arXiv Daily Digest - 2026-06-23

Deep reinforcement learning is pivotal for closed-loop autonomous driving yet remains constrained by severe bottlenecks in sampling efficiency. Standard parallel sampling mitigates this but suffers from the straggler effect, where the premature termination of a single environment necessitates a synchronized batch re-initialization, leading to suboptimal sample utilization and prohibitive re-initialization latency. To address this, we propose FAST, a synchronous parallel framework tailored for closed-loop simulation. Specifically, FAST employs Dynamic Parallel Sampling Alignment (DPSA) to maintain vectorization synchronization by extending terminated episodes via virtual continuation, thereby decoupling the sampling loop from individual terminations. By dynamically triggering global truncation based on the termination rate of parallel clips, FAST effectively eliminates the bottleneck of premature resets without sacrificing data diversity. Furthermore, to strictly preserve theoretical consistency, we incorporate a Scaled Mask-Padding Optimization (SMPO) that leverages validity masking and adaptive loss normalization to nullify the bias from auxiliary padding data. Empirical evaluations demonstrate that FAST achieves at least a 1.78 times wall-clock speedup over the single-clip baseline while preserving statistical unbiasedness.

We equip the space of beliefs with a cost geometry (what it costs to pass from one belief to another): optimal transport in Wasserstein space, reweighted conformally by Fisher information (the price of the precision at stake), distinct from the Fisher-Rao metric. In the setting we consider, a finite machine maintains a digital twin of a system; observing the territory through finite, noisy sensors, we model its coherent output as a belief: a probability density over states, the Bayes posterior. Certainty (the perfect twin) is denied twice, by observation and by physics, both read off the Fisher information. On the conformal class, essentially location-scale, three results emerge, all invariants of one change of cost unit. A wall: a well-posed inference rejects certainty to infinite distance as soon as the cost dominates the Fisher information (necessity conjectured beyond power laws). An honesty: an honest (eikonal) cost, each nat the same length everywhere, selects the geometries proportional to the Fisher information. A rigidity: these geometries are hyperbolic, and the Stam bound crowns the Gaussian, the most hyperbolic location-scale belief. Changing the unit dilates the geometry yet preserves the wall, the curvature ordering, and the extremality of the Gaussian: an absolute cost says nothing, only relative cost carries meaning, the value -1/4 being one of its images. The cost of reaching a given precision then has a geometric floor diverging at certainty. Thermodynamics fixes the cost unit and motivates this framework; the results are geometric, in nats.

Procedural mistake detection is important for quality control and user assistance across many disciplines. Recent work in this field has achieved significant gains by using the reasoning capabilities of Video-Language Models (VLMs) as components within multi-stage pipelines, which consist of separate modules for supervised temporal action segmentation, error detection, and explainability. Consequently, they remain dependent on tailored training datasets and require task-specific training, limiting their wider applicability. To remedy this, we introduce zero-shot procedural mistake detection and propose a unified Zero-shot Procedural Mistake detection (ZeProM) framework that jointly solves procedural mistake detection and temporal action segmentation with a single pre-trained VLM. By evaluating our framework on two canonical mistake detection benchmarks, EgoPER and CaptainCook4D, we find that ZeProM can perform these tasks successfully, while approaching, or even outperforming, the performance of fully supervised methods. For instance, we achieve a 4.4 point improvement in EDA and a 2.0 point improvement in F1@.5 on average over all five EgoPER tasks compared to the strongest supervised methods. Overall, our results show the potential of unified methods for procedural mistake detection, and we hope this will steer the field away from highly complex pipelines and toward more generally applicable solutions.

Agentic AI systems orchestrate multiple LLM-based agents through workflow architectures that coordinate decisions, tools, and external actions. While current platforms emphasize runtime safeguards, little support exists for verifying workflows during system design. From a Modeling \& Simulation perspective, this gap is analogous to composing conceptual models without verifying whether their building blocks interact coherently. We propose a design-time verification approach that models agentic workflows as compositions of reusable building blocks and checks their compatibility through twelve structural rules. We implemented these rules in a software prototype and evaluated them using two openly released datasets: 48 workflows with known design flaws and 168 variants that preserve workflow logic but alter graph structure. Results show that our verifier reliably detects violations even when flawed designs are obscured through structural transformations such as splitting tasks between agents. Future works could combine our verification with community repositories of building blocks to compose safe agentic workflows.

