative Policy Optimization (GRPO) to improve spatial grounding and sequential planning. To balance computational efficiency with agentic reasoning capacity, OmegaUse is built on a Mixture-of-Experts (MoE) backbone. To evaluate cross-terminal capabilities in an offline setting, we introduce OS-Nav, a benchmark suite spanning multiple operating systems: ChiM-Nav, targeting Chinese Android mobile environments, and Ubu-Nav, focusing on routine desktop interactions on Ubuntu. Extensive experiments show that OmegaUse is highly competitive across established GUI benchmarks, achieving a state-of-the-art (SOTA) score of 96.3% on ScreenSpot-V2 and a leading 79.1% step success rate on AndroidControl. OmegaUse also performs strongly on OS-Nav, reaching 74.24% step success on ChiM-Nav and 55.9% average success on Ubu-Nav.

Large language models (LLMs) struggle with complex, long-horizon reasoning due to instability caused by their frozen policy assumption. Current test-time scaling methods treat execution feedback merely as an external signal for filtering or rewriting trajectories, without internalizing it to improve the underlying reasoning strategy. Inspired by Popper's epistemology of "conjectures and refutations," we argue that intelligence requires real-time evolution of the model's policy through learning from failed attempts. We introduce Policy of Thoughts (PoT), a framework that recasts reasoning as a within-instance online optimization process. PoT first generates diverse candidate solutions via an efficient exploration mechanism, then uses Group Relative Policy Optimization (GRPO) to update a transient LoRA adapter based on execution feedback. This closed-loop design enables dynamic, instance-specific refinement of the model's reasoning priors. Experiments show that PoT dramatically boosts performance: a 4B model achieves 49.71% accuracy on LiveCodeBench, outperforming GPT-4o and DeepSeek-V3 despite being over 50 smaller.

Large Language Models (LLMs) can be fine-tuned on domain-specific data to enhance their performance in specialized fields. However, such data often contains numerous low-quality samples, necessitating effective data processing (DP). In practice, DP strategies are typically developed through iterative manual analysis and trial-and-error adjustment. These processes inevitably incur high labor costs and may lead to privacy issues in high-privacy domains like healthcare due to direct human access to sensitive data. Thus, achieving automated data processing without exposing the raw data has become a critical challenge. To address this challenge, we propose LLM-AutoDP, a novel framework that leverages LLMs as agents to automatically generate and optimize data processing strategies. Our method generates multiple candidate strategies and iteratively refines them using feedback signals and comparative evaluations. This iterative in-context learning mechanism enables the agent to converge toward high-quality processing pipelines without requiring direct human intervention or access to the underlying data. To further accelerate strategy search, we introduce three key techniques: Distribution Preserving Sampling, which reduces data volume while maintaining distributional integrity; Processing Target Selection, which uses a binary classifier to identify low-quality samples for focused processing; Cache-and-Reuse Mechanism}, which minimizes redundant computations by reusing prior processing results. Results show that models trained on data processed by our framework achieve over 80% win rates against models trained on unprocessed data. Compared to AutoML baselines based on LLM agents, LLM-AutoDP achieves approximately a 65% win rate. Moreover, our acceleration techniques reduce the total searching time by up to 10 times, demonstrating both effectiveness and efficiency.

Recently, a novel peer sampling protocol, Elevator, was introduced to construct network topologies tailored for emerging decentralized applications such as federated learning and blockchain. Elevator builds hub-based topologies in a fully decentralized manner, randomly selecting hubs among participating nodes. These hubs, acting as central nodes connected to the entire network, can be leveraged to accelerate message dissemination. Simulation results have shown that Elevator converges rapidly (within 3--4 cycles) and exhibits robustness against crash failures and churn. However, its resilience to Byzantine adversaries has not been investigated. In this work, we provide the first evaluation of Elevator under Byzantine adversaries and show that even a small fraction (2%) of Byzantine nodes is sufficient to subvert the network. As a result, we introduce LIFT, a new protocol that extends Elevator by employing a cryptographically secure pseudo-random number generator (PRNG) for hub selection, thereby mitigating Byzantine manipulation. In contrast, LIFT withstands adversarial infiltration and remains robust with up to 10% Byzantine nodes. These results highlight the necessity of secure randomness in decentralized hub formation and position LIFT as a more reliable building block for Byzantine-resilient decentralized systems.

Identifying safety-critical scenarios is essential for autonomous driving, but the rarity of such events makes supervised labeling impractical. Traditional rule-based metrics like Time-to-Collision are too simplistic to capture complex interaction risks, and existing methods lack a systematic way to verify whether statistical anomalies truly reflect physical danger. To address this gap, we propose an unsupervised anomaly detection framework based on a multi-agent Transformer that models normal driving and measures deviations through prediction residuals. A dual evaluation scheme has been proposed to assess both detection stability and physical alignment: Stability is measured using standard ranking metrics in which Kendall Rank Correlation Coefficient captures rank agreement and Jaccard index captures the consistency of the top-K selected items; Physical alignment is assessed through correlations with established Surrogate Safety Measures (SSM). Experiments on the NGSIM dataset demonstrate our framework's effectiveness: We show that the maximum residual aggregator achieves the highest physical alignment while maintaining stability. Furthermore, our framework identifies 388 unique anomalies missed by Time-to-Collision and statistical baselines, capturing subtle multi-agent risks like reactive braking under lateral drift. The detected anomalies are further clustered into four interpretable risk types, offering actionable insights for simulation and testing.

Can standard continuous-time generative models represent distributions whose support is an extremely sparse, globally constrained discrete set? We study this question using completed Sudoku grids as a controlled testbed, treating them as a subset of a continuous relaxation space. We train flow-matching and score-based models along a Gaussian probability path and compare deterministic (ODE) sampling, stochastic (SDE) sampling, and DDPM-style discretizations derived from the same continuous-time training. Unconditionally, stochastic sampling substantially outperforms deterministic flows; score-based samplers are the most reliable among continuous-time methods, and DDPM-style ancestral sampling achieves the highest validity overall. We further show that the same models can be repurposed for guided generation: by repeatedly sampling completions under clamped clues and stopping when constraints are satisfied, the model acts as a probabilistic Sudoku solver. Although far less sample-efficient than classical solvers and discrete-geometry-aware diffusion methods, these experiments demonstrate that classic diffusion/flow formulations can assign non-zero probability mass to globally constrained combinatorial structures and can be used for constraint satisfaction via stochastic search.

Recent neural audio compression models often rely on residual vector quantization for high-fidelity coding, but using a fixed number of per-frame codebooks is suboptimal for the wide variability of audio content-especially for signals that are either very simple or highly complex. To address this limitation, we propose SwitchCodec, a neural audio codec based on Residual Experts Vector Quantization (REVQ). REVQ combines a shared quantizer with dynamically routed expert quantizers that are activated according to the input audio, decoupling bitrate from codebook capacity and improving compression efficiency. This design ensures full training and utilization of each quantizer. In addition, a variable-bitrate mechanism adjusts the number of active expert quantizers at inference, enabling multi-bitrate operation without retraining. Experiments demonstrate that SwitchCodec surpasses existing baselines on both objective metrics and subjective listening tests.

Physics-informed neural networks (PINNs) solve time-dependent partial differential equations (PDEs) by learning a mesh-free, differentiable solution that can be evaluated anywhere in space and time. However, standard space--time PINNs take time as an input but reuse a single network with shared weights across all times, forcing the same features to represent markedly different dynamics. This coupling degrades accuracy and can destabilize training when enforcing PDE, boundary, and initial constraints jointly. We propose Time-Induced Neural Networks (TINNs), a novel architecture that parameterizes the network weights as a learned function of time, allowing the effective spatial representation to evolve over time while maintaining shared structure. The resulting formulation naturally yields a nonlinear least-squares problem, which we optimize efficiently using a Levenberg--Marquardt method. Experiments on various time-dependent PDEs show up to $4\times$ improved accuracy and $10\times$ faster convergence compared to PINNs and strong baselines.

Speculative decoding (SD) has proven effective for accelerating LLM inference by quickly generating draft tokens and verifying them in parallel. However, SD remains largely unexplored for Large Vision-Language Models (LVLMs), which extend LLMs to process both image and text prompts. To address this gap, we benchmark existing inference methods with small draft models on 11 datasets across diverse input scenarios and observe scenario-specific performance fluctuations. Motivated by these findings, we propose Test-time Adaptive Batched Ensemble Drafting (TABED), which dynamically ensembles multiple drafts obtained via batch inference by leveraging deviations from past ground truths available in the SD setting. The dynamic ensemble method achieves an average robust walltime speedup of 1.74x over autoregressive decoding and a 5% improvement over single drafting methods, while remaining training-free and keeping ensembling costs negligible through parameter sharing. With its plug-and-play compatibility, we further enhance TABED by integrating advanced verification and alternative drafting methods. Code and custom-trained models are available at https://github.com/furiosa-ai/TABED.

Scene graphs provide structured abstractions for scene understanding, yet they often overfit to spurious correlations, severely hindering out-of-distribution generalization. To address this limitation, we propose CURVE, a causality-inspired framework that integrates variational uncertainty modeling with uncertainty-guided structural regularization to suppress high-variance, environment-specific relations. Specifically, we apply prototype-conditioned debiasing to disentangle invariant interaction dynamics from environment-dependent variations, promoting a sparse and domain-stable topology. Empirically, we evaluate CURVE in zero-shot transfer and low-data sim-to-real adaptation, verifying its ability to learn domain-stable sparse topologies and provide reliable uncertainty estimates to support risk prediction under distribution shifts.

The rapid evolution of Large Language Model (LLM) agents has necessitated robust memory systems to support cohesive long-term interaction and complex reasoning. Benefiting from the strong capabilities of LLMs, recent research focus has shifted from simple context extension to the development of dedicated agentic memory systems. However, existing approaches typically rely on rigid retrieval granularity, accumulation-heavy maintenance strategies, and coarse-grained update mechanisms. These design choices create a persistent mismatch between stored information and task-specific reasoning demands, while leading to the unchecked accumulation of logical inconsistencies over time. To address these challenges, we propose Adaptive Memory via Multi-Agent Collaboration (AMA), a novel framework that leverages coordinated agents to manage memory across multiple granularities. AMA employs a hierarchical memory design that dynamically aligns retrieval granularity with task complexity. Specifically, the Constructor and Retriever jointly enable multi-granularity memory construction and adaptive query routing. The Judge verifies the relevance and consistency of retrieved content, triggering iterative retrieval when evidence is insufficient or invoking the Refresher upon detecting logical conflicts. The Refresher then enforces memory consistency by performing targeted updates or removing outdated entries. Extensive experiments on challenging long-context benchmarks show that AMA significantly outperforms state-of-the-art baselines while reducing token consumption by approximately 80% compared to full-context methods, demonstrating its effectiveness in maintaining retrieval precision and long-term memory consistency.

Ransomware has become one of the most serious cybersecurity threats causing major financial losses and operational disruptions worldwide.Traditional detection methods such as static analysis, heuristic scanning and behavioral analysis often fall short when used alone. To address these limitations, this paper presents multimodal multi agent ransomware analysis framework designed for ransomware classification. Proposed multimodal multiagent architecture combines information from static, dynamic and network sources. Each data type is handled by specialized agent that uses auto encoder based feature extraction. These representations are then integrated through a fusion agent. After that fused representation are used by transformer based classifier. It identifies the specific ransomware family. The agents interact through an interagent feedback mechanism that iteratively refines feature representations by suppressing low confidence information. The framework was evaluated on large scale datasets containing thousands of ransomware and benign samples. Multiple experiments were conducted on ransomware dataset. It outperforms single modality and nonadaptive fusion baseline achieving improvement of up to 0.936 in Macro-F1 for family classification and reducing calibration error. Over 100 epochs, the agentic feedback loop displays a stable monotonic convergence leading to over +0.75 absolute improvement in terms of agent quality and a final composite score of around 0.88 without fine tuning of the language models. Zeroday ransomware detection remains family dependent on polymorphism and modality disruptions. Confidence aware abstention enables reliable real world deployment by favoring conservativeand trustworthy decisions over forced classification. The findings indicate that proposed approach provides a practical andeffective path toward improving real world ransomware defense systems.

Diffusion Language Models (DLMs) offer order-agnostic generation that can explore many possible decoding trajectories. However, current decoding methods commit to a single trajectory, limiting exploration in trajectory space. We introduce Order-Token Search to explore this space through jointly searching over generation order and token values. Its core is a likelihood estimator that scores denoising actions, enabling stable pruning and efficient exploration of diverse trajectories. Across mathematical reasoning and coding benchmarks, Order-Token Search consistently outperforms baselines on GSM8K, MATH500, Countdown, and HumanEval (3.1%, 3.8%, 7.9%, and 6.8% absolute over backbone), matching or surpassing diffu-GRPO post-trained d1-LLaDA. Our work establishes joint search as a key component for advancing decoding in DLMs.

Cryptocurrency projects articulate value propositions through whitepapers, making claims about functionality and technical capabilities. This study investigates whether these narratives align with observed market behavior. We construct a pipeline combining zero-shot NLP classification (BART-MNLI) with CP tensor decomposition to compare three spaces: (1) a claims matrix from 24 whitepapers across 10 semantic categories, (2) market statistics for 49 assets over two years of hourly data, and (3) latent factors from tensor decomposition (rank 2, 92.45% variance explained). Using Procrustes rotation and Tucker's congruence coefficient, we test alignment across 23 common entities. Results show weak alignment: claims-statistics (phi=0.341, p=0.332), claims-factors (phi=0.077, p=0.747), and statistics-factors (phi=0.197, p<0.001). The statistics-factors significance validates our methodology, confirming the pipeline detects relationships when present. Inter-model validation with DeBERTa-v3 yields 32% exact agreement but 67% top-3 agreement. Cross-sectional analysis reveals heterogeneous contributions: NEAR, MKR, ATOM show positive alignment while ENS, UNI, Bitcoin diverge most. Excluding Bitcoin confirms results are not driven by market dominance. We interpret findings as weak alignment between whitepaper narratives and market factor structure. Limited power (n=23) precludes distinguishing weak from no alignment, but strong alignment (phi>=0.70) can be confidently rejected. Implications for narrative economics and investment analysis are discussed.

Recent advances in mobile Graphical User Interface (GUI) agents highlight the growing need for comprehensive evaluation benchmarks. While new online benchmarks offer more realistic testing than offline ones, they tend to focus on the agents' task instruction-following ability while neglecting their reasoning and exploration ability. Moreover, these benchmarks do not consider the random noise in real-world mobile environments. This leads to a gap between benchmarks and real-world environments. To addressing these limitations, we propose MobileBench-OL, an online benchmark with 1080 tasks from 80 Chinese apps. It measures task execution, complex reasoning, and noise robustness of agents by including 5 subsets, which set multiple evaluation dimensions. We also provide an auto-eval framework with a reset mechanism, enabling stable and repeatable real-world benchmarking. Evaluating 12 leading GUI agents on MobileBench-OL shows significant room for improvement to meet real-world requirements. Human evaluation further confirms that MobileBench-OL can reliably measure the performance of leading GUI agents in real environments. Our data and code will be released upon acceptance.

Robotic manipulation has increasingly adopted vision-language-action (VLA) models, which achieve strong performance but typically require task-specific demonstrations and fine-tuning, and often generalize poorly under domain shift. We investigate whether general-purpose large language model (LLM) agent frameworks, originally developed for software engineering, can serve as an alternative control paradigm for embodied manipulation. We introduce FAEA (Frontier Agent as Embodied Agent), which applies an LLM agent framework directly to embodied manipulation without modification. Using the same iterative reasoning that enables software agents to debug code, FAEA enables embodied agents to reason through manipulation strategies. We evaluate an unmodified frontier agent, Claude Agent SDK, across the LIBERO, ManiSkill3, and MetaWorld benchmarks. With privileged environment state access, FAEA achieves success rates of 84.9%, 85.7%, and 96%, respectively. This level of task success approaches that of VLA models trained with less than 100 demonstrations per task, without requiring demonstrations or fine-tuning. With one round of human feedback as an optional optimization, performance increases to 88.2% on LIBERO. This demonstration-free capability has immediate practical value: FAEA can autonomously explore novel scenarios in simulation and generate successful trajectories for training data augmentation in embodied learning. Our results indicate that general-purpose agents are sufficient for a class of manipulation tasks dominated by deliberative, task-level planning. This opens a path for robotics systems to leverage actively maintained agent infrastructure and benefit directly from ongoing advances in frontier models. Code is available at https://github.com/robiemusketeer/faea-sim

Deep topological data analysis (TDA) offers a principled framework for capturing structural invariants such as connectivity and cycles that persist across scales, making it a natural fit for anomaly segmentation (AS). Unlike thresholdbased binarisation, which produces brittle masks under distribution shift, TDA allows anomalies to be characterised as disruptions to global structure rather than local fluctuations. We introduce TopoOT, a topology-aware optimal transport (OT) framework that integrates multi-filtration persistence diagrams (PDs) with test-time adaptation (TTA). Our key innovation is Optimal Transport Chaining, which sequentially aligns PDs across thresholds and filtrations, yielding geodesic stability scores that identify features consistently preserved across scales. These stabilityaware pseudo-labels supervise a lightweight head trained online with OT-consistency and contrastive objectives, ensuring robust adaptation under domain shift. Across standard 2D and 3D anomaly detection benchmarks, TopoOT achieves state-of-the-art performance, outperforming the most competitive methods by up to +24.1% mean F1 on 2D datasets and +10.2% on 3D AS benchmarks.

Diffusion language models (DLMs) generate text through iterative denoising, but inference requires full-sequence attention at every iteration, resulting in substantial redundant computation on masked tokens. Block-wise diffusion can reduce this cost, yet it typically relies on retraining and constrained update orders, limiting its direct applicability to pretrained DLMs. Our token-level analysis reveals pronounced structural locality in DLM inference. Decoding is driven by a small set of prefix-localized active tokens; the influence of distant undecoded context diminishes rapidly, and decoded tokens exhibit stage-wise temporal stability, enabling reuse of intermediate representations except for a brief post-decode transient. Motivated by these observations, we propose \textbf{\placeholder}\footnote{The source code is available at https://github.com/vhicrgit/Window-Diffusion.}, a window-based token pruning and caching method for inference. We maintain a local computation window that slides rightward as denoising progresses, and partition undecoded tokens into: (i) \textit{active tokens} that are computed online, (ii) \textit{buffer tokens} whose KV states are cached and periodically refreshed, and (iii) \textit{far-field tokens} that are pruned outside the window. Computation is restricted to active and buffer tokens within the window, while far-field tokens are omitted at each stage. Experiments on LLaDA and Dream show that, under matched compute budgets, our method achieves up to $99\times$ inference speedup while largely preserving generation performance.

While large language models show promise in mental healthcare, evaluating their therapeutic competence remains challenging due to the unstructured and longitudinal nature of counseling. We argue that current evaluation paradigms suffer from an unanchored defect, leading to two forms of instability: process drift, where unsteered client simulation wanders away from specific counseling goals, and standard drift, where static pointwise scoring lacks the stability for reliable judgment. To address this, we introduce Ps, a unified framework that calibrates the therapeutic competence of LLMs via trajectory-anchored tournaments. We first anchor the interaction trajectory in simulation, where clients precisely control the fluid consultation process to probe multifaceted capabilities. We then anchor the battle trajectory in judgments through an efficient Swiss-system tournament, utilizing dynamic pairwise battles to yield robust Elo ratings. Beyond ranking, we demonstrate that tournament trajectories can be transformed into credible reward signals, enabling on-policy reinforcement learning to enhance LLMs' performance. Extensive experiments validate the effectiveness of PsychePass and its strong consistency with human expert judgments.

Automatic evaluation is crucial yet challenging for open-ended natural language generation, especially when rule-based metrics are infeasible. Compared with traditional methods, the recent LLM-as-a-Judge paradigms enable better and more flexible evaluation, and show promise as generative reward models for reinforcement learning. However, prior work has revealed a notable gap between their seemingly impressive benchmark performance and actual effectiveness in RL practice. We attribute this issue to some limitations in existing studies, including the dominance of pairwise evaluation and inadequate optimization of evaluation criteria. Therefore, we propose CE-RM-4B, a pointwise generative reward model trained with a dedicated two-stage rollout method, and adopting unified query-based criteria. Using only about 5.7K high-quality data curated from the open-source preference dataset, our CE-RM-4B achieves superior performance on diverse reward model benchmarks, especially in Best-of-N scenarios, and delivers more effective improvements in downstream RL practice.

KV caches, typically used only to speed up autoregressive decoding, encode contextual information that can be reused for downstream tasks at no extra cost. We propose treating the KV cache as a lightweight representation, eliminating the need to recompute or store full hidden states. Despite being weaker than dedicated embeddings, KV-derived representations are shown to be sufficient for two key applications: \textbf{(i) Chain-of-Embedding}, where they achieve competitive or superior performance on Llama-3.1-8B-Instruct and Qwen2-7B-Instruct; and \textbf{(ii) Fast/Slow Thinking Switching}, where they enable adaptive reasoning on Qwen3-8B and DeepSeek-R1-Distil-Qwen-14B, reducing token generation by up to $5.7\times$ with minimal accuracy loss. Our findings establish KV caches as a free, effective substrate for sampling and reasoning, opening new directions for representation reuse in LLM inference. Code: https://github.com/cmd2001/ICLR2026_KV-Embedding.

Recent advances in Multimodal Large Language Models have rapidly expanded to electrocardiograms, focusing on classification, report generation, and single-turn QA tasks. However, these models fall short in real-world scenarios, lacking multi-turn conversational ability, on-device efficiency, and precise understanding of ECG measurements such as the PQRST intervals. To address these limitations, we introduce ECG-Agent, the first LLM-based tool-calling agent for multi-turn ECG dialogue. To facilitate its development and evaluation, we also present ECG-Multi-Turn-Dialogue (ECG-MTD) dataset, a collection of realistic user-assistant multi-turn dialogues for diverse ECG lead configurations. We develop ECG-Agents in various sizes, from on-device capable to larger agents. Experimental results show that ECG-Agents outperform baseline ECG-LLMs in response accuracy. Furthermore, on-device agents achieve comparable performance to larger agents in various evaluations that assess response accuracy, tool-calling ability, and hallucinations, demonstrating their viability for real-world applications.

Large Language Models (LLMs) are increasingly deployed for personalized product recommendations, with practitioners commonly assuming that longer user purchase histories lead to better predictions. We challenge this assumption through a systematic benchmark of four state of the art LLMs GPT-4o-mini, DeepSeek-V3, Qwen2.5-72B, and Gemini 2.5 Flash across context lengths ranging from 5 to 50 items using the REGEN dataset. Surprisingly, our experiments with 50 users in a within subject design reveal no significant quality improvement with increased context length. Quality scores remain flat across all conditions (0.17--0.23). Our findings have significant practical implications: practitioners can reduce inference costs by approximately 88\% by using context (5--10 items) instead of longer histories (50 items), without sacrificing recommendation quality. We also analyze latency patterns across providers and find model specific behaviors that inform deployment decisions. This work challenges the existing ``more context is better'' paradigm and provides actionable guidelines for cost effective LLM based recommendation systems.

Existing self-evolution methods overlook the influence of fine-grained reasoning steps, which leads to the reasoner-verifier gap. The computational inefficiency of Monte Carlo (MC) process supervision further exacerbates the difficulty in mitigating the gap. Motivated by the Error-Related Negativity (ERN), which the reasoner can localize error following incorrect decisions, guiding rapid adjustments, we propose a Self-Adaptive Process Optimization (SAPO) method for self-improvement in Small Language Models (SLMs). SAPO adaptively and efficiently introduces process supervision signals by actively minimizing the reasoner-verifier gap rather than relying on inefficient MC estimations. Extensive experiments demonstrate that the proposed method outperforms most existing self-evolution methods on two challenging task types: mathematics and code. Additionally, to further investigate SAPO's impact on verifier performance, this work introduces two new benchmarks for process reward models in both mathematical and coding tasks.

The global shortage and uneven distribution of medical expertise continue to hinder equitable access to accurate diagnostic care. While existing intelligent diagnostic system have shown promise, most struggle with dual-user interaction, and dynamic knowledge integration -- limiting their real-world applicability. In this study, we present DiagLink, a dual-user diagnostic assistance system that synergizes large language models (LLMs), knowledge graphs (KGs), and medical experts to support both patients and physicians. DiagLink uses guided dialogues to elicit patient histories, leverages LLMs and KGs for collaborative reasoning, and incorporates physician oversight for continuous knowledge validation and evolution. The system provides a role-adaptive interface, dynamically visualized history, and unified multi-source evidence to improve both trust and usability. We evaluate DiagLink through user study, use cases and expert interviews, demonstrating its effectiveness in improving user satisfaction and diagnostic efficiency, while offering insights for the design of future AI-assisted diagnostic systems.

Latent-based watermarks, integrated into the generation process of latent diffusion models (LDMs), simplify detection and attribution of generated images. However, recent black-box forgery attacks, where an attacker needs at least one watermarked image and black-box access to the provider's model, can embed the provider's watermark into images not produced by the provider, posing outsized risk to provenance and trust. We propose SemBind, the first defense framework for latent-based watermarks that resists black-box forgery by binding latent signals to image semantics via a learned semantic masker. Trained with contrastive learning, the masker yields near-invariant codes for the same prompt and near-orthogonal codes across prompts; these codes are reshaped and permuted to modulate the target latent before any standard latent-based watermark. SemBind is generally compatible with existing latent-based watermarking schemes and keeps image quality essentially unchanged, while a simple mask-ratio parameter offers a tunable trade-off between anti-forgery strength and robustness. Across four mainstream latent-based watermark methods, our SemBind-enabled anti-forgery variants markedly reduce false acceptance under black-box forgery while providing a controllable robustness-security balance.

Large Language Model (LLM) serving faces a fundamental tension between stringent latency Service Level Objectives (SLOs) and limited GPU memory capacity. When high request rates exhaust the KV cache budget, existing LLM inference systems often suffer severe head-of-line (HOL) blocking. While prior work explored PCIe-based offloading, these approaches cannot sustain responsiveness under high request rates, often failing to meet tight Time-To-First-Token (TTFT) and Time-Between-Tokens (TBT) SLOs. We present SuperInfer, a high-performance LLM inference system designed for emerging Superchips (e.g., NVIDIA GH200) with tightly coupled GPU-CPU architecture via NVLink-C2C. SuperInfer introduces RotaSched, the first proactive, SLO-aware rotary scheduler that rotates requests to maintain responsiveness on Superchips, and DuplexKV, an optimized rotation engine that enables full-duplex transfer over NVLink-C2C. Evaluations on GH200 using various models and datasets show that SuperInfer improves TTFT SLO attainment rates by up to 74.7% while maintaining comparable TBT and throughput compared to state-of-the-art systems, demonstrating that SLO-aware scheduling and memory co-design unlocks the full potential of Superchips for responsive LLM serving.

The prediction objectives of online advertisement ranking models are evolving from probabilistic metrics like conversion rate (CVR) to numerical business metrics like post-click gross merchandise volume (GMV). Unlike the well-studied delayed feedback problem in CVR prediction, delayed feedback modeling for GMV prediction remains unexplored and poses greater challenges, as GMV is a continuous target, and a single click can lead to multiple purchases that cumulatively form the label. To bridge the research gap, we establish TRACE, a GMV prediction benchmark containing complete transaction sequences rising from each user click, which supports delayed feedback modeling in an online streaming manner. Our analysis and exploratory experiments on TRACE reveal two key insights: (1) the rapid evolution of the GMV label distribution necessitates modeling delayed feedback under online streaming training; (2) the label distribution of repurchase samples substantially differs from that of single-purchase samples, highlighting the need for separate modeling. Motivated by these findings, we propose RepurchasE-Aware Dual-branch prEdictoR (READER), a novel GMV modeling paradigm that selectively activates expert parameters according to repurchase predictions produced by a router. Moreover, READER dynamically calibrates the regression target to mitigate under-estimation caused by incomplete labels. Experimental results show that READER yields superior performance on TRACE over baselines, achieving a 2.19% improvement in terms of accuracy. We believe that our study will open up a new avenue for studying online delayed feedback modeling for GMV prediction, and our TRACE benchmark with the gathered insights will facilitate future research and application in this promising direction. Our code and dataset are available at https://github.com/alimama-tech/OnlineGMV .

Unified Multimodal Models (UMMs) exhibit strong understanding, yet this capability often fails to effectively guide generation. We identify this as a Cognitive Gap: the model lacks the understanding of how to enhance its own generation process. To bridge this gap, we propose Endogenous Reprompting, a mechanism that transforms the model's understanding from a passive encoding process into an explicit generative reasoning step by generating self-aligned descriptors during generation. To achieve this, we introduce SEER (Self-Evolving Evaluator and Reprompter), a training framework that establishes a two-stage endogenous loop using only 300 samples from a compact proxy task, Visual Instruction Elaboration. First, Reinforcement Learning with Verifiable Rewards (RLVR) activates the model's latent evaluation ability via curriculum learning, producing a high-fidelity endogenous reward signal. Second, Reinforcement Learning with Model-rewarded Thinking (RLMT) leverages this signal to optimize the generative reasoning policy. Experiments show that SEER consistently outperforms state-of-the-art baselines in evaluation accuracy, reprompting efficiency, and generation quality, without sacrificing general multimodal capabilities.

The application of iodinated contrast media (ICM) improves the sensitivity and specificity of computed tomography (CT) for a wide range of clinical indications. However, overdose of ICM can cause problems such as kidney damage and life-threatening allergic reactions. Deep learning methods can generate CT images of normal-dose ICM from low-dose ICM, reducing the required dose while maintaining diagnostic power. However, existing methods are difficult to realize accurate enhancement with incompletely paired images, mainly because of the limited ability of the model to recognize specific structures. To overcome this limitation, we propose a Structure-constrained Language-informed Diffusion Model (SLDM), a unified medical generation model that integrates structural synergy and spatial intelligence. First, the structural prior information of the image is effectively extracted to constrain the model inference process, thus ensuring structural consistency in the enhancement process. Subsequently, semantic supervision strategy with spatial intelligence is introduced, which integrates the functions of visual perception and spatial reasoning, thus prompting the model to achieve accurate enhancement. Finally, the subtraction angiography enhancement module is applied, which serves to improve the contrast of the ICM agent region to suitable interval for observation. Qualitative analysis of visual comparison and quantitative results of several metrics demonstrate the effectiveness of our method in angiographic reconstruction for low-dose contrast medium CT angiography.

Estimating object mass from visual input is challenging because mass depends jointly on geometric volume and material-dependent density, neither of which is directly observable from RGB appearance. Consequently, mass prediction from pixels is ill-posed and therefore benefits from physically meaningful representations to constrain the space of plausible solutions. We propose a physically structured framework for single-image mass estimation that addresses this ambiguity by aligning visual cues with the physical factors governing mass. From a single RGB image, we recover object-centric three-dimensional geometry via monocular depth estimation to inform volume and extract coarse material semantics using a vision-language model to guide density-related reasoning. These geometry, semantic, and appearance representations are fused through an instance-adaptive gating mechanism, and two physically guided latent factors (volume- and density-related) are predicted through separate regression heads under mass-only supervision. Experiments on image2mass and ABO-500 show that the proposed method consistently outperforms state-of-the-art methods.

Sharpness-Aware Minimization (SAM) has recently emerged as an effective technique for improving DNN robustness to input variations. However, its interplay with the compactness requirements of on-device DNN deployments remains less explored. Simply pruning a SAM-trained model can undermine robustness, since flatness in the continuous parameter space does not necessarily translate to robustness under the discrete structural changes induced by pruning. Conversely, applying SAM after pruning may be fundamentally constrained by architectural limitations imposed by an early, robustness-agnostic pruning pattern. To address this gap, we propose Compression-aware ShArpness Minimization (C-SAM), a framework that shifts sharpness-aware learning from parameter perturbations to mask perturbations. By explicitly perturbing pruning masks during training, C-SAM promotes a flatter loss landscape with respect to model structure, enabling the discovery of pruning patterns that simultaneously optimize model compactness and robustness to input variations. Extensive experiments on CelebA-HQ, Flowers-102, and CIFAR-10-C across ResNet-18, GoogLeNet, and MobileNet-V2 show that C-SAM consistently achieves higher certified robustness than strong baselines, with improvements of up to 42%, while maintaining task accuracy comparable to the corresponding unpruned models.

Self-supervised learning (SSL) has greatly advanced speech representation learning, but multilingual SSL models remain constrained to languages encountered during pretraining. Retraining from scratch to incorporate new languages is computationally expensive, while sequential training without migitation strategies often leads to catastrophic forgetting. To address this, we propose MiLorE-SSL, a lightweight framework that combines LoRA modules with a soft mixture-of-experts (MoE) mechanism for efficient continual multilingual training. LoRA provides efficient low-rank adaptation, while soft MoE promotes flexible expert sharing across languages, reducing cross-lingual interference. To further mitigate forgetting, we introduce limited replay data from existing languages, avoiding reliance on large historical corpora. Experiments on ML-SUPERB demonstrate that MiLorE-SSL achieves strong performance in new languages and improves the ability in existing ones with only 2.14% trainable parameters.

The evaluation and post-training of large language models (LLMs) rely on supervision, but strong supervision for difficult tasks is often unavailable, especially when evaluating frontier models. In such cases, models are demonstrated to exploit evaluations built on such imperfect supervision, leading to deceptive results. However, underutilized in LLM research, a wealth of mechanism design research focuses on game-theoretic incentive compatibility, i.e., eliciting honest and informative answers with weak supervision. Drawing from this literature, we introduce the peer prediction method for model evaluation and post-training. It rewards honest and informative answers over deceptive and uninformative ones, using a metric based on mutual predictability and without requiring ground truth labels. We demonstrate the method's effectiveness and resistance to deception, with both theoretical guarantees and empirical validation on models with up to 405B parameters. We show that training an 8B model with peer prediction-based reward recovers most of the drop in truthfulness due to prior malicious finetuning, even when the reward is produced by a 0.135B language model with no finetuning. On the evaluation front, in contrast to LLM-as-a-Judge which requires strong and trusted judges, we discover an inverse scaling property in peer prediction, where, surprisingly, resistance to deception is strengthened as the capability gap between the experts and participants widens, enabling reliable evaluation of strong models with weak supervision. In particular, LLM-as-a-Judge become worse than random guess when facing deceptive models 5-20x the judge's size, while peer prediction thrives when such gaps are large, including in cases with over 100x size difference.

Bridging the sim2real gap between computationally inexpensive models and complex physical systems remains a central challenge in machine learning applications to engineering problems, particularly in multi-scale settings where reduced-order models typically capture only dominant dynamics. In this work, we present Cheap2Rich, a multi-scale data assimilation framework that reconstructs high-fidelity state spaces from sparse sensor histories by combining a fast low-fidelity prior with learned, interpretable discrepancy corrections. We demonstrate the performance on rotating detonation engines (RDEs), a challenging class of systems that couple detonation-front propagation with injector-driven unsteadiness, mixing, and stiff chemistry across disparate scales. Our approach successfully reconstructs high-fidelity RDE states from sparse measurements while isolating physically meaningful discrepancy dynamics associated with injector-driven effects. The results highlight a general multi-fidelity framework for data assimilation and system identification in complex multi-scale systems, enabling rapid design exploration and real-time monitoring and control while providing interpretable discrepancy dynamics. Code for this project is is available at: github.com/kro0l1k/Cheap2Rich.

Photoacoustic computed tomography (PACT) is a promising imaging modality that combines the advantages of optical contrast with ultrasound detection. Utilizing ultrasound transducers with larger surface areas can improve detection sensitivity. However, when computationally efficient analytic reconstruction methods that neglect the spatial impulse responses (SIRs) of the transducer are employed, the spatial resolution of the reconstructed images will be compromised. Although optimization-based reconstruction methods can explicitly account for SIR effects, their computational cost is generally high, particularly in three-dimensional (3D) applications. To address the need for accurate but rapid 3D PACT image reconstruction, this study presents a framework for establishing a learned SIR compensation method that operates in the data domain. The learned compensation method maps SIR-corrupted PACT measurement data to compensated data that would have been recorded by idealized point-like transducers. Subsequently, the compensated data can be used with a computationally efficient reconstruction method that neglects SIR effects. Two variants of the learned compensation model are investigated that employ a U-Net model and a specifically designed, physics-inspired model, referred to as Deconv-Net. A fast and analytical training data generation procedure is also a component of the presented framework. The framework is rigorously validated in virtual imaging studies, demonstrating resolution improvement and robustness to noise variations, object complexity, and sound speed heterogeneity. When applied to in-vivo breast imaging data, the learned compensation models revealed fine structures that had been obscured by SIR-induced artifacts. To our knowledge, this is the first demonstration of learned SIR compensation in 3D PACT imaging.

Neural ranking models (NRMs) achieve strong retrieval effectiveness, yet prior work has shown they are vulnerable to adversarial perturbations. We revisit this robustness question with a minimal, query-aware attack that promotes a target document by inserting or substituting a single, semantically aligned word - the query center. We study heuristic and gradient-guided variants, including a white-box method that identifies influential insertion points. On TREC-DL 2019/2020 with BERT and monoT5 re-rankers, our single-word attacks achieve up to 91% success while modifying fewer than two tokens per document on average, achieving competitive rank and score boosts with far fewer edits under a comparable white-box setup to ensure fair evaluation against PRADA. We also introduce new diagnostic metrics to analyze attack sensitivity beyond aggregate success rates. Our analysis reveals a Goldilocks zone in which mid-ranked documents are most vulnerable. These findings demonstrate practical risks and motivate future defenses for robust neural ranking.

While explainable artificial intelligence (XAI) for large language models (LLMs) remains an evolving field with many unresolved questions, increasing regulatory pressures have spurred interest in its role in ensuring transparency, accountability, and privacy-preserving machine unlearning. Despite recent advances in XAI have provided some insights, the specific role of attention layers in transformer based LLMs remains underexplored. This study investigates the memory mechanisms instantiated by attention layers, drawing on prior research in psychology and computational psycholinguistics that links Transformer attention to cue based retrieval in human memory. In this view, queries encode the retrieval context, keys index candidate memory traces, attention weights quantify cue trace similarity, and values carry the encoded content, jointly enabling the construction of a context representation that precedes and facilitates memory retrieval. Guided by the Encoding Specificity Principle, we hypothesize that the cues used in the initial stage of retrieval are instantiated as keywords. We provide converging evidence for this keywords-as-cues hypothesis. In addition, we isolate neurons within attention layers whose activations selectively encode and facilitate the retrieval of context-defining keywords. Consequently, these keywords can be extracted from identified neurons and further contribute to downstream applications such as unlearning.

Time series forecasting has long been dominated by advances in model architecture, with recent progress driven by deep learning and hybrid statistical techniques. However, as forecasting models approach diminishing returns in accuracy, a critical yet underexplored opportunity emerges: the strategic use of post-processing. In this paper, we address the last-mile gap in time-series forecasting, which is to improve accuracy and uncertainty without retraining or modifying a deployed backbone. We propose $δ$-Adapter, a lightweight, architecture-agnostic way to boost deployed time series forecasters without retraining. $δ$-Adapter learns tiny, bounded modules at two interfaces: input nudging (soft edits to covariates) and output residual correction. We provide local descent guarantees, $O(δ)$ drift bounds, and compositional stability for combined adapters. Meanwhile, it can act as a feature selector by learning a sparse, horizon-aware mask over inputs to select important features, thereby improving interpretability. In addition, it can also be used as a distribution calibrator to measure uncertainty. Thus, we introduce a Quantile Calibrator and a Conformal Corrector that together deliver calibrated, personalized intervals with finite-sample coverage. Our experiments across diverse backbones and datasets show that $δ$-Adapter improves accuracy and calibration with negligible compute and no interface changes.

Long-context LLM agents must access the right evidence from large environments and use it faithfully. However, the popular Needle-in-a-Haystack (NIAH) evaluation mostly measures benign span localization. The needle is near-unique, and the haystack is largely irrelevant. We introduce EverMemBench-S (EMB-S), an adversarial NIAH-style benchmark built on a 326M-token MemoryBank. While the full MemoryBank spans 326M tokens for retrieval-based (RAG) evaluation, we evaluate native long-context models only at scales that fit within each model's context window (up to 1M tokens in this work) to ensure a fair comparison. EMB-S pairs queries with collision-tested near-miss hard negatives and gold evidence sets spanning one or more documents, validated via human screening and LLM verification. We also propose a decoupled diagnostic protocol that reports evidence access (document-ID localization) separately from end-to-end QA quality under full-context prompting. This enables consistent diagnosis for both native long-context prompting and retrieval pipelines. Across a reference-corpus ladder from domain-isolated 64K contexts to a globally shared 326M-token environment, we observe a clear reality gap. Systems that saturate benign NIAH degrade sharply in evidence access under semantic interference. These results indicate that semantic discrimination, not context length alone, is the dominant bottleneck for long-context memory at scale.