Transformers achieve strong performance, but their internal computations remain opaque. We view each Transformer layer as a dynamic graph whose nodes are token representations and per-head attention outputs, with Multi-Head Attention (ATT) and MLP as module boundaries. On this graph we use LIG (Layer-wise Integrated Gradients), which applies set-to-set Integrated Gradients (IG) at nonlinear module boundaries. Set-to-set IG applies IG to a map from a set of input token representations to a set of output representations, evaluating token-to-token contributions, which is not standard in prior IG applications. This extends IG from the usual scalar-objective setting to set-to-set maps via an L2 scalarization, and composes within-layer contributions in the spirit of Layer-wise Relevance Propagation (LRP), with IG completeness playing the role of LRP-style conservation at each boundary. We use LIG to analyze (i) the agreement between module-wise composition and layer-whole attribution under an L2 criterion, and (ii) within-layer information flow by tracing separated ATT and MLP contributions. On BERT-base and PTB, configurations that best preserved within-layer consistency used the target token's embedding as the ATT baseline and either the ATT output at a=0 or Zero as the MLP baseline. We therefore present LIG as a diagnostic XAI tool at module-boundary granularity, without model-specific retraining or per-operation interpreter design. Code is available at https://github.com/eightsuzuki/layer-wise-integrated-gradients.

Transformers are AI's workhorse with strong performance in modeling sequential data, but their computational cost becomes prohibitive when processing long sequences. We target long-horizon streaming vision and robotics applications like map-free pose estimation, where it is particularly impractical to store and maintain a history of observations. Recurrent Transformers address this limitation by maintaining fixed-size memory but their performance lags behind that of transformers operating over the full observation history. We argue that this gap does not stem from architectural limitations, but from differences in how these models learn to compress past information. Without access to an observation history, recurrent models must explicitly decide what to retain in memory at each step, a significantly harder learning problem. In this work, we propose a distillation approach that transfers the compression strategy of a classical full-history transformer to a recurrent variant. We enable this by designing a teacher model that explicitly compresses its observation history into a fixed-size bottleneck representation. By directly supervising the student's memory with this bottleneck representation, we align the two compression mechanisms. We show that this approach allows to train a recurrent latent robotic memory with linear-time complexity while substantially narrowing the performance gap to full-history transformers.

Large language models (LLMs) have shown strong potential in fine-grained translation quality evaluation (QE), yet existing MQM-based approaches typically rely on fixed rubric configurations shared across all translation samples. However, translation instances often differ substantially in error complexity, ambiguity, and required evaluation granularity, making static rubric allocation suboptimal for span-level error detection. We find that larger MQM subtype spaces improve error coverage but also introduce more false positives, while different translation instances prefer different rubric granularities, suggesting that evaluation spaces should be allocated dynamically for each case. Motivated by these observations, we propose a case-specific dynamic rubric framework that adaptively constructs MQM evaluation spaces for individual translation instances. Unlike fully free-form rubric generation methods, our framework remains grounded in the predefined MQM taxonomy while dynamically selecting suitable subtype spaces and evaluation granularity for different cases. Experiments on WMT span-level QE benchmarks across multiple model scales demonstrate that the proposed framework consistently improves MCC and produces cleaner span-level error localization compared with static rubric settings. Our results suggest that combining structured MQM rubrics with case-specific adaptive allocation is an effective strategy for fine-grained LLM-based translation evaluation.

Verification of multi-agent systems requires the ability to check meticulous topological properties when it comes to agents that can move through space in continuous time. This demands a logic with sufficient expressiveness to capture these dynamics. MuTGL logic has interesting properties for expressing entangled space-time properties. However, this logic lacks the expressivity needed to analyse reachability within specific distance bounds, or to track the length or the cost of communication chains: these are fundamental for decentralized monitoring, or graph-theoretic analysis of distributed protocols, where algorithmic complexities often relates with the system's communication graph diameter. We then introduce an extension of muTGL, including a new operator called the space horizon. This addition allows us to bound the distance of communication chains, hence enhancing the logic's expressiveness. We show that this operator allows to encode modalities from other logics, such as reachability or escaping which were not available in vanilla muTGL, while allowing a deeper entanglement of spatial and temporal properties. We provide a centralized offline monitoring algorithm for this logic and illustrate it on several examples on simulations of Consensus-Based Bundle Algorithms, distributed protocols for task allocation.

Collaborative Problem Solving (CPS) is a core skill in education, where the process of peer interaction is highly important. However, existing educational dialogue datasets mostly focus on classroom instruction or tutoring (i.e., teacher/tutor-student interaction), yet datasets centering small-group, student-student interaction are limited. This thus leaves research with limited resources for studying how students interact, coordinate, and solve problems together in real educational settings. To address this, we introduce PeerMathDial, the first dataset of peer CPS dialogues collected from authentic middle school math classrooms. It contains 55 dialogues from 27 students, totaling 6,406 turns. To facilitate research on CPS discourse analysis, we further build a corpus-grounded dialogue act taxonomy assisted by LLMs. Using the dataset and the dialogue act taxonomy, we demonstrate the practical applications of PeerMathDial across three use cases. First, we track how dialogues evolve over time and measure the impact of teacher interventions. Second, we align dialogue actions with student surveys to reveal the connection between students' traits (e.g., confidence, leadership) and their actual behaviors. Third, by evaluating LLMs on dialogue act prediction, we glimpse at the potential of LLMs for student simulation in educational applications. Our dataset and source code will be released to the community.

Agentic retrieval-augmented generation (RAG) systems combine iterative reasoning loops, query decomposition, and adaptive retrieval to tackle multi-hop question answering. However, the contribution of each component remains poorly understood, particularly under resource-constrained settings using only local language models. Many agentic designs add adaptive retrieval routing and deeper retrieval loops on the assumption that the added complexity helps. To test whether it does, we run a controlled ablation study of a full agentic RAG pipeline evaluated on 5,000 questions from the HotpotQA distractor development set using a local 7B parameter model (Qwen2.5-7B-Instruct). Our full pipeline achieves EM=53.2% and F1=61.6%, compared to a single-pass dense-retrieval baseline of EM=43.1% and F1=54.0%. Across eight ablation conditions, we find that: (1) fixed hybrid retrieval via reciprocal rank fusion consistently outperforms rule-based adaptive routing (+1.8 EM, +1.9 F1), as the routing heuristic over-routes to BM25 by firing on named entities present in nearly all multi-hop sub-questions; (2) two retrieval iterations over the decomposed sub-questions capture 95% of the gains of five, with no meaningful benefit from deeper loops; and (3) query decomposition and cross-encoder reranking each contribute statistically significant but smaller gains (p<0.01 and p<0.001 respectively). Taken together, on a fixed local-model budget, the simpler and fixed choices turn out to be competitive with or better than their adaptive versions: most of the gain comes from running a short retrieval loop, not from adaptive routing or from many iterations. We use no proprietary APIs or large-scale compute.

Alignment from human feedback uses human judgments about model outputs to steer the behavior of language models after pretraining. When those judgments reflect conflicting views of desirable behavior, the learned objective becomes an aggregate determination of what the model should prefer. We survey recent work that has studied this aggregation problem through the lens of social choice theory. We illustrate how the social choice perspective helps identify failure modes in the feedback aggregation layer and reveals a broader design space for handling disagreement in explicit and principled ways.