Defining psycholinguistic characteristics in written texts is a task gaining increasing attention from researchers. One of the most widely used tools in the current field is Linguistic Inquiry and Word Count (LIWC) that originally was developed to analyze English texts and translated into multiple languages. Our approach offers the adaptation of LIWC methodology for the Russian language, considering its grammatical and cultural specificities. The suggested approach comprises 96 categories, integrating syntactic, morphological, lexical, general statistical features, and results of predictions obtained using pre-trained language models (LMs) for text analysis. Rather than applying direct translation to existing thesauri, we built the dictionary specifically for the Russian language based on the content from several lexicographic resources, semantic dictionaries and corpora. The paper describes the process of mapping lemmas to 42 psycholinguistic categories and the implementation of the analyzer as part of RusLICA web service.

Diffusion Transformers (DiTs) have gained increasing adoption in high-quality image and video generation. As demand for higher-resolution images and longer videos increases, single-GPU inference becomes inefficient due to increased latency and large activation sizes. Current frameworks employ sequence parallelism (SP) techniques such as Ulysses Attention and Ring Attention to scale inference. However, these implementations have three primary limitations: (1) suboptimal communication patterns for network topologies on modern GPU machines, (2) latency bottlenecks from all-to-all operations in inter-machine communication, and (3) GPU sender-receiver synchronization and computation overheads from using two-sided communication libraries. To address these issues, we present StreamFusion, a topology-aware efficient DiT serving engine. StreamFusion incorporates three key innovations: (1) a topology-aware sequence parallelism technique that accounts for inter- and intra-machine bandwidth differences, (2) Torus Attention, a novel SP technique enabling overlapping of inter-machine all-to-all operations with computation, and (3) a one-sided communication implementation that minimizes GPU sender-receiver synchronization and computation overheads. Our experiments demonstrate that StreamFusion outperforms the state-of-the-art approach by an average of $1.35\times$ (up to $1.77\times$).

Phishing continues to be one of the most prevalent attack vectors, making accurate classification of phishing URLs essential. Recently, large language models (LLMs) have demonstrated promising results in phishing URL detection. However, their reasoning capabilities that enabled such performance remain underexplored. To this end, in this paper, we propose a Least-to-Most prompting framework for phishing URL detection. In particular, we introduce an "answer sensitivity" mechanism that guides Least-to-Most's iterative approach to enhance reasoning and yield higher prediction accuracy. We evaluate our framework using three URL datasets and four state-of-the-art LLMs, comparing against a one-shot approach and a supervised model. We demonstrate that our framework outperforms the one-shot baseline while achieving performance comparable to that of the supervised model, despite requiring significantly less training data. Furthermore, our in-depth analysis highlights how the iterative reasoning enabled by Least-to-Most, and reinforced by our answer sensitivity mechanism, drives these performance gains. Overall, we show that this simple yet powerful prompting strategy consistently outperforms both one-shot and supervised approaches, despite requiring minimal training or few-shot guidance. Our experimental setup can be found in our Github repository github.sydney.edu.au/htri0928/least-to-most-phishing-detection.

Machine learning models in high-stakes applications, such as recidivism prediction and automated personnel selection, often exhibit systematic performance disparities across sensitive subpopulations, raising critical concerns regarding algorithmic bias. Fairness auditing addresses these risks through two primary functions: certification, which verifies adherence to fairness constraints; and flagging, which isolates specific demographic groups experiencing disparate treatment. However, existing auditing techniques are frequently limited by restrictive distributional assumptions or prohibitive computational overhead. We propose a novel empirical likelihood-based (EL) framework that constructs robust statistical measures for model performance disparities. Unlike traditional methods, our approach is non-parametric; the proposed disparity statistics follow asymptotically chi-square or mixed chi-square distributions, ensuring valid inference without assuming underlying data distributions. This framework uses a constrained optimization profile that admits stable numerical solutions, facilitating both large-scale certification and efficient subpopulation discovery. Empirically, the EL methods outperform bootstrap-based approaches, yielding coverage rates closer to nominal levels while reducing computational latency by several orders of magnitude. We demonstrate the practical utility of this framework on the COMPAS dataset, where it successfully flags intersectional biases, specifically identifying a significantly higher positive prediction rate for African-American males under 25 and a systemic under-prediction for Caucasian females relative to the population mean.

Recent advances in stochastic differential equations (SDEs) have enabled robust modeling of real-world dynamical processes across diverse domains, such as finance, health, and systems biology. However, parameter estimation for SDEs typically relies on accurately timestamped observational sequences. When temporal ordering information is corrupted, missing, or deliberately hidden (e.g., for privacy), existing estimation methods often fail. In this paper, we investigate the conditions under which temporal order can be recovered and introduce a novel framework that simultaneously reconstructs temporal information and estimates SDE parameters. Our approach exploits asymmetries between forward and backward processes, deriving a score-matching criterion to infer the correct temporal order between pairs of observations. We then recover the total order via a sorting procedure and estimate SDE parameters from the reconstructed sequence using maximum likelihood. Finally, we conduct extensive experiments on synthetic and real-world datasets to demonstrate the effectiveness of our method, extending parameter estimation to settings with missing temporal order and broadening applicability in sensitive domains.

Decision Transformer (DT) shows promise for generative auto-bidding by capturing temporal dependencies, but suffers from two critical limitations: insufficient cross-correlation modeling among state, action, and return-to-go (RTG) sequences, and indiscriminate learning of optimal/suboptimal behaviors. To address these, we propose C2, a novel framework enhancing DT with two core innovations: (1) a Cross Learning Block (CLB) via cross-attention to strengthen inter-sequence correlation modeling; (2) a Constraint-aware Loss (CL) incorporating budget and Cost-Per-Acquisition (CPA) constraints for selective learning of optimal trajectories. Extensive offline evaluations on the AuctionNet dataset demonstrate consistent performance gains (up to 3.23\% over state-of-the-art GAVE) across diverse budget settings; ablation studies verify the complementary synergy of CLB and CL, confirming C2's superiority in auto-bidding. The code for reproducing our results is available at: https://github.com/Dingjinren/C2.

Online hate on social media ranges from overt slurs and threats (\emph{hard hate speech}) to \emph{soft hate speech}: discourse that appears reasonable on the surface but uses framing and value-based arguments to steer audiences toward blaming or excluding a target group. We hypothesize that current moderation systems, largely optimized for surface toxicity cues, are not robust to this reasoning-driven hostility, yet existing benchmarks do not measure this gap systematically. We introduce \textbf{\textsc{SoftHateBench}}, a generative benchmark that produces soft-hate variants while preserving the underlying hostile standpoint. To generate soft hate, we integrate the \emph{Argumentum Model of Topics} (AMT) and \emph{Relevance Theory} (RT) in a unified framework: AMT provides the backbone argument structure for rewriting an explicit hateful standpoint into a seemingly neutral discussion while preserving the stance, and RT guides generation to keep the AMT chain logically coherent. The benchmark spans \textbf{7} sociocultural domains and \textbf{28} target groups, comprising \textbf{4,745} soft-hate instances. Evaluations across encoder-based detectors, general-purpose LLMs, and safety models show a consistent drop from hard to soft tiers: systems that detect explicit hostility often fail when the same stance is conveyed through subtle, reasoning-based language. \textcolor{red}{\textbf{Disclaimer.} Contains offensive examples used solely for research.}

SWE-bench has emerged as the premier benchmark for evaluating Large Language Models on complex software engineering tasks. While these capabilities are fundamentally acquired during the mid-training phase and subsequently elicited during Supervised Fine-Tuning (SFT), there remains a critical deficit in metrics capable of guiding mid-training effectively. Standard metrics such as Perplexity (PPL) are compromised by the "Long-Context Tax" and exhibit weak correlation with downstream SWE performance. In this paper, we bridge this gap by first introducing a rigorous data filtering strategy. Crucially, we propose the Entropy Compression Hypothesis, redefining intelligence not by scalar Top-1 compression, but by the capacity to structure uncertainty into Entropy-Compressed States of low orders ("reasonable hesitation"). Grounded in this fine-grained entropy analysis, we formulate a novel metric, HE-SNR (High-Entropy Signal-to-Noise Ratio). Validated on industrial-scale Mixture-of-Experts (MoE) models across varying context windows (32K/128K), our approach demonstrates superior robustness and predictive power. This work provides both the theoretical foundation and practical tools for optimizing the latent potential of LLMs in complex engineering domains.

Open-ended question answering (QA) evaluates a model's ability to perform contextualized reasoning beyond factual recall. This challenge is especially acute in practice-based domains, where knowledge is procedural and grounded in professional judgment, while most existing LLM benchmarks depend on pre-existing human exam datasets that are often unavailable in such settings. We introduce a framework for automated benchmark generation from expert-authored guidelines informed by Bloom's Taxonomy. It converts expert practices into implicit violation-based scenarios and expands them into auto-graded multiple-choice questions (MCQs) and multi-turn dialogues across four cognitive levels, enabling deterministic, reproducible, and scalable evaluation. Applied to three applied domains: teaching, dietetics, and caregiving, we find differences between model and human-like reasoning: LLMs sometimes perform relatively better on higher-order reasoning (Analyze) but fail more frequently on lower-level items (Remember). We produce large-scale, psychometrically informed benchmarks that surface these non-intuitive model behaviors and enable evaluation of contextualized reasoning in real-world settings.

Exhaustively evaluating many large language models (LLMs) on a large suite of benchmarks is expensive. We cast benchmarking as finite-population inference and, under a fixed query budget, seek tight confidence intervals (CIs) for model accuracy with valid frequentist coverage. We propose Factorized Active Querying (FAQ), which (a) leverages historical information through a Bayesian factor model; (b) adaptively selects questions using a hybrid variance-reduction/active-learning sampling policy; and (c) maintains validity through Proactive Active Inference -- a finite-population extension of active inference (Zrnic & Candes, 2024) that enables direct question selection while preserving coverage. With negligible overhead cost, FAQ delivers up to $5\times$ effective sample size gains over strong baselines on two benchmark suites, across varying historical-data missingness levels: this means that it matches the CI width of uniform sampling while using up to $5\times$ fewer queries. We release our source code and our curated datasets to support reproducible evaluation and future research.

Recently, flow-based generative models have shown superior efficiency compared to diffusion models. In this paper, we study rectified flow models, which constrain transport trajectories to be linear from the base distribution to the data distribution. This structural restriction greatly accelerates sampling, often enabling high-quality generation with a single Euler step. Under standard assumptions on the neural network classes used to parameterize the velocity field and data distribution, we prove that rectified flows achieve sample complexity $\tilde{O}(\varepsilon^{-2})$. This improves on the best known $O(\varepsilon^{-4})$ bounds for flow matching model and matches the optimal rate for mean estimation. Our analysis exploits the particular structure of rectified flows: because the model is trained with a squared loss along linear paths, the associated hypothesis class admits a sharply controlled localized Rademacher complexity. This yields the improved, order-optimal sample complexity and provides a theoretical explanation for the strong empirical performance of rectified flow models.

AI assistance produces significant productivity gains across professional domains, particularly for novice workers. Yet how this assistance affects the development of skills required to effectively supervise AI remains unclear. Novice workers who rely heavily on AI to complete unfamiliar tasks may compromise their own skill acquisition in the process. We conduct randomized experiments to study how developers gained mastery of a new asynchronous programming library with and without the assistance of AI. We find that AI use impairs conceptual understanding, code reading, and debugging abilities, without delivering significant efficiency gains on average. Participants who fully delegated coding tasks showed some productivity improvements, but at the cost of learning the library. We identify six distinct AI interaction patterns, three of which involve cognitive engagement and preserve learning outcomes even when participants receive AI assistance. Our findings suggest that AI-enhanced productivity is not a shortcut to competence and AI assistance should be carefully adopted into workflows to preserve skill formation -- particularly in safety-critical domains.

Background: The Node Package Manager (npm) ecosystem plays a vital role in modern software development by providing a vast repository of packages and tools that developers can use to implement their software systems. However, recent vulnerabilities in third-party packages have led to serious security breaches, compromising the integrity of applications that depend on them. Objective: This study investigates how npm package developers perceive and handle security in their work. We examined developers' understanding of security risks, the practices and tools they use, the barriers to stronger security measures, and their suggestions for improving the npm ecosystem's security. Method: We conducted an online survey with 75 npm package developers and undertook a mixed-methods approach to analyzing their responses. Results: While developers prioritize security, they perceive their packages as only moderately secure, with concerns about supply chain attacks, dependency vulnerabilities, and malicious code. Only 40% are satisfied with the current npm security tools due to issues such as alert fatigue. Automated methods such as two-factor authentication and npm audit are favored over code reviews. Many drop dependencies due to abandonment or vulnerabilities, and typically respond to vulnerabilities in their packages by quickly releasing patches. Key barriers include time constraints and high false-positive rates. To improve npm security, developers seek better detection tools, clearer documentation, stronger account protections, and more education initiatives. Conclusion: Our findings will benefit npm package contributors and maintainers by highlighting prevalent security challenges and promoting discussions on best practices to strengthen security and trustworthiness within the npm landscape.

The scaling law, which indicates that model performance improves with increasing dataset and model capacity, has fueled a growing trend in expanding recommendation models in both industry and academia. However, the advent of large-scale recommenders also brings significantly higher computational costs, particularly under the long-sequence dependencies inherent in the user intent of recommendation systems. Current approaches often rely on pre-storing the intermediate states of the past behavior for each user, thereby reducing the quadratic re-computation cost for the following requests. Despite their effectiveness, these methods often treat memory merely as a medium for acceleration, without adequately considering the space overhead it introduces. This presents a critical challenge in real-world recommendation systems with billions of users, each of whom might initiate thousands of interactions and require massive memory for state storage. Fortunately, there have been several memory management strategies examined for compression in LLM, while most have not been evaluated on the recommendation task. To mitigate this gap, we introduce MALLOC, a comprehensive benchmark for memory-aware long sequence compression. MALLOC presents a comprehensive investigation and systematic classification of memory management techniques applicable to large sequential recommendations. These techniques are integrated into state-of-the-art recommenders, enabling a reproducible and accessible evaluation platform. Through extensive experiments across accuracy, efficiency, and complexity, we demonstrate the holistic reliability of MALLOC in advancing large-scale recommendation. Code is available at https://anonymous.4open.science/r/MALLOC.

We study black-box optimization of Lipschitz functions under noisy evaluations. Existing adaptive discretization methods implicitly avoid suboptimal regions but do not provide explicit certificates of optimality or measurable progress guarantees. We introduce \textbf{Certificate-Guided Pruning (CGP)}, which maintains an explicit \emph{active set} $A_t$ of potentially optimal points via confidence-adjusted Lipschitz envelopes. Any point outside $A_t$ is certifiably suboptimal with high probability, and under a margin condition with near-optimality dimension $α$, we prove $\Vol(A_t)$ shrinks at a controlled rate yielding sample complexity $\tildeO(\varepsilon^{-(2+α)})$. We develop three extensions: CGP-Adaptive learns $L$ online with $O(\log T)$ overhead; CGP-TR scales to $d > 50$ via trust regions with local certificates; and CGP-Hybrid switches to GP refinement when local smoothness is detected. Experiments on 12 benchmarks ($d \in [2, 100]$) show CGP variants match or exceed strong baselines while providing principled stopping criteria via certificate volume.

Full-duplex voice interaction is crucial for natural human computer interaction. We present a framework that decomposes complex dialogue into minimal conversational units, enabling the system to process each unit independently and predict when to transit to the next. This framework is instantiated as a semi-cascaded full-duplex dialogue system built around a multimodal large language model, supported by auxiliary modules such as voice activity detection (VAD) and text-to-speech (TTS) synthesis. The resulting system operates in a train-free, plug-and-play manner. Experiments on the HumDial dataset demonstrate the effectiveness of our framework, which ranks second among all teams on the test set of the Human-like Spoken Dialogue Systems Challenge (Track 2: Full-Duplex Interaction). Code is available at the GitHub repository https://github.com/yu-haoyuan/fd-badcat.

Service Function Chaining (SFC) requires efficient placement of Virtual Network Functions (VNFs) to satisfy diverse service requirements while maintaining high resource utilization in Data Centers (DCs). Conventional static resource allocation often leads to overprovisioning or underprovisioning due to the dynamic nature of traffic loads and application demands. To address this challenge, we propose a hybrid forecast-driven Deep reinforcement learning (DRL) framework that combines predictive intelligence with SFC provisioning. Specifically, we leverage DRL to generate datasets capturing DC resource utilization and service demands, which are then used to train deep learning forecasting models. Using Optuna-based hyperparameter optimization, the best-performing models, Spatio-Temporal Graph Neural Network, Temporal Graph Neural Network, and Long Short-Term Memory, are combined into an ensemble to enhance stability and accuracy. The ensemble predictions are integrated into the DC selection process, enabling proactive placement decisions that consider both current and future resource availability. Experimental results demonstrate that the proposed method not only sustains high acceptance ratios for resource-intensive services such as Cloud Gaming and VoIP but also significantly improves acceptance ratios for latency-critical categories such as Augmented Reality increases from 30% to 50%, while Industry 4.0 improves from 30% to 45%. Consequently, the prediction-based model achieves significantly lower E2E latencies of 20.5%, 23.8%, and 34.8% reductions for VoIP, Video Streaming, and Cloud Gaming, respectively. This strategy ensures more balanced resource allocation, and reduces contention.

Accurate simulation of nuclear quantum effects is essential for molecular modeling but expensive using path integral molecular dynamics (PIMD). We present GG-PI, a ring-polymer-based framework that combines generative modeling of the single-bead conditional density with Gibbs sampling to recover quantum statistics from classical simulation data. GG-PI uses inexpensive standard classical simulations or existing data for training and allows transfer across temperatures without retraining. On standard test systems, GG-PI significantly reduces wall clock time compared to PIMD. Our approach extends easily to a wide range of problems with similar Markov structure.

Inferring physical fields from sparse observations while strictly satisfying partial differential equations (PDEs) is a fundamental challenge in computational physics. Recently, deep generative models offer powerful data-driven priors for such inverse problems, yet existing methods struggle to enforce hard physical constraints without costly retraining or disrupting the learned generative prior. Consequently, there is a critical need for a sampling mechanism that can reconcile strict physical consistency and observational fidelity with the statistical structure of the pre-trained prior. To this end, we present ProFlow, a proximal guidance framework for zero-shot physics-consistent sampling, defined as inferring solutions from sparse observations using a fixed generative prior without task-specific retraining. The algorithm employs a rigorous two-step scheme that alternates between: (\romannumeral1) a terminal optimization step, which projects the flow prediction onto the intersection of the physically and observationally consistent sets via proximal minimization; and (\romannumeral2) an interpolation step, which maps the refined state back to the generative trajectory to maintain consistency with the learned flow probability path. This procedure admits a Bayesian interpretation as a sequence of local maximum a posteriori (MAP) updates. Comprehensive benchmarks on Poisson, Helmholtz, Darcy, and viscous Burgers' equations demonstrate that ProFlow achieves superior physical and observational consistency, as well as more accurate distributional statistics, compared to state-of-the-art diffusion- and flow-based baselines.

We present two machine learning frameworks for forecasting aggregated curves and optimizing storage in the EPEX SPOT day-ahead market. First, a fast parametric model forecasts hourly demand and supply curves in a low-dimensional and grid-robust representation, with minimum and maximum volumes combined with a Chebyshev polynomial for the elastic segment. The model enables daily use with low error and clear interpretability. Second, for a more comprehensive analysis, though less suited to daily operation, we employ generative models that learn the joint distribution of 24-hour order-level submissions given weather and fuel variables. These models generate synthetic daily scenarios of individual buy and sell orders, which, once aggregated, yield hourly supply and demand curves. Based on these forecasts, we optimize a price-making storage strategy, quantify revenue distributions, and highlight the price-compression effect with lower peaks, higher off-peak levels, and diminishing returns as capacity expands.

We introduce control models for LLM-powered code completion in JetBrains IDEs: ML classifiers which trigger inference and filter the generated suggestions to better align them with users and reduce unnecessary requests. To this end, we evaluate boosting- and transformer-based architectures on an offline dataset of real code completions with n=98 users. We further evaluate the offline classification performance of our boosting-based approach on a range of syntactically diverse languages; and perform an A/B study in a production environment where they improve completion efficiency and quality metrics. With this study, we hope to demonstrate the potential in using auxiliary models for smarter in-IDE integration of LLM-driven features, highlight fruitful future directions, and open problems.

Large language models have achieved strong performance on medical reasoning benchmarks, yet their deployment in clinical settings demands rigorous verification to ensure factual accuracy. While reward models offer a scalable approach for reasoning trace verification, existing methods face two limitations: they produce only scalar reward values without explicit justification, and they rely on single-pass retrieval that precludes adaptive knowledge access as verification unfolds. We introduce $\method$, an agentic framework that addresses these limitations by training medical reasoning verifiers to iteratively query external medical corpora during evaluation. Our approach combines tool-augmented verification with an iterative reinforcement learning paradigm that requires only trace-level supervision, alongside an adaptive curriculum mechanism that dynamically adjusts training data distribution. Across four medical reasoning benchmarks, $\method$ achieves substantial gains over existing methods, improving MedQA accuracy by 23.5% and MedXpertQA by 32.0% relative to the base generator in particular. Crucially, $\method$ demonstrates an $\mathbf{8\times}$ reduction in sampling budget requirement compared to prior reward model baselines. These findings establish that grounding verification in dynamically retrieved evidence offers a principled path toward more reliable medical reasoning systems.

We derive an accounting identity for predictive models that links accuracy with common fairness criteria. The identity shows that for globally calibrated models, the weighted sums of miscalibration within groups and error imbalance across groups is equal to a "total unfairness budget." For binary outcomes, this budget is the model's mean-squared error times the difference in group prevalence across outcome classes. The identity nests standard impossibility results as special cases, while also describing inherent tradeoffs when one or more fairness measures are not perfectly satisfied. The results suggest that accuracy and fairness are best viewed as complements in binary prediction tasks: increasing accuracy necessarily shrinks the total unfairness budget and vice-versa. Experiments on benchmark data confirm the theory and show that many fairness interventions largely substitute between fairness violations, and when they reduce accuracy they tend to expand the total unfairness budget. The results extend naturally to prediction tasks with non-binary outcomes, illustrating how additional outcome information can relax fairness incompatibilities and identifying conditions under which the binary-style impossibility does and does not extend to regression tasks.

Reinforcement learning has empowered large language models to act as intelligent agents, yet training them for long-horizon tasks remains challenging due to the scarcity of high-quality trajectories, especially under limited resources. Existing methods typically scale up rollout sizes and indiscriminately allocate computational resources among intermediate steps. Such attempts inherently waste substantial computation budget on trivial steps while failing to guarantee sample quality. To address this, we propose \textbf{Spark} (\textbf{S}trategic \textbf{P}olicy-\textbf{A}ware explo\textbf{R}ation via \textbf{K}ey-state dynamic branching), a novel framework that selectively branches at critical decision states for resource-efficient exploration. Our key insight is to activate adaptive branching exploration at critical decision points to probe promising trajectories, thereby achieving precise resource allocation that prioritizes sampling quality over blind coverage. This design leverages the agent's intrinsic decision-making signals to reduce dependence on human priors, enabling the agent to autonomously expand exploration and achieve stronger generalization. Experiments across diverse tasks (e.g., embodied planning), demonstrate that \textsc{Spark} achieves superior success rates with significantly fewer training samples, exhibiting robust generalization even in unseen scenarios.

As an important part of urbanization, the development monitoring of newly constructed parks is of great significance for evaluating the effect of urban planning and optimizing resource allocation. However, traditional change detection methods based on remote sensing imagery have obvious limitations in high-level and intelligent analysis, and thus are difficult to meet the requirements of current urban planning and management. In face of the growing demand for complex multi-modal data analysis in urban park development monitoring, these methods often fail to provide flexible analysis capabilities for diverse application scenarios. This study proposes a multi-modal LLM agent framework, which aims to make full use of the semantic understanding and reasoning capabilities of LLM to meet the challenges in urban park development monitoring. In this framework, a general horizontal and vertical data alignment mechanism is designed to ensure the consistency and effective tracking of multi-modal data. At the same time, a specific toolkit is constructed to alleviate the hallucination issues of LLM due to the lack of domain-specific knowledge. Compared to vanilla GPT-4o and other agents, our approach enables robust multi-modal information fusion and analysis, offering reliable and scalable solutions tailored to the diverse and evolving demands of urban park development monitoring.

Mixture-of-Experts (MoE) layers have emerged as an important tool in scaling up modern neural networks by decoupling total trainable parameters from activated parameters in the forward pass for each token. However, sparse MoEs add complexity to training due to (i) new trainable parameters (router weights) that, like all other parameter groups, require hyperparameter (HP) tuning; (ii) new architecture scale dimensions (number of and size of experts) that must be chosen and potentially taken large. To make HP selection cheap and reliable, we propose a new parameterization for transformer models with MoE layers when scaling model width, depth, number of experts, and expert (hidden) size. Our parameterization is justified by a novel dynamical mean-field theory (DMFT) analysis. When varying different model dimensions trained at a fixed token budget, we find empirically that our parameterization enables reliable HP transfer across models from 51M to over 2B total parameters. We further take HPs identified from sweeping small models on a short token horizon to train larger models on longer horizons and report performant model behaviors.

Recent work has shown that machine-learned predictions can provably improve the performance of classic algorithms. In this work, we propose the first minimum-cost network flow algorithm augmented with a dual prediction. Our method is based on a classic minimum-cost flow algorithm, namely $\varepsilon$-relaxation. We provide time complexity bounds in terms of the infinity norm prediction error, which is both consistent and robust. We also prove sample complexity bounds for PAC-learning the prediction. We empirically validate our theoretical results on two applications of minimum-cost flow, i.e., traffic networks and chip escape routing, in which we learn a fixed prediction, and a feature-based neural network model to infer the prediction, respectively. Experimental results illustrate $12.74\times$ and $1.64\times$ average speedup on two applications.

Recent advances align diffusion models with human preferences to increase aesthetic appeal and mitigate artifacts and biases. Such methods aim to maximize a conditional output distribution aligned with higher rewards whilst not drifting far from a pretrained prior. This is commonly enforced by KL (Kullback Leibler) regularization. As such, a central issue still remains: how does one choose the right regularization strength? Too high of a strength leads to limited alignment and too low of a strength leads to "reward hacking". This renders the task of choosing the correct regularization strength highly non-trivial. Existing approaches sweep over this hyperparameter by aligning a pretrained model at multiple regularization strengths and then choose the best strength. Unfortunately, this is prohibitively expensive. We introduce DeRaDiff, a denoising time realignment procedure that, after aligning a pretrained model once, modulates the regularization strength during sampling to emulate models trained at other regularization strengths without any additional training or finetuning. Extending decoding-time realignment from language to diffusion models, DeRaDiff operates over iterative predictions of continuous latents by replacing the reverse step reference distribution by a geometric mixture of an aligned and reference posterior, thus giving rise to a closed form update under common schedulers and a single tunable parameter, lambda, for on the fly control. Our experiments show that across multiple text image alignment and image-quality metrics, our method consistently provides a strong approximation for models aligned entirely from scratch at different regularization strengths. Thus, our method yields an efficient way to search for the optimal strength, eliminating the need for expensive alignment sweeps and thereby substantially reducing computational costs.

Finite mixture models are widely used in econometric analyses to capture unobserved heterogeneity. This paper shows that maximum likelihood estimation of finite mixtures of parametric densities can suffer from substantial finite-sample bias in all parameters under mild regularity conditions. The bias arises from the influence of outliers in component densities with unbounded or large support and increases with the degree of overlap among mixture components. I show that maximizing the classification-mixture likelihood function, equipped with a consistent classifier, yields parameter estimates that are less biased than those obtained by standard maximum likelihood estimation (MLE). I then derive the asymptotic distribution of the resulting estimator and provide conditions under which oracle efficiency is achieved. Monte Carlo simulations show that conventional mixture MLE exhibits pronounced finite-sample bias, which diminishes as the sample size or the statistical distance between component densities tends to infinity. The simulations further show that the proposed estimation strategy generally outperforms standard MLE in finite samples in terms of both bias and mean squared errors under relatively weak assumptions. An empirical application to latent group panel structures using health administrative data shows that the proposed approach reduces out-of-sample prediction error by approximately 17.6% relative to the best results obtained from standard MLE procedures.

Robust reinforcement learning methods typically focus on suppressing unreliable experiences or corrupted rewards, but they lack the ability to reason about the reliability of their own learning process. As a result, such methods often either overreact to noise by becoming overly conservative or fail catastrophically when uncertainty accumulates. In this work, we propose a meta-cognitive reinforcement learning framework that enables an agent to assess, regulate, and recover its learning behavior based on internally estimated reliability signals. The proposed method introduces a meta-trust variable driven by Value Prediction Error Stability (VPES), which modulates learning dynamics via fail-safe regulation and gradual trust recovery. Experiments on continuous-control benchmarks with reward corruption demonstrate that recovery-enabled meta-cognitive control achieves higher average returns and significantly reduces late-stage training failures compared to strong robustness baselines.

X-Codec-2.0 has shown strong performance in neural audio compression and multilingual speech modeling, operating at a 50 Hz latent rate and a 16 kHz sampling rate using frozen HuBERT features. While effective, this configuration limits temporal efficiency and audio fidelity. In this work, we explore a simple and effective modification by introducing additional pooling and increasing the decoder hop size. This reduces the latent rate from 50 Hz to 25 Hz and simultaneously raises the output sampling rate from 16 kHz to 24 kHz, improving efficiency and perceptual quality without altering the core architecture. Evaluated on the multilingual Common Voice 17 test set, the proposed configuration achieves a 0.29 MOS improvement over the original X-Codec-2.0 baseline based on UTMOSv2, and attains the best reported performance among all codecs operating at 25 Hz. The source code, checkpoints, and generation comparisons are released at \href{https://huggingface.co/Scicom-intl/xcodec2-25TPS-24k}{https://huggingface.co/Scicom-intl/xcodec2-25TPS-24k}.

The increasing integration of AI agents into cyber-physical systems (CPS) introduces new security risks that extend beyond traditional cyber or physical threat models. Recent advances in generative AI enable deepfake and semantic manipulation attacks that can compromise agent perception, reasoning, and interaction with the physical environment, while emerging protocols such as the Model Context Protocol (MCP) further expand the attack surface through dynamic tool use and cross-domain context sharing. This survey provides a comprehensive review of security threats targeting AI agents in CPS, with a particular focus on environmental interactions, deepfake-driven attacks, and MCP-mediated vulnerabilities. We organize the literature using the SENTINEL framework, a lifecycle-aware methodology that integrates threat characterization, feasibility analysis under CPS constraints, defense selection, and continuous validation. Through an end-to-end case study grounded in a real-world smart grid deployment, we quantitatively illustrate how timing, noise, and false-positive costs constrain deployable defenses, and why detection mechanisms alone are insufficient as decision authorities in safety-critical CPS. The survey highlights the role of provenance- and physics-grounded trust mechanisms and defense-in-depth architectures, and outlines open challenges toward trustworthy AI-enabled CPS.

Performative prediction captures the phenomenon where deploying a predictive model shifts the underlying data distribution. While simple retraining dynamics are known to converge linearly when the performative effects are weak ($ρ< 1$), the complexity in the regime $ρ> 1$ was hitherto open. In this paper, we establish a sharp phase transition: computing an $ε$-performatively stable point is PPAD-complete -- and thus polynomial-time equivalent to Nash equilibria in general-sum games -- even when $ρ= 1 + O(ε)$. This intractability persists even in the ostensibly simple setting with a quadratic loss function and linear distribution shifts. One of our key technical contributions is to extend this PPAD-hardness result to general convex domains, which is of broader interest in the complexity of variational inequalities. Finally, we address the special case of strategic classification, showing that computing a strategic local optimum is PLS-hard.

Out-of-distribution (OOD) generalization remains a fundamental challenge in real-world classification, where test distributions often differ substantially from training data. Most existing approaches pursue domain-invariant representations, implicitly assuming that invariance implies reliability. However, features that are invariant across domains are not necessarily causally effective for prediction. In this work, we revisit OOD classification from a causal perspective and propose to evaluate learned representations based on their necessity and sufficiency under distribution shift. We introduce an explicit segment-level framework that directly measures causal effectiveness across domains, providing a more faithful criterion than invariance alone. Experiments on multi-domain benchmarks demonstrate consistent improvements in OOD performance, particularly under challenging domain shifts, highlighting the value of causal evaluation for robust generalization.

Numerical techniques for solving partial differential equations (PDEs) are integral for many fields across science and engineering. Such techniques usually involve solving large, sparse linear systems, where preconditioning methods are critical. In recent years, neural methods, particularly graph neural networks (GNNs), have demonstrated their potential through accelerated convergence. Nonetheless, to extract connective structures, existing techniques aggregate discretized system matrices into graphs, and suffer from rank inflation and a suboptimal convergence rate. In this paper, we articulate NeuraLSP, a novel neural preconditioner combined with a novel loss metric that leverages the left singular subspace of the system matrix's near-nullspace vectors. By compressing spectral information into a fixed low-rank operator, our method exhibits both theoretical guarantees and empirical robustness to rank inflation, affording up to a 53% speedup. Besides the theoretical guarantees for our newly-formulated loss function, our comprehensive experimental results across diverse families of PDEs also substantiate the aforementioned theoretical advances.

We present a new nonlinear dimensionality reduction method, MAPLE, that enhances UMAP by improving manifold modeling. MAPLE employs a self-supervised learning approach to more efficiently encode low-dimensional manifold geometry. Central to this approach are maximum manifold capacity representations (MMCRs), which help untangle complex manifolds by compressing variances among locally similar data points while amplifying variance among dissimilar data points. This design is particularly effective for high-dimensional data with substantial intra-cluster variance and curved manifold structures, such as biological or image data. Our qualitative and quantitative evaluations demonstrate that MAPLE can produce clearer visual cluster separations and finer subcluster resolution than UMAP while maintaining comparable computational cost.

We study how neural emulators of partial differential equation solution operators internalize physical symmetries by introducing an influence-based diagnostic that measures the propagation of parameter updates between symmetry-related states, defined as the metric-weighted overlap of loss gradients evaluated along group orbits. This quantity probes the local geometry of the learned loss landscape and goes beyond forward-pass equivariance tests by directly assessing whether learning dynamics couple physically equivalent configurations. Applying our diagnostic to autoregressive fluid flow emulators, we show that orbit-wise gradient coherence provides the mechanism for learning to generalize over symmetry transformations and indicates when training selects a symmetry compatible basin. The result is a novel technique for evaluating if surrogate models have internalized symmetry properties of the known solution operator.

As software engineering moves toward SE3.0, AI agents are increasingly used to carry out development tasks and contribute changes to software projects. It is therefore important to understand the extent of these contributions and how human developers review and intervene, since these factors shape the risks of delegating work to AI agents. While recent studies have examined how AI agents support software development tasks (e.g., code generation, issue resolution, and PR automation), their role in documentation tasks remains underexplored-even though documentation is widely consumed and shapes how developers understand and use software. Using the AIDev, we analyze 1,997 documentation-related pull requests (PRs) authored by AI agents and human developers, where documentation PRs are those that create or modify project documentation artifacts. We find that AI agents submit substantially more documentation-related PRs than humans in the studied repositories. We further observe that agent-authored documentation edits are typically integrated with little follow-up modification from humans, raising concerns about review practices and the reliability of agent-generated documentation. Overall, while AI agents already contribute substantially to documentation workflows, our results suggest concerns for emerging challenges for documentation quality assurance and human-AI collaboration in SE3.0.

Due to the rise of cardinality minimization in optimization, sparse support vector machines (SSVMs) have attracted much attention lately and show certain empirical advantages over convex SVMs. A common way to derive an SSVM is to add a cardinality function such as $\ell_0$-norm to the dual problem of a convex SVM. However, this process lacks theoretical justification. This paper fills the gap by developing a local duality theory for such an SSVM formulation and exploring its relationship with the hinge-loss SVM (hSVM) and the ramp-loss SVM (rSVM). In particular, we prove that the derived SSVM is exactly the dual problem of the 0/1-loss SVM, and the linear representer theorem holds for their local solutions. The local solution of SSVM also provides guidelines on selecting hyperparameters of hSVM and rSVM. {Under specific conditions, we show that a sequence of global solutions of hSVM converges to a local solution of 0/1-loss SVM. Moreover, a local minimizer of 0/1-loss SVM is a local minimizer of rSVM.} This explains why a local solution induced by SSVM outperforms hSVM and rSVM in the prior empirical study. We further conduct numerical tests on real datasets and demonstrate potential advantages of SSVM by working with locally nice solutions proposed in this paper.

Prior work suggests that language models, while trained on next token prediction, show implicit planning behavior: they may select the next token in preparation to a predicted future token, such as a likely rhyming word, as supported by a prior qualitative study of Claude 3.5 Haiku using a cross-layer transcoder. We propose much simpler techniques for assessing implicit planning in language models. With case studies on rhyme poetry generation and question answering, we demonstrate that our methodology easily scales to many models. Across models, we find that the generated rhyme (e.g. "-ight") or answer to a question ("whale") can be manipulated by steering at the end of the preceding line with a vector, affecting the generation of intermediate tokens leading up to the rhyme or answer word. We show that implicit planning is a universal mechanism, present in smaller models than previously thought, starting from 1B parameters. Our methodology offers a widely applicable direct way to study implicit planning abilities of LLMs. More broadly, understanding planning abilities of language models can inform decisions in AI safety and control.

Large Language Model (LLM)-based mobile agents have made significant performance advancements. However, these agents often follow explicit user instructions while overlooking personalized needs, leading to significant limitations for real users, particularly without personalized context: (1) inability to interpret ambiguous instructions, (2) lack of learning from user interaction history, and (3) failure to handle personalized instructions. To alleviate the above challenges, we propose Me-Agent, a learnable and memorable personalized mobile agent. Specifically, Me-Agent incorporates a two-level user habit learning approach. At the prompt level, we design a user preference learning strategy enhanced with a Personal Reward Model to improve personalization performance. At the memory level, we design a Hierarchical Preference Memory, which stores users' long-term memory and app-specific memory in different level memory. To validate the personalization capabilities of mobile agents, we introduce User FingerTip, a new benchmark featuring numerous ambiguous instructions for daily life. Extensive experiments on User FingerTip and general benchmarks demonstrate that Me-Agent achieves state-of-the-art performance in personalization while maintaining competitive instruction execution performance.

Software development agents such as Claude Code, GitHub Copilot, Cursor Agent, Devin, and OpenAI Codex are being increasingly integrated into developer workflows. While prior work has evaluated agent capabilities for code completion and task automation, there is little work investigating how these agents perform Java refactoring in practice, the types of changes they make, and their impact on code quality. In this study, we present the first analysis of agentic refactoring pull requests in Java, comparing them to developer refactorings across 86 projects per group. Using RefactoringMiner and DesigniteJava 3.0, we identify refactoring types and detect code smells before and after refactoring commits. Our results show that agent refactorings are dominated by annotation changes (the 5 most common refactoring types done by agents are annotation related), in contrast to the diverse structural improvements typical of developers. Despite these differences in refactoring types, we find Cursor to be the only model to show a statistically significant increase in refactoring smells.

Most of the current software security analysis tools assess vulnerabilities in isolation. However, sophisticated software supply chain security threats often stem from cascaded vulnerability and security weakness chains that span dependent components. Moreover, although the adoption of Software Bills of Materials (SBOMs) has been accelerating, downstream vulnerability findings vary substantially across SBOM generators and analysis tools. We propose a novel approach to SBOM-driven security analysis methods and tools. We model vulnerability relationships over dependency structure rather than treating scanner outputs as independent records. We represent enriched SBOMs as heterogeneous graphs with nodes being the SBOM components and dependencies, the known software vulnerabilities, and the known software security weaknesses. We then train a Heterogeneous Graph Attention Network (HGAT) to predict whether a component is associated with at least one known vulnerability. Since documented multi-vulnerability chains are scarce, we model cascade discovery as a link prediction problem over CVE pairs using a multi-layer perceptron neural network. This way, we produce ranked candidate links that can be composed into multi-step paths. The HGAT component classifier achieves an Accuracy of 91.03% and an F1-score of 74.02%.

Pairwise-constrained clustering augments unsupervised partitioning with side information by enforcing must-link (ML) and cannot-link (CL) constraints between specific samples, yielding labelings that respect known affinities and separations. However, ML and CL constraints add an extra layer of complexity to the clustering problem, with current methods struggling in data scalability, especially in niche applications like quantum or quantum-hybrid clustering. We propose PASS, a pairwise-constraints and ambiguity-driven subset selection framework that preserves ML and CL constraints satisfaction while allowing scalable, high-quality clustering solution. PASS collapses ML constraints into pseudo-points and offers two selectors: a constraint-aware margin rule that collects near-boundary points and all detected CL violations, and an information-geometric rule that scores points via a Fisher-Rao distance derived from soft assignment posteriors, then selects the highest-information subset under a simple budget. Across diverse benchmarks, PASS attains competitive SSE at substantially lower cost than exact or penalty-based methods, and remains effective in regimes where prior approaches fail.

Self-supervised learning (SSL) have improved empirical performance by unleashing the power of unlabeled data for practical applications. Specifically, SSL extracts the representation from massive unlabeled data, which will be transferred to a plenty of down streaming tasks with limited data. The significant improvement on diverse applications of representation learning has attracted increasing attention, resulting in a variety of dramatically different self-supervised learning objectives for representation extraction, with an assortment of learning procedures, but the lack of a clear and unified understanding. Such an absence hampers the ongoing development of representation learning, leaving a theoretical understanding missing, principles for efficient algorithm design unclear, and the use of representation learning methods in practice unjustified. The urgency for a unified framework is further motivated by the rapid growth in representation learning methods. In this paper, we are therefore compelled to develop a principled foundation of representation learning. We first theoretically investigate the sufficiency of the representation from a spectral representation view, which reveals the spectral essence of the existing successful SSL algorithms and paves the path to a unified framework for understanding and analysis. Such a framework work also inspires the development of more efficient and easy-to-use representation learning algorithms with principled way in real-world applications.

Spin pumping provides a fundamental route for dynamical spin transport, yet in its conventional form it produces only linear spin responses at the driving frequency. Recent studies have shown that spin-orbit coupling (SOC) can lift these restrictions and enable highly nonlinear spin and charge currents. Here we propose a distinct mechanism for high-harmonic spin pumping that does not rely on spin-orbit interactions. We show that nonlinearities in the adiabatic energy spectrum--rather than SOC itself--constitute the essential ingredient for high-harmonic generation in pumped spin currents. Such nonlinearities can arise in purely magnetic systems when a secondary magnetic order parameter is introduced perpendicular to the cone axis of a precessing magnetic moment. As a result, spin pumping and its higher harmonics emerge even in the complete absence of SOC. Our findings establish a new route to ultrafast spin pumping based solely on magnetic structure and dynamics.

Star-like microgels have recently emerged as a promising class of thermoresponsive soft colloids, that have an internal architecture similar to that of star polymers. Here, we perform extensive monomer-resolved simulations to theoretically establish this analogy. First, we characterize the effective potential between star-like microgels, finding that it is Gaussian for an extended range of distances, in stark contrast to the Hertzian-like one of standard microgels, but almost identical to that of star polymers with a core partially covered by chains. Next, we investigate the ratio between gyration and hydrodynamic radii across the volume-phase transition, showing qualitative agreement with both star polymers and experimental data. Finally, we estimate the bulk modulus, finding star-like microgels significantly softer than standard microgels and comparable to star polymers. The present work thus demonstrates that star-like microgels behave as ultrasoft particles, akin to star polymers, paving the way for their exploration at high concentrations.

Directed and heterogeneous hypergraphs capture directional higher-order interactions with intrinsically asymmetric functional dependencies among nodes. As a result, damage to certain nodes can suppress entire hyperedges, whereas failure of others only weakens interactions. Metabolic reaction networks offer an intuitive example of such asymmetric dependencies. Here we develop a message-passing and statistical mechanics framework for percolation in directed hypergraphs that explicitly incorporates directionality and node heterogeneity. Remarkably, we show that these hypergraph features have a fundamental effect on the critical properties of hypergraph percolation, reshaping criticality in a way that depends on network structure. Specifically, we derive anomalous critical exponents that depend on whether node or hyperedge percolation is considered in maximally correlated, heavy-tailed regimes. These theoretical predictions are validated on synthetic hypergraph models and on a real directed metabolic network, opening new perspectives for the characterization of the robustness and resilience of real-world directed, heterogeneous higher-order networks.

Phagocytosis is a fundamental process of the innate immune system, yet the physical determinants that govern the engulfment of soft, deformable targets remain poorly understood. Existing theoretical models typically approximate targets as rigid particles, overlooking the fact that both immune cells and many biological targets undergo significant membrane deformation during contact. Here, we develop a Monte Carlo-based membrane simulation framework to model the interactions of multiple vesicles, enabling us to explore phagocytosis-like processes in systems where both the phagocyte and the target possess flexible, thermally fluctuating membranes. We first validate our approach against established observations for the engulfment of rigid objects. We then investigate how the mechanical properties of a soft target -- specifically membrane bending rigidity govern the outcome of phagocytic interactions. Our simulations reveal three distinct mechanical regimes: (i) biting or trogocytosis, in which the phagocyte extracts a portion of the target vesicle; (ii) pushing, where the target is displaced rather than engulfed; and (iii) full engulfment, in which the target is completely internalized. Increasing membrane tension via internal pressure produces analogous transitions, demonstrating a unified mechanical origin for these behaviours. Qualitative comparison with experiments involving Giant Unilamellar Vesicles (GUVs, deformable microparticles) and lymphoma cells supports the relevance of these regimes to biological phagocytosis. Together, these results highlight how target deformability fundamentally shapes phagocytic success and suggest that immune cells may exploit mechanical cues to recognize among different classes of soft targets.

The recent discovery of type-II multiferroicity in monolayer NiI${_2}$ indicated a new pathway for intrinsic magnetoelectric coupling in the two-dimensional limit. However, determining whether this phenomenon is a unique anomaly or a general, chemically tunable property of the material class remains unresolved. Here, we demonstrate the universality of type-II multiferroicity in the transition metal dihalides by visualizing the ferroelectric order in monolayer NiBr${_2}$. Using scanning tunneling microscopy (STM), we resolve atomic-scale ferroelectric domains and confirm their magnetoelectric origin through reciprocal manipulation experiments: reorienting magnetic order via electric fields and suppressing the electric polarization with external magnetic fields. Furthermore, we find that the multiferroic state in NiBr${_2}$ is energetically less robust than in its iodide counterpart, consistent with modified superexchange interactions and the reduced spin-orbit coupling. Our results establish the transition metal dihalides as a versatile platform where the stability of magnetoelectric phases can be engineered through chemical substitution.

Species populations often modify their environment as they grow. When environmental feedback operates more slowly than population growth, the system can undergo boom-bust dynamics, where the population overshoots its carrying capacity and subsequently collapses. In extreme cases, this collapse leads to total extinction. While deterministic models typically fail to capture these finite-time extinction events, we propose a stochastic framework, derived from an individual-based model, to describe boom-bust-extirpation dynamics. We identify a noise-driven, threshold-like behavior where, depending on initial conditions, the population either undergoes a "boom" or is extirpated before the expansion occurs. Furthermore, we characterize a transition between an excitable regime, where most trajectories are captured by the absorbing state immediately after the first bust, and a persistent regime, where most populations reach a metastable state. We show that this transition is governed by the diffusion strength and the ratio of environmental-to-population timescales. This framework provides a theoretical basis for understanding irreversible transitions in invasive species, plant succession, and microbial dynamics.

Spin-wave nonreciprocity, manifested as a frequency difference between counterpropagating modes, underpins many proposed magnonic devices. While this effect is commonly attributed to dipolar interactions or interfacial chirality, the microscopic origin of nonreciprocal dispersion in magnetic multilayers remains under debate. Here, we analyze nonreciprocal spin-wave dispersion in planar multilayer heterostructures without Dzyaloshinskii-Moriya interaction. Using a frequency-shift dynamic matrix and an interaction-resolved dynamic energy-density formalism, we show that the frequency asymmetry cannot generally be ascribed to dipolar effects alone. Instead, once counterpropagating modes differ in their geometric structure along the thickness, interlayer exchange dominates the frequency shift. Applied to representative multilayer systems, we find that the interlayer exchange contribution exceeds dipolar and intralayer exchange effects by up to two to three orders of magnitude over a broad wave-vector range. Our results establish interlayer exchange as the primary mechanism controlling nonreciprocal dispersion in multilayer magnonic systems and provide a quantitative framework for engineering large frequency shifts in nonreciprocal magnonic devices.

Solving the quantum many-body ground state problem remains a central challenge in computational physics. In this context, the Variational Monte Carlo (VMC) framework based on Projected Entangled Pair States (PEPS) has witnessed rapid development, establishing itself as a vital approach for investigating strongly correlated two-dimensional systems. However, standard PEPS-VMC algorithms predominantly rely on sequential local updates. This conventional approach often suffers from slow convergence and critical slowing down, particularly in the vicinity of phase transitions or within frustrated landscapes. To address these limitations, we propose an efficient autoregressive row-wise sampling algorithm for PEPS that enables direct, rejection-free sampling via single-layer contractions. By utilizing autoregressive single-layer row updates to generate collective, non-local configuration proposals, our method significantly reduces temporal correlations compared to local Metropolis moves. We benchmark the algorithm on the two-dimensional transverse-field Ising model and the quantum spin glass. Our results demonstrate that the row-wise scheme effectively mitigates critical slowing down near the Ising critical point. Furthermore, in the rugged landscape of the quantum spin glass, it yields improved optimization stability and lower ground-state energies. These findings indicate that single-layer autoregressive row updates provide a flexible and robust improvement to local PEPS-VMC sampling and may serve as a basis for more advanced sampling schemes.

Searching a database is a central task in computer science and is paradigmatic of transport and optimization problems in physics. For an unstructured search, Grover's algorithm predicts a quadratic speedup, with the search time $τ(N)=Θ(\sqrt{N})$ and $N$ the database size. Numerical studies suggest that the time complexity can change in the presence of feedback, injecting information during the search. Here, we determine the time complexity of the quantum analog of a randomized algorithm, which implements feedback in a simple form. The search is a continuous-time quantum walk on a complete graph, where the target is continuously monitored by a detector. Additionally, the quantum state is reset if the detector does not click within a specified time interval. This yields a non-unitary, non-Markovian dynamics. We optimize the search time as a function of the hopping amplitude, detection rate, and resetting rate, and identify the conditions under which time complexity could outperform Grover's scaling. The overall search time does not violate Grover's optimality bound when including the time budget of the physical implementation of the measurement. For databases of finite sizes monitoring can warrant rapid convergence and provides a promising avenue for fault-tolerant quantum searches.

The formation of complexes between like-charged polyelectrolytes challenges conventional electrostatic intuition and highlights the central role of ions in mediating macromolecular organization. Here, we investigate the salt-dependent association of DNA with poly(acrylic acid) (PAA) using atomistic molecular dynamics simulations in NaCl, MgCl$_2$, and CaCl$_2$ solutions. A time-resolved state classification scheme, based on heavy-atom distance and hydrogen-bond formation, was applied to distinguish bound and unbound configurations, enabling quantitative analysis of how ion valency modulates complex stability and structure. The results reveal a clear hierarchy of association strength with Ca$^{2+}$ promoting persistent complex formation through direct inner-sphere coordination between DNA phosphates and PAA carboxylates, Mg$^{2+}$ mediating weaker, transient bridging interactions and Na$^+$ exhibiting only electrostatic screening action with negligible bridge formation. Structural analysis shows that multivalent ions not only enhance complex stability but also reshape the molecular organization of both macromolecules. Ca$^{2+}$ induces expansion of DNA and compaction of PAA within a strongly bridged complex characterized by directional alignment and backbone-dominated binding, whereas Mg$^{2+}$ promotes more transient groove associations and Na$^+$ supports flexible, weakly correlated contacts. Our findings provide molecular-level insight into ion-specific mechanisms underlying polyelectrolyte organization and inform the design of responsive biomaterials and nucleic acid-based assemblies in multivalent ionic environments.

The remarkable advances achieved in two-dimensional materials are now being directly transposed to low-dimensional oxides. Here we show using first-principles-based atomistic simulations that ultrathin freestanding ferroelectric layers host a rich variety of polar states, from liquid-like ferroelectric domains with long-range orientational order to helix-wave and chiral bubbles configurations reminiscent of those observed in twisted freestanding oxide layers. Time-dependent electric fields enable reversible control, revealing freestanding oxide layers as ideal platforms to explore complex polar states and their potential applications in future ferroic devices.

Anderson localization admits a complete symmetry classification given by the Altland-Zirnbauer (AZ) tenfold scheme, whereas an analogous framework for interacting many-body localization (MBL) has remained elusive. Here we develop a symmetry-based classification of static MBL phases formulated at the level of local integrals of motion (LIOMs). We show that a symmetry is compatible with stable MBL if and only if its action can be consistently represented within a quasi-local LIOM algebra, without enforcing extensive degeneracies or nonlocal operator mixing. This criterion sharply distinguishes symmetry classes: onsite Abelian symmetries are compatible with stable MBL and can host distinct symmetry-protected topological MBL phases, whereas continuous non-Abelian symmetries generically preclude stable MBL. By systematically combining AZ symmetries with additional onsite symmetries, we construct a complete classification table of MBL phases, identify stable, fragile, and unstable classes, and provide representative lattice realizations. Our results establish a unified and physically transparent framework for understanding symmetry constraints on MBL.

When macroscopic quantum condensates -- superconductors, superfluids, cold atoms and ions, polaritons etc. -- are put in rotation, a quantum vortex lattice forms inside. In homogeneous type-II superconductors, each vortex has a tiny core where the superconducting gap $Δ(r)$ is known to smoothly vanish towards the core centre on the scale of the coherence length $ξ$. The cores host quantized quasiparticle energy levels known as Caroli-de Gennes-Matricon (CdGM) bound states [Caroli {\it et al.,} Phys. Lett. v. 9, 307 (1964)]. In pure materials, the spectrum of the low-lying CdGM states has the characteristic level spacing $\sim Δ_0^2/E_F$, where $E_F$ is the Fermi energy and $Δ_0$ is the bulk gap. In disordered ones, the CdGM states shift and broaden due to scattering. Here, we show, both experimentally and theoretically, that the situation is completely different in granular Nb films, which are commonly used in superconducting electronics. In these films, in which the grains are smaller than $ξ$, the gap $Δ$ in the quasiparticle spectrum reduces towards the vortex core centres by discrete jumps at the grain boundaries. The bound states adapt to the local environment and appear at unexpectedly high energies. Both $Δ(r)$ and bound states form a puzzle-like spatial structure of the core, elements of which are whole grains. Our discovery shakes up the established understanding of the quantum vortex and encourages a reconsideration of the vortex motion and pinning mechanisms in granular superconductors.

In thermodynamic equilibrium, the fluctuation-dissipation theorem links thermal fluctuations and dissipation. Biological systems, however, are driven out of equilibrium by internal processes that produce additional, active fluctuations. Despite being relevant for biological functions such as intracellular transport, predicting the statistical properties of active fluctuations remains challenging. Here, we address this challenge in a minimal model of an active gel as a network of elastic elements connected by transient crosslinks. The crosslinkers' binding and unbinding rates break detailed balance, which drives the system out of equilibrium. Through coarse-graining, we derive fluctuating hydrodynamic equations including an active noise term, which emerges explicitly from the breaking of detailed balance. Finally, we provide predictions for the stochastic motion of a tracer particle embedded in the active gel, which enables comparisons with microrheology experiments both in synthetic active gels and in cells. Overall, our work provides an explicit link between the statistical properties of active fluctuations and the molecular breaking of detailed balance. Thus, it paves the way toward complementing the fluctuation-dissipation theorem with a fluctuation-activity relation in active systems.

We investigate bias-driven non-equilibrium quantum phase transitions in a paradigmatic quantum-transport setup: an interacting quantum dot coupled to non-interacting metallic leads. Using the Random Phase Approximation, which is exact in the limit of a large number of dot levels, we map out the zero-temperature non-equilibrium phase diagram as a function of interaction strength and applied bias. We focus our analysis on the behavior of the charge susceptibility and the current noise in the vicinity of the transition. Remarkably, despite the intrinsically non-equilibrium nature of the steady state, critical charge fluctuations admit an effective-temperature description, $T_{\text{eff}}(T,V)$, that collapses the steady-state behavior onto its equilibrium form. In sharp contrast, current fluctuations exhibit genuinely non-equilibrium features: the fluctuation-dissipation ratio becomes negative in the ordered phase, corresponding to a negative effective temperature for the current degrees of freedom. These results establish current noise as a sensitive probe of critical fluctuations at non-equilibrium quantum phase transitions and open new directions for exploring voltage-driven critical phenomena in quantum transport systems.

Layered Ising metamagnets are antiferromagnetic (AF) materials consisting of monolayer Ising ferromagnets coupled to each other via interlayer AF interactions. They exhibit rich magnetic phase diagrams, featuring tricritical and critical end points, due to the competing magnetic interactions and the Ising anisotropy. While conventional thermodynamic probes can identify these critical points in bulk Ising metamagnets, achieving this in the two-dimensional (2D) limit, where enhanced fluctuation effects can substantially modify critical phenomena, remains to be realized. Here, we combine specific heat capacity (C_V) and magnetic circular dichroism measurements to identify these critical points, extract a tricritical exponent, and map out the complete magnetic phase diagram of 2D Ising metamagnetic CrI3. This is achieved in a nanomechanical device of 6-layer CrI3, in which a direct measurement of the temperature derivative of its mechanical resonance frequency gives C_V. The tricritical point is identified by the onset of an abrupt spin-flip transition on one side and, on the other side, by a vanishing specific heat λ-anomaly for a continuous AF phase transition. In contrast, only the spin-flip transition remains near the critical end point. Our results establish nanomechanical calorimetry as a general route to classify metamagnetic phase transitions and to study multicritical phenomena in 2D magnets.

When polydisperse granular systems are sheared, the transverse dynamics is characterized by the interplay of size segregation and diffusion. Segregation in nonuniform and confined shearing flows is studied using annular shear cell experiments complemented with discrete numerical simulations of bidisperse, inelastic, and frictional spheres under gravity. We explored the role of shear localization, granular temperature, boundaries, and mixture properties in the evolution of the segregation rate and the maximum degree of segregation achieved by a bidisperse granular system in the steady state. A faster segregation process and a more developed degree of segregation is observed for bidisperse mixtures with a larger size ratio and a higher proportion of large particles. Normally, in the presence of gravity, size segregation induces large particles to rise and small particles to sink. However, two additional complex segregation patterns were found: inverse segregation and horizontal segregation. The first might be related to the kinematics of the flow, while the second is a geometrical effect. This additional segregation mechanism, in addition to diffusion fluxes and high confining pressure, hampers complete segregation in the steady state, where some degree of mixing always persists.

We examine the Ramo current in a nanoscale planar vacuum diode undergoing field emission in the presence of a DC voltage supply and an external resistor. We describe a simple mechanism for generating persistent current oscillations in the diode due to the voltage drop across the external resistor (beam loading) which reduces the total field and inhibits the emission. The amplitude and the frequency, which is in the THz domain, depend on the operating parameters of the diode. Molecular dynamics simulations are used to find the characteristics and physical basis of the mechanism, and a simple analytical model is presented, in good agreement with the simulation.

Coupled-wire description has been developed as a powerful framework for providing bosonic descriptions of strongly correlated quantum matter, with early applications to systems such as the cuprates and the integer and fractional quantum Hall states. In this topical review, we discuss recent developments of coupled-wire description in nanoscale systems, where it emerges not only as a theoretical tool but also as a highly tunable physical platform. In these nanoscale realizations, coupled-wire networks are formed by one-dimensional channels embedded in two-dimensional materials, most prominently in moiré and twisted structures. Such networks host a broad range of unconventional states of matter, including superconductivity, charge density waves, spin density waves, Mott insulating phases, Anderson insulating phases, quantum spin Hall states, quantum anomalous Hall states, and their fractionalized counterparts. The ability to electrically control interaction strength, confinement, and coupling between wires makes these systems qualitatively different from earlier realizations and allows continuous tuning between competing phases. Notably, recent work has demonstrated that the coupled-wire framework in moiré networks completes the trio of quantum Hall phenomena, encompassing quantum Hall, quantum spin Hall, and quantum anomalous Hall states, together with their fractional analogues. This development highlights coupled-wire networks in nanoscale materials as a versatile and experimentally relevant setting for exploring the interplay of topology, strong correlations, and low-dimensional physics.

Scanning transmission X-ray microscopy and ptychography have become mature tools for high-resolution, element-specific imaging of nanoscale structures. However, transmission geometries impose stringent constraints on sample thickness and preparation, thereby limiting investigations of extended or bulk specimens, especially in the soft X-ray region. Here, we demonstrate reflection geometry soft X-ray ptychography as a robust imaging mode. Instrumental feasibility and spatial resolution are established using a lithographically defined Siemens star and barcode test pattern on a multilayer substrate. We empirically demonstrate a full-pitch spatial resolution of ca. 45 nm from Fourier ring correlation analysis of the reconstructed object. The results highlight the potential of the reflection geometry for nondestructive X-ray studies of materials without the need for transmissive samples.

An array of quantum dots coupled via superconductivity provides a new platform for creating Kitaev chains with Majorana zero modes, offering a promising avenue toward topological quantum computing. In this work, we theoretically study the quantum capacitance of a minimal Kitaev chain weakly coupled to an external normal lead. We find that in the open regime, charge stability diagrams of quantum capcaitance can help to identify the sweet spot of a Kitaev chain, consistent with tunnel spectroscopy. Moreover, the quantum capacitance of a single quantum dot coupled to Andreev bound states reveals the interplay between two distinct parity switching mechanisms: coupling to an external normal lead and intrinsic quasiparticle poisoning. Our work provides useful physical insights into the quantum capacitance and parity dynamics in a quantum-dot-based Kitaev chain device.

Complex nonequilibrium self-assembly enables the formation of materials with specific patterns and functions from the bottom up. How to directionally control the assembly to form the target configuration is a challenge. Here, we propose a dissipation bias principle for targeted assembly, which highlights that controlling the dissipation tendency can play an important role by modulating the frequency and intensity of local rearrangements. Following this principle, one can induce ordered target configurations from disordered structures and also achieve directional selection among multiple assembly pathways. We use the assembly of active colloids as a platform to show our results.

Motivated by the recent interest in the criticality of open quantum many-body systems, we study nonlinear sigma models with complexified couplings as a general framework for nonunitary field theory. Applying the perturbative renormalization-group analysis to the tenfold symmetric spaces, we demonstrate that fixed points with complex scaling dimensions and critical exponents arise generically, without counterparts in conventional nonlinear sigma models with real couplings. We further clarify the global phase diagrams in the complex-coupling plane and identify both continuous and discontinuous phase transitions. Our work elucidates universal aspects of critical phenomena in complexified field theory.

Periodic assemblies of nanoparticles are central to surface patterning, with applications in biosensing, energy conversion, and nanofabrication. Evaporation of colloidal droplets on substrates provides a simple yet effective route to achieve such assemblies. This work reports the first experimental demonstration of patterns formed through stick-slip dynamics of the three-phase contact line during evaporation of gold nanoparticle suspensions on inclined substrates. Variation in nanoparticle concentration and substrate inclination alter the balance of interfacial and gravitational forces, producing multiple stick-slip events that generate periodic stripes. Stripe density exhibits a sinusoidal dependence on inclination angle, while inter-stripe spacing remains nearly invariant. Independent control over inter-stripe spacing is achieved through adjustment of nanoparticle or surfactant concentration. These results highlight the complex interplay of gravitational and interfacial forces in directing periodic nanoparticle assembly and establish a versatile, programmable framework for surface patterning with tunable nano/microscale dimensions.

Understanding how water wets graphene is critical for predicting and controlling its behavior in nanofluidic, sensing, and energy applications. A key measure of wetting is the contact angle made by a liquid droplet against the surface, yet experimental measurements for graphene span a wide range, and no consensus has emerged for free-standing graphene. Here, we use a machine learning potential with approaching ab initio accuracy to perform nanosecond-scale molecular dynamics and provide an atomistic first-principles benchmark for this unsolved problem. We find the contact angle of water on free-standing graphene, after finite-size correction, to be $72.1 \pm 1.5 °$. We also show that the three-phase contact line of a nanoscale water droplet couples strongly to the intrinsic thermal ripples of free-standing graphene, and that graphene's wetting properties are highly sensitive to mechanical strain. Tensile strain makes graphene significantly more hydrophobic, while compressive strain induces coherent ripples that the droplet "surfs", resulting in pronounced anisotropic wetting and contact angle hysteresis. Our results demonstrate that graphene's wetting properties are governed not only by its chemistry but also by its dynamic morphology, offering an additional explanation for the variability of experimental measurements. Furthermore, mechanical strain may be a practical route to controlling wetting in graphene-based technologies, with promising consequences for nanofluidic and nano-filtration applications.

We investigate first-hitting location (FHL) statistics induced by drift--diffusion processes in domains with absorbing boundaries, and examine how such boundary laws give rise to intrinsic information observables. Rather than introducing explicit encoding or decoding mechanisms, information is viewed as emerging directly from the geometry and dynamics of stochastic transport through first-passage events. Treating the FHL as the primary observable, we characterize how geometry and drift jointly shape the induced boundary measure. In diffusion-dominated regimes, the exit law exhibits scale-free, heavy-tailed spatial fluctuations along the boundary, whereas a nonzero drift component introduces an intrinsic length scale that suppresses these tails and reorganizes the exit statistics. Within a generator-based formulation, the FHL arises naturally as a boundary measure induced by an elliptic operator, allowing explicit boundary kernels to be derived analytically in canonical geometries. Planar absorbing boundaries in different ambient dimensions are examined as benchmark cases, illustrating how directed transport regularizes diffusion-driven fluctuations and induces qualitative transitions in boundary statistics. Overall, the present work provides a unified structural framework for first-hitting location laws and highlights FHL statistics as natural probes of geometry, drift, and irreversibility in stochastic transport.

Flat-band materials have attracted substantial interest for their intriguing quantum geometric effects. Here we investigate how singular flat bands (SFBs) respond to a static, uniform electric field and whether they can support single-particle dc transport. By constructing a minimal two-band lattice model, we show that away from the singular band crossing point (BCP), the Wannier-Stark (WS) spectrum of the flat band is well captured by an intraband Berry phase $Φ_{\mathrm{B}}$. The associated WS eigenstates are exponentially localized along the field direction, precluding dc transport. In contrast, near the BCP the interband Berry connection becomes prominent and drives Landau-Zener tunneling, which bends the flat-band WS ladder and delocalizes the SFB wavefunctions. Remarkably, this regime is governed solely by the maximal quantum distance $d$ through two geometric phases $(θ,\varphi)$: $θ$ characterizes the tunneling rate and $\varphi$ acts as a generalized Berry phase. These results highlight the essential role of quantum geometry in enabling nontrivial transport signatures in SFBs.

We study a gas of $N$ diffusing particles on the line subject to batch resetting: at rate $r$, a uniformly random subset of $m$ particles is reset to the origin. Despite the absence of interactions, the dynamics generates a nonequilibrium stationary state (NESS) with long-range correlations. We obtain exact results, both for the NESS and for the time dependence of the correlations, which are valid for arbitrary $m$ and $N$. By varying $m$, the system interpolates between an uncorrelated regime ($m=1$) and the fully synchronous resetting case ($m=N$). For all $1<m<N$, correlations exhibit a non-monotonic time dependence due to the emergence of an intrinsic decorrelation mechanism. In the stationary state, the correlation strength can be tuned by varying $m$, and it displays a transition at a critical value $N_c=6$. Our predictions extend straightforwardly to any spatial dimension $d$ and the critical value $N_c=6$ remains the same in all dimensions. Our predictions are testable in existing experimental setups on optically trapped colloidal particles.

In trapped-ion quantum simulators, effective spin-1/2 XY interactions can be engineered via laser-induced coupling between internal atomic states and collective phonon modes. In the simplest one-dimensional ($1d$) traps, these interactions decay as a power-law with distance $1/r^α$, with a tunable exponent $α$. For small $α$, the resulting long-range $1d$ XY model exhibits continuous symmetry breaking, in marked contrast to its nearest neighbor counterpart. In this paper, we examine this model near the phase transition at $α_c$ from the lens of the spin stiffness, or superfluid density. We develop a stochastic series expansion (SSE) quantum Monte Carlo (QMC) simulation and a generalized winding number estimator to measure the superfluid density in the presence of power-law interactions, which we test against exact diagonalization for small lattice sizes. Our results show how conventional superfluidity in the $1d$ XY model is enhanced in the long-range interacting regime. This is observed as a diverging superfluid density as $α\rightarrow 0$ in the thermodynamic limit, which we show is consistent with linear spin-wave theory. Finally, we define a normalized superfluid density estimator that clearly distinguishes the short, medium, and long-range interacting regimes, providing a novel QMC probe of the critical value $α_c$.

We present a deep learning framework for modeling and analyzing the small-angle scattering data of polydisperse hard-rod systems, a widely used models for anisotropic colloidal particles. We use a variational autoencoder-based neural network to learn the mapping from the system parameters such as the volume fraction, rod length, and polydispersity, to the scattering function. The dataset for training and testing such neural network model is obtained from Markov chain Monte Carlo simulation of 20,000 hard spherocylinders using the hard particle Monte Carlo package from the HOOMD-blue. Four datasets were generated, each with 5,500 pairs of system parameters and corresponding scattering functions. We use one of the dataset to investigate the feasibility of the learning, and three additional datasets with different polydisperse distribution to demonstrate the generality of our approach. The neural network model transcends the fundamental limitations of the Percus-Yevick approximation by accurately capturing anisotropic interactions and high-concentration effects that analytical models often fail to resolve. This framework achieves significantly higher accuracy in reproducing scattering functions and enables a least-square fitting routine for quantitative data analysis.

We develop a Floquet transfer matrix method to solve scattering in extended 1D Floquet lattices, uncovering an underlying conjugate symplectic structure that enforces current conservation across sidebands. By engineering a spatial adiabatic boundary, we suppress multi-channel sideband interference, allowing us to establish a direct mapping between the bulk winding number $C$ and a rigid shift in the transmission energy windows--quantified as $C\hbarω$. We further propose two experimental realizations: cold-atom Bragg scattering to directly verify the transmission shift, and surface-acoustic-wave-induced transport demonstrating the quantized zero-bias current plateau.

Two-photon light-matter interactions exhibit distinctive features such as spectral collapse. The two-photon Dicke model has been reported to exhibit a superradiant phase which could be useful in quantum applications. Here we show that this superradiant phase is not a genuine thermodynamic phase but a finite-size effect. Combining analytical and numerical analyses, we demonstrate that the superradiant region shrinks with increasing system size and disappears in the thermodynamic limit, while spectral collapse remains. Our results clarify the nature of superradiant conditions in two-photon systems and constrain its realization in quantum platforms.

We investigate timelike entanglement measures derived from the spacetime density kernel in the Rosenzweig-Porter model and show that they sharply diagnose both eigenvector ergodicity and spectral chaos. For several Hilbert-space bipartitions, we compute the second Tsallis entropy, the entanglement imagitivity that quantifies non-Hermiticity, and Schatten-norm diagnostics of the kernel. The imagitivity and Frobenius norm exhibit rapid growth and high late-time plateaus in the ergodic regime, are suppressed in the localized regime, and show intermediate behavior in the fractal phase. The real part of the second Tsallis entropy displays a spectral form factor-like dip-ramp-plateau throughout the chaotic window and a suppressed ramp in the localized regime. We further introduce a kernel negativity, defined as the negative spectral weight of the Hermitian part of the kernel. This negativity equals the trace-norm distance to the set of positive semidefinite operators and the maximal witnessable negative quasiprobability, and its time-averaged value decreases across the ergodic-fractal-localized crossover in close correspondence with the fractal dimension.

Half-quantum vortices in spin-triplet superconductors are predicted to harbor Majorana zero modes and may provide a viable avenue to topological quantum computation. Here, we introduce a novel approach for directly measuring the half-integer-quantized magnetic fluxes, $Φ= h / (4e)$, carried by such half-quantum vortices via spin-qubit relaxometry. We consider a superconducting strip with a narrow pinch point at which vortices cross quasi-periodically below a spin qubit as a result of a bias current. We demonstrate that the relaxation rate of the spin qubit exhibits a pronounced peak if the vortex-crossing frequency matches the transition frequency of the spin qubit and conclude that the magnetic flux $Φ$ of a single vortex can be obtained by dividing the corresponding voltage along the strip with the transition frequency. We discuss experimental constraints on implementing our proposed setup in spin-triplet candidate materials like UTe$_2$, UPt$_3$, and URhGe.

The Wilson-Fisher fixed point defines a continuous family of interacting conformal field theories in non-integer dimensions. In integer dimensions, it is widely believed to lie in the same universality class as the critical Ising model. In this work, we revisit the identification between the Wilson-Fisher fixed point at integer dimensions and the Ising CFT. We argue that a literal equality between the two theories is incompatible with the emergence of Virasoro symmetry in two dimensions. Instead, we propose that the Ising model emerges only as a subsector of the Wilson-Fisher fixed point. We support this scenario through a detailed study of the two-dimensional $O(n)$ model and by examining operators transforming in irreducible representations of the orthogonal group whose multiplicities become negative for integer values of the spacetime dimension. Finally, we comment on the implications of these results for attempts to construct a $d=2+ε$ expansion starting from exact two-dimensional data.

Fluorescent spin qubits are central building blocks of quantum technologies. Placing these qubits at surfaces maximizes coupling to nearby spins and fields, enabling nanoscale sensing and facilitating integration with photonic and superconducting devices. However, reducing the dimensions or size of established qubit systems without sacrificing the qubit performance or degrading the coherence lifetime remains challenging. Here, we introduce a surface molecular qubit formed by pentacene molecules scaffolded on a two-dimensional (2D) material, hexagonal boron nitride (hBN). The qubit exhibits stable fluorescence and optically detected magnetic resonance (ODMR) from cryogenic to ambient conditions. With fully deuterated pentacene, the Hahn-echo coherence reaches 22 $μ$s and further extends to 214 $μ$s under dynamical decoupling, outperforming state-of-the-art shallow NV centers in diamond, despite being positioned directly on the surface. We map the local spin environment, resolving couplings to nearby nuclear and electron spins that can serve as auxiliary quantum resources. This platform combines true surface integration, long qubit coherence, and scalable fabrication, opening routes to quantum sensing, quantum simulation, and hybrid quantum devices. It also paves the way for a broader family of 2D material-supported molecular qubits.

Anomalous KPZ spin transport is well established in integrable non-Abelian lattice models but has not been investigated in continuum field theories as discretization in numerics generally break the continuum theory's integrability. We show that finite temperature acts as a regulator that can restore anomalous transport over a broad time window. In a family of spin field theories labeled by integer $n$, the $n = 1$ case is the Landau-Lifshitz model, whose numerical data shows spin superdiffusion with Kardar-Parisi-Zhang (KPZ) scaling and, at lower temperature ballistic energy transport, whereas both observables are diffusive at high temperature. The non-integrable $n = 2$ case shows the same crossover. While Lyapunov analysis confirms the model's non-integrability, the structure of spin-density space-time profiles suggests that long-lived soliton-like trajectories exist at low temperature.

We report an approach to obtain effective pair potentials which describe the structure of two-dimensional systems of active Brownian particles. The pair potential is found by an inverse method, which matches the radial distribution function found from two different schemes. The inverse method, previously demonstrated via simulated equilibrium configurations of passive particles, has now been applied to a suspension of active particles. Interestingly, although active particles are inherently not in equilibrium, we still obtain effective interaction potentials which accurately describe the structure of the active system. Treating these effective potentials as if they were those of equilibrium systems, furthermore allows us to measure effective chemical potentials and pressures. Both the passive interactions and active motion of the active Brownian particles contribute to their effective interaction potentials.

The band structure of Bernal-stacked bilayer graphene can be tuned using double-gated transistors to apply a perpendicular electric field that generates an interlayer potential energy difference $Δ$. Dielectric breakdown limits the operation of conventional devices to the $Δ\ll t_\perp \simeq 360$ meV regime. We employ double ionic gating to reach fields past $ 1$ V/nm, for which $Δ> t_\perp$. We find that for $Δ\simeq t_\perp$, the evolution of the longitudinal resistance ($R_{xx}$) peak as a function of applied gate voltages undergoes a sharp change in slope, exhibiting a pronounced "knee". Increasing $Δ$ past the "knee" results in an unusual evolution transport properties: the peak in $R_{xx}$ decreases in magnitude, it exhibits a splitting concomitant with multiple sign reversals of the Hall resistance, and hysteresis in the peak position emerges. We explain the observed phenomenology in terms of in-gap bound states, whose energy strongly depends on the perpendicular electric field, and crosses the mid-gap level for sufficiently large $Δ> t_\perp$. The phenomenon causes large changes in the electronic density of in-gap states that profoundly affect the evolution of the chemical potential. Our experimental results and their interpretation reveal unique aspects of the physics of in-gap states in Bernal bilayer graphene and demonstrate that double ionic gating enables investigating the large-$Δ$ regime, which has remained experimentally inaccessible so far.

We discuss a possibility of superconductivity in the p-wave magnets. These are recently discovered materials that have zero net magnetization by symmetry and finite non-relativistic spin splitting of electron bands, like in altermagnets. Similarly, the spin polarizations is collinear in the momentum space. Yet, as opposed to altermagnets, the magnetization is noncollinear in the real space, and the spin splitting obeys time-reversal symmetry in the momentum space. As a result, if such material harbors superconductivity (due to phonons, or any other mechanism), the only supported superconducting symmetry is Ising superconductivity, an exotic symmetry where any Cooper pair is a 50:50 mix of singlet and triplet. This unusual behavior is also in stark contrast to regular antiferromagnet, which can support Cooper pairs of any parity, and altermagnets, which can only support nonunitary triplet pairs. The presence of large triplet component and enhanced resilience against pair breaking is inherent to the p-wave magnets and as such is unconventional as it does not materialize in conventional spin-orbit coupling induced Ising superconductors.

Racism remains a persistent societal issue, increasingly amplified by the structure and dynamics of online social networks. In this work, we propose a three-state compartmental model to study the spreading and suppression of racist content, drawing from epidemic-like dynamics and interaction-driven transitions. We analyze the model on fully-connected (homogeneous mixing) networks using a set of coupled differential equations, and on Barabási-Albert (BA) scale-free and Watts-Strogatz (WS) small-world networks through agent-based simulations. The system exhibits three distinct stationary regimes: two racism-free absorbing states and one active phase with persistent racist content. We identify and characterize the phase transitions between these regimes, discuss the role of network topology, and highlight the emergence of absorbing states. Our findings illustrate how statistical physics tools can help uncover the macroscopic consequences of microscopic social interactions in digital environments.

Transition metal di-iodides such as FeI2, NiI2 and CoI2 are an emerging class of 2D magnets exhibiting rich and diverse magnetic behaviour, but their study at the monolayer limit has been severely hindered by fabrication challenges due to their air-sensitivity. Here, we introduce a polymer-free method for clean, rapid, and high-yield assembly of hermetically encapsulated suspended samples of air-sensitive monolayers. Applying it to di-iodides enables atomic resolution characterisation of thin samples - down to the monolayer limit - for the first time. Our imaging, combined with complementary first-principles calculations, reveals an unusually small energy barrier between alternate stable stacking polytypes in few-layer films, enabling extrinsic control of the stacking phase. We also observe stable isolated iodine vacancies that do not aggregate to form extended structures, and identify and verify the stability of the various edge configurations of thin samples. These results establish the unique structural characteristics of these materials in the thin limit, and more broadly demonstrate the utility of our transfer platform for creating atomically clean suspended vdW heterostructures.

Non-Hermitian systems display spectral and topological phenomena absent in Hermitian physics; yet, their geometric characterization can be hindered by an intrinsic ambiguity rooted in the eigenspace of non-Hermitian Hamiltonians, which becomes especially pronounced in the pure quantum regime. Because left and right eigenvectors are not related by conjugation, their norms are not fixed, giving rise to a biorthogonal ${\rm GL}(N,{\mathbb C})$ gauge freedom. Consequently, the standard Berry connection admits four inequivalent definitions depending on how left and right eigenvectors are paired, giving rise to distinct Berry phases and generally complex-valued holonomies. Here we show that these ambiguities and the emergence of complex phases are fully resolved by introducing a covariant-derivative formalism built from the metric tensor of the Hilbert space of the underlying non-Hermitian Hamiltonian. The resulting unique Berry connection remains real-valued under an arbitrary ${\rm GL}(N,{\mathbb C})$ frame change, and transforms as an affine gauge potential, while reducing to the conventional Berry (or Wilczek-Zee) connection in the Hermitian limit. This establishes an unambiguous and gauge-consistent geometric framework for Berry phases, non-Abelian holonomies, and topological invariants in quantum systems described by non-Hermitian Hamiltonians.

Localization is a pervasive phenomenon across physics, shaping transport from electrons in solids to light and sound in engineered media. In traditional settings, disorder strongly impedes transport, resulting in dynamical localization or, at best, sub-ballistic or diffusive dynamics. A distinct and previously unobserved regime, erratic non-Hermitian skin localization (ENHSL), can arise in globally reciprocal non-Hermitian lattices with disorder. It features macroscopic, disorder-dependent localization at irregular bulk positions with subexponential decay, linked to stochastic interfaces governed by the universal order statistics of random walks. We realize this regime experimentally in an acoustic lattice implementing a disordered Hatano-Nelson chain with imaginary gauge fields. Using Green's-function-based spectroscopy together with time-resolved measurements on the same platform, we reconstruct the full complex spectrum and eigenstates, and directly observe wave-packet dynamics. Remarkably, we observe ballistic transport despite strong spectral localization. We develop a transport theory that connects the dominant propagation site to the maximal random-walk excursion within an expanding light cone and predicts a universal Levy-arcsine statistics, in quantitative agreement with experiment. Our results decouple eigenstate localization from transport and establish ENHSL as a new paradigm for wave dynamics.

Advances in quantum technologies are accelerating the demand for optical quantum state sensors that combine high precision, versatility, and scalability within a unified hardware platform. Quantum reservoir computing offers a powerful route toward this goal by exploiting the nonlinear dynamics of quantum systems to process and interpret quantum information efficiently. Photonic neural networks are particularly well suited for such implementations, owing to their intrinsic sensitivity to photon-encoded quantum information. However, the practical realisation of photonic quantum reservoirs remains constrained by the inherently weak optical nonlinearities of available materials and the technological challenges of fabricating densely coupled quantum networks. To address these limitations, we introduce a hybrid quantum-classical detection protocol that integrates the advantages of quantum reservoirs with the adaptive learning capabilities of analogue neural networks. This synergistic architecture substantially enhances information-extraction accuracy and robustness, enabling low-cost performance improvements of quantum light sensors. Based on the proposed approach, we achieved significant improvements in quantum state classification, tomography, and feature regression, even for reservoirs with a relatively small nonlinearity-to-losses ratio $U/γ\approx 0.02$ in a network of only five nodes. By reducing reliance on material nonlinearity and reservoir size, the proposed approach facilitates the practical deployment of high-fidelity photonic quantum sensors on existing integrated platforms, paving the way toward chip-scale quantum processors and photonic sensing technologies.

In recent years, nonlinear transport phenomena have garnered significant interest in both theoretical explorations and experiments. In this work, we utilize the semi-classical wave packet theory to calculate disorder-induced second-order transport coefficients: second-order electrical ($σ$), thermoelectric ($α$), and thermal ($κ$) coefficients, capturing the interplay between side-jump and skew-scattering contributions in systems with time-reversal symmetry. Using a topological insulator model, we quantitatively characterize the Fermi-level dependence of these second-order transport coefficients by explicitly including Coulomb impurity potentials. Furthermore, we elucidate the relationships between these coefficients, establishing the second-order Mott relation and the Wiedemann-Franz law induced by disorder. This study develops a comprehensive theoretical framework elucidating the nonlinear thermoelectric transport mechanisms in quantum material systems.

We study the formation of hybrid plasmon-magnon modes in a heterostructure comprising a monolayer semiconductor with strong Rashba spin-orbit coupling -- specifically, Janus transition-metal dichalcogenides (TMDs) -- and an insulating ferrimagnet, such as yttrium iron garnet-based compounds. Using a combined microscopic-macroscopic framework for plasmon-magnon coupling, we show that plasmons and magnons strongly hybridize over both GHz and THz frequency ranges, enabling experimental access well above cryogenic temperatures. Moreover, the developed approach provides an efficient and natural classification of the topology of the hybrid modes, rooted in the phase winding of the plasmon-magnon coupling induced by spin-momentum locking and the associated chiral winding of the electronic spin along the Fermi contours. Finally, we identify experimentally accessible manifestations of the hybridization, such as topological interface modes and an anomalous thermal Hall response.

Moiré patterns in magnetic bilayers generate spatially modulated interlayer exchange interactions that can give rise to nonuniform magnetic textures. We study a minimal classical bilayer Ising model with a moiré-modulated interlayer coupling, generated either by relative twist or differential strain between the layers. Using large-scale classical Monte Carlo simulations, we show that the ordering transition remains in the conventional two-dimensional Ising universality class, even when the low-temperature state is domain-textured. At low temperatures, we find a smooth crossover between a uniform ferromagnet and domain-textured state, in which the spins locally follow the sign of the interlayer exchange. We demonstrate that there is no breaking of layer symmetry for twisted bilayers. The location of the crossover is determined by a simple geometric energy balance between bulk interlayer exchange and intralayer domain-wall costs. Our results provide a minimal framework for understanding how moiré-modulated magnetic textures can emerge from geometric energetics without requiring a thermodynamic phase transition.

We study Langevin dynamics of $N$ Brownian particles on two-dimensional Riemannian manifolds, interacting through pairwise potentials linear in geodesic distance with quenched random couplings. These \emph{frustrated Brownian particles} experience competing demands of random attractive and repulsive interactions while confined to curved surfaces. We consider three geometries: the sphere $S^2$, torus $T^2$, and bounded cylinder. Our central finding is disorder-induced dimension reduction with spontaneous rotational symmetry breaking: order emerges from two sources of randomness (thermal noise and quenched disorder), with manifold topology determining the character of emerging structures. Glassy relaxation drives particles from 2D distributions to quasi-1D structures: bands on $S^2$, rings on $T^2$, and localized clusters on the cylinder. Unlike conventional symmetry breaking, the symmetry-breaking direction is not frozen but evolves slowly via thermal noise. On the sphere, the structure normal precesses diffusively on the Goldstone manifold with correlation time $τ_c \approx 18$, a classical realization of type-A dissipative Nambu-Goldstone dynamics. The model requires no thermodynamic gradients, no fine-tuning, and no slow external input. We discuss connections to spin glass theory, quantum field theory, astrophysical structure formation, and self-organizing systems. The model admits a large-$N$ limit yielding statistical field theory on Riemannian surfaces, while remaining experimentally realizable in colloidal and soft matter systems.

Reaction-diffusion processes play an important role in a variety of physical, chemical, and biological systems. Conventionally, the kinetics of these processes are described by the law of mass action. However, there are various cases where these equations are insufficient. A fundamental challenge lies in accurately accounting for the microscopic correlations that inevitably arise in bimolecular reactions. While approaches to describe microscopic correlations in many specific cases exist, no general theory for multistage reactions has been established. In this article, we apply the quantum field theory approach to derive kinetic equations for general multistage reactive systems termed CMET (complete modified encounter theory). CMET can be formulated as a set of coupled partial differential equations that can be easily integrated numerically, thereby serving as a versatile tool for investigating reaction-diffusion processes. Across multiple case studies, we demonstrated that CMET reproduces the kinetics predicted by many other theories within their respective scopes of applicability.

Beta Rank Function (BRF) is a two-sided distribution characterized by a smooth peak and double powerlaw decay, widely used to model empirical data exhibiting deviations from pure power laws. In this paper, we introduce a novel two-step generative process that produces data exactly following the BRF distribution. The first step involves any mechanism generating a power-law distribution, while the second step applies a regressive redistribution process that reallocates resources from poorer to richer entities, thereby amplifying inequality. This approach represents the first analytic derivation of an exact BRF distribution from a generative mechanism. We validate the model through applications to income and urban population distributions. Beyond exact generation, this framework offers new insights into the systemic origins of deviations from power laws frequently observed in complex systems, linking rank distributions to underlying feedback and redistribution dynamics.

In this work, we present a comprehensive numerical framework that couples numerical solutions of Maxwell's equations using the Finite-Difference Time-Domain (FDTD) approach, Molecular Dynamics (MD), and the Two-Temperature Model (TTM) to describe ultrafast laser-matter interactions in metallic systems at the atomic scale. The proposed Maxwell-Two-Temperature Model-Molecular Dynamics (M-TTM-MD) bridges the gap between electromagnetic field propagation, electron-phonon energy exchange, and atomic motion, allowing for a self-consistent treatment of energy absorption, transport, and structural response within a unified simulation environment. The calculated electromagnetic fields incorporate dispersive dielectric properties derived using the Auxiliary Differential Equation (ADE) technique, while the electronic and lattice subsystems are dynamically coupled through spatially and temporally resolved energy exchange terms. The changes in the material topography are then reflected in the updated grid for the FDTD scheme. The developed M-TTM-MD model provides a self-consistent numerical framework that offers insights into laser-induced phenomena in metals, including energy transport and surface dynamics under extreme nonequilibrium conditions.

The Chesapeake Bay, one of the largest estuaries in the United States, is an ecological system of great complexity and relevance. The food web is composed of thirty-six trophic components, all of which are functionally connected. In this work, the interactions among these components are numerically analyzed using complex network methods. An energy flow cutoff paradigm is applied to a weighted ecological network. The results reveal patterns characteristic of connectivity dynamics, evidencing both the initial robustness of the system and its tendency to fragmentation at higher values of the cutoff. From an applied perspective, the findings underscore the importance of conservation strategies that protect keystone species, such as carnivorous fish, which act as crucial connectors between the two main subnetworks. Although they are positioned at the top of the food web and are often assumed to be less critical to network stability, these species play a pivotal role in regulating populations of lower-level organisms, thereby maintaining the overall integrity of the ecosystem.

Exchange between tissue compartments is crucial for interpretation of diffusion MRI measurements in brain gray matter. Reported values of exchange time are broadly dispersed, about two orders of magnitude. We analyze the measurement technique called Filtered Exchange Imaging (FEXI) using numerical solution of Bloch--Torrey equation in digitalized neurons downloaded from NeuroMorpho.org. The FEXI outcome, the recovery of diffusion coefficient in cells with impermeable membrane is multiexponential, tightly related to the eigenvalues of the Laplace operator. Fitting the commonly used exponential recovery function results in a strong dependence of the apparent exchange time on the involved mixing time interval. For short mixing times, exchange time is of the order of 10 ms. It gets an order of magnitude larger for mixing times of a few hundreds of milliseconds. To obtain an estimate of membrane permeability, we reinterpret previously published data on preexchange lifetime in neuronal cell culture. It results in the permeability 0.005 micrometer/ms. The corresponding exchange time is about 140 ms. We conclude that essentially shorter exchange times are due to fast geometric exchange inside the ramified cells.

Label-free optical microscopy through absorption or scattering spectroscopy provides fundamental insights across biology and materials science, yet its sensitivity remains fundamentally limited by photon shot noise. While recent demonstrations of quantum nonlinear microscopy show sub-shot-limited sensitivity, they are intrinsically limited by availability of high peak-power squeezed light sources. Here, we introduce squeezing-enhanced photothermal (SEPT) microscopy, a quantum imaging technique that leverages twin-beam quantum correlations to detect absorption induced signals with unprecedented sensitivity. SEPT achieves 3.5 dB noise suppression beyond the standard quantum limit, enabling a 2.5-fold increase in imaging throughput or 31% reduction in pump power, while providing an unmatched versatility through the intrinsic compatibility between continuous-wave squeezing and photothermal modulation. We showcase SEPT applications by providing high-precision characterization of nanoparticles and revealing subcellular structures, such as cytochrome c, that remain undetectable under shot-noise-limited imaging. By combining label-free contrast, quantum-enhanced sensitivity, and compatibility with existing microscopy platforms, SEPT establishes a new paradigm for molecular absorption imaging with far-reaching implications in cellular biology, nanoscience, and materials characterization.

This study examines the statistical nature of instantaneous scalar concentration in an elevated point-source plume (neutral or buoyant) dispersing within a turbulent boundary layer. Using high-frequency long-duration experimental measurements, we extensively validate the gamma distribution as the appropriate probability density function of concentration, particularly at large scalar magnitudes. The two-parameter gamma distribution is shown to capture the PDF at all locations across the plume. The classical similarity of the mean and root-mean-square (RMS) concentration, often expressed through a Gaussian form, is recovered through similarity of the scale and shape parameters of the gamma distribution. In addition, statistics of extreme events, such as the 99th percentile of the instantaneous concentration signal, are also well predicted, and their observed invariance near the plume centreline is reasoned. Further, similarity is observed for the third- and higher-order central moments and standardised central moments from the experimental data. The framework of the gamma distribution is also analytically extended to higher-order statistics. The experimental data are in good agreement with the predicted central moments up to the eighth order. The results emphasise the importance of achieving statistical convergence for the intermittent concentration signal, directly influenced by finite sampling times in a measurement. A secondary result is obtained for the ratio of plume half-widths based on the mean and the RMS concentration to be $1/\sqrt{2}$, consistent with experimental observations. The results establish the gamma distribution as a consistent and unified model for all scalar concentration statistics in elevated point source plumes within a turbulent boundary layer.

Generative artificial intelligence (GAI) plays a fundamental role in high-impact AI-based systems such as SORA and AlphaFold. Currently, GAI shows limited capability in the specialized domains due to data scarcity. In this paper, we develop a continuum mechanics-based theoretical framework to generalize the optimal transport theory from pure mathematics, which can be used to describe the dynamics of data, realizing the generative tasks with a small amount of data. The developed theory is used to solve three typical problem involved in many mechanical designs and engineering applications: at material level, how to generate the stress-strain response outside the range of experimental conditions based on experimentally measured stress-strain data; at structure level, how to generate the temperature-dependent stress fields under the thermal loading; at system level, how to generate the plastic strain fields under transient dynamic loading. Our results show the proposed theory can complete the generation successfully, showing its potential to solve many difficult problems involved in engineering applications, not limited to mechanics problems, such as image generation. The present work shows that mechanics can provide new tools for computer science. The limitation of the proposed theory is also discussed.

Prediction markets mobilize financial incentives to forecast binary event outcomes through the aggregation of dispersed beliefs and heterogeneous information. Their growing popularity and demonstrated predictive accuracy in political elections have raised speculation and concern regarding their susceptibility to manipulation and the potential consequences for democratic processes. Using agent-based simulations combined with an analytic characterization of price dynamics, we study how high-budget agents can introduce price distortions in prediction markets. We explore the persistence and stability of these distortions in the presence of herding or stubborn agents, and analyze how agent expertise affects market-price variance. Firstly we propose an agent-based model of a prediction market in which bettors with heterogeneous expertise, noisy private information, variable learning rates and budgets observe the evolution of public opinion on a binary election outcome to inform their betting strategies in the market. The model exhibits stability across a broad parameter space, with complex agent behaviors and price interactions producing self-regulatory price discovery. Second, using this simulation framework, we investigate the conditions under which a highly resourced minority, or ''whale'' agent, with a biased valuation can distort the market price, and for how long. We find that biased whales can temporarily shift prices, with the magnitude and duration of distortion increasing when non-whale bettors exhibit herding behavior and slow learning. Our theoretical analysis corroborates these results, showing that whales can shift prices proportionally to their share of market capital, with distortion duration depending on non-whale learning rates and herding intensity.

We study the disequilibrium dynamics of a stylised model of production networks in which firms use perishable and non-substitutable intermediate inputs, so that adverse idiosyncratic productivity shocks can trigger downstream shortages and output losses. To protect against such disruptions, firms hold precautionary inventories that act as buffer stocks. We show that, for a given dispersion of firm-level productivity shocks, there exists a critical level of inventories above which the economy remains in a stable stochastic steady state. Below this critical level, the system becomes fragile, i.e., it becomes prone to system-wide crises. As this resilience-fragility boundary is approached from above, aggregate output volatility rises sharply and diverges, even though shocks are purely idiosyncratic. Because inventories are costly, competitive pressures induce firms to economize on buffers. Although we do not explicitly model such costs, we argue that the resulting behaviour of individual firms drives the system close to criticality, generating persistent excess macroeconomic volatility -- in other words, ``small shocks, large cycles'' -- in line with other settings where efficiency and resilience are in tension with each other. In the language of phase transitions, the resilient-to-fragile transition is continuous (supercritical): the economy exhibits a well-defined stochastic equilibrium with finite volatility on one side of the boundary, while beyond it the probability of a collapse in finite time tends to one. We characterize this transition primarily through numerical simulations and derive an analytical description in a high-perishability, high-connectivity limit.

Planktonic organisms such as copepods sense swimming prey and sinking food particles through the hydrodynamic disturbances they generate. However, because these flow fields are often highly symmetric, they provide little directional information, making accurate localization of the source challenging. Here, we introduce the steer'n roll sensing and response strategy. This strategy combines stereoscopic flow sensing and a roll motion. Stereoscopic sensing allows plankton to disambiguate flow signals by integrating two spatially separated flow measurements, while a roll about the swimming axis enhances exploration of the three-dimensional space. We show that steer'n roll is efficient, achieving a 100% success rate, versatile across signal type, and robust to flow sensing noise, orientational diffusion, and turbulence. Together, these findings identify a biologically plausible mechanism for prey detection and capture via flow sensing, and offer testable insights into the sensory-motor strategies of planktonic organisms.

We extend the data-assimilation approach of Ling and Lozano-Durán (AIAA 2025-1280) to develop machine-learning-based subgrid-scale stress (SGS) models for large-eddy simulation (LES) that are consistent with the numerical scheme of the flow solver. The method accounts for configurations with two inhomogeneous directions and is applied to turbulent boundary layers (TBL) under adverse pressure gradients (APG). To overcome the limitations of linear eddy-viscosity closures in complex flows, we adopt a non-Boussinesq SGS formulation along with a dissipation-matching training loss. A second improvement is the integration of a multi-task learning strategy that explicitly promotes monotonic convergence with grid refinement, a property that is often absent in conventional SGS models. A posteriori tests show that the proposed model improves predictions of the mean velocity and wall-shear stress relative to the Dynamic Smagorinsky model (DSM), while also achieving monotonic convergence with grid refinement.

We developed methods for protein synthesis and performed two-dimensional electronic spectroscopy (2D-ES) to examine the influence of pigment arrangement on the photoexcitation dynamics of light-harvesting proteins in phycobilisome. We synthesized allophycocyanin (APC), C-phycocyanin (CPC), and mutant CPC lacking the \b{eta}153 phycocyanobilin (PCB) pigment by an Escherichia coli expression system. The number of pigments in the mutant CPC is identical to that in the wild-type APC, and their spatial arrangements are similar. The absorption and fluorescence spectra of the mutant CPC closely resemble those of the wild-type CPC rather than the wild-type APC, indicating that pigment spatial arrangement is not a primary factor in determining the excited-state energy structure. The 2D-ES measurements show that the wild-type CPC retains broad positive signals at 1 ps, signifying incomplete relaxation and persistence of excited vibronic states, unlike APC, which vibrationally relaxes to the bottom of the potential energy surface within the same timeframe. The mutant CPC behaves similarly to the wild-type CPC in the 2D-ES, reinforcing that the pigment number or arrangement is not a dominant factor. Instead, the local pigment-protein interaction governs the electronic structure and relaxation dynamics. Structural analysis reveals that the bent structure of PCB in CPC's α-chain versus the planar structure of PCB in APC. The bent PCB in CPC reduces the degree of π-conjugation, and exhibits excited-state properties distinct from those of the planar structure of PCB in APC. This finding highlights a critical role of the electronic structure governed by the local interaction in ultrafast energy relaxation.

The growing volume of scientific simulation data presents a significant challenge for storage and transfer. Error-bounded lossy compression has emerged as a critical solution for mitigating these challenges, providing a means to reduce data size while ensuring that reconstructed data remains valid for scientific analysis. In this paper, we present a data-driven scientific data compressor, called Discontinuous Data-informed Local Subspaces (Discontinuous DLS), to improve compression-to-error ratios over data-agnostic compressors. This error-bounded compressor leverages localized spatial and temporal subspaces, informed by the underlying data structure, to enhance compression efficiency and preserve key features. The presented technique is flexible and applicable to a wide range of scientific data, including fluid dynamics, environmental simulations, and other high-dimensional, time-dependent datasets. We describe the core principles of the method and demonstrate its ability to significantly reduce storage requirements without compromising critical data fidelity. The technique is implemented in a distributed computing environment using MPI, and its performance is evaluated against state-of-the-art error-bounded compression methods in terms of compression ratio and reconstruction accuracy. This study highlights discontinuous DLS as a promising approach for large-scale scientific data compression in high-performance computing environments, providing a robust solution for managing the growing data demands of modern scientific simulations.

Heat-pipe microreactors (HPMRs) are compact and transportable nuclear power systems exhibiting inherent safety, well-suited for deployment in remote regions where access is limited and reliance on costly fossil fuels is prevalent. In prior work, we developed a design optimization framework that incorporates techno-economic considerations through surrogate modeling and reinforcement learning (RL)-based optimization, focusing solely on minimizing the levelized cost of electricity (LCOE) by using a bottom-up cost estimation approach. In this study, we extend that framework to a multi-objective optimization that uses the Pareto Envelope Augmented with Reinforcement Learning (PEARL) algorithm. The objectives include minimizing both the rod-integrated peaking factor ($F_{Δh}$) and LCOE -- subject to safety and operational constraints. We evaluate three cost scenarios: (1) a high-cost axial and drum reflectors, (2) a low-cost axial reflector, and (3) low-cost axial and drum reflectors. Our findings indicate that reducing the solid moderator radius, pin pitch, and drum coating angle -- all while increasing the fuel height -- effectively lowers $F_{Δh}$. Across all three scenarios, four key strategies consistently emerged for optimizing LCOE: (1) minimizing the axial reflector contribution when costly, (2) reducing control drum reliance, (3) substituting expensive tri-structural isotropic (TRISO) fuel with axial reflector material priced at the level of graphite, and (4) maximizing fuel burnup. While PEARL demonstrates promise in navigating trade-offs across diverse design scenarios, discrepancies between surrogate model predictions and full-order simulations remain. Further improvements are anticipated through constraint relaxation and surrogate development, constituting an ongoing area of investigation.

We present a new adaptive particle-based data assimilation scheme for cryospheric applications that leverages promising developments in importance sampling. The proposed approach seeks to combine some of the advantages of two widely used classes of schemes: particle methods and iterative ensemble Kalman methods. Specifically, it extends the PBS that is commonly used in cryospheric data assimilation, with the AMIS algorithm. This adaptive formulation transforms the PBS into an iterative scheme with improved resilience against ensemble collapse and the ability to implement early-stopping strategies. As such, computational cost is automatically adapted to the complexity of the problem at hand, even down to the grid-cell and water year level in distributed multiyear simulations. In homage to the schemes that it builds on, we coin this new algorithm the Adaptive Particle Batch Smoother (AdaPBS) and we test it across a range of scenarios. First, we conducted an intercomparison of some of the most commonly used cryospheric data assimilation algorithms using MCMC simulation as a costly gold-standard benchmark in a simplified temperature index model assimilating snow depth observations. We further evaluated AdaPBS by assimilating snow depth observations from the ESMSnowMIP project at 6 different sites spanning 3 continents, using an ensemble of simulations generated with the more complex FSM2. Our results demonstrate that AdaPBS is a robust and reliable tool, outperforming or at least matching the performance of other commonly used algorithms and successfully handling complex cases with dense observational datasets. All experiments were carried out using the open-source MuSA toolbox, which now includes AdaPBS and MCMC among the growing list of available cryospheric data assimilation methods.

We study fast and reliable generative transport for the 3D KS (Keller-Segel) and KPP (Kolmogorov-Petrovsky-Piskunov) equations in the presence of fluid flows with the goal to approximate the map between initial and terminal distributions for a range of physical parameters $σ$ under the Wasserstein metric. To minimize the inaccuracy of direct Wasserstein solver, we propose a two-stage pipeline that retains one-step efficiency while reinstating an explicit $W_2$ objective where it is tractable. In Stage I, a Meanflow-style regressor yields a deterministic, one-step global transport that moves particles close to their terminal states. In Stage II, we freeze this initializer and train a near-identity corrector (Deep Particle, DP) that directly minimizes a mini-batch $W_2$ objective using warm-started optimal transport couplings computed on the Meanflow outputs. Crucially, after the one-step transport (from Stage I) concentrating mass on the approximated correct support, the induced geometry stabilizes high-dimensional $W_2$ computation of the direct Wasserstein solver. We validate our construction in the 3D KS and KPP equations subject to fluid flows with ordered and chaotic streamlines.

Purpose: Hemolysis is a key issue in the design of blood-handling medical devices. Computational prediction of this phenomenon is challenging due to the complex multiscale nature of blood. As a result, conventional approaches often fail to predict hemolysis accurately, commonly showing deviations of multiple orders of magnitude compared to experimental data. More accurate models are typically computationally expensive and thus impractical for real-world applications. This work aims to fill this gap by presenting accurate yet simple and efficient computational hemolysis models. Methods: Hemolysis modeling relies on two key components: a red blood cell model and a hemoglobin release model. In this work, we compare three red blood cell models: a common stress-based model (Bludszuweit), a simple strain-based model based on the Kelvin-Voigt constitutive law, and a more complex tensor-based model (TTM). Further, we compare two hemoglobin release models: the widely used power-law approach and a biophysical pore formation model. Results: We evaluate these models in two benchmark cases: the FDA blood pump and the FDA nozzle. In both benchmarks, the simple strain-based model combined with the pore formation model achieves absolute predictions of hemolysis within the standard deviation of experimental measurements. In contrast, stress-based power law models deviate by several orders of magnitude. Conclusion: The strain-based pore modeling approach takes into account the biophysical properties of red blood cell membranes, in particular their viscoelastic deformation behavior and hemoglobin release through membrane pores. This leads to significantly improved hemolysis predictions in a framework that can easily be integrated into common CFD workflows.

Intricate interactions among firms, institutions, and spatial structures shape urban economic systems. In this study, we propose a framework based on three structural dimensions -- abundance, diversity, and longevity (ADL) of economic units -- as proxies of urban economic complexity and resilience. Using a decade of georeferenced firm-level data from Mexico City, we analyze the relationships among ADL variables using regression, spatial correlation, and time-series clustering. Our results reveal nonlinear dynamics across urban space, with powerlaw behavior in central zones and logarithmic saturation in peripheral areas, suggesting differentiated growth regimes. Notably, firm longevity modulates the relationship between abundance and diversity, particularly in periurban transition zones. These spatial patterns point to an emerging polycentric restructuring within a traditionally monocentric metropolis. By integrating economic complexity theory with spatial analysis, our approach provides a scalable method to assess the adaptive capacity of urban economies. This has implications for understanding informality, designing inclusive urban policies, and navigating structural transitions in rapidly urbanizing regions.

In crystal plasticity finite element (CPFE) simulations, accurately quantifying geometrically necessary dislocations (GNDs) is critical for capturing strain gradients in polycrystals. We compare different methods for quantifying GNDs, all of which originate from the Nye tensor, which is computed as the curl of the plastic deformation gradient. The projection technique directly decomposes the Nye tensor onto individual screw and edge dislocation components to compute GNDs. This approach requires converting a nine-component Nye tensor into densities for a larger number of dislocation systems, a fundamentally underdetermined (non-unique) process, which is resolved using $L2$ minimization. In contrast, when employing CPFE analysis, one could directly compute dislocation densities on each slip system using shear gradients. Projection and slip gradient methods are compared with respect to their prediction of GNDs with changing grain size, strain, and grain neighborhoods, including multigrain junctions. Although these techniques match analytical GND densities for single slip, single crystal deformation, and are consistent with anticipated overall GND trends, we find that the GND densities from projection techniques are significantly lower than those predicted from CPFE-based slip gradients in polycrystals. A suggested improvement of only using the active dislocation systems in the projection technique almost entirely resolved this mismatch.

We introduce ELECTRAFI, a fast, end-to-end differentiable model for predicting periodic charge densities in crystalline materials. ELECTRAFI constructs anisotropic Gaussians in real space and exploits their closed-form Fourier transforms to analytically evaluate plane-wave coefficients via the Poisson summation formula. This formulation delegates non-local and periodic behavior to analytic transforms, enabling reconstruction of the full periodic charge density with a single inverse FFT. By avoiding explicit real-space grid probing, periodic image summation, and spherical harmonic expansions, ELECTRAFI matches or exceeds state-of-the-art accuracy across periodic benchmarks while being up to $633 \times$ faster than the strongest competing method, reconstructing crystal charge densities in a fraction of a second. When used to initialize DFT calculations, ELECTRAFI reduces total DFT compute cost by up to ~20%, whereas slower charge density models negate savings due to high inference times. Our results show that accuracy and inference cost jointly determine end-to-end DFT speedups, and motivate our focus on efficiency.

Nitinol (NiTi shape-memory alloy) is an interesting candidate in various medical applications like dental, orthopedic, and cardiovascular devices, owing to its unique mechanical behaviors and proper biocompatibility. The aim of this work is the local controlled delivery of a cardiovascular drug, heparin, loaded onto nitinol treated by electrochemical anodizing and chitosan coating. In this regard, the structure, wettability, drug release kinetics, and cell cytocompatibility of the specimens were analyzed in vitro. The two-stage anodizing process successfully developed a regular nanoporous layer of Ni-Ti-O on nitinol, which considerably decreased the sessile water contact angle and induced hydrophilicity. The application of the chitosan coatings controlled the release of heparin mainly by a diffusional mechanism, where the drug release mechanisms were evaluated by the Higuchi, first-order, zero-order, and Korsmeyer-Pepass models. Human umbilical cord endothelial cells (HUVECs) viability assay also showed the non-cytotoxicity of the samples, so that the best performance was found for the chitosan-coated samples. It is concluded that the designed drug delivery systems are promising for cardiovascular, particularly stent applications.

The use of bioceramics as hard tissue substitutes is extensive due to their excellent biocompatible and osteogenic behaviors. Among various bioceramics, Ca-Mg silicates are unique from the viewpoints of osteoinductive and mechanical properties, as well as their outstanding osteoconductive and angiogenic behaviors owing to the release of Si, Ca and Mg. Despite these distinct advantages, different compositions of these bioceramics still require mechanical and biological enhancements for further applications. For this purpose, doping with some ions like F-, Sr2+, Cu2+, Eu2+, Ba+, Ce3+ and some alkali cations has been proved to be a valued approach. This review attempts to bring together areas for the performance improvement of the further researched Ca-Mg silicates (i.e., diopside, akermanite, bredigite and monticellite) and the alteration of these compositions via ion-doping. It is concluded that a correct choice of dopants incorporated at the optimal concentration makes these silicates ideal bone substitutes competing or even superior to calcium phosphates (apatites) and bioglasses which are known as the most prominent bioceramics.

The spatial structure of populations may promote the emergence and maintenance of cooperation. Cooperation in the prisoner's dilemma is favored under specific update rules in evolutionary graph theory models with one individual per node of a graph, but this effect vanishes in models with well-mixed demes connected by migrations under soft selection. In contrast, experiments and models involving cycles of growth, merging and dilution have shown that spatial structure can favor cooperation. Here, we reconcile these findings by studying deme-structured populations under growth-merging-dilution dynamics, corresponding to a clique (fully connected graph) under hard selection. We obtain analytical conditions for the cooperator fraction to increase during deterministic logistic growth, and to increase on average under dilution-growth-merging cycles, in the weak selection regime. Furthermore, we analytically express the fixation probability of cooperators under weak selection, yielding a criterion for cooperative mutants to have a higher fixation probability than neutral ones. Finally, numerical simulations show that stochastic growth further promotes cooperation. Overall, hard selection is essential for cooperation to be promoted in deme-structured populations.

Each T cell typically carries a specific T-cell receptor (TCR) that determines its specificity against an epitope presented by the HLA complex on a target cell. Antigenic challenge triggers the expansion of reactive cells within a diverse pool of T cells with randomly generated receptors, a process that results in epitope-driven shifts of TCR frequencies over time. Here, we analyze the effects of SARS-CoV-2 vaccination on the TCR populations in peripheral blood drawn from seven COVID-naive individuals, before vaccines were widely available. To identify SARS-CoV-2 vaccine-associated TCR sequences among the $\sim 10^{5}-10^{6}$ TCR sequences sampled before and after vaccination, we develop statistical criteria to detect significant increases in abundance of positive TCR clones. Application of our statistical methods shows a robust identification of TCR sequences that respond to SARS-CoV-2 vaccination in vivo, illustrating the feasibility of quantifying the clone-specific dynamics of T-cell abundance changes following immunological perturbations.

Significance: Voltage imaging microscopy has emerged as a powerful tool to investigate neural activity both in vivo and in vitro. Various imaging approaches have been developed, including point-scanning, line-scanning and wide-field microscopes, however the effects of their different temporal sampling methods on signal fidelity have not yet been fully investigated. Aim: To provide an analysis of the inherent advantages and disadvantages of temporal sampling in scanning and wide-field microscopes and their effect on the fidelity of voltage spike detection. Approach: We develop a mathematical framework based on a mixture of analytical modeling and computer simulations with Monte-Carlo approaches. Results: Scanning microscopes outperform wide-field microscopes in low signal-to-noise conditions and when only a small subset of spikes needs to be detected. Wide-field microscopes outperform scanning microscopes when the measurement is temporally undersampled and a large fraction of the spikes needs to be detected. Both modalities converge in performance as sampling increases and the frame rate reaches the decay rate of the voltage indicator. Conclusions: Our work provides guidance for the selection of optimal temporal sampling parameters for voltage imaging. Most importantly it advises against using scanning voltage imaging microscopes at frame rates below 500 Hz.

This paper gives an overview of the use of control systems engineering in synthetic biology, motivated by applications such as cell therapy and cell fate reprogramming for regenerative medicine. A ubiquitous problem in these and other applications is the ability to control the concentration of specific regulatory factors in the cell accurately despite environmental uncertainty and perturbations. The paper describes the origin of these perturbations and how they affect the dynamics of the biomolecular ``plant'' to be controlled. A variety of biomolecular control implementations are then introduced to achieve robustness of the plant's output to perturbations and are grouped into feedback and feedforward control architectures. Although sophisticated control laws can be implemented in a computer today, they cannot be necessarily implemented inside the cell via biomolecular processes. This fact constraints the set of feasible control laws to those realizable through biomolecular processes that can be engineered with synthetic biology. After reviewing biomolecular feedback and feedforward control implementations, mostly focusing on the author's own work, the paper illustrates the application of such control strategies to cell fate reprogramming. Within this context, a master regulatory factor needs to be controlled at a specific level inside the cell in order to reprogram skin cells to pluripotent stem cells. The article closes by highlighting on-going challenges and directions of future research for biomolecular control design.

While businesses are typically more profitable if their workers and communities are minimally exposed to diseases, the same is not true for daycare centers. Here it is shown that a daycare center could maximize its profits by maintaining a population of sick children within the center, with the intention to infect more children who then do not attend. Through a modification of the Susceptible-Infected-Recovered (SIR) model for disease spread we find the optimal number of sick children who should be kept within the center to maximize profits. We show that as disease infectiousness increases, the optimal attendance rate of sick children approaches zero, while the potential profit increases.

We investigate the process of quantum teleportation in an expanding universe modeled by Friedmann-Robertson-Walker spacetime, focusing on two cosmologically relevant scenarios: a power-law expansion and the de Sitter universe. Adopting a field-theoretical approach, we analyze the quantum correlations between two comoving observers who share an entangled mode of a scalar field. Using the Bogoliubov transformation, we compute the teleportation fidelity and examine its dependence on the expansion rate, initial entanglement, and the mode frequency. Our findings indicate that spacetime curvature and the underlying cosmological background significantly affect the efficiency of quantum teleportation, particularly through mode mixing and vacuum structure. We also compare our results with the flat Minkowski case to highlight the role of cosmic expansion in degrading or preserving quantum information.

We explore the derivation of the Friedmann equations from a thermodynamic perspective, applying the unified first law of thermodynamics to the apparent horizon of a flat Friedmann-Lemaître-Robertson-Walker (FLRW) universe. We extend this framework to incorporate gravitationally induced particle creation, treating the region enclosed by the apparent horizon as an open thermodynamic system. A crucial aspect of our analysis is the recognition that the apparent horizon volume is not comoving; this requires a consistent accounting of particle exchange across the moving boundary. We demonstrate that the evolution of the particle number, and explicitly the matter entropy, can be decomposed into two distinct physical contributions: genuine bulk particle production and a net flux induced by the dynamics of the horizon itself. Finally, we derive the Generalized Second Law (GSL) in this setting, showing transparently how the total entropy budget is balanced by horizon thermodynamics, bulk creation, and boundary fluxes.

Computing Gaussian ground states via variational optimization is challenging because the covariance matrices must satisfy the uncertainty principle, rendering constrained or Riemannian optimization costly, delicate, and thus difficult to scale, particularly in large and inhomogeneous systems. We introduce a symplectic optimization framework that addresses this challenge by parameterizing covariance matrices directly as positive-definite symplectic matrices using unit-triangular factorizations. This approach enforces all physical constraints exactly, yielding a globally unconstrained variational formulation of the bosonic ground-state problem. The unconstrained structure also naturally supports solution reuse across nearby Hamiltonians: warm-starting from previously optimized covariance matrices substantially reduces the number of optimization steps required for convergence in families of related configurations, as encountered in crystal lattices, molecular systems, and fluids. We demonstrate the method on weakly dipole-coupled lattices, recovering ground-state energies, covariance matrices, and spectral gaps accurately. The framework further provides a foundation for large-scale approximate treatments of weakly non-quadratic interactions and offers potential scaling advantages through tensor-network enhancements.

Standard approaches to quantum error correction (QEC) require active maintenance using measurements and classical processing. The possibility of passive QEC has so far only been established in an unphysical number of spatial dimensions. In this work, we present a simple method for autonomous QEC in two spatial dimensions, formulated as a quantum cellular automaton with a fixed, local and translation-invariant update rule. The construction uses hierarchical, self-simulating control elements based on the classical schemes from the seminal results of Gács (1986, 1989) together with a measurement-free concatenated code. We analyze the system under a local noise model and prove a noise threshold below which the logical errors are suppressed arbitrarily with increasing system size and the memory lifetime diverges in the thermodynamic limit. The scheme admits a continuous-time implementation as a time-independent, translation-invariant local Lindbladian with engineered dissipative jump operators. Further, the recursive nature of our protocol allows for the fault-tolerant encoding of arbitrary quantum circuits and thus constitutes a self-correcting universal quantum computer.

We develop a semiclassical framework for studying quantum particles constrained to curved surfaces using the momentous quantum mechanics formalism, which extends classical phase-space to include quantum fluctuation variables (moments). In a spherical geometry, we derive quantum-corrected Hamiltonians and trajectories that incorporate quantum back-reaction effects absent in classical descriptions. For the free particle, quantum fluctuations induce measurable phase shifts in azimuthal precession of approximately 8-12%, with uncertainty growth rates proportional to initial moment correlations. When a non-central Makarov potential is introduced, quantum corrections dramatically amplify its asymmetry. For strong coupling ($γ$ = -1.9), the quantum-corrected force drives trajectories preferentially toward the southern hemisphere on timescales 40% shorter than classical predictions, with trajectory densities exhibiting up to 3-fold enhancement in the preferred region. Throughout evolution, the solutions rigorously satisfy Heisenberg uncertainty relations, validating the truncation scheme. These results demonstrate that quantum effects fundamentally alter semiclassical dynamics in curved constrained systems, with direct implications for charge transport in carbon nanostructures, exciton dynamics in curved quantum wells, and reaction pathways in cyclic molecules.

Long spin coherence times of carriers are essential for implementing quantum technologies using semiconductor devices for which, however, a possible obstacle is spin relaxation. For the spin dynamics, decisive features are the band structure, crystal symmetry, and quantum confinement. Perovskite semiconductors recently have come into focus of studies of their spin states, notivated by efficient optical access and potentially long-living coherence. Here, we report an electron spin coherence time $T_2$ of the order of 1 ms, measured for a bulk FA$_{0.95}$Cs$_{0.05}$PbI$_3$ lead halide perovskite crystal. Using periodic laser pulses, we synchronize the electron spin Larmor precession about an external magnetic field in an inhomogeneous ensemble, the effect known as spin mode locking. It appears as a decay of the optically created ensemble spin polarization within the dephasing time $T_2^*$ of up to 20 ns and its revival during the spin coherence time $T_2$ reaching the millisecond range. This exceptionally long spin coherence time in a bulk crystal is complemented by millisecond-long longitudinal spin relaxation times $T_1$ for electrons and holes, measured by optically-detected magnetic resonance. These long-lasting spin dynamics highlight perovskites as promising platform for the quantum devices with all-optical control.

In this note we discuss the Efimov effect emerging in a three-particle quantum system with zero-range interactions. In particular, we consider two non-interacting identical bosons plus a different lighter particle such that the interaction between a boson and the light particle is resonant. We also assume the validity of the Born-Oppenheimer approximation. Under these conditions, we show that the three-particle system exhibits infinitely many negative eigenvalues which accumulate at zero and satisfy the universal geometrical law characterising the Efimov effect. The result we find is a generalisation of previous results recently obtained in [13, 24].

We study the quantum Pais-Uhlenbeck oscillator at the resonant (equal-frequency) point, where the dynamics becomes non-diagonalisable and the conventional Fock-space construction collapses. At the classical level, the degenerate system admits more than one Hamiltonian formulation generating the same equations of motion, leading to a nontrivial quantisation ambiguity. Working first in the ghostly two-dimensional Hamiltonian formulation, we construct differential intertwiners that generate a spectrum-generating algebra acting on the generalised eigenspaces of the Hamiltonian. This algebra organises the generalised eigenvectors into finite Jordan chains and closes into a hidden $su(2)$ Lie algebra that exists only at resonance. We then show that quantising a classically equivalent Hamiltonian yields a radically different quantum theory, with a fully diagonalisable spectrum and genuine degeneracies. Our results demonstrate that the resonant Pais-Uhlenbeck oscillator provides a concrete example in which classically equivalent Hamiltonians define inequivalent quantum theories.

In molecular aggregates, multiple delocalized exciton states interact with phonons, making the state-resolved spectroscopic monitoring of dynamics challenging. We propose a protocol that combines photon-entanglement-enhanced narrowband excitation of two-exciton states with time-frequency-filtered two-photon coincidence counting. It can alleviate bottlenecks associated with probing exciton dynamics spread across multiple spectral and temporal windows. We demonstrate that non-classical correlations of entangled photon pairs can be used to prepare narrowband two-exciton population distributions, circumventing transport in mediating states. The distributions thus created can be monitored using time-frequency-filtered photon coincidence counting, and the pathways contributing to photon emission events can be classified by tuning filtering parameters. Numerical simulations for a light-harvesting aggregate highlight the ability of this protocol to achieve selectivity by suppressing or amplifying specific pathways. Combining entangled photonic sources and multidimensional photon counting allow promising applications to spectroscopy and sensing.

We propose a co-designed variant of ADAPT-VQE (Co-ADAPT-VQE) where the quantum hardware is taken into account in the construction of the ansatz. This framework can be readily used to optimize state preparation circuits for any device, addressing shortcomings such as limited connectivity, short coherence times, and variable gate errors. We exemplify the impact of Co-ADAPT-VQE by creating state preparation circuits for devices with linear nearest-neighbor (LNN) connectivity. We show a reduction of the CNOT count of the final circuits by up to 97% for 12-14 qubit systems, with the impact being greater for larger and more strongly correlated systems. Surprisingly, the circuits created by Co-ADAPT-VQE provide an over 70% CNOT count reduction with respect to the original ADAPT-VQE in all-to-all connectivity, despite being restricted to LNN qubit interactions.

A mildly inhomogeneous universe with a cosmological constant may look like it contains evolving dark energy. We show that could be the case by modelling the inhomogeneities and their effects in three different ways: as clumped matter surrounded by voids, as back-reaction of small-scale structure on the overall expansion of the Universe, and, finally, as a large-scale curvature inhomogeneity. In all of these cases, the propagation of light is affected, and differs from that in a homogeneous and isotropic universe. The net result is that cosmological observables, such as angular diameter and luminosity distances, become distorted. We find, in all three models, that the inclusion of these effects pushes the distance-redshift relation towards closer agreement with recent data from both supernovae Ia from the Dark Energy Survey, and from baryon acoustic oscillations from the Dark Energy Spectroscopic Instrument. The amount of inhomogeneity in these models might not be enough to explain the entirety of the deviation from a cosmological constant, but is found to be of a similar order of magnitude, hinting that these data may be consistent with a universe dominated by a cosmological constant.

Rydberg-atom sensors convert radiofrequency, microwave and terahertz fields into optical signals with SI-traceable calibration, high sensitivity, and broad tunability. This review assesses their potential for space applications by comparing five general architectures (Autler-Townes, AC-Stark, superheterodyne, radiofrequency-to-optical conversion, and fluorescence) against space application needs. We identify promising roles in radiometry, radar, terahertz sensing, and in-orbit calibration, and outline key limitations, including shot noise, sparse terahertz transitions, and currently large Size, Weight, Power and Cost. A staged roadmap highlights which uncertainties should be resolved first and how research organisations, industry and space agencies could take the lead for the different aspects.

Nonclassical states of light and their correlations lie at the heart of quantum optics, serving as fundamental resources that underpin both the exploration of quantum phenomena and the realisation of quantum information protocols. These lecture notes provide an accessible yet rigorous introduction to the foundations of quantum optics, emphasising their relevance to quantum information science and technology. Starting from the quantisation of the electromagnetic field and the bosonic formalism of Fock space, the notes develop a unified framework for describing and analysing quantum states of light. Key families of states -- thermal, coherent, and squeezed -- are introduced as paradigmatic examples illustrating the transition from classical to nonclassical behaviour. The concepts of convexity, classicality, and quasiprobability representations are presented as complementary tools for characterising quantumness and defining operational notions such as P-nonclassicality. The discussion extends naturally to Gaussian states, composite systems, and continuous-variable entanglement, highlighting how nonclassicality serves as a resource for generating and quantifying quantum correlations. Theoretical developments are complemented by computational and experimental perspectives, including simulations of optical states using the Python library Strawberry Fields and data analysis from simulated data. Together, these notes aim to bridge the foundational concepts of quantum optics and modern quantum information, offering both conceptual insight and practical tools for students and researchers entering the field.

We investigate the absorption and scattering by a Schwarzschild black hole in scalar--tensor theories of gravity, where the coupling between matter and the scalar field induces different models for the effective mass of the scalar field. In model~I, a Bondi-type mass model described by the asymptotic mass $μ_c$, horizon mass $μ_H$, and profile slope $λ$, it is found that the absorption cross section increases with steeper $λ$, larger $μ_c$ (especially at higher frequencies), or smaller $μ_H$. The differential scattering cross section in this model shows the strongest dependence on the horizon mass $μ_H$. When $μ_H$ exceeds a critical value for a fixed incoming wave frequency $ω$, no partial wave transmits into the black hole, flattening the differential scattering cross section as a function of angle before it increases again with further increase of $μ_H$. Model~II, which considers a truncated accretion region outside some radius $r_0$, contains a potential well in its effective scattering potential. Its absorption cross section decreases in the low-frequency region as the accretion radius $r_0$ decreases, and more importantly, it shows resonance peaks at the quasibound wave frequencies due to resonances induced by the potential well. The differential scattering cross sections show dips around intermediate scattering angles when the parameters (mainly $μ_H$ and $ω$) are such that the resonantly scattered and non-resonant waves interfere destructively around these angles. In both models, absorption exhibits a zero-absorption band as $ω$ approaches $μ_c$ from above, and in both absorption and scattering, the effects of the parameters are found to diminish in the high-frequency limit.

Statistical paradoxes such as the Hardy paradox and the enhancement of phase estimation via post-selection both draw upon the same non-classical features of quantum statistics described by non-positive quasi-probabilities. In this paper, we introduce a post-selected quantum metrology scenario where the initial state, the dynamics associated with the phase shift, and the post-selection are all inspired by the Hardy paradox. Specifically, we identify an anomalous weak value that is characteristic of both the Hardy paradox and the potential enhancement of sensitivity by the post-selection. We find that the efficiency of the enhancement is reduced when the expectation value associated with the anomalous weak value is different from the inverse of this value. We conclude that the relation between enhanced phase estimation and the Hardy paradox requires a detailed understanding of the relation between weak values and expectation values.

Hexagonal boron nitride (hBN) has emerged as a promising two-dimensional host for stable single-photon emission owing to its wide bandgap, high photostability, and compatibility with nanophotonic integration. We present a simulation-based study of temperature-dependent spectral dynamics and optical coherence in a mechanically decoupled quantum emitter in hBN. Employing a hybrid stochastic framework that combines Ornstein--Uhlenbeck detuning fluctuations with temperature-dependent, Gaussian-distributed discrete frequency jumps, motivated by experimentally observed spectral diffusion and blinking, we reproduce the measured evolution of inhomogeneous linewidth broadening and the progressive degradation of photon coherence across the relevant cryogenic range (5-30K). The model captures phonon-related spectral diffusion with a cubic temperature dependence and the onset of jump-like spectral instabilities at higher temperatures. By calibrating the hybrid diffusion, jump parameters to the experimentally measured full width at half maximum (FWHM) of the emission line and analyzing the second-order correlation function $g^{(2)}(τ)$ under resonant driving, we establish a unified phenomenological description that links stochastic detuning dynamics to the decay of optical coherence in a resonantly driven emitter. Analysis of $g^{(2)}(τ)$ under resonant driving reveals an additional dephasing rate $γ_{\mathrm{sd+j}}$ that rises monotonically with temperature and drive strength, leading to a predicted critical crossover to overdamped dynamics at $T_{\mathrm{crit}} \approx 25.91$~K. This hybrid framework provides a quantitative connection between accessible spectroscopic observables and the dominant noise mechanisms limiting coherent optical control in mechanically decoupled quantum emitters, exemplified in hBN and generalizable to similar emitters in other materials.

We observe that the transition probability in a static two-level quantum detector interacting with a coherent Rindler photon is different from the same of the Rindler detector which is in interaction with a coherent Minkowski photon. Situation does not change in the response of quantum detector for the classical limit of the photon state. This we investigate in $(1+1)$ and $(3+1)$-spacetime dimensions. Interestingly, the transition probabilities of the ``classical'' detector in the classical limit of the photon state in $(1+1)$-dimensions, for these two scenarios, appear to be identical when the frequencies of photon mode and detector are taken to be same. However, our obtained detector's transition probabilities in $(3+1)$-dimensions, which are calculated under the large acceleration condition, do not show such signature. The implications of these observations are discussed as well.

We present a fully quantum field-theoretic framework for gravitational wave (GW) detection in which the interaction is described as photon-graviton scattering. In this picture, the GW acts as a coherent background that induces inelastic energy exchanges with the electromagnetic field - analogous to the Stokes and anti-Stokes shifts in Raman spectroscopy. We propose a detection scheme sensitive to this microscopic mechanism based on Hong-Ou-Mandel interference. We show that the scattering-induced phase shifts render frequency-entangled photon pairs distinguishable, spoiling their destructive quantum interference. GW signal is thus encoded in the modulation of photon coincidence rates rather than classical field intensity, offering a complementary quantum probe of the gravitational universe that recovers the standard classical response in the macroscopic limit.

This paper presents a comprehensive stability analysis of the black hole solution within Modified General Relativity (MGR), a theory proposing a unified geometric description of dark matter (DM) and dark energy (DE). A rigorous gauge-invariant formalism is employed to analyze gravitational perturbations of the extended Schwarzschild metric. The central finding is a critical pathology within the polar perturbation sector, where metric fluctuations couple to the theory's fundamental line element field. This coupling is governed by a factor that, while well-behaved at the horizon, diverges powerfully in the far-field limit as a direct consequence of the theory's non-asymptotically flat nature. This indicates a strong infrared instability that overwhelms perturbations at large distances. In stark contrast, the axial perturbation sector is found to be completely stable. This dichotomy proves that the instability is not inherent to the background metric but is specifically generated by the novel coupling mechanism encoding MGR's unified dark sectors. The results reveal a fundamental strong-coupling problem within the MGR framework, challenging its physical viability as an alternative to Einstein's General Relativity (EGR).

The concept of quantum particles derives from quantum field theory. Accepting that quantum mechanics is valid all the way implies that not only composite particles (such as protons and neutrons) would be derived from a field theory, but also the center of mass of bodies as heavy as rocks. Despite the fabulous success of quantum mechanics, it is unreasonable to assume the existence of annihilation and creation operators for rocks, and so on. Fortunately, there are strong reasons to doubt that wave mechanics can describe the center of mass of systems at or above the Planck scale, thereby jeopardizing the construction of the corresponding Fock space. As a result, systems with masses exceeding the Planck mass would have their center of mass described through classical mechanics, regardless of being able to harbor macroscopic quantum phenomena as observed in the laboratory. Here, we briefly revisit (i) the arguments for the need for a Heisenberg cut delimitating the boundary between the quantum and classical realms and (ii) the kind of new physics expected at (the uncharted region of) the Heisenberg cut.''

We investigate the tidal response of general five-dimensional (5D) black holes of STU supergravity, which include as special cases important solutions such as the Myers-Perry, BMPV, 5D Reissner-Nordström, Kerr-Newman and dyonic black holes. Solutions are parameterized by their mass, two angular momenta and up to three $U(1)$ charges. Love numbers and dissipation coefficients are obtained in the static and dynamic cases. In the latter scenario, we find new, nontrivial conditions, realized in important limiting cases of the theory, such as the BPS limit, where frequency-independent vanishing conditions are obtained. We also develop a ladder formalism for static solutions and derive the conserved charges. To the best of our knowledge, this formalism had not been previously derived for 5D black holes, including neutral ones. Finally, we show the emergence of Love symmetry in the near-zone regime, and derive the generators of the associated $sl(2,\mathbb{R})$ algebra. It is shown that all conditions for Love-number vanishing can be explained by this algebra in terms of the highest-weight property.

In this paper, we describe the procedure implemented in the Multi-Band Template Analysis (MBTA) search pipeline to produce online posterior distributions of compact binary coalescence (CBC) gravitational-wave parameters. This procedure relies on an SNR optimizer technique, which consists of filtering dense local template banks. We present how these banks are constructed using information from the initial detection and detail how the results of the filtering are used to estimate source parameters and provide posterior distributions. We demonstrate the performance of our procedure on simulations and compare our source parameter estimates with the results from the first part of the fourth observing run (O4a) recently released by the LIGO-Virgo-KAGRA (LVK) collaboration.

We study the wave function $φ^{(3)}$ of three identical bosons scattering at zero energy, zero total momentum, and zero orbital angular momentum in two dimensions, interacting via short-range potentials with a finite two-body scattering length $a$. We derive asymptotic expansions of $φ^{(3)}$ in two regimes: the 111-expansion, where all three pairwise distances are large, and the 21-expansion, where one particle is far from the other two. In the 111-expansion, the leading term grows as $\ln^3(B/a)$ at large hyperradius $B=\sqrt{(s_1^2+s_2^2+s_3^2)/2}$. At order $B^{-2}\ln^{-3}(B/a)$, we identify a three-body parameter $D$ with dimension of length squared, which we term the three-body scattering area. This quantity should be contrasted with the three-body scattering area previously studied for infinite or vanishing two-body scattering length. If the two-body interaction is attractive and supports bound states, $D$ acquires a negative imaginary part, and we derive its relation to the probability amplitudes for the production of two-body bound states in three-body collisions. Under weak modifications of the interaction potentials, we derive the corresponding shift of $D$ in terms of $φ^{(3)}$ and the changes of the two-body and three-body potentials. We also study the effects of $D$ and $φ^{(3)}$ on three-body and many-body physics, including the three-body ground-state energy in a large periodic volume, the many-body energy and the three-body correlation function of the dilute two-dimensional Bose gas, and the three-body recombination rates of two-dimensional ultracold atomic Bose gases.

This note presents a concise and non-polemical comparison of several major interpretations of quantum mechanics, with a particular emphasis on the distinction between FAPP-solutions ("For All Practical Purposes'') versus ontological solutions to the measurement problem. Building on this distinction, we argue that the Contexts-Systems-Modalities (CSM) framework, supplemented by the operator-algebraic description of macroscopic contexts, provides a conceptually complete, non-FAPP ontology that naturally incorporates irreversibility and the physical structure of measurement devices. This approach differs significantly from other ontological interpretations such as Bohmian mechanics, spontaneous collapse, or many-worlds, and highlights the major role of contextual quantization in shaping quantum theory.

The localization property of a non-Hermitian Su-Schrieffer-Heeger (SSH) chain with quasi-periodic on-site potential is investigated. In contrast to the preceding investigations, the quantum phase transition between localized state and extended one is achieved by adjusting the strength of intracellular or intercellular hopping. The energy spectra and eigenstate distributions of the system's Hamiltonian near the boundary of the phase transition exhibit different behaviors when the Hermiticity, non-Hermiticity and mosaic modulation of the quasi-periodic potential are considered, respectively. The existence of the mobility ring is revealed in the non-Hermitian SSH chain by studying of the critical behaviors near the boundary. More interestingly, the multiple mobility rings emerge when the period number of the mosaic modulation is increased. The result is helpful for the investigation of the localization-delocalization transition in the SSH-type system under the combined action of the non-Hermiticity and quasi-periodicity.

Magnetic fields play a central role in black hole astrophysics, powering relativistic jets and other energetic phenomena. While near-horizon magnetic field is usually assumed to originate from the accretion flow, additional large-scale magnetic fields - such as those supplied by a companion neutron star in stellar-mass binaries or by galactic fields around supermassive black holes - may also affect the horizon-threading flux. In this work, we study the superposition of a weak arbitrarily inclined external uniform magnetic field with the internal Blandford-Znajek split-monopole field around a Schwarzschild black hole. This setup generically gives rise to magnetic null points, where the total field vanishes. We compute the magnetic flux through an arbitrarily tilted hemisphere of the event horizon and show that the flux can be substantially suppressed by the external field. In the axisymmetric case, the flux can even vanish completely. However, with nonzero inclination, complete cancellation becomes impossible, despite significant reduction. We further explore the ionization and subsequent particle acceleration from a Keplerian accretion disk, finding that efficient collimated outflows persist even under significant field inclination. We show that the acceleration is critically dependent on the external field orientation, with the escape fraction maximized at non-zero inclinations due to the destabilization of trapping zones and minimized in the anti-aligned configuration, where closed magnetic loops effectively suppress the outflow. We discuss the astrophysical implications of these findings, proposing that geometric flux cancellation can serve as a mechanism for jet quenching in compact binaries and offering an explanation for the lack of a prominent large-scale jet in Sgr A*.

High fidelity quantum operations are key to enabling fault-tolerant quantum computation. Superconducting quantum processors have demonstrated high-fidelity operations, but on larger devices there is commonly a broad distribution of qualities, with the low-performing tail affecting near-term performance and applications. Here we present an error budgeting procedure for the native two-qubit operation on a 32-qubit superconducting-qubit-based quantum computer, the OQC Toshiko gen-1 system. We estimate the prevalence of different forms of error such as coherent error and control qubit leakage, then apply error suppression strategies based on the most significant sources of error, making use of pulse-shaping and additional compensating gates. These techniques require no additional hardware overhead and little additional calibration, making them suitable for routine adoption. An average reduction of 3.7x in error rate for two qubit operations is shown across a chain of 16 qubits, with the median error rate improving from 4.6$\%$ to 1.2$\%$ as measured by interleaved randomized benchmarking. The largest improvements are seen on previously under-performing qubit pairs, demonstrating the importance of practical error suppression in reducing the low-performing tail of gate qualities and achieving consistently good performance across a device.

The Leggett-Garg inequality (LGI) is a temporal analogue of Bell's inequality and provides a quantitative test of the nonclassicality of a system through its violation. We analytically investigate the violation of the LGI in the context of photon-graviton conversion in a magnetic field background, motivated by its potential applications to testing the nonclassicality of gravity. When gravitational perturbations are quantized as gravitons, the conversion of an initial single photon state gives rise to a superposition of photon and graviton states. We show that the temporal correlations obtained from successive projective measurements on the photon-graviton system violate the LGI. Observation of such a violation would provide a novel avenue for probing the quantum nature of gravity.

This work investigates network nonlocality sharing for a broad class of bipartite Bell inequalities in a generalized star network with an $(n,m,k)$ configuration, comprising $n$ independent branches, $m$ sequential Alices per branch, and $k$ measurement settings per party. On each branch, the intermediate Alices implement optimal weak measurements, whereas the final Alice and the central Bob perform sharp projective measurements. Network nonlocality sharing is witnessed when the quantum values of the network correlations associated with relevant parties simultaneously violate a star-network Bell inequality generated from the given class of bipartite Bell inequalities. We streamline the calculation of the quantum values of the network correlations and derive an analytical expression for the bipartite quantum correlator, valid for arbitrary measurement settings and weak-measurement strengths. The network nonlocality sharing for Vértesi inequalities has been studied within the framework, and simultaneous violations are found in $(2,2,6)$ and $(2,2,465)$ cases, with the latter exhibiting greater robustness. Our approach suggests a practical route to studying network nonlocality sharing by utilizing diverse bipartite Bell inequalities beyond the commonly used CHSH-type constructions.

Preparing large-qubit Dicke states is of broad interest in quantum computing and quantum metrology. However, the number of qubits available on a single quantum processing unit (QPU) is limited -- motivating the distributed preparation of such states across multiple QPUs as a practical approach to scalability. In this article, we investigate the distributed preparation of $n$-qubit $k$-excitation Dicke states $D(n,k)$ across a general number $p$ of QPUs, presenting a distributed quantum circuit (each QPU hosting approximately $\lceil n/p \rceil$ qubits) that prepares the state with communication complexity $O(p \log k)$, circuit size $O(nk)$, and circuit depth $O\left(p^2 k + \log k \log (n/k)\right)$. To the best of our knowledge, this is the first construction to simultaneously achieve logarithmic communication complexity and polynomial circuit size and depth. We also establish a lower bound on the communication complexity of $p$-QPU distributed state preparation for a general target state. This lower bound is formulated in terms of the canonical polyadic rank (CP-rank) of a tensor associated with the target state. For the special case $p = 2$, we explicitly compute the CP-rank corresponding to the Dicke state $D(n,k)$ and derive a lower bound of $\lceil\log (k + 1)\rceil$, which shows that the communication complexity of our construction matches this fundamental limit.

We presents a unified and concise exposition of key topics in the mathematical theory of open quantum systems, developed within the framework of operator algebras. The manuscript consolidates and extends a series of invited articles originally prepared for the Modern Encyclopedia of Mathematical Physics, combining foundational material with modern perspectives on non-equilibrium quantum statistical mechanics. After introducing the C*- and W*-algebraic formulation of quantum mechanics, the paper reviews quantum dynamical systems, KMS states, and Tomita-Takesaki modular theory, as well as CCR and CAR algebras for bosonic and fermionic systems. Particular emphasis is placed on infinite systems, non-equilibrium steady states, entropy production, and linear response theory. The later sections develop a systematic treatment of small systems coupled to reservoirs, open lattice quantum spin systems, culminating in a detailed discussion of competing notions of quantum entropy production. The presentation highlights structural insights, conceptual clarity, and connections between equilibrium and non-equilibrium phenomena, providing a self-contained reference for researchers and graduate students in mathematical physics.

In this work, we use the generalised Klein-Gordon Equation in curved spacetime with an electromagnetic field to investigate the tunnelling phenomenon of scalar particles originating from the hot NUT Kerr Newman Kasuya Anti de Sitter (KNKNK AdS) black hole. Using the tunnelling formalism, we obtain a modified Hawking temperature different from previous works due to the quantum gravity effect for the charged Dirac particle at the HNKNK AdS black hole's horizon. We find that the modified Hawking temperature is affected by the cosmological constant, magnetic mass, electric and magnetic charges. We demonstrate that a significant number of discontinuities exist in the heat capacities of the HNKNK AdS, indicating that the black hole system becomes unstable as the black hole size decreases.

We study topological defects with a general structure in higher-dimensional cosmological backgrounds described by a set of angle deficit parameters. As special cases, they include higher-dimensional generalizations of cosmic strings and global monopoles. The corresponding complete set of mode functions is presented for a massive scalar field with a general curvature coupling parameter. For general scale factors and radial functions in the line element, the angular parts of the scalar modes are expressed in terms of associated Legendre functions. De Sitter and Milne universes are considered as examples of cosmological expansion. For the de Sitter bulk, we present the time-dependent parts of the mode functions in inflationary, hyperbolic, and global coordinates.

In this paper, we generalized the quantum frequency mixing technology to a ladder-type four-level system and studied its effect on electromagnetically induced transparency spectra. We found a secondary splitting of Autler-Townes splitting in the probing field transmission spectra, which could be understood by the effective Hamiltonian derived with multi-mode Floquet theory. The Frequency mixing scheme developed here enables continuous tunablity of the resonant frequency between upper levels, which facilitates the broad band sensing of AC field. Furthermore, by introducing an additional periodic driving, we realize an effective model that two distinct quantum interference effects coexist: interference among Floquet channels and loop interference arising from closed coherent pathways. Both interference effects could be read out from the transmission spectra independently. The changing of the distance between double splitting peaks represents the interference of Floquet channels, while their asymmetric linewidth broadening is linked with the total effective phase of the loop. This not only provides complementary readout for extracting the phase of AC field, but also establishes a new paradigm for coherent control in multi-level quantum systems.

We present a simple framework for classical and quantum ``memory'' in which the Hamiltonian at time $t$ depends on past values of a control Hamiltonian through a causal kernel. This structure naturally describes finite-bandwidth or filtered control channels and provides a clean way to distinguish between memory in the control and genuine non-Markovian dynamics of the state. We focus on models where $H(t)=H_0+\int_{-\infty}^{t}K(t-s)\,H_1(s)\,ds$, and illustrate the framework on single-qubit examples such as $H(t)=σ_z+Φ(t)σ_x$ with $Φ(t)=\int_{-\infty}^{t}K(t-s)\,u(s)\,ds$. We derive basic properties of such dynamics, discuss conditions for unitarity, give an equivalent time-local description for exponential kernels, and show explicitly how hysteresis arises in the response of a driven qubit.

Building upon our recently established correspondence between quantum cosmology and the hydrogen atom [1], we investigate the specific sector of a negative cosmological constant ($Λ< 0$) in a flat FLRW universe with dust. While the positive $Λ$ sector [1] yields a continuous spectrum and a single bounce, we show here that the negative $Λ$ sector leads to a discrete spectrum of energy eigenvalues, effectively quantizing the cosmological constant. Within this dual description, the operator-ordering ambiguity parameter appears as the azimuthal quantum number of the hydrogen atom. A skewed Bohr correspondence emerges for the bound states, matching classical evolution at large volumes but deviating near the bounce. By constructing wave packets from these bound states, we demonstrate that the classical Big Bang and Big Crunch singularities are resolved, and the universe oscillates between quantum bounces and classical turnaround points. The expectation values of the observables indicate a cyclic universe -- with vanishing Hubble parameter at turnarounds -- undergoing quantum bounces. This exactly solvable model offers a tractable setting to explore quantum gravitational effects in cosmology. We analyze the properties of this cyclic universe, contrasting its bound-state dynamics with the scattering states of the de Sitter case.

Gaussian Boson Sampling (GBS) is a quantum computational model that leverages linear optics to solve sampling problems believed to be classically intractable. Recent experimental breakthroughs have demonstrated quantum advantage using GBS, motivating its application to real-world combinatorial optimization problems. In this work, we reformulate the graph coloring problem as an integer programming problem using the independent set formulation. This enables the use of GBS to identify cliques in the complement graph, which correspond to independent sets in the original graph. Our method is benchmarked against classical heuristics and exact algorithms on two sets of instances: Erdős-Rényi random graphs and graphs derived from a smart-charging use case. The results demonstrate that GBS can provide competitive solutions, highlighting its potential as a quantum-enhanced heuristic for graph-based optimization.

Quantum computing is entering a period in which progress will be shaped as much by advances in computer science as by improvements in hardware. The central thesis of this report is that early fault-tolerant quantum computing shifts many of the primary bottlenecks from device physics alone to computer-science-driven system design, integration, and evaluation. While large-scale, fully fault-tolerant quantum computers remain a long-term objective, near- and medium-term systems will support early fault-tolerant computation with small numbers of logical qubits and tight constraints on error rates, connectivity, latency, and classical control. How effectively such systems can be used will depend on advances across algorithms, error correction, software, and architecture. This report identifies key research challenges for computer scientists and organizes them around these four areas, each centered on a fundamental question.

We prove that almost-linear-time high-fidelity state transfer is achievable in a quantum spin chain using loop weights at the second and second-to-last nodes. We provide specific parameter values, and using a careful analysis of the eigenvectors we make precise quantitative estimates of the transfer time and strength.

Although classical nonlinear dynamics suggests that sufficiently strong nonlinearity can sustain oscillations, quantization of such model typically yields a time-independent steady state that respects time-translation symmetry and thus precludes time-crystal behavior. We identify dephasing as the primary mechanism enforcing this symmetry, which can be suppressed by intercomponent phase correlations. Consequently, a sufficient condition for realizing a continuous time crystal is a nonlinear quantum self-sustaining system exhibiting quantum synchronization among its constituents. As a concrete example, we demonstrate spontaneous oscillations in a synchronized array of van der Pol oscillators, corroborated by both semiclassical dynamics and the quantum Liouville spectrum. These results reduce the identification of time crystals in many-body systems to the evaluation of only two-body correlations and provide a framework for classifying uncorrelated time crystals as trivial.

Contextuality is a central feature distinguishing quantum from classical probability theories, yet its operational meaning remains subject to interpretation. We reconsider contextuality from an information-theoretic perspective, focusing on operational models constrained to maintain a single internal state with fixed semantics across multiple contexts. Under this constraint, we show that contextual statistics certify an unavoidable obstruction to classical probabilistic descriptions. Specifically, any classical model that reproduces such statistics must either embed contextual dependence into the internal state or introduce additional external labels carrying nonzero information. This result identifies contextuality as a witness of irreducible information cost in classical representations, rather than as a purely nonclassical anomaly. From this viewpoint, quantum probability emerges as a canonical framework that accommodates contextual operations without requiring explicit contextual encoding.

We study a string-inspired dilaton cosmology in the Damour--Polyakov (DP) regime using dynamical-systems methods, aiming to make explicit how cosmological relaxation controls deviations from the equivalence principle. Working in the Einstein frame, we consider a spatially flat FLRW universe filled with pressureless matter and a universally coupled dilaton. Expanding the conformal coupling function and the scalar potential around the least-coupling point, we obtain a closed and self-consistent autonomous system governing the late-time evolution of the scalar-matter sector. The resulting phase space contains a stable fixed point associated with least coupling, approached only asymptotically along cosmological trajectories. Therefore, at any finite epoch the solution typically retains a small displacement from the fixed point. In the DP regime this finite-epoch displacement sets the ambient coupling, and thus determines the magnitude of fifth-force effects and deviations from the equivalence principle in the nonrelativistic limit. By linearising the system around a finite-epoch reference state, we show that the damping of the displacement is controlled by the Jacobian eigenvalues of the DP fixed point. This yields a direct dynamical estimate of how rapidly deviations from the equivalence principle are reduced during cosmological evolution. The mechanism is global and cosmological in origin, and is conceptually distinct from local environmental screening as in chameleon or symmetron scenarios. Overall, our results illustrate how phase-space techniques provide a clear bridge between cosmological dynamics and weak-field departures from General Relativity.

The minimum amount of thermodynamic work required in order to implement a quantum computation or a quantum state transformation can be quantified using frameworks based on the resource theory of thermodynamics, deeply rooted in the works of Landauer and Bennett. For instance, the work we need to invest in order to implement $n$ independent and identically distributed (i.i.d.) copies of a quantum channel is quantified by the thermodynamic capacity of the channel when we require the implementation's accuracy to be guaranteed in diamond norm over the $n$-system input. Recent work showed that work extraction can be implemented universally, meaning the same implementation works for a large class of input states, while achieving a variable work cost that is optimal for each individual i.i.d. input state. Here, we revisit some techniques leading to derivation of the thermodynamic capacity, and leverage them to construct a thermodynamic implementation of $n$ i.i.d. copies of any time-covariant quantum channel, up to some process decoherence that is necessary because the implementation reveals the amount of consumed work. The protocol uses so-called thermal operations and achieves the optimal per-input work cost for any i.i.d. input state; it relies on the conditional erasure protocol in our earlier work, adjusted to yield variable work. We discuss the effect of the work-cost decoherence. While it can significantly corrupt the correlations between the output state and any reference system, we show that for any time-covariant i.i.d. input state, the state on the output system faithfully reproduces that of the desired process to be implemented. As an immediate consequence of our results, we recover recent results for optimal work extraction from i.i.d. states up to the error scaling and implementation specifics, and propose an optimal preparation protocol for time-covariant i.i.d. states.

We investigate the arrow of time problem in the context of gravitational collapse of radiating stars in higher dimensions for both neutral and charged matter. The interior spacetime is described by a shear-free spherically symmetric metric filled with a dissipative fluid. The exterior spacetime of the radiating star is taken as the higher dimensional Vaidya metric. We establish that the arrow of time associated with gravitational entropy is opposite to the thermodynamic arrow of time for all dimensions. The physical consequences of our results are considered. Our result conforms with previous studies on shear-free spherical collapse, which suggests, avoidance of the naked singularity as the end state results in a wrong arrow of time, indicating a fundamental problem with the local application of the Weyl curvature hypothesis.

We propose and systematically analyze a practical scheme for implementing a one-dimensional non-Hermitian Su-Schrieffer-Heeger model using individually addressable Rydberg atom arrays. Our setup consists of an atomic chain with three-atom unit cells, in which a synthetic gauge field is generated by applying multi-color laser fields. By engineering fast dissipative channels for one auxiliary atom in each unit cell, the adiabatic elimination effectively gives rise to a non-Hermitian skin effect. We examine how fluctuations in the experimental parameters influence both the skin effect and the topological invariant under open and periodic boundary conditions in real space and find that both features remain highly robust. This work establishes a versatile, controllable, and programmable open-system quantum simulator with neutral atoms, providing a clear route for exploring rich non-Hermitian topological phenomena.

Superconducting qubits based on tantalum (Ta) thin films have demonstrated the highest-performing microwave resonators and qubits. This makes Ta an attractive material for superconducting quantum computing applications, but, so far, direct deposition has largely relied on high substrate temperatures exceeding \SI{400}{\celsius} to achieve the body-centered cubic phase, BCC (\textalpha-Ta). This leads to compatibility issues for scalable fabrication leveraging standard semiconductor fabrication lines. Here, we show that changing the sputter gas from argon (Ar) to krypton (Kr) promotes BCC Ta synthesis on silicon (Si) at temperatures as low as \SI{200}{\celsius}, providing a wide process window compatible with back-end-of-the-line fabrication standards. Furthermore, we find these films to have substantially higher electronic conductivity, consistent with clean-limit superconductivity. We validated the microwave performance through coplanar waveguide resonator measurements, finding that films deposited at \SI{250}{\celsius} and \SI{350}{\celsius} exhibit a tight performance distribution at the state of the art. Higher temperature-grown films exhibit higher losses, in correlation with the degree of Ta/Si intermixing revealed by cross-sectional transmission electron microscopy. Finally, with these films, we demonstrate transmon qubits with a relatively compact, \SI{20}{\micro\meter} capacitor gap, achieving a median quality factor up to 14 million.

This paper introduces and rigorously analyzes a new class of one-dimensional discrete-time quantum walks whose dynamics are governed by a parametrized family of extended CMV matrices. The model generalizes the unitary almost Mathieu operator (UAMO) and exhibits a richer spectral phase diagram, closely resembling the extended Harper's model. It provides the first example of a solvable quasi-periodic quantum walk that exhibits a stable region of purely singular continuous spectrum.

In a seminal work [PRL85.4972], Walgate, Short, Hardy, and Vedral prove in finite dimensions that for every pair of pure multipartite orthogonal quantum states, there exists a one-way local operations and classical communication (LOCC) protocol that perfectly distinguishes the pair. We extend this result to infinite dimensions with a simpler proof. For states on $\mathbb{C}^{d_A \times d_A} \otimes \mathbb{C}^{d_B \times d_B}$, we strengthen this existence result by constructing an $O(d_A^2 d_B^2)$-time algorithm that specifies such a perfect one-way LOCC protocol. Finally, we establish the equivalence between Walgate et al.'s result and the fact that the one-shot environment-assisted classical capacity of every quantum channel is at least 1 bit per channel use, thereby clarifying the literature on these notions. At the core of all of these results is the fact that every operator with vanishing trace admits a basis where its diagonal entries are all zero.

Despite rapid advances in quantum hardware, noise remains a central obstacle to deploying quantum algorithms on near-term devices. In particular, random coherent errors that accumulate during circuit execution constitute a dominant and fundamentally challenging noise source. We introduce a noise-resilient framework for Quantum Signal Processing (QSP) that mitigates such coherent errors without increasing circuit depth or ancillary qubit requirements. Our approach uses ensembles of noisy QSP circuits combined with measurement-level averaging to suppress random phase errors in Z rotations. Building on this framework, we develop robust QSP algorithms for implementing polynomial functions of Hermitian matrices and for estimating observables, with applications to Hamiltonian simulation, quantum linear systems, and ground-state preparation. We analyze the trade-off between approximation error and hardware noise, which is essential for practical implementation under the stringent depth and coherence constraints of current quantum hardware. Our results establish a practical pathway for integrating error mitigation seamlessly into algorithmic design, advancing the development of robust quantum computing, and enabling the discovery of scientific applications with near- and mid-term quantum devices.

In this work we use the recently introduced concept of self-interacting dark matter with scale-dependent equation of state, and we provide an analytic model of dark matter that can produce viable rotation curves even for low-surface-brightness galaxies, irregular galaxies, low-luminosity spirals and dwarf galaxies, all known to challenge the cold dark matter description. The radius dependent effective equation of state of the self-interacting dark matter model we shall introduce is assumed to be an isothermal equation of state of the form $P(r)=K(r)\left(\frac{ρ(r)}{ρ_{\star}}\right)$, where the energy density will have the form $ρ(r)=\frac{ρ_0}{\left( 1+\frac{r^2}{α^2}\right)^{5/2}}$, while the entropy function $K(r)$ is $K(r)=\frac{K_0}{\left( 1+\frac{r^2}{α^2}\right)^{1/2}}$. The resulting model is confronted in detail with the SPARC galaxy data and 175 galaxies are used and tested. It proves that the analytic model can successfully produce the rotation curves of 116 galaxies, most of which are small mass spirals, irregular galaxies, low-surface-brightness and low-luminosity spirals and dwarf galaxies. On the other hand, 59 galaxies cannot be successfully described by our analytic model. We tested statistically the correlation between the parameter $K_0$ of the entropy function corresponding to the viable galaxies, and the flat rotation velocity $V_{flat}$ and the maximum rotation velocity $V_{max}$ of the galaxies from the SPARC data. We also examined the baryon mass $M_b$-$K_0$ relation and the luminosity $L$-$K_0$ relation. We have been able to produce the baryonic Tully-Fisher relation for the viable galaxies, directly from the correlation $K_0$-$M_b$ and $K_0$-$V_{flat}$, with the resulting relation being $M_b\sim V_{flat}^{4.026 \pm 0.371}$, however we failed to produce the canonical Tully-Fisher relation.

We discuss the consequence of excluding allowed RF-transition between substates of a field-dressed Rydberg manifold when predicting the spectrum that will be observed if the dressed system is probed in an optical EIT scheme.

It is known that if the standard slow-roll inflation is followed by a power-law inflationary regime, then the trans-Planckian modes may be safely be contained in the Hubble horizon and never exit it during inflation. In this work we investigate how to realize a smooth transition between a slow-roll and a power-law inflationary regime in the context of single scalar field inflation. As we show it is possible to realize such a smooth transition by generalizing the kinetic energy of single scalar field in the form $\dotφ^2=β(φ)V(φ)$, where $β(φ)$ is some appropriate function of the scalar field. Using two distinct approaches we show that it is possible to realize a smooth transition from a slow-roll to a power-law inflationary regime, and the two approaches produce identical results regarding the slow-roll regime. Also we show that the slow-roll regime is quite short, about $N\sim 30$ $e$-foldings, with the flatness and horizon problems being solved with the synergistic effect of the two inflationary patches. The slow-roll era is found to be compatible with the Atacama Cosmology Telescope data.

The ACT data have severely constrained the single scalar field models. Known models of inflation, like the Starobinsky model, the Higgs model and the $a$-attractors are at least $2σ$ off the ACT data. In this work we aim to provide a top-to-bottom approach in single scalar field inflationary cosmology compatible with the ACT data. Specifically, inspired by the fact that the Starobinsky model, the Higgs model and the $a$-attractors, all being plateau potentials, result to the same attractor relation between the spectral index of scalar perturbations and the tensor-to-scalar ratio, which is of the form $n_s(r)=1-αr^{1/2}$, in this work we seek for attractors of the form $n_s(r)=f(r)$ that may lead to ACT-compatible inflation. Specifically, we fix the function $f(r)$ to have a specific desirable form and then solve the differential equation $n_s(r)=f(r)$ to find the potential which results to the relation $n_s(r)=f(r)$. We discovered analytically three classes of potentials which are variants of the general form $n_s(r)=γ\pm βr \pm r^{1/2}$ and all these models are found to be compatible with the ACT data.

Quantum transport on structured networks is strongly influenced by interference effects, which can dramatically modify how information propagates through a system. We develop a quantum-information-theoretic framework for scattering on graphs in which a full network of connected scattering sites is treated as a quantum channel linking designated input and output ports. Using the Redheffer star product to construct global scattering matrices from local ones, we identify resonant concatenation, a nonlinear composition rule generated by internal back-reflections. In contrast to ordinary channel concatenation, resonant concatenation can suppress noise and even produce super-activation of the quantum capacity, yielding positive capacity in configurations where each constituent channel individually has zero capacity. We illustrate these effects through models exhibiting resonant-tunneling-enhanced transport. Our approach provides a general methodology for analyzing coherent information flow in quantum graphs, with relevance for quantum communication, control, and simulation in structured environments.

Bound states, known as particles tied together and moving as a whole, are profound correlated effects induced by particle-particle interactions. While dimer-monomer bound states are manifested as a single particle attached to dimer bound pair, it is still unclear about quantum walks and Bloch oscillations of dimer-monomer bound states. Here, we revisit three-particle bound states in the Bose-Hubbard model and find that interaction-induced impurities adjacent to bound pair and boundaries cause two kinds of bound states: one is dimer-monomer bound state and the other is bound edge states. In quantum walks, the spread velocity of dimer-monomer bound state is determined by the maximal group velocity of their energy band, which is much smaller than that in the single-particle case. In Bloch oscillations, the period of dimer-monomer bound states is one third of that in the single-particle case. Our works provide new insights to the collective dynamics of three-particle bound states.

The quantum approximate optimization algorithm (QAOA) is a hybrid quantum-classical algorithm for solving combinatorial optimization problems. Multi-angle QAOA (MA-QAOA), which assigns independent parameters to each Hamiltonian operator term, achieves superior approximation performance even with fewer layers than standard QAOA. Unfortunately, this increased expressibility can raise the classical computational cost due to a greater number of parameters. The recently proposed Layerwise MA-QAOA (LMA-QAOA) reduces this overhead by training one layer at a time, but it may suffer from obtaining the precise solution due to the previously fixed parameters. This work addresses two questions for efficient MA-QAOA training: (i) What is the optimal granularity for parameter updates per epoch, and (ii) How can we get precise final cost function results while only partially updating the parameters per epoch? Despite the benefit of reducing the parameters that update per epoch can reduce the classical computation overhead, too fine or coarse a granularity of Hamiltonian update can degrade the MA-QAOA training efficiency. We find that optimizing one complete layer per epoch is an efficient granularity. Moreover, selectively retraining each layer by tracking gradient variations can achieve a final cost function equivalent to the standard MA-QAOA while lowering the parameter update overhead. Based on these insights, we propose Orbit-QAOA, which cyclically revisits layers and selectively freezes stabilized parameters. Across diverse graph benchmarks, Orbit-QAOA reduces training steps by up to 81.8%, reduces approximation ratio error by up to 72x compared to the unified stop condition-applied enhanced LMA-QAOA, and achieves equivalent approximation performance compared to the standard MA-QAOA.

Quantum technologies promise secure communication networks and powerful new forms of information processing, but building these systems at scale remains a major challenge. Diamond is an especially attractive material for quantum devices because it can host atomic-scale defects that emit single photons and store quantum information with exceptional stability. However, fabricating the optical structures needed to control light in diamond typically relies on slow, bespoke processes that are difficult to scale. In this work, we introduce a manufacturing approach that brings diamond quantum photonics closer to industrial production. Instead of sequentially defining each device by lithography written directly on diamond, we fabricate high-precision silicon masks using commercial semiconductor foundries and transfer them onto diamond via microtransfer printing. These masks define large arrays of nanoscale optical structures, shifting the most demanding pattern-definition steps away from the diamond substrate, improving uniformity, yield, and throughput. Using this method, we demonstrate hundreds of diamond "quantum microchiplets" with improved optical performance and controlled interaction with quantum emitters. The chiplet format allows defective devices to be replaced and enables integration with existing photonic and electronic circuits. Our results show that high-quality diamond quantum devices can be produced using scalable, foundry-compatible techniques. This approach provides a practical pathway toward large-scale quantum photonic systems and hybrid quantum-classical technologies built on established semiconductor manufacturing infrastructure.

We present a novel quantum circuit extraction scheme that tightly integrates graph-like ZX diagrams with hardware-adaptive routing. The method utilizes the degrees of freedom during the conversion from a ZX diagram to a quantum circuit (extraction). It alternates between generating multiple extraction options and evaluating them based on hardware constraints, allowing the routing algorithm to inform and guide the extraction process. This feedback loop extends existing graph-like ZX extraction and supports modular integration of different extraction algorithms, routing strategies, and target hardware, making it a versatile building block during compilation. To perform numerical evaluations, a reference instance of the scheme is implemented with SWAP-based routing for neutral atom hardware and evaluated using various benchmark collections on small-to mid-scale circuits. The reference code is available as open-source, allowing fast integration of other extraction and/or routing tools to stimulate further research and foster improvements of the proposed scheme.

We investigate a static, spherically symmetric black hole arising in Einstein gravity coupled to a Kalb-Ramond field and ModMax nonlinear electrodynamics, both of which are independently well motivated extensions of standard electrovacuum gravity. The solution depends, beyond mass and charge, on a Lorentz-violating parameter, a ModMax deformation parameter, and a discrete branch selector $ζ=\pm1$. We show that the ordinary branch admits extremal and non-extremal configurations, while the phantom branch generically supports a single-horizon geometry. Black-hole thermodynamics is analyzed within the Tsallis non-extensive framework, revealing branch-dependent stability and Joule-Thomson behavior. Weak gravitational lensing, photon propagation in plasma, and tidal forces are then studied, revealing clear optical and strong-field signatures that distinguish the two branches. In particular, the ordinary branch exhibits finite-radius tidal inversion, absent in the phantom sector. Our results demonstrate that the combined Kalb-Ramond and ModMax effects lead to a rich and observationally distinguishable black-hole phenomenology.

Explicit breaking of diffeomorphism symmetry with nondynamical background fields in gravitational theories can lead to inconsistencies between the equations of motion and the underlying pseudo-Riemannian geometry. These theories produce a constraint equation that follows from the modified Einstein's equations and the contracted Bianchi identity which has no simple solution and raises questions about the consistency of backgrounds structures in extensions to general relativity. In contrast, spontaneous symmetry breaking where fields acquire a vacuum expectation value are shown to avoid these problems. Nevertheless, there are some approaches that consistently implement explicit diffeomorphism breaking by: i) introducing geometrical restrictions which may dynamically restore diffeomorphism invariance, ii) using the Stückelberg formalism where auxiliary scalar fields can carry degrees of freedom associated to spontaneously broken symmetries, and more recently by iii) exploiting the isometries of a given gravitational configuration in the presence of a background; which ultimately requires the background fields to be Lie-dragged along the Killing vectors. In this work, we further develop and apply the latter approach focusing on a concrete example involving the Gödel metric and the minimal gravitational Standard-Model Extension. We show that the resulting dynamics is fully consistent with the constraint equation and produce new Noether's identities. Furthermore we investigate causality violations by analyzing the critical radius, which is found to depend explicitly on the background fields. This dependence allows for the emergence of both causal and non-causal regions in spacetime.

Optically addressable spins at material surfaces have represented a long-standing ambition in quantum sensing, providing atomic resolution and quantum-limited sensitivity. However, they are constrained by a finite depth at which the quantum spins can be stabilized. Here, we demonstrate a hybrid molecular-2D architecture that realizes quantum spin sensors directly on top of the surface. By anchoring spin-active molecules onto hexagonal boron nitride (hBN), we eliminate the depth of the quantum sensor while also exhibiting robust spin properties from 4~K to room temperature (RT). The Hahn-echo spin coherence time exceeds \(T_2 = 3.4~\upmu\text{s}\) at 4~K, outperforming values in bulk organic crystals and overturning the prevailing expectation that spin inevitably deteriorates upon approaching the surface. By chemically tuning the molecule through deuteration, \(T_2\) improves by more than 10-fold, and under dynamic decoupling, coherence is prolonged to the intrinsic lifetime limit, exceeding 300~\(\upmu\text{s}\). Proximal proton spins and the magnetic response of two-dimensional magnets beneath the hBN layer have been detected at RT. These molecular spins form surface quantum sensors with long coherence, optical addressability, and interfacial versatility, enabling a scalable, adaptable architecture beyond what conventional solid-state platforms offer.

We study the formation, stability, and decay of macrovortices in a rotating Bose gas confined by a Mexican-hat potential with a multiconfigurational ansatz. By systematically including correlations beyond the mean-field level, we map the equilibrium phase diagram and identify regimes of coexistence between vortex lattices and multiply charge central vortices. Quench dynamics reveals that macrovortices are robust under changes in rotation or interaction strength, sustaining clean monopole oscillations with well-separated, vorticity-dependent breathing frequencies. In contrast, trap quenches trigger a universal decay process mediated by vortex-phonon coupling, in which rotational energy is progressively transferred to compressible modes until the macrovortex splits into singly quantized vortices. Our results demonstrate that macrovortex lifetimes and decay pathways can be tuned by trap confinement, providing experimentally accessible signatures of vortex-phonon interactions and collective energy transfer in correlated quantum fluids.

We develop a reinforcement learning algorithm to study the holographic entropy cone. Given a target entropy vector, our algorithm searches for a graph realization whose min-cut entropies match the target vector. If the target vector does not admit such a graph realization, it must lie outside the cone, in which case the algorithm finds a graph whose corresponding entropy vector most nearly approximates the target and allows us to probe the location of the facets. For the $\sf N=3$ cone, we confirm that our algorithm successfully rediscovers monogamy of mutual information beginning with a target vector outside the holographic entropy cone. We then apply the algorithm to the $\sf N=6$ cone, analyzing the 6 "mystery" extreme rays of the subadditivity cone from arXiv:2412.15364 that satisfy all known holographic entropy inequalities yet lacked graph realizations. We found realizations for 3 of them, proving they are genuine extreme rays of the holographic entropy cone, while providing evidence that the remaining 3 are not realizable, implying unknown holographic inequalities exist for $\sf N=6$.

We present a time-domain iteration scheme for solving the Dynamical Mean-Field Theory (DMFT) self-consistent equations using retarded Green's functions in real time. Unlike conventional DMFT approaches that operate in imaginary time or frequency space, our scheme operates directly with real-time quantities. This makes it particularly suitable for near-term quantum computing hardware with limited Hilbert spaces, where real-time propagation can be efficiently implemented via Trotterization or variational quantum algorithms. We map the effective impurity problem to a finite one-dimensional chain with a small number of bath sites, solved via exact diagonalization as a proof-of-concept. The hybridization function is iteratively updated through time-domain fitting until self-consistency. We demonstrate stable convergence across a wide range of interaction strengths for the half-filled Hubbard model on a Bethe lattice, successfully capturing the metal-to-insulator transition. Despite using limited time resolution and a minimal bath discretization, the spectral functions clearly exhibit the emergence of Hubbard bands and the suppression of spectral weight at the Fermi level as interaction strength increases. This overcomes major limitations of two-site DMFT approximations by delivering detailed spectral features while preserving efficiency and compatibility with quantum computing platforms through real-time dynamics.

Distinct monitoring or intervention schemes can produce different conditioned stochastic quantum trajectories while sharing the same unconditional (ensemble-averaged) dynamics. This is the essence of unravelings of a given Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) master equation: any trajectory-ensemble average of a function that is linear in the conditional state is completely determined by the unconditional density matrix, whereas applying a nonlinear function before averaging can yield unraveling-dependent results beyond the average evolution. A paradigmatic example is resonance fluorescence, where direct photodetection (jump/Poisson) and homodyne or heterodyne detection (diffusive/Wiener) define inequivalent unravelings of the same GKSL dynamics. In earlier work, we showed that nonlinear trajectory averages can distinguish such unravelings, but observing the effect in that optical setting requires demanding experimental precision. Here we translate the same idea to a digital setting by introducing synthetic unravelings implemented as quantum circuits acting on one and two qubits. We design two unravelings - a projective measurement unraveling and a random-unitary "kick" unraveling - that share the same ensemble-averaged evolution while yielding different nonlinear conditional-state statistics. We implement the protocols on superconducting-qubit hardware provided by IBM Quantum to access trajectory-level information. We show that the variance across trajectories and the ensemble-averaged von Neumann entropy distinguish the unravelings in both theory and experiment, while the unconditional state and the ensemble-averaged expectation values that are linear in the state remain identical. Our results provide an accessible demonstration that quantum trajectories encode information about measurement backaction beyond what is fixed by the unconditional dynamics.

Clusters of galaxies have been used to measure a subtle effect predicted by Einstein: gravitational redshift. This signal encodes pristine information about our Universe, since it is sensitive to the depth of the clusters' gravitational potential wells. In this work, we show how gravitational redshift can be used to test a fundamental physical principle: the weak equivalence principle. This principle stipulates that all matter falls in the same way in a gravitational potential, regardless of its nature. By comparing the amplitude of the gravitational redshift signal with the velocity dispersion in galaxy clusters, we build a novel test of this principle targeted to the unknown dark matter. Our test is sensitive to any additional interaction that would alter the way dark matter falls in gravitational potentials, hence leading to a violation of the equivalence principle. We show that currently available data can constrain the presence of a fifth force in clusters at the level of 7-14%, while the newest surveys will reach a precision of a few percents. This demonstrates the crucial role played by galaxy clusters in testing fundamental properties of dark matter.

Recently, a novel dictionary relating solutions of the Einstein--Scalar--Maxwell theory to solutions of gauged Skyrme--Maxwell--Einstein models in $(3+1)$ dimensions has been established. This development provides a clear and systematic route to constructing new configurations with nontrivial Baryonic charge and magnetic field, leveraging the fact that the Einstein--Scalar--Maxwell system is considerably more tractable, thanks to powerful solution-generating techniques. In this work, we exploit the framework that allows compact sources dressed by scalar fields to be consistently embedded in external electromagnetic backgrounds, and we construct their dual counterparts carrying Baryonic charge in the Skyrme sector. The resulting Baryonic charge is expressed directly in terms of the parameters characterizing the seed spacetime, and a corresponding quantization condition involving these parameters is explicitly derived. Consequently, the mass and the Baryonic charge are not independent parameters. These results provide a closed analytic formula for the black hole mass parameter in terms of the Baryonic charge and the magnetic field. This relation between the mass parameter and the Baryonic charge is linear for large values of the mass, while significant deviations from linearity arise if the mass takes intermediate values.

Graph states provide a powerful framework for describing multipartite entanglement in quantum information science. In their standard formulation, graph states are generated by controlled-$Z$ interactions and naturally encode symmetric exchange properties. Here we establish a precise correspondence between graph topology and exchange symmetry by proving that a graph state is fully symmetric under particle permutations if and only if the underlying graph is complete. We then introduce a generalized graph-based construction using a non-commutative two-qudit gate, denoted $GR$, which requires directed edges and an explicit vertex ordering. We show that complete directed graphs endowed with appropriate orientations, for an odd number of qudits generate fully antisymmetric multipartite states. Together, these results provide a unified graph-theoretic description of bosonic and fermionic exchange symmetry based on graph completeness and edge orientation.

Low check weight is practically crucial code property for fault-tolerant quantum computing, which underlies the strong interest in quantum low-density parity-check (qLDPC) codes. Here, we explore the theory of weight-constrained stabilizer codes from various foundational perspectives including the complexity of computing code weight and the explicit boundary of feasible low-weight codes in both theoretical and practical settings. We first prove that calculating the optimal code weight is an $\mathsf{NP}$-hard problem, demonstrating the necessity of establishing bounds for weight that are analytical or efficiently computable. Then we systematically investigate the feasible code parameters with weight constraints. We provide various explicit analytical lower bounds and in particular completely characterize stabilizer codes with weight at most 3, showing that they have distance 2 and code rate at most 1/4. We also develop a powerful linear programming (LP) scheme for setting code parameter bounds with weight constraints, which yields exact optimal weight values for all code parameters with $n\leq 9$. We further refined this constraint from multiple perspectives by considering the generator weight distribution and overlap. In particular, we consider practical architectures and demonstrate how to apply our methods to e.g.~the IBM 127-qubit chip. Our study brings the weight as a crucial parameter into coding theory and provide guidance for code design and utility in practical scenarios.

Qubit shuttling has become an indispensable ingredient for scaling leading quantum computing platforms, including semiconductor spin, neutral-atom, and trapped-ion qubits, enabling both crosstalk reduction and tighter integration of control hardware. Cai et al. (2023) proposed a scalable architecture that employs short-range shuttling to realize effective three-dimensional connectivity on a strictly two-dimensional device. Building on recent advances in quantum error correction, we show that this architecture enables the native implementation of folded surface codes on 2D hardware, reducing the runtime of all single-qubit logical Clifford gates and logical CNOTs within subsets of qubits from $\mathcal{O}(d)$ in conventional surface code lattice surgery to constant time. We present explicit protocols for these operations and demonstrate that access to a transversal $S$ gate reduces the spacetime volume of 8T-to-CCZ magic-state distillation by more than an order of magnitude compared with standard 2D lattice surgery approaches. Finally, we introduce a new "virtual-stack" layout that more efficiently exploits the quasi-three-dimensional structure of the architecture, enabling efficient multilayer routing on these two-dimensional devices.

The minimum error probability for distinguishing between two quantum states is bounded by the Helstrom limit, derived under the assumption that measurement strategies are restricted to positive operator-valued measurements. We explore scenarios in which the error probability for discriminating two quantum states can be reduced below the Helstrom bound under some constrained access of resources, indicating the use of measurement operations that go beyond the standard positive operator-valued measurements framework. We refer to such measurements as non-positive operator-valued measurements. While existing literature often associates these measurements with initial entanglement between the system and an auxiliary, followed by joint projective measurement and discarding the auxiliary, we demonstrate that initial entanglement between system and auxiliary is not necessary for the emergence of such measurements in the context of state discrimination. Interestingly, even initial product states can give rise to effective non-positive measurements on the subsystem, and achieve sub-Helstrom discrimination error when discriminating quantum states of the subsystem.

In a recent preprint [arXiv:2601.14134v1], Rubin argues that the arrow of time originates from the monotonic growth of the volume of extra dimensions. While the identification of a geometric origin for time's arrow is compelling in the case of brane-world models, we point out a possible tension between the proposed volume growth and the observational stability of the effective four-dimensional Newton's gravitational constant, G, that may arise in Kaluza-Klein (KK) theory. In standard KK approaches, such volume growth induces a time-variation of G that exceeds Big Bang Nucleosynthesis (BBN) and Lunar Laser Ranging (LLR) bounds by many orders of magnitude. To resolve this tension while preserving the author's key insight in the Kaluza-Klein case, we propose an extension: the "shape-dynamic arrow of time". By utilizing the scale-invariant monotonicity of Perelman's nu-entropy under normalized Ricci flow, we demonstrate how an arrow of time can emerge from the geometric smoothing of extra dimensions at fixed volume, thereby satisfying observational constraints on fundamental constants.

Broadband detection of a weak oscillatory field with unknown carrier frequency underlies magnetometry, axion searches and gravitational-wave sensing. We show that the Grover-like integration-time lower bound for this task is a geometric corollary an upper bound on the integrated quantum Fisher information, a metrological constraint. We further give an all-analog multi-resonant protocol based on a randomized Su-Schrieffer-Heeger control Hamiltonian and an m-register GHZ probe and verify near-optimal scaling through simulation.

The evolution of large-scale structure within the standard model of cosmology is well posed only up to the onset of shell crossing, where particle trajectories appear to intersect. Beyond this point, the evolution equations become non-predictive and perturbative approaches break down. We show that in General Relativity, a matter horizon forms before caustics develop for a well-defined initial over-density on an expanding FLRW spacetime. The matter horizon was first identified by Ellis and Stoeger in 2010 as a dynamical causal boundary that encloses a sub-region of spacetime where structure formation actually takes place. We construct a multi-scale hierarchical framework for the propagation of geodesic congruences that avoids the shell-crossing singularity by cutting the spacetime at the matter horizon and glueing to another spacetime with opposite orientation. We identify a relationship between the multi-scale hierarchical framework and the cosmological zoom-in N-body simulation approach, and relate the local sub-region that decouples from the Hubble flow to the region of interest in cosmological zoom-in N-body simulations. Most importantly, the multi-scale hierarchical framework provides a more robust way of implementing boundary conditions, which could benefit cosmological zoom-in N-body simulation approaches.

We present a new geometric perspective on quantum error correction based on spectral triples in noncommutative geometry. In this approach, quantum error correcting codes are reformulated as low energy spectral projections of Dirac type operators that separate global logical degrees of freedom from local, correctable errors. Locality, code distance, and the Knill Laflamme condition acquire a unified spectral and geometric interpretation in terms of the induced metric and spectrum of the Dirac operator. Within this framework, a wide range of known error correcting codes including classical linear codes, stabilizer codes, GKP type codes, and topological codes are recovered from a single construction. This demonstrates that classical and quantum codes can be organized within a common geometric language. A central advantage of the spectral triple perspective is that the performance of error correction can be directly related to spectral properties. We show that leakage out of the code space is controlled by the spectral gap of the Dirac operator, and that code preserving internal perturbations can systematically increase this gap without altering the encoded logical subspace. This yields a geometric mechanism for enhancing error correction thresholds, which we illustrate explicitly for a stabilizer code. We further interpret Berezin Toeplitz quantization as a mixed spectral code and briefly discuss implications for holographic quantum error correction. Overall, our results suggest that quantum error correction can be viewed as a universal low energy phenomenon governed by spectral geometry.

We treat inter-branch message transfer in a Wigner's-friend circuit as a practical benchmark for near-term superconducting quantum processors. Implementing Violaris' unitary message-transfer primitive, we compare performance across IBM Eagle, Nighthawk, and Heron (r2/r3) processors for message sizes up to $n=32$, without error mitigation. We study three message families -- sparse (one-hot), half-weight, and dense -- and measure conditional string success $p_{\mathrm{all}}=\Pr(P=μ\mid R=0)$, memory erasure after uncomputation, and correlation diagnostics (branch contrast and bitwise mutual information). The sparse family compiles to essentially constant two-qubit depth, yielding a depth-controlled probe of device noise: at $n=32$ we observe $p_{\mathrm{all}}$ spanning $\approx0.07$ to $\approx0.68$ across backends. In contrast, half and dense messages incur rapidly growing routing overhead, and transpiler-seed variability becomes a practical limitation near the coherence frontier. We further report an amplitude sweep (no-amplification test) and a divergence ``cousins'' sweep that quantifies degradation with branch-conditioned complexity. All data and figure-generation scripts are released.

Pulsar timing arrays' hint for a stochastic gravitational-wave background (SGWB) leverages the expectations of a future detection in the millihertz band, particularly with the LISA space mission. However, finding an SGWB with a single orbiting detector is challenging: It calls for cautious modelling of instrumental noise, which is also mainly stochastic. It was shown that agnostic noise reconstruction methods provide robustness in the detection process. We build on previous work to include more realistic instrumental simulations and additional degrees of freedom in the noise inference model and analyze the impact of LISA's sensitivity to SGWBs. Particularly, we model the two main types of noise sources with separate transfer functions and power spectral density spline fitting. We assess the detectability bounds and their dependence on the flexibility of the noise model and on the prior probability, allowing us to refine previously reported results.

Compiling quantum circuits into Clifford+$T$ gates is a central task for fault-tolerant quantum computing using stabilizer codes. In the near term, $T$ gates will dominate the cost of fault tolerant implementations, and any reduction in the number of such expensive gates could mean the difference between being able to run a circuit or not. While exact synthesis is exponentially hard in the number of qubits, local synthesis approaches are commonly used to compile large circuits by decomposing them into substructures. However, composing local methods leads to suboptimal compilations in key metrics such as $T$-count or circuit depth, and their performance strongly depends on circuit representation. In this work, we address this challenge by proposing \textsc{Q-PreSyn}, a strategy that, given a set of local edits preserving circuit equivalence, uses a RL agent to identify effective sequences of such actions and thereby obtain circuit representations that yield a reduced $T$-count upon synthesis. Experimental results of our proposed strategy, applied on top of well-known synthesis algorithms, show up to a $20\%$ reduction in $T$-count on circuits with up to 25 qubits, without introducing any additional approximation error prior to synthesis.

We develop and analyze a two-oscillator model for black hole evaporation in which an effective geometric degree of freedom and a representative Hawking radiation mode are described by coupled harmonic oscillators with opposite signs in their free Hamiltonians. The normal-mode structure is obtained analytically and the corresponding modal amplitudes determine the pattern of energy exchange between the two sectors. To bridge the discrete and semiclassical pictures, we introduce smooth envelope functions that provide a continuous effective description along the geometric variable. Numerical simulations in a truncated Fock space show that the two oscillators exchange quanta in an approximately out-of-phase manner, consistent with an effective conservation of $\langle n_x\rangle - \langle n_y\rangle$. The reduced entropy $S_x(t)$ exhibits periodic growth, indicating entanglement generation. These results demonstrate that even a minimal two-mode framework can capture key qualitative features of energy transfer and information flow during evaporation.

A single geometric invariant fixes the relative normalization and structure of gravity, Yang-Mills theory, and fermion kinetic terms -- including ghost freedom in the gravitational sector -- without tuning. Our results establish a minimal geometric route to unification that does not rely on extra dimensions or symmetry breaking by hand. Unlike previous gauge-gravity constructions, the relative normalizations and ghost freedom emerge from a single Clifford-algebraic invariant, without explicit symmetry breaking.

The applicability and performance of qubits dressed by classical fields are limited because their control protocols give rise to an undesired counter-rotating term (CRT). This in turn forces operation in a regime where a (dressed) rotating-wave approximation (RWA) is valid, thereby restricting key aspects of their operation. Here, using only a single coupling axis in the laboratory frame, we introduce a dressed-qubit control protocol that optimally removes the CRT, eliminating the need for the RWA and delivering substantial improvements in multiple performance metrics, including single-qubit gate speed, two-qubit gate fidelity, spectroscopic range, clock stability, and coherence preservation. In addition, we provide a general parameterization together with a Floquet-based coherence-time expression, which elucidates the protocol's working principles and lowers the barrier to adoption. Collectively, these advances position our scheme as the state-of-the-art strategy for qubit control, paving the way for a wider class of quantum technologies to be realized using dressed-qubit architectures.

We develop a theoretical framework for fast, robust and high-fidelity topological quantum state transfer in one-dimensional systems with long-range couplings, motivated by chains of Rydberg atoms with dipole-dipole interactions. Such long-range interactions naturally give rise to extended Su-Schrieffer-Heeger and Rice-Mele models supporting topologically protected edge states. We show that these edge states enable high-fidelity edge-to-edge excitation transfer using both time-independent protocols, based on coherent edge state dynamics, and time-dependent protocols, based on adiabatic modulation of system parameters. Long-range couplings play a central role by enhancing the relevant energy gaps, leading to a substantial improvement in transfer efficiency compared to nearest neighbour models. The resulting transfer is robust against positional disorder, reflecting its topological origin and highlighting the potential of long-range interacting platforms for reliable quantum state transfer.

Linear oscillators contribute to most branches of contemporary quantum science. They have already successfully served as quantum sensors and memories, found applications in quantum communication, and hold promise for cluster-state-based quantum computing. To master universal quantum processing with linear oscillators, an unconditional nonlinear operation is required. We propose such an operation using light-mediated interaction with another system that possesses a nonlinearity equivalent to more than a quadratic potential. Such a potential grants access to a nonlinear operation that can be broadcast to the target linear system. The nonlinear character of the operation can be verified by observing adequate negative values of the target system's Wigner function and the squeezing of the variance of a certain nonlinear combination of the quadratures below the thresholds attainable by Gaussian states. We explicitly evaluate an optically levitated mechanical oscillator as a flexible source of nonlinearity for a proof-of-principle demonstration of the nonlinearity broadcasting to linear systems, for example, mechanical oscillators or macroscopic atomic spin ensembles.

This chapter introduces the fundamental principles of gravitational wave detectors in a simple and comprehensive manner. Because these instruments aim for extremely high sensitivity, it is essential to understand their various noise sources, how such noise couples to the detector output, and the strategies used to mitigate them. These noises contributions are computed in the frame of the Virgo detector and a sensitivity curve is calculated. Although a simplified layout of a gravitational wave detector is considered, it takes into account the most dominant effects and yields in a sensitivity estimate close to the what is observed in real detectors.

We apply the wavelet formalism of quantum field theory to investigate nonperturbative dynamics within the Hamiltonian framework. In particular, we employ Daubechies wavelets in momentum space, whose basis functions are labeled by resolution and translation indices, providing a natural nonperturbative truncation of both infrared and ultraviolet truncation of quantum field theories. As an application, we compute the energy spectra of a free scalar field theory and the interacting $1+1$-dimensional $φ^4$ theory. This approach successfully reproduces the well-known strong-coupling phase transition in the $m^2 > 0$ regime. We find that the extracted critical coupling systematically converges toward its established value as the momentum resolution is increased, demonstrating the effectiveness of the wavelet-based Hamiltonian formulation for nonperturbative field-theoretic calculations.

We present a further theoretical study of the orbital $P$-wave bottom-charmed baryon within the framework of nonrelativistic QCD (NRQCD), considering both the direct photoproduction channel $γ+γ\rightarrow Ξ_{bc} +\bar{c}+\bar{b}$ and the resolved photoproduction channel $γ+g \rightarrow Ξ_{bc} +\bar{c}+\bar{b}$. At future linear colliders, ILC and CLIC, the initial photons can be emitted from the laser back-scattering (LBS) and then the parton gluon can be emitted from the photon. The formation of $Ξ_{bc}$ can be modeled in two-step: a compact diquark state $\langle bc\rangle[n]$ is formed first and subsequently captures a light quark from the vacuum to hadronize into the baryon $Ξ_{bc}$. The color and spin quantum number $[n]$ of $\langle bc\rangle$-diquark can be $[{}^3S_1]_{\bar{\textbf{3}}/\textbf{6}}$, $[{}^1S_0]_{\bar{\textbf{3}}/\textbf{6}}$, $[{}^1P_1]_{\bar{\textbf{3}}/\textbf{6}}$ or $[{}^3P_J]_{\bar{\textbf{3}}/\textbf{6}}$ with $J=0,1,2$. Based on the collision energies and design luminosity of ILC and CLIC, the cross sections, the differential distributions and the estimated produced events of $Ξ_{bc}$ baryon have been analyzed. The results show that the contribution of the orbital excited $P$-wave $Ξ_{bc}$ baryon can reach 7%-9% of the $S$-wave, providing a non-negligible contributions.

We show that, under very general hypotheses, topological quantum field theories (TQFTs) cannot detect homotopy spheres bounding parallelisable manifolds, such as Milnor's exotic 7-dimensional sphere. The result holds for a wide variety of target categories (or $(\infty,n)$-categories) and arbitrary tangential structures. An appendix contains results on the mapping class groups of (stably-) framed manifolds that may be of independent interest.

We investigate the ultraviolet completion of an $O(N)$ scalar field theory non-minimally coupled to gravity using the Wilsonian functional renormalization group in the proper-time formulation. Focusing on the spontaneously broken phase, we study the RG flow of the scalar potential and the non-minimal curvature coupling expanded around a running minimum. We identify two distinct classes of fixed-point solutions, one of which is ultraviolet attractive and characterized by a vanishing quartic coupling together with finite, interacting gravitational couplings. For a finite region of infrared initial conditions, the RG trajectories remain regular at all scales and approach this fixed point. This mechanism renders the theory asymptotically safe and leads to a flat scalar potential in the ultraviolet. We show that this mechanism is robust under changes of cutoff scheme and truncation, allowing the ultraviolet completion requirement to constrain the infrared values of the scalar couplings and the mass scale in the broken phase.

The $f_0(500)$ meson is often labeled a "non-Breit-Wigner" resonance due to the large background phase required for its description. By fitting elastic phase shifts of the $f_0(500)$, $ρ(770)$, and $Δ(1232)$, we extract the effective background phase $φ_\mathrm{B}$ and uncover an empirical regularity within our formalism: $φ_\mathrm{B}$ is equal to the angle subtended by the pole and the threshold $φ_0$. This implies the residue phase is geometrically constrained to be $θ\approx 2φ_0$. The anomalous nature of the $f_0(500)$ thus arises from the proximity of its pole to the $ππ$ threshold. A unified scaling relation confirms the $f_0(500)$ behaves as a standard Breit-Wigner resonance.

The first observation of the $χ_{c1}(3872)\rightarrow J\mskip -3mu/\mskip -2muψμ^+μ^-$ decay is reported using proton-proton collision data recorded with the LHCb detector corresponding to an integrated luminosity of $9fb^{-1}$. The decay mode is observed for the first time, with a significance of $6.5σ$. Its branching fraction is measured relative to the $χ_{c1}(3872)\rightarrow J\mskip -3mu/\mskip -2muψπ^+π^-$ decay mode \begin{align*} \frac{\cal{BF}(χ_{c1}(3872)\rightarrow J\mskip -3mu/\mskip -2muψμ^+μ^-)}{\cal{BF}(χ_{c1}(3872)\rightarrow J\mskip -3mu/\mskip -2muψπ^+π^-)} = \left(1.64\pm 0.32\pm 0.05\right)\times10^{-3}, \end{align*} where the first uncertainty includes both statistical contributions and systematic contributions which are uncorrelated between data-taking periods, and the second represents the systematic contributions that are correlated between data-taking periods.

A search for the lepton-flavour-violating decay $τ^-\to μ^-μ^+μ^-$ is carried out using data collected by the LHCb experiment between 2016 and 2018 in proton-proton collisions at the LHC at a centre-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 5.4 fb$^{-1}$. An upper limit of $1.9\,(2.3)\times 10^{-8}$ is set at the 90% (95%) confidence level on the branching fraction of the decay.

Axion, the goldstone boson arising from the spontaneous breaking of a global $U(1)$ Peccei-Quinn symmetry, provides a dynamical solution to the strong CP problem and is an excellent dark matter candidate. Various experiments are designed to search for the axion, however no confirmative signal has been observed. On the other hand, there are also hypothetical heavy particles in axion models, such as the heavy scalar $s$, which is the CP-even component of the complex scalar that carries $U(1)_{PQ}$ charge, and the vector-like heavy quark (VLQ) in the Kim-Shifman-Vainshtein-Zakharov~(KSVZ) axion model. Studying signals induced by them are helpful for axion searches. In this paper, we calculate the graviton bremsstrahlung energy spectrum arising from the decay of the heavy scalar or VLQ in the KSVZ model. The result shows that these heavy particles can emit ultrahigh-frequency gravitational waves (GWs), with the peak frequency depending on the model's parameter inputs. In addition, the graviton spectrum is distinguished from the thermal GW background at high frequencies if there is an early matter-dominated era induced by these heavy particles. Future measurements of ultrahigh-frequency GWs may provide indirect evidence for the KSVZ axion.

In this work, we perform a systematic investigation of the hidden-charm tetraquark states with $I^G(J^{PC})=1^+(1^{+-})$ within the hadronic molecular picture, placing particular emphasis on systems composed of an $S$-wave $(D, D^*)$ meson and a $P$-wave $(D_0^*(2300), D_1(2430), D_1(2420), D_2^*(2460))$ meson. Adopting the One-Boson Exchange potential, we solve the Schrödinger equation in momentum space via the Complex Scaling Method. A crucial feature of our approach is the rigorous treatment of the unstable nature of the $P$-wave constituents by incorporating three-body decay effects arising from self-energy corrections and the static limit approximation. Our results demonstrate that these three-body dynamics play a crucial role in determining the pole positions, specifically in reproducing the large decay widths observed experimentally. We identify several broad resonances in the $D^*\bar{D}_1(2420)$ and $D^*\bar{D}_2^*(2460)$ systems as candidates for the $Z_c(4430)$, while the significantly broader resonances in the $D\bar{D}_0^*(2300)$ and $D\bar{D}_1(2430)$ sectors are suggested as candidates for the $Z_c(4200)$. Focusing on the $D^*\bar{D}_2^*(2460)$ assignment as a specific case study, we further analyze the line shape of the $Z_c(4430)$ candidate using a Flatté-like parametrization with energy-dependent self-energy terms, providing predictions for its open-charm decay modes to guide future experimental searches.

This contribution presents a overview of a recent CMS-based determination of the effective leptonic weak mixing angle, $\sin^2θ^\ell_{\mathrm{eff}}$, derived from forward-backward asymmetry measurements in Drell-Yan events at 13 TeV. Although the CMS analysis achieved a major reduction in uncertainties, its overall precision is ultimately limited by residual parton distribution function (PDF) uncertainties. This proceeding highlights the role of complementary CMS observables, which probe distinct parton-density combinations and provide additional constraints beyond those obtained from the original asymmetry measurement alone. The improved analysis yields a substantially reduced total uncertainty, resulting in $\sin^2θ^\ell_{\mathrm{eff}} = 0.23156\pm0.00024$. This result is consistent with the Standard Model prediction and represents the highest precision achieved so far in an individual determination of this parameter.

Following a nonperturbative formulation of strong-field QED developed in our earlier works, we consider photon emission accompanying vacuum instability under the action of a quasi-constant strong electric field of finite duration T. We construct closed formulas for the total probabilities and study the photon emission accompanying an electron-positron pair creation from a vacuum. We establish the domain of the applicability of the locally constant field approximation (LCFA) for the photon emission. We study angular and polarization distribution of the emission as well as emission characteristics in a high-frequency approximations with respect of 1/T. The results presented in this work is suitable to a further development of the LCFA proposed in [Phys. Rev. D 95, 076013 (2017)].

We propose the $Λ_b\toφη_c Λ$ reaction to observe a $P_{cs}$ state around $4200$ MeV, predicted at lower masses than expected from comparison with the $P_c$ states, stemming as a consequence of the important role played by coupled channels in the $P_{cs}$ case, which does not appear in the $P_c$ case. That state decays to $η_c Λ$ with a width of about $200$ keV. The reaction is related to $Λ_b^0\toφD_s^- Λ_c^+$, which has already been observed. We predict a branching fraction for $Λ_b\toφP_{cs}(4200)$; $P_{cs}\toη_c Λ$ of the order of $10^{-5}$, which is within present capabilities of the LHCb collaboration. The observation of this state would bring valuable light on the nature of the $P_c$ and $P_{cs}$ states and the role played by coupled channels in hadron structure and hadron reactions.

We numerically investigate the matrix model of two-color one-flavor adjoint QCD (matrix-QCD$_{2,1}^{\text{adj}}$) in the weak coupling regime (small $g$) and in the chiral limit. The Yang-Mills potential has two distinct gauge invariant minima: one at $A_i=0$ and the other at $A_i = \frac{σ_i}{2g}$. We show that when the chiral chemical potential $c \leq \frac{3}{2}$, there is a quantum phase transition at $g_0^\ast \simeq 0.143$: for $g<g_0^\ast$, the ground state wavefunction is localized near $A_i=0$, while for $g>g_0^\ast$, the ground state is delocalized over the gauge configuration space. The transition between these two phases is singular, with the ground state at $g_0^\ast$ being distinctly different from that of $g_0^\ast \pm|ε|$. At $g_0^\ast$, we show that the square of the chromoelectric field vanishes, strongly suggesting that the system is in a ``dual superconductor" phase. Numerical evidence shows that the localization-delocalization phenomenon holds for the 1st and 2nd excited states as well, leading us to conjecture that there are an infinite number of isolated singular points $g_0^\ast> g_1^\ast>g_2^\ast> \cdots$ accumulating to $g=0$. For $c=1$, the model formally possesses $\mathcal{N}=1$ supersymmetry. We show that in the localized phase (i.e. for $g<g_0^\ast$) the supermultiplet structure is disrupted and SUSY is spontaneously broken.

The complex Langevin method (CLM) is a promising tool to address the sign problem in quantum field theories with complex actions. However, it can converge to incorrect results even when simulations appear stable, highlighting the need for robust diagnostics. Existing checks, such as monitoring drift distributions, are useful but indirect. We propose a complementary test based on the configurational temperature, constructed from the gradient and Hessian of the complex action. Unlike drift-based criteria, this estimator directly probes thermodynamic consistency and provides a physically interpretable cross-check of CLM dynamics. Using one-dimensional PT-symmetric models, we show that it reproduces the input temperature with high precision and sensitively detects algorithmic errors, step-size artifacts, and incomplete thermalization. While demonstrated in simple systems, the method extends naturally to higher-dimensional scalar and gauge theories. Since temperature is tied to the bare coupling in many lattice theories, configurational monitoring can also provide an independent check on coupling-dependent observables. Our results indicate that configurational temperature can enhance CLM reliability across a broad range of applications, including lattice QCD at finite density.

We present a systematic study of the Bjorken initial energy density in Xe-Xe collisions at $\sqrt{s_{NN}} = 5.44$ TeV, estimated using charged-particle multiplicity data and a generalized transverse overlap geometry applicable beyond the most central collisions. The dependence of the extracted energy density is examined by adopting both a constant formation time and a centrality-dependent formation time derived from Pb-Pb collisions at $\sqrt{s_{NN}} = 5.02$ TeV. Corresponding Bjorken energy density estimates for Pb-Pb collisions are also presented for comparison. Taking the Bjorken energy density and formation time as initial conditions, the subsequent longitudinal evolution of the quark-gluon plasma (QGP) formed in these collisions is studied. Both ideal and first-order viscous boost-invariant hydrodynamics are employed to assess the influence of dissipation. We observe that viscous effects slow the longitudinal expansion and lead to entropy production dominated by early-time dynamics. The lifetime of the QGP is observed to increase with centrality and is substantially enhanced by viscous effects. These effects are highly sensitive to the choice of formation time, particularly in peripheral collisions. A comparative analysis of Xe-Xe and Pb-Pb collisions demonstrates that the longitudinal evolution is primarily controlled by the initial energy density scale set by the Bjorken prescription. Consequently, when this scale is comparable, both systems exhibit nearly identical evolution patterns, while appreciable distinctions emerge in peripheral collisions due to system-size and geometric effects.

An overview of tau lepton physics is presented, using the tau lepton discovery and its precision measurements as examples to illustrate the importance of the energy region to be covered by the super tau-charm factory. By presenting the current measurement status of the major physics topics, the emphasize is put on pointing out a few open issues and possibilities for the super tau-charm factory.

This analysis utilizes simulated data privately generated based on the High Luminosity Large Hadron Collider (HL-LHC) configuration to investigate the angular distribution of high-mass dimuon pairs produced during the foreseen proton-proton collisions at a center-of-mass energy of 14 TeV. The study focuses on the cos$θ_{CS}$ variable, which is defined in the Collins-Soper frame. In the Standard Model, the production of high-mass dimuon pairs is primarily governed by the Drell-Yan process, which demonstrates a significant forward-backward asymmetry. However, scenarios beyond the Standard Model suggest different shapes for the cos$θ_{CS}$ distribution. By observing excess events not predicted by the Standard Model, the angular distribution can help differentiate among these alternative models. Furthermore, we used a simplified Einstein-Cartan gravity model to analyze the simulated data. This analysis established upper limits at the 95\% confidence level regarding the masses of various particles within the model, including a spin-2 dark neutral gauge boson and the torsion field.

The exact seesaw relation in a type-I seesaw framework puts constraints on the relations between active and sterile neutrino sectors in terms of their masses and mixing angles. In such a setup, we employ a model-independent approach to investigate the signature of sterile neutrinos in the half-life of the neutrinoless double beta ($0νββ$) decay process. In particular, we aim to study the contribution of sterile neutrinos in the mass range $\sim$~keV that is motivated by the dark matter constituent of the Universe. Further, the masses of the sterile neutrinos are determined by the active neutrino masses, mixing angles, and phases, and active-sterile mixing angles and $CP$-violating phases. The parameter space is constrained by the exact seesaw relation, thereby making the analysis constrained. After capturing the parameter space that can account for $\sim~$keV scale masses for the sterile neutrinos, we adopt the chiral effective field theory approach to calculate the half-life and effective mass in the $0νββ$ decay. As the study transitions from the TeV scale to scenarios involving at least one sterile neutrino in the keV mass range, it reveals a significant modification of the effective mass. In particular, the cancellation region associated with the normal mass hierarchy for TeV-scale sterile neutrinos no longer persists when a keV-scale sterile neutrino is introduced, resulting in a finite effective mass that future experiments can probe. Likewise, the involvement of keV-scale sterile neutrino in the inverted mass hierarchy case makes the band distorted and scattered points appear around the main band.

We propose a conjectural picture of baryon structure in which gluon knots, a type of topologically nontrivial configuration of color--magnetic monopole condensates, forms the dynamical core of the baryon. Within this framework, quarks interact with the gluon knot via abelian-dominated color-electric fields, which are squeezed into flux tubes by the dual Meissner effect, leading naturally to quark confinement. The color--magnetic fields associated with the gluon knot also induce a local chiral condensate, contributing to spontaneous chiral symmetry breaking and the baryon mass. Extending this conjecture to heavy-flavor mesons, we argue that stable flux tubes and gluon knots may also play a role in their internal structure, whereas light-flavor mesons are dominated by alternative confinement mechanisms. Our approach provides a unified, topologically motivated picture linking confinement, chiral symmetry breaking, and the internal dynamics of baryons and certain mesons, suggesting that gluon knots constitute fundamental infrared degrees of freedom of Yang--Mills theory.

We calculate Gaussian averages of arbitrary exponentials of the matrix variable $X$ with the help of superintegrability, which provides explicit expressions for Schur averages. As in the simpler cases the answer is expressed in terms of Laguerre polynomials, but in a somewhat sophisticated way. It involves triangular sum over partitions, with simple exponential factor and a complicated polynomial prefactor. Some ingredients of the formula are not found in full generality and there is still a room for further work.

In this talk, the electromagnetic transition form factors from the nucleon ground state to twelve $N^*$ and $Δ$ states are exhibited and discussed. Those results are extracted through a comprehensive coupled-channel approach -- the Juelich-Bonn-Washington model, with the center-of-mass energy ranging from $1.13$ GeV to $1.8$ GeV, and the photon virtuality up to $8$ GeV$^2$. The extraction is based on $10^5$ electroproduction data points of $πN$, $ηN$, and $KΛ$ channels, with additionally about $5\times 10^4$ data points in the hadronic sector as well as photoproductions as boundary conditions. The form factors are defined from the residues at the corresponding resonance poles in the multipole amplitudes. Uncertainties are also estimated by the exploration of the parameter space. The qualitative behavior of the transition form factors of $N(1440)$ and $Δ(1232)$ here are in agreement with the previous studies, while for the other states, there has not been literature results that are also defined at the resonance poles.

Tornadoes and other severe weather hazards affect thousands of people every year. Despite this, the details surrounding tornadic processes including formation, decay, and longevity are not well understood, partially due to limitations of available instrumentation. Measurements of atmospheric pressure within tornadic systems currently rely almost entirely on in-situ instrumentation, and no existing techniques can provide two-dimensional spatial information of the atmospheric density field. Atmospheric muons may hold a solution to this problem: muons are attenuated by matter, and tornadic storms are large regions of low atmospheric density, suggesting that tornadic storms induce a directional perturbation on the atmospheric muon flux. Measurements of this perturbation could then be used to infer the density field associated with severe weather. Simulations of these systems indicate that a robust measurement of the atmospheric density field would require a relatively large muon detector, however smaller detectors may be able to detect ambient muon flux perturbations if the storm is large and intense enough. This paper presents results from a pilot field study that measured the atmospheric muon flux near tornadic storms during May 2025, including directional measurements of the muon flux near tornadic mesocyclones and a measurement of the muon flux near the base of a forming tornado.

Reliable predictions of many high-energy strong interaction processes rely heavily on the non-perturbative parton fragmentation functions (FFs) extracted from existing experimental data. Conventional methods often require parameterized forms of FFs and additional scale evolution according to the Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) evolution equations. We introduce a novel approach to determining parton FFs using a Physics-Informed Neural Network (PINN). Unlike traditional methods, our approach does not require prior parameterized forms and directly integrates the DGLAP evolution equations into the neural network architecture, allowing the FFs to automatically satisfy these equations. We present new sets of parton FFs extracted from hadron spectra in electron-positron annihilation processes at next-to-leading order (NLO) in pQCD using this new technique. To validate our approach, we calculate charged hadron spectra in proton-(anti)proton collisions using the extracted FFs and demonstrate that the results align well with experimental data across a large range of colliding energies ($\sqrt{s}$ = 130, 200, 500, 630, 900, 1800, 2760, 5020, 5440, 7000 GeV). Our findings indicate that the PINN method not only simplifies the extraction process but also enhances the universal applicability of FFs across different energy scales. By eliminating the need for parameterized forms and additional DGLAP evolution, our approach represents a significant step forward toward fast and accurate extractions of non-perturbative quantities such as parton fragmentations functions and parton distribution functions.

The feasibility of studying, numerically, properties of infinite volume QCD-like theories in the large $N$ limit using coherent state variational methods is reassessed. An entirely new implementation of this approach is described, applicable to SU($N$) lattice gauge theories, with or without fundamental representation fermions, on cubic lattices of up to four dimensions. In addition to various test cases, initial results are presented for Hamiltonian Yang-Mills theory on an infinite two-dimensional spatial lattice.

The hadronic phase in heavy-ion collisions plays a crucial role in shaping the final-state hadron abundances. In this work, we study Au+Au collisions at $\sqrt{s_{\rm NN}}$ = 7.7-200 GeV using the Hadron Resonance Gas model in Partial Chemical Equilibrium (HRG-PCE). By fitting the yields of stable hadrons and short-lived resonances such as K$^*(892)^0$, we extract both chemical and kinetic freeze-out temperatures as functions of center-of-mass energy and centrality. The analysis, performed using the Thermal-FIST package, avoids assumptions about radial flow profile or freeze-out hypersurfaces. Furthermore, we estimate the baryon annihilation freeze-out temperature from the experimentally measured $\bar{\rm p}/$p ratio, using the HRG-PCE framework extended to include $B\bar{B} \leftrightarrow nπ$ reactions. The inferred annihilation freeze-out temperature lies between the chemical and kinetic freeze-out temperatures, suggesting that baryon annihilation remains active in the early hadronic phase but ceases prior to kinetic freeze-out. These results provide a consistent picture of the sequential decoupling of hadronic processes and demonstrate that inelastic hadronic interactions significantly influence the chemical composition of the system between chemical and kinetic freeze-outs at RHIC energies.

A search for heavy long-lived charged particles at the LHC is presented. Particles interacting with the CMS muon detector across several bunch crossings are searched for using a data sample of proton-proton collisions at $\sqrt{s}$ = 13.6 TeV collected with the CMS detector in 2024, corresponding to an integrated luminosity of 3.7 fb$^{-1}$. This is the first search relying on the novel level-1 trigger scouting data set collected without any trigger selection, allowing correlations between bunch crossings to be analyzed. The results are interpreted as upper limits on the cross sections of several benchmark processes with pair production of heavy long-lived charged particles. Upper limits on the fiducial cross section of a heavy long-lived charged particle with $p_\mathrm{T}$ $\gt$ 500 GeV and $\lvertη\rvert$ $\lt$ 0.83 are also set in different ranges of $β=v/c$. This analysis is a crucial proof of concept for the level-1 trigger data scouting system and complements existing searches for heavy long-lived charged particles by extending the sensitivity to lower $β$ values.

The production cross sections of the $Υ$(1S), $Υ$(2S), and $Υ$(3S) mesons are measured in proton-proton collisions at $\sqrt{s}$ = 13.6 TeV, using a data sample collected in 2022 by the CMS experiment and corresponding to an integrated luminosity of 37.4 fb$^{-1}$. The measurement is performed in the $μ^+μ^-$ decay channels, differentially as a function of transverse momentum in the 20$-$200 GeV range, in the $\lvert y\rvert$ $\lt$ 0.6 and 0.6 $\lt$ $\lvert y\rvert$ $\lt$ 1.2 rapidity intervals.

A search is presented for massive narrow-width resonances in the mass range of 1$-$4.5 TeV, decaying into pairs of Higgs bosons (HH). The search uses proton-proton collision data at a center-of-mass energy of 13 TeV collected with the CMS detector at the CERN LHC during 2016$-$2018, corresponding to an integrated luminosity of 138 fb$^{-1}$. The analysis targets final states where one Higgs boson decays into a pair of bottom quarks and the other into a pair of tau leptons, X $\to$ HH $\to$ $\mathrm{b\bar{b}}τ^+τ^-$. It uses a single large jet to reconstruct the H $\to$ $\mathrm{b\bar{b}}$ decay, while the H $\to$ $τ^+τ^-$ decay products can either be contained within a single large jet or appear as two isolated tau leptons. The observed data are consistent with standard model background expectations. Upper limits at 95% confidence level are set on the production cross section for resonant HH production for masses between 1 and 4.5 TeV. This analysis sets the most sensitive limits to date on X $\to$ HH $\to$ $\mathrm{b\bar{b}}τ^+τ^-$ decays in the mass range of 1.4 to 4.5 TeV.

A review of the current status of the field of Ultra-High-Energy Cosmic Ray (UHECR) including a summary of remaining open questions was presented in the white paper "Ultra-High Energy Cosmic Rays: at the Intersection of the Cosmic and Energy Frontiers" (Astropart. Phys. 147 (2023) 102794; arXiv:2205.05845). The authors concluded that two types of next-generation detectors are needed to answer these questions: high-accuracy instruments and detectors that maximize exposure at the highest energies. The Probe Of Extreme Multi-Messenger Astrophysics (POEMMA), a proposed dual-satellite observatory, exemplifies the latter class and is designed to increase statistics of the highest-energy cosmic rays and to detect very-high-energy neutrinos following multi-messenger alerts. POEMMA-Balloon with Radio (PBR) implements a compact, balloon-borne version of the POEMMA concept, adapted for a Super-Pressure Balloon flight from Wanaka, New Zealand, with an expected campaign exceeding 20 days. PBR couples a wide field-of-view Schmidt telescope and a hybrid optical focal surface with a dedicated radio instrument to deliver simultaneous, complementary measurements of extensive air showers. The mission will validate the fluorescence detection strategy from space and raise technology readiness for a POEMMA-like space mission by observing UHECR-induced fluorescence light from suborbital altitudes, obtaining the first simultaneous optical Cherenkov and radio observations of high-altitude horizontal air showers above the cosmic-ray knee (E>3PeV), enabling energy-spectrum and composition studies at the PeV scale, and performing follow-ups of multi-messenger alerts to search for very-high-energy neutrinos via upward-going air showers. This paper summarizes the PBR payload and its expected performance.

This article provides a detailed and rigorous study of $4d$ semi-holomorphic Chern-Simons theories and their associated $2d$ integrable field theories from the homological perspective of $L_\infty$-algebras. Through the use of homotopy transfer techniques, it is shown precisely how both the integrable field theory and its corresponding Lax connection emerge from the $4d$ theory, which results in a novel perspective on Lax connections in terms of $L_\infty$-morphisms.

Higgs-boson pair production is of fundamental importance for probing the Higgs potential. At hadron colliders, the dominant production channel proceeds via gluon-gluon fusion (ggF) mediated by a top-quark loop. We report the first fully differential predictions for Higgs-boson pair production through ggF at next-to-next-to-next-to-leading order (N$^3$LO) in the strong coupling $α_s$ in the heavy-top-quark limit (HTL). Fiducial cross section and selected differential distributions are presented at a center-of-mass energy of $\sqrt{s}$ = 14 TeV, under realistic experimental selection cuts. The N$^3$LO QCD corrections reduce the scale uncertainties of the next-to-next-to-leading order fiducial and differential predictions by approximately a factor of three, bringing the theoretical uncertainty to the percent level in the HTL. After incorporating top-quark-mass effects at next-to-leading order in $α_s$, we provide one of the most precise parton-level differential predictions to date for ongoing experimental searches for Higgs-boson pair production at the LHC.

We construct twisted noncommutative gauge theories on twistor space and show that they are equivalent to four-dimensional twist-noncommutative gauge theories. In particular, we study twists of the Poincaré algebra. We explain how such a twist leads to twisted noncommutative twistor space and how to construct noncommutative versions of BF theory and holomorphic Chern-Simons theory on noncommutative supertwistor space. We show how those theories are equivalent to noncommutative versions of Yang-Mills theory and supersymmetric Yang-Mills theory, respectively.

We present FORM 5, a major release of the symbolic-manipulation system FORM. Version 5 introduces an integrated diagram generator, based on the GRACE graph-generator, to produce Feynman diagrams directly from FORM scripts. This release also adds support for arbitrary precision floating point coefficients, together with statements for the numerical evaluation of common mathematical functions as well as multiple zeta values and Euler sums. In addition, FORM 5 provides an interface to the FLINT library, offering substantially faster polynomial arithmetic. Various further functions and commands have been added alongside these major features, as well as performance improvements for TFORM and improved compression of FORM's temporary files. Compatibility with the previous release, FORM 4.3.1, is retained except where prior behaviour contradicted the manual or was experimental.

The next-to-leading order (NLO) Standard Model Effective Field Theory (SMEFT) renormalization group equations are needed to account for phenomenologically relevant operator mixing and ensure renormalization scale independence in NLO calculations of observables. For the first time, we present the renormalization group equations of the baryon-number-conserving sector of the dimension-six SMEFT up to two-loop order. Our calculations have been performed using functional methods with an anticommuting $ γ_5 $-scheme. A variety of strategies are employed to mitigate the reading-point ambiguities inherent to this scheme choice. We also describe how a local version of the $ \boldsymbol{R}^\ast $-method is adapted to handle the evanescent operators arising in dimensional regularization. The results are provided in various supplementary files to make them accessible for both human inspection and numerical implementation.

Coherent elastic neutrino-nucleus scattering (CE$ν$NS) stands out as a pivotal process for precision tests of the Standard Model electroweak sector, investigations of neutrino properties, and searches for new physics (NP). Recent experimental measurements by COHERENT, CONUS+, and ton-scale xenon detectors--including PandaX-4T and XENONnT--underscore the need for a systematic theoretical framework to bridge high-energy physics scenarios with low-energy observational data. In this work, we develop a comprehensive end-to-end effective field theory (EFT) framework for CE$ν$NS, encompassing the complete energy scale hierarchy spanning the ultraviolet (UV) regime down to the nuclear sector. We consider the low-energy EFT (LEFT) operators up to dimension 8, incorporating their QCD renormalization group running effects, and employ the systematic spurion method to achieve the matching between these operators and the chiral Lagrangian. A full power counting analysis is performed, extending to nuclear response functions, which evaluates contributions from LEFT operators up to dimension 8 while accounting for the nucleon number enhancement effect intrinsic to CE$ν$NS. Moreover, we match the relevant LEFT operators for CE$ν$NS onto operators up to dimension 8 within the Standard Model EFT. By also providing their complete tree-level ultraviolet completions, this procedure establishes a consistent top-down theoretical workflow. Leveraging a broad suite of CE$ν$NS experimental data, this framework enables a combined analysis to extract constraints on the scales of EFT operators and neutrino non-standard interaction parameters.

We discuss interrelations between eigenfunctions of the Hamiltonians associated with the commutative (integer ray) subalgebras of the Ding-Iohara-Miki algebra and those of the twisted Cherednik system. In the case of $t=q^{-m}$ with natural $m$, eigenfunctions of the first system of Hamiltonians are the twisted Baker-Akhiezer functions (BAFs) introduced by O. Chalykh, while eigenfunctions of the twisted Cherednik Hamiltonians are twisted non-symmetric Macdonald polynomials. Actually, the twisted Cherednik ground state is symmetric and coincides with a peculiar symmetric BAF. We lift this correspondence to excited states, and claim that both Cherednik eigenfunctions and BAF's can be combined to produce symmetric functions, which coincide with each other and are eigenfunctions of the both DIM Hamiltonians and power sums of the twisted Cherednik Hamiltonians at once. This reflects the correspondence between the DIM algebra and the spherical DAHA explicitly.

Recent simulations have identified long-lived ``prompt cusps'' -- compact remnants of early density peaks with inner profiles $ρ\propto r^{-3/2}$. They can survive hierarchical assembly and potentially enhance signals of dark matter annihilation. In this work, we incorporate prompt cusps into the semi-analytic substructure framework \textsc{SASHIMI}, enabling a fully hierarchical, environment-dependent calculation of the annihilation luminosity that consistently tracks subhalos, sub-subhalos, and tidal stripping. We assign prompt cusps to first-generation microhalos and propagate their survival through the merger history, including an explicit treatment of cusps associated with stripped substructure. We find that the substructure hierarchy converges rapidly once a few levels are included, and that prompt cusps can raise the total annihilation boost of Milky-Way--size hosts at $z=0$ to $B\sim O(10)$ for fiducial cusp-occupation assumptions, compared to a subhalo-only baseline of $B_{\rm sh}\sim\mathrm{few}$. Across a wide range of host masses and redshifts, prompt cusps increase the normalization of $B(M_{\rm host},z)$ while largely preserving its mass and redshift trends. Compared to universal-average, peak-based estimates, our fiducial boosts are lower by about an order of magnitude, primarily reflecting a correspondingly smaller inferred cusp abundance in host halos, highlighting the importance of unifying peak-based cusp formation with merger-tree evolution and environmental dependence.

In models where ultralight fields constitute dark matter, the misalignment mechanism leads to coherent, low-amplitude oscillations in fundamental constants. This effect arises from effective operators that couple dark matter to Standard Model fields. We present different models that can induce these effective operators by integrating out a mediator field. For mediator masses within the reach of collider searches, an alternative way to discover ultralight dark matter is to search for a resonance. Due to being a mediator to dark matter, the mass of the mediator oscillates. The resonance therefore, should not appear as a single isolated peak, but is smeared out once data is averaged over an oscillation cycle or more. Remarkably, the oscillation period and amplitude are in the range of current and future collider searches, even though constraints from atomic clocks probing variations of the fine-structure constant and the electron mass are very strong. We recast existing searches and projections from Belle II, LHCb and SHiP for an `oscillating resonance', and discuss how the periodicity of the signal can be used to reconstruct the peak from mass-binned data. We further show that time-stamped data would allow to unfold the signal via a fast Fourier transform and determine the significance of the signal for different background levels. The discovery of an oscillating resonance at a collider, with characteristics as predicted in ultralight dark matter scenarios, would constitute a powerful probe of dark matter's underlying nature.

We present a unified description of dense matter and neutron-star structure based on simple but physically motivated models. Starting from the thermodynamics of degenerate Fermi gases, we construct an equation of state for cold, catalyzed matter by combining relativistic fermion statistics with the liquid drop model of nuclear binding. The internal stratification of matter in the outer crust is described by $β$-equilibrium, neutron drip and a gradual transition to supranuclear matter. Short-range repulsive interactions inspired by Quantum Hadrodynamics are incorporated at high densities in order to ensure stability and causality. The resulting equation of state is used as input to the Tolman--Oppenheimer--Volkoff equations, yielding self-consistent neutron-star models. We compute macroscopic stellar properties including the mass-radius relation, compactness and surface redshift that can be compared with recent observational data. Despite the simplicity of the underlying microphysics, the model produces neutron-star masses and radii compatible with current observational constraints from X-ray timing and gravitational-wave measurements. This work demonstrates that physically transparent models can already capture the essential features of neutron-star structure and provide valuable insight into the connection between dense-matter physics and astrophysical observables while they can also be used as easy to handle models to test the impact of more complicated phenomena and variations in neutron stars.

We study tropicalisations of quasi-automorphisms of cluster algebras and show that their induced action on the g-vectors can be realized by tropicalising their action on the homogeneous $\hat{y}$ (or $\mathcal{X}$) variables of a chosen initial cluster. This perspective allows us to interpret the action on g-vectors as a change of coordinates in the tropical setting. Focusing on Grassmannian cluster algebras, we analyse tropicalisations of quasi-automorphisms in detail. We derive tropical analogues of the braid group action and the twist map on both g-vectors and tableaux. We introduce the notions of unstable and stable fixed points for quasi-automorphisms, which prove useful for constructing cluster monomials and non-real modules, respectively. As an application, we demonstrate that the counts of prime non-real tableaux with a fixed number of columns in $\mathrm{SSYT}(3, [9])$ and $\mathrm{SSYT}(4, [8])$, arising from the braid group action on stable fixed points, are governed by Euler's totient function. Furthermore, we apply our findings to scattering amplitudes in physics, providing a novel interpretation of the square root associated with the four-mass box integral via stable fixed points of quasi-automorphisms of the Grassmannian cluster algebra $\CC[\Gr(4,8)]$.

These lecture notes provide a self-contained, graduate-level introduction to the strong $CP$ problem and QCD axion physics from an effective field theory (EFT) viewpoint. We review the construction of the chiral EFT of QCD yielding a $θ$-dependent potential, from which vacuum alignment, $θ$ periodicity and branch structure follow. We further show how the framework leads to the Witten-Veneziano relation highlighting the role of the pure-glue topological susceptibility in organizing $θ$-dependent hadronic observables. Using these tools, we show how to extract representative $CP$-odd mesonic and baryonic amplitudes, including the chiral estimate underlying the neutron EDM bound, and how to generalize the effective framework to confining SU(N) theories with fermions in arbitrary representations. We further show how to employ the Veneziano-Yankielowicz effective Lagrangian for N=1 supersymmetric Yang-Mills theory to extract salient information on the $θ$-dependent physics of one-flavour QCD via orientifold planar equivalence. We also revisit a recent no strong $CP$ claim based on an ordering of limits in the sum over topological sectors and show, in the EFT language, that it amounts to introducing an extra non-propagating axion-like degree of freedom not required by QCD. We then present the standard dynamical resolution to the strong $CP$ problem, i.e. the Peccei-Quinn mechanism, the resulting axion potential and mass from chiral EFT and briefly review associated time-honored UV completions, and the axion quality problem from gravitational corrections.

The Belle II experiment at the SuperKEKB accelerator in Tsukuba, Japan, searches for physics beyond the Standard Model, with a focus on precise measurements of flavor physics observables. Highly accurate Monte Carlo simulations are essential for this endeavor, as they must correctly model the variations in detector conditions and beam backgrounds that occur during data collection. To meet this requirement, the "run-dependent" Monte Carlo has been developed. This approach incorporates time-dependent detector conditions and beam-induced backgrounds collected via random triggers, allowing for different conditions with a granularity of just a few hours. In this document, we will discuss the procedures and challenges associated with producing run-dependent Monte Carlo simulations for Belle II. We will also highlight the improvements these simulations offer over traditional "run-independent" Monte Carlo methods.

We formulate and solve a Euclidean Bethe-Salpeter equation for the lightest scalar glueball (0++) in pure Yang-Mills theory, using the refined Gribov-Zwanziger gluon tree-level propagator as an infrared-complete input. In a minimal ladder truncation with an effective constant kernel strength g_C^2 and the dominant s-wave component, we extract scalar glueball masses in the range 1.7-2.3 GeV for representative values of g_C^2, with a preferred value around 1.9 GeV near g_C^2 = 0.54. The result is consistent with RGZ correlator-based infrared moment analyses and with lattice expectations, providing a cross-check of RGZ-inspired infrared gluodynamics from a bound-state viewpoint.

We develop a heat kernel method to compute the one-loop effective action for a general class of nonlinear electrodynamic (NLED) theories in four dimensional Minkowski spacetime. Working in the background field formalism, we extract the logarithmically divergent part of the effective action, the so-called induced action, corresponding to the DeWitt $a_2$ coefficient of the heat kernel. In NLED, quantisation yields non-minimal differential operators, for which standard heat kernel techniques are not immediately applicable. Considering the weak-field regime, we calculate the $a_0$, $a_1$ and $a_2$ contributions to leading order in the background electromagnetic field strength. Finally, we consider conformal NLED theories and compute the $a_0$ contribution to all orders. For this class, we comment on the role of causality being necessary and sufficient for the convergence of the exact $a_1$ and $a_2$ contributions.

JWST observations have revealed that massive galaxies formed and evolved far faster than predicted by galaxy formation models, with many having already assembled a large mass in stars $\sim12$ billion years ago [1-7]. However, masses of distant galaxies are highly uncertain, as they assume a distribution of stellar birth masses (the initial mass function [IMF]) similar to that in the Milky Way (MW). Specifically, the contribution from low-mass stars, which make up the bulk of stellar mass, is not directly observed, but inferred based on an extrapolation of the MW IMF. Here, we provide the first robust measurements of the IMF beyond the local Universe. Using ultra-deep spectra of nine massive, quiescent galaxies at $z\sim0.7$ from the ambitious JWST-IMFERNO program, extended to bluer wavelengths with deep spectra from LEGA-C [8], we find that these distant galaxies have excess low-mass stars. In other words, they have more bottom-heavy IMFs than the MW. For the oldest two galaxies, which are direct descendants of JWST's "impossibly early" galaxies, the bottom-heavy IMFs increase their stellar masses by a factor of $3-4$. These galaxies thus amplify the tension with galaxy formation models.

Understanding the variability of Sagittarius A* (Sgr A*) requires coordinated, multi-wavelength observations that span the electromagnetic spectrum. In this work, we focus on data from four key observatories: Chandra in the X-ray (2-8 keV), GRAVITY on the Very Large Telescope in the near-infrared (2.2 microns), Spitzer in the infrared (4.5 microns), and ALMA in the submillimeter (340 GHz). These multi-band observations are essential for probing the physics of accretion and emission near the black hole's event horizon, yet they suffer from irregular sampling, band-dependent noise, and substantial data gaps. These limitations complicate efforts to robustly identify flares and measure cross-band time lags, key diagnostics of the physical processes driving variability. To address this challenge, we introduce a diffusion-based generative model, for interpolating sparse, multivariate astrophysical time series. This represents the first application of score-based diffusion models to astronomical time series. We also present the first transformer-based model for light curve reconstruction that includes calibrated uncertainty estimates. The models are trained on simulated light curves constructed to match the statistical and observational characteristics of real Sgr A* data. These simulations capture correlated multi-band variability, realistic observation cadences, and wavelength-specific noise. We compare our models against a multi-output Gaussian Process. The diffusion model achieves superior accuracy and competitive calibration across both simulated and real datasets, demonstrating the promise of diffusion models for high-fidelity, uncertainty-aware reconstruction of multi-wavelength variability in Sgr A*.

We present a NOEMA survey of CO(3-2), CO(4-3), and [C I]($^3$P$_1$-$^3$P$_0$) in eight faint (average $S_{\rm 850 μm} = 2.3$ mJy) dusty star-forming galaxies (DSFGs) at $z = 1.2-2.5$. We used a NIR flux-color cut to match faint SCUBA-2 sources to red stellar counterparts with existing spectroscopic redshifts, allowing us to target CO lines at known frequencies. We obtained seven new CO detections and a serendipitous [C I] detection in an off-axis source, and measured molecular gas masses of $M_{\rm mol} = (6-22)\times10^{10}~(α_{\rm CO}/3.6)~{\rm M}_\odot$ from these lines. We performed UV-to-mm SED fits to measure the SFRs and stellar masses of our sample, and compared these with two other $z = 1-3$ CO samples from the literature. The CO detections have constant depletion times of $t_{\rm dep} \sim 500$ Myr, with no evidence for correlation between $t_{\rm dep}$ and redshift or main-sequence offset. We find that low-mass ($M_\star \lesssim 10^{11}~{\rm M}_\odot$), starbursting galaxies have gas fractions and depletion times twice as high as predicted by molecular gas scaling relations, which may indicate that $M_{\rm mol}$ is systematically over-estimated in this population, possibly due to decreased $α_{\rm CO}$ or increased CO excitation compared to the well-studied massive and/or main-sequence DSFG population.

This work investigates whether two known Andromeda (M31) satellites, Pisces (LGS 3) and Andromeda XVI, have interacted with M33, M31's most massive satellite. $Λ$CDM predictions imply a handful of satellite galaxies around M33, yet few M33 satellites have been found and confirmed despite its high mass. We use proper motions combined with backward orbit integration in a semi-analytic potential to constrain plausible interaction scenarios for Pisces and And XVI. Both dwarfs are currently M31 satellites, defined as being inside its virial radius. However, our results show that, in our fiducial mass models, 42% (And XVI) and 60% (Pisces) of dwarf orbits support that they were previously satellites of M33 (i.e., once inside its virial radius). Both dwarfs had fly-by encounters with M33 at relative velocities greater than M33's escape speed within the past 1-2 Gyr. In over 70% of orbits, Pisces and And XVI also had a close approach with each other post-M33 interaction and share an orbital plane, suggesting possible past group accretion. We explore a range of mass combinations for M31 and M33, finding that this primarily regulates the likelihood that the dwarfs were satellites of M33 in the past, while upholding conclusions of recent flybys about M33. These close interactions provide new evidence for past satellite exchange and/or group infall scenarios between M31 and M33. Such interactions also affect comparisons to observational surveys that define satellites primarily by their distance relative to host galaxies.

Accurate modelling of redshift-space distortions (RSD) is essential for maximizing the cosmological information extracted from large galaxy redshift surveys. In preparation for the forthcoming analysis of the Euclid spectroscopic data, we investigate three approaches to modelling RSD effects on the power spectrum multipoles of mock H$α$ emission line galaxies. We focus on two one-loop perturbation theory models -- the effective field theory (EFT) and velocity difference generator (${\rm VDG_ \infty}$) -- which differ in their treatment of the real-to-redshift space mapping on small scales, and a third approach, the BACCO emulator, which adopts a hybrid strategy combining perturbation theory with high-resolution N-body simulations. We assess the ability of these models to recover key cosmological parameters, including the expansion rate $h$, the cold dark matter density parameter $ω_{\rm c}$, and the scalar amplitude $A_{\rm s}$, across four redshift bins spanning $0.9 \leq z \leq 1.8$. In each bin, we find that ${\rm VDG_ \infty}$ and BACCO outperform the EFT model across all scales up to $k_{max} \lesssim 0.35 h\,Mpc^{-1} $. While BACCO saturates in constraining power at intermediate scales and higher redshift, the ${\rm VDG_ \infty}$ model continues to improve parameter constraints beyond $k_{max} \gtrsim 0.30 h\,Mpc^{-1}$. The EFT model, although robust on large scales, exhibits significant parameter biases for $k_{max} \gtrsim 0.25 h\,Mpc^{-1}$, limiting its applicability to Euclid-like H$α$ samples. Among the full perturbation theory-based models, the enhanced treatment of small-scale RSD effects in ${\rm VDG_ \infty}$ improves cosmological parameter constraints by up to a factor of two.

This paper discusses the broad science case for obtaining milliarcsecond to microarcsecond astronomical imaging resolution in the soft to medium-energy X-ray band (~0.5 to ~8 keV). Astronomy across much of the electromagnetic spectrum has been fundamentally transformed with a rapid increase in ground-based and space-based capabilities to examine celestial objects on small scales that relate directly to their relevant physical processes. X-ray imaging capabilities, however, have fallen far behind observations at longer wavelengths. As such, without decisive advances in X-ray imaging, we will be unable to uncover key phenomena on the smallest astrophysical scales, leaving entire classes of high-energy discoveries beyond our reach. Here we describe several science goals for which high quality X-ray imaging is crucial and the status of some current technologies or mission concepts that would be required for these advances. In particular, we discuss the Accretion Explorer, a mission architecture under current study for a dispersed aperture X-ray interferometer.

Knowledge of the actual density distribution of matter in the local universe is needed for a variety of purposes -- for instance, as a baseline model for ultrahigh energy cosmic ray sources in the continuum limit and for predicting the diffuse Dark Matter annihilation signal. Determining the local mass density and velocity distribution is the aim of the CosmicFlows project. An alternate approach is based on catalogs of galaxies, supplemented with some scheme for filling in for unseen galaxies. Here, we compare the density field proposed by Biteau (2021) with the quasi-linear density field of CosmicFlows2 (Hoffman et al. 2018) and the mean posterior field of CosmicFlows4 (Valade 2026). We find factor-two level differences in some regions and even greater in regions toward the Galactic center zone of avoidance (ZoA) (|l| < 30°, |b| < 20°) as filled by Biteau using "cloning". Within 11 Mpc the density field is well-determined by the Local Volume catalog (Karachentsev et al. 2018) which Biteau directly incorporates; at larger distances, Biteau (2021) should not be used in the ZoA where "galaxies" are entirely fictitious but otherwise is to be preferred over CosmicFlows emphasizing the direction and integrated mass of structures; the radial distribution of mass in Biteau (2021) is less robust due to line-of-sight peculiar velocities. The angular positions of structures in CosmicFlows are sometimes not congruent with evidence in the galaxy catalog.

The presence of massive O-type stars with surfaces enriched by CNO-cycle products has been known since the early 1980s. For many years, internal rotational mixing was assumed to be the dominant mechanism responsible for this chemical contamination. However, accumulating evidence now suggests that binary interaction -- particularly mass-transfer episodes -- may play an equally important, if not dominant, role. We aim to carry out a large-scale investigation of surface helium (He) abundances in Galactic O-type stars, based on the results from the analysis of high-quality spectroscopic data from the IACOB project. We perform a homogeneous spectroscopic analysis of 318 Galactic O-type stars with the IACOB-BROAD and FASTWIND/IACOB-GBAT tools, deriving rotational velocities, atmospheric parameters, and He abundances. We also account for the influence of binarity, runaway status, and parameter degeneracies (e.g., microturbulence, wind properties, diagnostic lines, and companion contamination) on the abundance determinations. We present homogeneously determined surface He abundances (YHe=N(He)/N(H)) for the so far largest, statistically significant sample of Galactic O-type stars. About 78% of the stars show He abundances consistent with the previously proposed cosmic abundance standard of YHe=0.098$\pm$0.002. The remaining 22% display clear He enrichment (YHe>0.13). We also provide observational evidence indicating that most of these He-enriched stars are likely the products of binary interaction. Our study highlights how large spectroscopic surveys are gradually opening robust observational avenues to identify the products of massive binary interaction. It also emphasizes the need for caution when interpreting the spectroscopic properties of apparently single O-type stars. A significant fraction may in fact be the outcome of binary evolution rather than isolated stellar birth.

We explore the ability of normalizing flow (NF) generative models to reproduce weak-lensing summary statistics when trained on a set of cosmological simulations. Our analysis focuses on how accurately NF models recover the mean, standard deviation, and covariance of key statistics derived from convergence ($κ$) maps: The angular power spectrum $C_{\ell}$, probability density function, and Minkowski functionals of weak lensing convergence $κ$-maps. We test two scenarios for training: (1) on the data vectors and (2) on the full $κ$-maps. In both cases, the NF models reproduce the mean and variance of the target statistics within percent-level accuracy. However, the accuracy of the off-diagonal elements of the covariance matrix is underestimated by up to $\sim25\%$. We study several mitigation strategies and find that data augmentation and training with noisy fields help improve covariance recovery to $\mathcal{O}(10\%)$. Our study demonstrates that while the means and variances of weak lensing statistics can be well modeled by NF, covariances can be significantly underestimated if mitigation strategies are not applied.

There is a fundamental lack of understanding as to why quasars that are otherwise very similar can have such a wide range of radio jet powers, and the large-scale environment is thought to play an important role. We investigate the spatial clustering properties of 225,382 quasars from the extended Baryon Oscillation Spectroscopic Survey (eBOSS) within the LOFAR Two-metre Sky Survey (LoTSS) Data Release 2 footprint, split by the statistically-calculated fraction of their radio flux densities contributed by jets (relative to the contribution from star formation). We find a positive correlation between the clustering strengths of quasars and their jet fraction, where quasars with a higher jet fraction have a higher clustering amplitude measured by their two-point correlation functions. We show that this correlation is unlikely related to differences in BH masses or bolometric luminosities. Quasars dominated by powerful jet activities generally reside in dark matter haloes $10-100$ times more massive than those without strong jets, with typical halo masses of $10^{13-14}\ h^{-1}M_\odot$, establishing a robust link between powerful AGN jets and rich cluster environments. Our results demonstrate that halo mass is important for determining the power of radio jets, but suggest that there is no minimum dark matter halo mass or BH mass required for the triggering of jets. The observed correlation suggests that BH spin is likely to play a minor role in jet production; instead, the key driver could be the presence of a strong magnetic flux.

We report new measurements of millimeter-wave temperature power spectra in the angular multipole range $1700 \le \ell \le 11,000$ (wavelengths $13^\prime \gtrsim λ\gtrsim 2^\prime$). We use two years of data in three observing bands centered near 95, 150, and 220 GHz from the SPT-3G receiver on the South Pole Telescope that cover a 1646 deg$^2$ region of the Southern sky. Using the measured power spectra, we present constraints on the thermal and kinematic Sunyaev-Zel'dovich (SZ) effects, radio galaxies, and cosmic infrared background (CIB). We find that inferred SZ powers are dependent on the detailed modeling of the thermal SZ-CIB correlation, and to a lesser extent on the assumed angular dependence of the SZ spectra. We report constraints for simulation-based model templates as well as fits where the angular dependencies of the SZ and CIB power spectra are allowed to vary. In the latter case at $\ell=3000$, we find thermal SZ power at 143 GHz of $D_{3000}^{\rm tSZ} = 4.91\pm0.37\, μ{\rm K}^2$ and kinematic SZ power of $D_{3000}^{\rm kSZ} =1.75\pm0.86\, μ{\rm K}^2$. We use the measured kinematic SZ power to estimate the duration of reionization, noting that the reionization inferences are sensitive to the model choices and assumed level of homogeneous kinematic SZ power from the late-time universe. We find a 95% limit on the duration from an ionization fraction of 25% to 75% of $Δ^{50} z_{\rm re} <\,3.8$ based on a semi-analytic model, or a limit on the duration from an ionization fraction of 5% to 95% of $Δ^{90} z_{\rm re} <\,6.1$ based on the AMBER simulations.

The physical mechanism producing Anomalous Microwave Emission (AME) has been an unresolved puzzle for close to 30 years. One candidate mechanism is rotational emission from polycyclic aromatic hydrocarbons (PAHs) which can have the necessary electric dipole moment and size distribution to account for the AME in representative interstellar environments. However, previous investigations have found that AME is better correlated with the far-infrared dust emission rather than the PAH emission. In this work we analyze the correlations between the AME and the PAH and far-infrared dust emission using the 3.3 $μ$m PAH emission feature as observed by band 3 of the Diffuse Infrared Background Experiment (DIRBE). This analysis builds on previous work conducted in individual molecular clouds and extends it into fainter, more diffuse structures. In addition, we utilize the COSMOGLOBE DIRBE reduction for this work, building on previous studies that used the original DIRBE data set. We find that the AME is better correlated with far-infrared dust emission ($ρ\sim$0.9) than the PAH emission ($ρ\sim$0.7) in the central $|b|\leq$ 10\degree\ region of the sky. This could indicate either that non-PAH dust grains or an alternative physical emission mechanism is primarily responsible for the AME in the Galactic Plane, or that the excitation conditions for mid-infrared emission and for AME from PAHs differ substantially.

Restricted by event horizon suppression, tidal disruption events (TDEs) provide a unique window into otherwise hidden supermassive black holes (SMBHs) at the lower end of the mass spectrum, allowing the connection between star formation and SMBH mass to be explored across a broad stellar mass range. We derive stellar masses and specific star formation rates using Prospector fits to UV-MIR broadband spectral energy distributions (SEDs) for 42 TDE hosts, together with a high-mass comparison sample, and combine these with SMBH mass estimates from the literature. We first verify our approach by reproducing the established result that quenched galaxies host more massive SMBHs than star-forming systems at fixed stellar mass, a result widely interpreted as evidence for SMBH growth driving the blue-to-red sequence transition. However, examining the TDE sample in isolation reveals a trend reversal at lower masses, uncovering a surprising population of low-mass ($10^{9.6} \lesssim M_{\rm gal} \lesssim 10^{10.5}$ M$_\odot$), quenched galaxies hosting SMBHs systematically less massive ($M_{\rm BH} \lesssim 10^{6.5}$ M$_\odot$) than those in star-forming galaxies of comparable stellar mass. After ruling out degeneracies in our SED fits, we conclude that this reflects a physical difference in the quenching mechanism between these TDE hosts and the more massive galaxies. This is unlikely to be driven by AGN feedback, and could instead result from environmental processes, which can end star formation and hinder SMBH growth. We also show that the quenched and post-starburst population within the TDE sample is likely under-represented due to selection biases, suggesting the true fraction could be even higher than observed.

We investigate the signature of the jet-ejecta interaction in early kilonova (t < 1day) using detailed atomic opacities developed in Banerjee et al. (2020, 2024), appropriate for early times (t~1hour after merger). We explore jets with different powers and opening angles. We find that the presence of the jet shifts the spectral peak to longer wavelengths, with the strongest effect near the polar viewing angle. This occurs because the jet creates a thin, low-density outer layer ahead of the bulk ejecta. The opacity of this layer can be as high as kappa ~200 cm2/g, causing photons to escape from cooler, faster-moving outer layer rather than from the hot inner ejecta. The bolometric light curves likewise exhibit a clear imprint of the jet-ejecta interaction, showing suppressed early-time luminosity near polar viewing angles compared to the equatorial one, as the photosphere resides in this thin layer where radioactive heating is lower than in the bulk ejecta. These signatures are also evident in multi-color light curves, particularly in the ultraviolet and u-bands. In the Swift-UVW2 band at t~= 0.15 days for a source at 100 Mpc, the ultraviolet luminosities can reach ~ 19.5 mag if no jet is present, while the presence of the jet can make it fainter by ~ 2.5 mag. The strongest observational signature occurs in the UVEX-NUV, Swift-UVW2, and UVM2 bands, which remains detectable out to viewing angles of ~ 60 deg for t <=1 days. Rapid follow-up with future ultraviolet facilities, such as ULTRASAT and UVEX, will provide powerful probes of jet-ejecta interaction through early-time kilonova observations.

We use ALMA CO(1-0) observations and VLT/MUSE rest-frame optical data of the ultraluminous infrared galaxy (ULIRG) IRAS20100-4156 at $z=0.1297$ to characterize its powerful outflow in multiple phases using tracers of cold molecular, ionized, and neutral atomic gas and dust as well. Our analysis uses the correspondence with the stellar velocity field to split the complex emission line profiles of the CO(1-0) line into components in gravitational and non-gravitational motion. We find a massive ($8\times10^{9}\,M_\odot$) molecular outflow containing about 40% of the total molecular gas mass in the system. The outflow shows a bi-conical morphology centered on the brightest galaxy in the merger, oriented along its minor axis and extending to $\sim5\,\mathrm{kpc}$. This outflow has a characteristic velocity of $170\,\mathrm{km/s}$, an outflow mass rate of $700\,M_\odot/\mathrm{yr}$, a depletion time of $16\,\mathrm{Myr}$, and energetics consistent with star formation as a driver. The neutral atomic and ionized gas phases traced by NaI absorption and H$α$ emission show counterparts to the blueshifted cold molecular outflow but are only 15% and 3% as massive. None of the three gas phases show any signs of slowing down over the extent at which we detected the outflow, suggesting an acceleration mechanism acting on the outflowing gas at kiloparsec scales. We also detect $3.5\times 10^7\,M_\odot$ of dust, traced by optical extinction in the MUSE data, in the blueshifted outflowing cold molecular gas. The ionization state of the non-outflowing gas is consistent with star formation, while the outflowing component shows shock-like ionization. We conclude that the multiphase outflow in IRAS20100-4156 originates in the southeast nucleus of the merger and is driven by the starburst activity there, with radiation pressure likely playing a significant role in its acceleration.

Recent JWST observations have unveiled a large number of quiescent galaxies at $z\gtrsim3$, bringing potential challenges to current galaxy formation models. Since star formation is expected to be fed by external gas accretion, the knowledge about the circumgalactic media (CGM) of these galaxies is essential to understanding how they quench. In this work, we present the discovery of a massive and passive galaxy ($M_\star\simeq10^{11}\,M_\odot$) within the MQN01 structure at z~3.25, containing one of the largest overdensities of galaxies and active galactic nuclei (AGN) found so far at $z\gtrsim3$. The passive galaxy has a star-formation rate of $4^{+6}_{-2}~M_\odot$/yr, placing it more than 1 dex below the star-forming main sequence, and has no detectable molecular gas ($M_\mathrm{H2}<7\times10^{9}\,M_\odot$). Surprisingly, it is located at the center of a large cool gas reservoir, as traced by bright Ly$α$ and H$α$ emission. By taking advantage of deep multi-wavelength information unique to this field, including deep Chandra X-ray data, we argue that the inefficient gas accretion from the CGM onto this galaxy over the last few hundreds of Myr, as suggested by the observations, could be caused by an AGN jet of a nearby star-forming galaxy located at a projected distance of 48 kpc. In particular, we argue that the jet feedback may have maintained a high level of CGM turbulence around the passive galaxy and thus caused a reduced gas accretion over the required time-scales. In addition, the elevated ionizing field provided by the AGN overdensity, including the nearby AGN, can illuminate the passive galaxy's cool CGM and make it visible through fluorescent emission. Our study demonstrates that the star formation rates of high-redshift galaxies could be substantially reduced and maintained at a low level even within gas-rich and overdense environments in particular situations.

We investigate year-scale quasi-periodic oscillations in the $γ$-ray emission of the high-redshift flat-spectrum radio quasar PKS~2052$-$47 using monthly binned \emph{Fermi}-LAT data spanning MJD~54727.99--58507.99. To assess the statistical significance of periodic features embedded in red-noise-dominated variability, we apply several complementary timing techniques, including the Lomb--Scargle periodogram, weighted wavelet $Z$-transform, date-compensated discrete Fourier transform, REDFIT assuming an AR(1) process, and damped random walk modelling. The analyses reveal a dominant quasi-periodic modulation on a timescale of $\sim600$--630~d, together with a secondary longer-timescale feature near $\sim1050$--1110~d. Monte Carlo simulations show that the shorter-period signal exceeds the highest local confidence levels, while the longer modulation reaches $\gtrsim99$ per cent local significance in several tests; independent DRW-based simulations place both peaks above the $4σ$ envelope in the Lomb--Scargle analysis. Spectral-window diagnostics indicate that the detected periodicities are not artefacts of uneven sampling, and a sliding-window analysis shows that the QPO power is episodic across the $\sim11$~yr baseline. We discuss possible physical interpretations in terms of geometric Doppler modulation associated with jet precession or helical motion, accretion-driven instabilities, and SMBBH-induced dynamics.

Type Ia supernovae (SNe), thermonuclear explosions of white dwarfs in binary systems, are widely used as standard candles owing to the empirical width-luminosity relation of their light curves. Recent theoretical and observational studies indicate a diversity of progenitor systems and explosion mechanisms. In the supernova remnant (SNR) phase, the diversity in Fe-K$α$ centroid energies and line luminosities suggests variations in the underlying explosion mechanisms. X-ray spectra of SNRs, which trace shocked ejecta and the surrounding medium, are crucial diagnostics of progenitor systems and explosion physics. Thanks to recent advances in spectroscopy with XRISM, high-resolution X-ray spectroscopy enables 3-D diagnostics, including line-of-sight velocities. In this study, we perform 3-D hydrodynamic simulations of SNRs from six Type Ia explosion models: two each of pure deflagration, delayed detonation, and double detonation. Each model is evolved for 1000 years in a uniform medium, consistently accounting for non-equilibrium ionization. Our efficient numerical scheme enables systematic parameter surveys in full 3-D. From these models, we synthesize X-ray spectra with $\sim$1 eV resolution, exceeding XRISM/Resolve's spectral resolution. This work presents the first calculation of X-ray spectra for Type Ia SNRs derived from 3-D hydrodynamic simulations that follow the evolution self-consistently from the SN phase into the SNR phase. Our results show inter-model diversity in the X-ray spectra. Asymmetric, red- and blueshifted line profiles arise from the 3-D ejecta distributions. These findings demonstrate that 3-D SNR modeling can reproduce the observed diversity of Type Ia SNRs and provide qualitative constraints on progenitor systems and explosion mechanisms.

We present a comprehensive study of perturbation theory for neutrino decoherence in power-law correlated turbulent matter, establishing rigorous convergence criteria and providing detailed numerical validation. Starting from the exact Laplace-space solution derived from the generalized master equation with memory kernel $K(t) \propto t^{-ν}e^{-t/τ_c}$, we develop a systematic perturbation expansion in the turbulence strength parameter $ξ$. The expansion is performed around the high-density matter basis where $\cos 2θ_m \approx 1$, rather than the vacuum basis, reflecting the physical conditions in supernova cores. We obtain explicit expressions for the first few perturbation terms and derive the general structure of higher-order terms, which involve Mittag-Leffler functions that emerge from the fractional dynamics. Using high-precision numerical methods (multiple-precision arithmetic and asymptotic expansions), we analyze the convergence radius and rate of the perturbation series. We find that for $ξ< ξ_c \approx 0.8$, the series converges with error decreasing geometrically. For $ξ= 0.16$ (typical supernova conditions), third-order perturbation theory achieves $\sim 1\%$ accuracy. We provide practical guidelines for applying perturbation theory to realistic astrophysical scenarios and validate all analytical results with extensive numerical computations. Our work clarifies several subtle aspects of the perturbative approach that were not addressed in the original literature, offering a complete toolkit for researchers studying neutrino-turbulence interactions. In particular, we systematically address the ultraviolet divergence of the correlation function at short time scales for spectral indices $ν\geq 1$, introducing a physical small-scale cutoff regularization that ensures mathematical rigor for the fractional master equation and perturbative expansion.

We perform a data-driven test of the FLRW metric and the flatness of the Universe, independently of any Dark Energy model, and in light of the latest DESI DR2 results. We use Pantheon+ and DESY5 SNIa data to reconstruct the distance modulus, dimensionless comoving distance and Hubble parameter, using an iterative smoothing algorithm. Then, combining the various reconstructions with the recent BAO measurements from DESI DR2, we perform the $\mathcal{O}_k$ diagnostic, a litmus test of the FLRW metric and the flatness of the Universe. We obtain robust results that do not depend on Dark Energy models and test some of the underlying hypotheses of the concordance model. We find that when the reconstructed $\mathcal{O}_k$ diagnostic is consistent with the FLRW metric, then the median value of $\Oko$ over all reconstructions that provide an improved fit relative to the flat $Λ$CDM model are: $Ω_{k,0}^\text{med} = 0.035 ^{+0.046}_{-0.079}\pm 0.037$ for the Pantheon+ \& DESI DR2 data combination, $Ω_{k,0}^\text{med} = 0.092 ^{+0.055}_{-0.132} \pm 0.064$ for the same data but with the Pantheon+ SNIa cut at redshift $z=1.13$, which is the maximum redshift of the DES~Y5 data, and $Ω_{k,0}^\text{med} = -0.119^{+0.113}_{-0.047}\pm 0.043$ for DES~Y5 \& DESI DR2. The first uncertainties correspond to the spread in $\Oko$ over all reconstructions, followed by the median 1$σ$ error.

We investigate longitude-velocity ($\ell$-$v$) diagrams as a diagnostic tool to study the metallicity structure of the Milky Way (MW) disk. The present-day metallicity structure encodes the imprint of the Galaxy's formation, assembly, and secular evolution. Using oxygen abundances from HII regions across the MW disk, together with MW-mass galaxies from the Feedback in Realistic Environments (FIRE-2) cosmological simulations, we show that $\ell$-$v$ diagrams trace radial metallicity gradients and non-axisymmetric azimuthal metallicity variations. Because they do not rely on distance measurements, $\ell$-$v$ diagrams complement face-on maps for studying metallicity structure. In the MW, we detect the radial metallicity gradient in $\ell$-$v$ space, but current HII region oxygen abundance errors are too high to reveal azimuthal variations. In the FIRE-2 MW-mass galaxies, the radial gradient is evident in $\ell$-$v$ diagrams regardless of observer location, but anomalous gas kinematics can mimic azimuthal metallicity variations. We term these "anomalous motions", which have an excess local standard of rest (LSR) velocity tail 3 times larger in the FIRE-2 simulations compared to the MW. Our results highlight $\ell$-$v$ diagrams as a largely unexplored tool for probing metallicity structure without requiring distances, and underscore discrepancies between the gas kinematics in the FIRE-2 simulations and those in the MW.

We present a detailed analysis of $γ$-ray emission from the middle-aged supernova remnant (SNR) S147 (G180.0$-$1.7) using approximately 16.5 years of Fermi-LAT data. Spatially, a new extended $γ$-ray component distinct from the emission associated with the H$α$ filaments of the SNR shell is identified. This new component exhibits a strong spatial correlation with dense molecular clouds (MCs) identified in CO emission at Local Standard of Rest velocities of $0$-$5\,\mathrm{km\,s^{-1}}$. Spectrally, the cloud-associated emission implies an underlying cosmic-ray (CR) proton population described by a hard power law with an index of $Γ\approx 2.1$, compatible with the standard diffusive shock acceleration prediction. We interpret the $γ$-ray emission in this region with a hadronic scenario involving two distinct CR populations: trapped CRs reaccelerated within the radiative SNR shell as proposed in previous work, and escaping CRs illuminating the nearby MCs. The derived CR proton intensity in the MC region significantly exceeds the local Galactic background measured by AMS-02, strongly suggesting that the cloud is illuminated by particles accelerated by S147. These findings provide observational evidence for CR escape during the earlier evolutionary phases of this middle-aged SNR and highlight S147 as a promising candidate for detection at TeV energies by LHAASO.

Observations of binary black hole (BBH) mergers provide a unique window into the lives of massive stars across cosmic time. Connecting redshift-dependent merger properties to massive star progenitors requires accurate models of cosmic star formation and chemical enrichment histories. Analytical fits for the metallicity-specific cosmic star formation rate density S(Z, z) are commonly used as proxies for the complex underlying star formation history, yet they remain unconstrained. Using the IllustrisTNG cosmological simulations, we evaluate the accuracy of these analytical S(Z, z) prescriptions and assess how simulation resolution and volume affect the inferred S(Z, z). By coupling the simulated and analytical S(Z, z) to the population synthesis code COMPAS, we investigate the resulting BBH merger rates and mass distributions. We find that analytical S(Z, z) prescriptions can overestimate BBH merger rates at high redshift ($z \gtrsim 6$) by up to a factor of $10$-$10^4$, depending on cosmological simulation resolution, and can introduce spurious features in the BBH mass distribution. For example, they can produce an artificial feature near $8\,M_\odot$ in the primary mass distribution at $z \lesssim 2$, which is absent when using the full simulation-based S(Z, z), while simultaneously suppressing high-mass features. These discrepancies arise because simple analytical models fail to capture a high-metallicity bump and a more flattened low-metallicity tail in the simulated S(Z, z) metallicity distribution. Our results highlight the importance of accurate star formation histories for modeling BBH populations, demonstrate the limitation of widely used analytical S(Z, z) fits, and underscore the need for careful integration of cosmological simulations, analytical fits, and population synthesis when interpreting gravitational-wave observations.

(Abridged) Complex organic molecules (COMs) are considered essential precursors to prebiotic species. While COMs were once expected to be efficiently destroyed under UV-irradiated conditions, detections in photodissociation regions (PDRs) have challenged this view. However, the mechanisms by which UV radiation contributes to their formation are still uncertain. Here, we present moderately resolved maps of simple and complex organic molecules at the UV-illuminated edge of the Horsehead nebula, obtained by combining ALMA and IRAM 30m single-dish observations at $\sim 15^{\prime\prime}$ resolution. We analyze the spatial distribution of species such as C$^{17}$O, CH$_2$CO, CH$_3$CHO, HNCO, CH$_3$CN, and HC$_3$N. By incorporating previous C$^{17}$O and C$^{18}$O single-dish data as well as PdBI maps of H$_2$CO and CH$_3$OH, we derive profiles of gas density, temperature, thermal pressure, and column densities of the organic species as a function of distance from the UV source. Our results show that most organic species$-$particularly H$_2$CO, CH$_2$CO, CH$_3$CHO, HNCO, and CH$_3$CN$-$exhibit enhanced column densities at the UV-illuminated edge compared to cloud interiors, possibly indicating efficient dust-grain surface chemistry driven by the diffusion of atomic C and radicals produced via photodissociation of CO and CH$_3$OH, as supported by recent laboratory experiments. The exceptions, HC$_3$N and CH$_3$OH, can be attributed to inefficient formation on dust grains and ineffective non-thermal desorption into the gas phase, respectively. Additionally, contributions from gas-phase hydrocarbon photochemistry$-$possibly seeded by grain-surface products$-$cannot be ruled out. Further chemical modeling is needed to confirm the efficiency of these pathways for the studied species, which could have important implications for other cold, UV-irradiated environments such as protoplanetary disks.

In addition to fast radio burst (FRB) searches carried out using dedicated surveys, a number of radio observatories take advantage of commensal opportunities with large facilities in which observations for other projects can be searched for FRBs and other transient sources. We present the results from one such effort, the first 24,186 hours of the GREENBURST search for dispersed radio pulses with the Green Bank Telescope (GBT). To date, GREENBURST has detected a total of 50 pulsars and three FRBs. One of the pulsars, PSR J0039+5407, has a period of 2.2 s and was previously unknown. Using follow-up observations with the Canadian Hydrogen Intensity Mapping Experiment, we found a timing solution for this pulsar which shows it to have a characteristic age of 2 Myr. Additional GBT observations show the pulsar has a very high nulling fraction ($\sim70-80\%$). All three of the FRBs are repeating sources that were previously known and were being monitored by the GBT as part of other projects. A major challenge for GREENBURST in the discovery of new FRBs is its single beam. This makes it hard to distinguish some of the pulses from sources of radio frequency interference. We highlight this problem with a case study of an FRB-like pulse that initially passed our interference filters. Upon closer inspection, the event appears to be part of a longer-duration narrow-band source of unknown origin. Further observations and monitoring are required to determine whether it is terrestrial or celestial.

In a dense star cluster core, a tidal disruption event (TDE) of a white dwarf (WD) can occur if the WD passes within the tidal radius of an intermediate-mass black hole (IMBH). Very close encounters cause extreme tidal compression in the WD, raising temperatures enough to induce runaway fusion and produce a thermonuclear supernova (SN). Using the hydrodynamics code AREPO augmented with a 55-isotope nuclear reaction network, we performed high-resolution simulations of the TDE of a $0.6$ Msun C/O WD by a $500$ Msun IMBH for different values of the scaled impact parameter $b$ (i.e., the ratio of periapsis distance to tidal radius). Closer encounters produce combined TDE+SN events, with a partial burning of $^{12}$C and $^{16}$O into heavier isotopes -- the $^{56}$Ni fractions of the disrupted WD material vary from 1% at $b = 0.19$ to 82% at $b = 0.10$, while wider ones ($b \gtrsim 0.20$) lead to standard TDEs. In all cases, the material away from the denser regions remains unburnt, spanning a wide range of radial velocities. Such WD TDEs also exhibit a central cavity, wherein little material is found below a radial velocity of several $1000 \,\mathrm{km s}^{-1}$. We also performed 1D and 2D radiative-transfer calculations for these WD-TDEs using the codes CMFGEN and LONGPOL, respectively, covering epochs from a few days to one hundred days. We recover the typical rise times and peak luminosities of SNe Ia, but with an extremely strong viewing-angle dependence of both light curves and spectra. At nebular times, isolated strong emission lines like [Ca ii] λλ 7291, 7323 may appear both displaced and skewed by many $1000 \,\mathrm{km s}^{-1}$ -- such extreme offsets are harder to identify at earlier times due to optical depth effects and line overlap. WD TDEs may produce a diverse set of transients with extreme asymmetry and peculiar composition.

We present a physically-grounded population model for optical tidal disruption events (TDEs) that combines first-principles hydrodynamic simulations of stellar disruption with statistical inference of the underlying stellar and black hole populations. The model's prediction of peak luminosity is based directly on recent global simulations that follow the disruption self-consistently and contains no tunable parameters related to the emission physics. We construct the predicted joint distribution of peak luminosity and black hole mass, including both full and partial disruptions, and compare it to a sample of observed TDEs using Bayesian inference and Markov chain Monte Carlo sampling. We find that the model reproduces the distribution in the ($M_{BH},L_{peak}$) plane for the bulk of the observed TDE population with good statistical consistency. The data strongly favor an old stellar population, with a sharp suppression of stars above $M_* \simeq 1.5 - 2 M_\odot$. They also indicate that, at fixed stellar mass, the volumetric TDE rate is nearly independent of black hole mass. Partial disruptions contribute a substantial fraction ($\sim 30\%$) of detected events in flux-limited samples and are essential for reproducing the observed distribution. The inferred population properties are robust to different approximations to the stellar mass-radius relation, although the event rate at high luminosity is sensitive to the form of this relation for massive stars. We predict a large population of difficult to detect low luminosity TDEs, implying that the true volumetric TDE rate may exceed that inferred from present samples by up to an order of magnitude.

Ultra-faint galaxies at high-$z$ are fundamental elements of the early galaxy assembly, and spectroscopic characterization of this population is essential to understand the earliest galaxy evolution. Leveraging the ultra-deep JWST/NIRCam and NIRSpec observations of a gravitational lensing field of Abell S1063, taken as part of the GLIMPSE survey, we present spectroscopic properties of 16 galaxies fainter than $M_{\rm UV}=-17$ mag, including the metallicity, dust attenuation, and the ionizing photon production efficiency. The emission lines are generally quite strong, roughly half of which cannot be replicated with standard stellar populations and require an extreme ionizing source. We also identify relatively strong [OIII] emission lines from all sample galaxies, which indicates that the low-mass end of the mass-metallicity relation is extended down to $M_\star\sim10^6\ M_\odot$ at $z\sim6$. The strong [OIII] line detection from the lowest-mass galaxy among the sample ($M_\star\sim10^{5.6}\ M_\odot$) stands in contrast to recent reports of extremely metal-poor galaxy candidates at similar mass and redshift, suggesting that there could be two distinct pathways of the earliest metal enrichment as simulations have predicted. Interestingly, we detect both dust attenuation and galactic outflow in one of the sample galaxies with $M_\star=10^{6.6}\ M_\odot$ at $z=5.5$. All the dust, metal, and outflow contents in this galaxy can be consistently explained by supernovae (SNe), indicative of the key roles of SNe in the earliest galaxy assembly such as dust production, metal enrichment, stellar feedback, and baryon cycle.

As observations have yet to constrain the ionizing properties of the faintest (M$_{\rm UV}$ > -16) galaxies, their contribution to cosmic reionization remains unclear. The rest-frame ultraviolet (UV) continuum slope ($β$) is a powerful diagnostic of stellar populations and one of the few feasible indicators of the escape fraction of ionizing photons (f$_{\rm esc}$) for such faint galaxies at high-redshift. Leveraging ultra-deep JWST/NIRCam GLIMPSE imaging of strong lensing field Abell S1063, we estimate UV continuum slopes of 555 galaxies at z $>$ 6 with absolute magnitudes down to M$_{\rm UV}$ $\simeq -$12.5. We find a modest evolution of $β$ with redshift and a flattening in the $β$-M$_{\rm UV}$ relation such that galaxies fainter than M$_{\rm UV}$ $\sim -$16.5 no longer exhibit the bluest UV slopes. The 138 ultra-faint galaxies with M$_{\rm UV}$ $> -$16 are a diverse population encompassing dusty (30\%), old (15\%), and low-mass (50\%) galaxies. We apply the empirical $β$-f$_{\rm esc}$ relation from local Lyman continuum leakers, finding the mean f$_{\rm esc}$ peaks at $\sim 20\%$ at M$_{\rm UV}=-$16.5 and declines towards fainter galaxies, while remaining consistent with f$_{\rm esc}$ = 14\% within uncertainties, in agreement with recent radiative transfer simulations. Incorporating GLIMPSE constraints on the UV luminosity function, ionizing photon production efficiency, and escape fractions produces a reionization history consistent with independent observational constraints. Our results indicate galaxies with M$_{\rm UV}$ between $-18$ and $-14$ supplied $\sim 60\%$ of the ionizing photons to cosmic reionization, while the lower f$_{\rm esc}$ of fainter galaxies produces a natural cutoff in the ionizing photon production rate density.

We report the measurement of the R3=[O III]5008/Hb ratios for 54 galaxies in the GLIMPSE-D survey. Thanks to gravitational lensing, our sample includes galaxies with -20 < Muv < -14 at z=6-9. We derive oxygen abundances using calibrated relationships. We observe a significant decline in R3 values below Muv > -18, which we interpret as evidence of decreasing metallicities in fainter regimes. We explore four prescription models of the evolution of R3 with UV emission based on the new measurements and results from previous surveys. Applying these models to the GLIMPSE [O III]+Hb luminosity functions, we measure and extrapolate the ionising photon production rate $\dot{N}_{ion}$ of galaxies down to very faint limits SFR(Ha) > 5e-3 Msun/yr. Our results support the dominant contribution of star-forming galaxies to reionisation, and are consistent with the recent discovery of ultra-faint metal-poor galaxies. Our measurements of the relative contribution of each luminosity bin show that galaxies with L(Ha)~1e41 to 1e42 erg/s dominate at 8<z<9, but the relative contributions become more uniform at $7<z<8$. Extreme models either under- or over-estimate the ionising photon budget, while intermediate models align with recent observational constraints.

Context: The radial drift and fragmentation of small dust grains in protoplanetary discs impedes their growth past centimetre sizes. Several mechanisms have been proposed to overcome these planet formation barriers, such as dust porosity or the streaming instability (SI), which is today regarded as the most promising mechanism to form planetesimals. Aims: Here, we examine whether the conditions for the SI to lead to strong clumping, the first step in planetesimal formation, are realised in protoplanetary discs containing porous grains. Methods: We used results from previous simulations of the evolution of porous grains subjected to growth, fragmentation, compaction and bouncing in protoplanetary discs. In the ensuing disc structures, we determined the regions where the dust-to-gas ratio exceeds the critical value for strong clumping found in simulations of the SI including external turbulence. Results: We find that the conditions for strong clumping are met within the first hundred thousand years in large regions of protoplanetary discs containing porous grains, provided that the CO snow line is taken into account. If the CO snow line is neglected, the conditions are met only very close to the inner disc edge early on, or over large areas well after 200,000 yr.

Galaxy clusters are the most massive gravitationally bound structures in the universe and serve as tracers of the assembly of large-scale structure. Studying their progenitors, proto-clusters, sheds light on the earliest stages of cluster formation. Yet, detecting proto-clusters is demanding: their member galaxies are loosely bound and the emerging hot intracluster medium (ICM) may only be in the initial stages of virialization. Recent JWST observations located several proto-cluster candidates by identifying overdensities of $z\gtrsim5$ galaxies. However, none of these candidates was detected by X-ray observations, which offer a powerful way to unveil the hot ICM. Here, we report the combined Chandra and JWST detection of a proto-cluster, JADES-ID1, at $z\approx5.68$, merely one billion years after the Big Bang. We measure a bolometric X-ray luminosity of $L_{\rm bol} = (1.5^{+0.5}_{-0.6}) \times10^{44} \ \rm{erg \ s^{-1}}$ and infer a total gravitating mass of $M_{500}= (1.8^{+0.6}_{-0.7}) \times 10^{13} \ \rm{M_{\odot}}$, making this system a progenitor of today's most massive galaxy clusters. The detection of extended, shock-heated gas indicates that substantial ICM heating can occur in massive halos as early as $z\approx5.7$. In addition, given the limited survey volume, the discovery of such a massive cluster is statistically unlikely, implying that the formation of the large-scale structure must have occurred more rapidly in some regions of the early universe than standard cosmological models predict.

Supernovae characterized by enduring narrow optical hydrogen emission lines (SNe IIn) are believed to result primarily from the core-collapse of massive stars undergoing sustained interaction with a dense circumstellar medium (CSM). While the properties of SN IIn progenitors have relatively few direct constraints, the ongoing ejecta-CSM interaction provides unique information about late-stage stellar mass-loss preceding core-collapse. We present late-time X-ray and radio observations of four $\geq$3000 day-old SNe IIn: SN 2013L, SN 2014ab, SN 2015da, and KISS15s. The radio and X-ray emission from KISS15s indicate a mass-loss rate of $\rm{\dot M\sim4\times 10^{-3}~{M_{\odot}\,yr^{-1}}}$ at $\sim$450 years pre-supernova -- 2 orders of magnitude below earlier optical estimates (which probed the mass-loss immediately preceding the supernova). We find hints of a spectral inversion in the radio SED of KISS15s; a possible signature of a secondary shock due to a binary system or the emergence of a pulsar wind. For SN 2013L, we obtain a mass-loss rate of $\rm{\dot M\sim2 \times 10^{-3}~\rm{M_{\odot}\,yr^{-1}}}$ at $\sim$400 years pre-explosion based on the X-ray detection. For SN 2014ab and SN 2015da, we find a upper limits on the mass-loss rates of $\rm{\dot M<2\times10^{-3}~M_{\odot}\,yr^{-1}}$ explosion at $\sim$ 250 and 300 years pre-explosion, respectively. All four objects display mass-loss rates lower than estimates from earlier optical analyses by at least 1-3 orders of magnitude, necessitating a rapidly evolving progenitor process over the last centuries pre-explosion. Our analysis reveals how X-ray and radio observations can elucidate progenitor evolution when these objects have faded at optical wavelengths.

Hypervelocity stars (HVSs) are stars ejected from the Galactic Centre (GC) through tidal interactions with the central supermassive black hole. Formed in the immediate vicinity of Sgr~A$^\ast$, these stars are accelerated to velocities high enough to escape the GC and be observable in the Galactic halo. Using spectroscopy from the Dark Energy Spectroscopic Instrument (DESI) and astrometry from Gaia, we conducted a six-dimensional search for HVSs and identified a compelling candidate, hereafter DESI-312, whose bound trajectory can be confidently traced back to the GC. The star resides in the inner halo and exhibits supersolar metallicity ([Fe/H] $= 0.27\pm 0.09$), distinct from other known stellar populations with radial orbits. Its inferred GC ejection velocity of $698^{+35}_{-27}$ is consistent with a Hills mechanism ejection, supporting an origin in the innermost regions of the Milky Way. We considered alternative origins for the star, including disk ejections from young clusters and globular clusters, but these scenarios fail to explain both its orbit and metallicity. Unlike previously identified A- and B-type HVSs, DESI-312 is a $\sim 1\,M_{\odot}$ star on the main sequence or early subgiant branch, thus enabling a detailed chemical analysis of its atmosphere and offering a rare window - unobscured by dust and crowding - into the composition of the central regions of the Galaxy.

We present observations of HCN and HCO$^+$ J = $3 - 2$ in the central $424'' \times 424''$ region of the W40 massive star forming region. The observations were taken as part of a pilot project for the MAJORS large program at the JCMT telescope. By incorporating prior knowledge of N(H$_2$) and $T_K$, assuming a constant density, and using the RADEX radiative transfer code we found that the HCN and HCO$^+$ abundances range from $X$(HCN) = $0.4-7.0 \times 10^{-8}$ and $X$(HCO$^+$) = $0.4-7.3 \times 10^{-9}$. Additional modelling using the NAUTILUS chemical evolution code, that takes H$_2$ density variations into account, however, suggests the HCN and HCO$^+$ abundances may be fairly constant. Careful modelling of three different positions finds $X$(HCN) = $1.3-1.7 \times 10^{-8}$, $X$(HCO$^+$) = $1.3-3.1 \times 10^{-9}$. Cross-comparison of the two models also provides a crude estimate of the gas density producing the HCN and HCO$^+$ emission, with H$_2$ densities in the range $5 \times 10^4 - 5 \times 10^5$ cm$^{-3}$, suggesting that the HCN and HCO$^+$ emission does indeed arise from dense gas. High UV intensity (e.g. $G_o >$ a few thousand) has no effect on the abundances in regions where the visual extinction is large enough to effectively shield the gas from the UV field. In regions where $A_V < 6$, however, the abundance of both species is lowered due to destructive reactions with species that are directly affected by the radiation field.

SNe Ia have been used to provide key constraints on the equation-of-state parameter of dark energy. They are generally standardized under the assumption that they belong to a single population, with luminosities standardized in a continuous (roughly linear) fashion using the observed light-curve timescale. We update the Union3+UNITY1.5 SN cosmology analysis in light of increasing evidence for at least two core populations of SNe Ia and apply this "UNITY1.8" model to the updated "Union3.1" compilation (Hoyt et al. 2026). In addition to finding evidence for two different light-curve-shape (x1) distributions, we also find that the color distributions are different, that the light-curve-shape/magnitude standardization relations are different, and that these populations have different distributions across host-galaxy stellar mass and redshift. Importantly, we find that the residual host-mass luminosity step found in prior SN Ia cosmology analyses is now consistent with zero for unreddened SNe. We report a significantly tightened constraint on the split in the red-color standardization between SNe in low- and high-mass galaxies. We find that the estimated uncertainties shrink on cosmological parameters when fitting the same SNe assuming two modes versus one mode. We confirm similar trends in simulated data when running both versions of UNITY on the same (two-mode) simulations. For a flat LambdaCDM cosmology, we find Om = 0.334+0.025-0.024 from SNe alone; for a flat w0-wa cosmology, we find w0 = -0.760+0.084-0.082 and wa = -0.79+0.28-0.30 when including SNe, BAO, and CMB. In the 2D w0-wa plane, adding SNe to BAO and compressed CMB increases the tension with flat LambdaCDM from 2.1 sigma to 2.6 sigma.

We present the first robust dynamical measurement of the supermassive black hole (SMBH) mass in the massive early-type galaxy NGC 4061 using high-spatial-resolution ALMA observations of the $^{12}$CO(2-1) emission. By combining archival Cycle 6 data with new Cycle 7 observations, we achieve a synthesized beam of $0''.16 \times 0''.13$, comparable to the expected sphere of influence of the central black hole. The molecular gas forms a regularly rotating circumnuclear disk aligned with the prominent dust lane seen in HST imaging. We model the full three-dimensional ALMA data cube using the KinMS forward-modeling framework, exploring both data-driven and analytic prescriptions for the gas surface brightness distribution. Our Bayesian analysis yields a best-fitting SMBH mass of $M_{\rm BH} = (1.17^{+0.08}_{-0.10}\,[{\rm stat.}] \pm 0.43\,[{\rm syst.}]) \times 10^{9}$ M$_\odot$ and an $I$-band stellar mass-to-light ratio of $M/L_{\rm F814W} = 3.46^{+0.07}_{-0.06}\,[{\rm stat.}] \pm 0.10\,[{\rm syst.}]$ M$_\odot$/L$_\odot$. The inferred black hole mass is fully consistent across different modeling assumptions and remains insensitive to plausible radial variations in the $M/L_{\rm F814W}$ profile. Our results resolve the long-standing discrepancy between previous indirect mass estimates based on conflicting stellar velocity dispersion measurements and demonstrate that the exceptionally large dispersion reported in the literature is likely spurious. This study highlights the power of high-resolution ALMA molecular gas kinematics for precision SMBH mass measurements at the high-mass end of the local black hole mass function.

Aims. We investigate the impact of diffuse ionised gas (DIG) on the determination of emission line ratios and gas-phase metallicities in two local analogues of high-redshift star-forming galaxies: UM 462 and IIZw 40. Understanding how DIG affects these quantities is essential for interpreting unresolved observations of distant galaxies, where integrated spectra are often used to trace their chemical evolution. Methods. Using archival Very Large Telescope, Multi-Unit Spectroscopic Explorer (MUSE) data, we spatially resolved the warm ionised medium of both galaxies. We derived oxygen abundances through the direct method and several HII-based strong-line calibrators, and we used the H$α$ surface brightness ($Σ$(H$α$)) to distinguish regions dominated by HII or DIG emission. Results. Oxygen abundances derived from the N2 and O3N2 indices show an inverse correlation with $Σ$(H$α$), ionisation parameter, and EW(H$α$), with DIG-dominated regions exhibiting higher 12+log(O/H) than the galaxy mean by $\sim$0.2 dex in UM 462 and $\sim$0.1 dex in IIZw 40. The metallicity differences between HII-dominated and DIG-dominated $Σ$(H$α$) bins reach $\sim$0.4 dex and $\sim$0.3 dex in UM 462 and IIZw 40, respectively. The observed trends with $Σ$(H$α$), metallicity, EW(H$α$), and ionisation parameter indicate smoothly varying ionisation conditions rather than true abundance variations. These effects reflect different ionisation sources and levels, and can produce spurious metallicity gradients in galaxies with extended DIG structures, potentially mimicking signatures of metal-poor gas infall. In our sample, DIG ionisation is most likely dominated by photon leakage from H II regions, with additional contributions from feedback-driven shocks.

The VISTA Variables in the Via Lactea eXtended (VVVX) ESO public survey observed the Galactic plane and the outer Galactic bulge in the near-infrared to mitigate the effects of extinction that severely limit optical observations of these regions. By significantly expanding the area covered by the original VVV survey, VVVX enables a deeper and broader exploration of the most obscured and crowded regions of the Milky Way. We aim to extend and complete our photometric catalogs of the entire Galactic plane region accessible from the southern hemisphere, focusing on the areas newly covered by the VVVX survey. Building on previous work, we applied point-spread function fitting techniques to detect point sources and extract their deep J, H, and Ks photometry across the VVVX footprint. The resulting catalogs were calibrated using astrometric and photometric reference data. Cross-matching between filters and epochs was used to ensure a high level of reliability and completeness. We produce a deep, highly complete near-infrared catalog of more than 700 million sources in the Galactic plane and outer Galactic bulge. When combined with our previous VVV atlas, the full catalog includes over 1.5 billion sources. The derived density maps and color-magnitude diagrams enable detailed studies of Galactic structure, extinction, and stellar populations, and highlight features such as the Carina arm tangency, the Sagittarius stream, and numerous star clusters. This extended atlas provides an unprecedented view of the innermost regions of the Milky Way. It is now publicly available through the VISTA Science Archive, offering a valuable resource for the astronomical community to investigate the structure and evolution of the Galactic disk and bulge.