Graph convolutional networks (GCNs) have demonstrated significant success in capturing complex user-item relationships for collaborative filtering (CF). However, due to their reliance on extensive model training, training-free graph filtering (GF)-based CF methods have emerged as a promising alternative, offering computational efficiency by smoothing graph signals via matrix operations. In particular, polynomial GF-based approaches demonstrate improved accuracy through their ability to design more expressive and flexible filtering functions. Despite these advantages, existing GF methods suffer from a critical memory bottleneck: they necessitate storing the full item similarity graph, incurring prohibitive memory costs for large-scale datasets, which limits their practical applicability. To tackle this challenge, we propose Mem-GF (Memory-efficient GF), a new GF-based CF method that departs from conventional designs by principally leveraging the structure of Krylov subspaces as a core mechanism for approximating polynomial graph filters without explicitly storing the item similarity graph. We theoretically analyze the minimum Krylov subspace size that guarantees lossless approximation. Through extensive experiments, we demonstrate that Mem-GF achieves up to 5.74$\times$ lower memory usage and 4.38$\times$ speedup in runtime, while consistently exceeding the recommendation accuracy of state-of-the-art GF and GCN-based methods. Mem-GF robustly scales to datasets with tens of millions of interactions, establishing itself as a practically viable and theoretically grounded solution for efficient CF.

We study first-order methods for solving monotone variational inequalities arising in min-max optimization. Classical approaches such as the extragradient method rely on two gradient queries per iteration, which limits their analysis and applicability in the online and stochastic settings. We propose a family of Generalized Optimistic Methods with Anchoring (GOMA), which combine two-time-scale optimistic updates with an anchoring term inspired by Halpern iteration. In the deterministic setting, GOMA achieves the optimal accelerated last-iterate rate $O(1/k^2)$ on the squared gradient norm for monotone Lipschitz operators. In the stochastic setting with unbounded variance, a simplified single-call variant of GOMA achieves a last-iterate convergence rate of $O(1/\sqrt{k})$ on the squared gradient norm. To the best of our knowledge, this is the first such guarantee for stochastic monotone Lipschitz variational inequalities in the unconstrained setting without variance reduction or growing batches.

Model-free reinforcement learning algorithms such as Proximal Policy Optimization (PPO) treat the environment as a black box, estimating policy gradients from sampled rewards; this process demands millions of interactions and relies on high-variance advantage estimates. When environment dynamics are differentiable, the return is an end-to-end differentiable function of the policy parameters, enabling exact gradient computation via backpropagation through simulation. We term this approach Analytic Policy Gradients (APG) and evaluate it against PPO on four continuous control tasks of increasing dynamical complexity: a one-dimensional point-mass target-reaching task, a 2D point-mass navigation task with obstacle avoidance, a 2D rigid-body T-block pushing task, and a 7-DOF Franka FR3 end-effector reaching task. Both algorithms share identical model architectures, observation normalization, and optimizer settings. To decouple sample efficiency from compute efficiency, we design a multi-axis evaluation protocol that records performance against environment steps and gradient steps. We report a segmented backpropagation scheme with MC and critic-based bootstrap modes that mitigates gradient degradation on long-horizon tasks, and present ablations over segment length and bootstrap strategy.

Detecting hallucinations in clinical text is increasingly framed as an explainability problem: systems should not merely flag an unreliable response but point to the offending span. Architectures built around knowledge-graph (KG) triple decomposition are marketed for exactly this auditability, yet their localization ability is typically assumed rather than measured. We introduce MedHal-Loc, a benchmark and metric for localization faithfulness -- whether a detector's top-ranked error unit actually overlaps the erroneous span. The controlled subset comprises 300 PubMedQA-derived statements with single, span-level errors injected across four localizable types (entity substitution, relation error, mechanism misattribution, invention), yielding gold spans by construction; a complementary natural subset documents that real hallucinations are dominated by diffuse conclusion-flips that resist span localization (a human expert accepted 1/18 candidate spans). Evaluating four fine-grained paradigms, we find that NLI-per-clause, consistency-per-sentence, and the dedicated span detector FAVA all localize well above chance, whereas an elaborate KG-triple pipeline localizes no better than chance (+3.3pp, n.s.), bottlenecked by ~59% entity-extraction coverage -- despite competitive detection F1 (0.609). Detection competence does not imply faithful localization; architectural explainability must be validated, not presumed.

Recently, Muon has gained substantial attention as an appealing alternative to Adam-like optimizers, with many works highlighting its advantages through spectral normalization and improved conditioning. Yet this positive theoretical narrative contrasts with its empirical performance in large language model (LLM) training, where Muon's gains over Adam/AdamW are often mixed, schedule-sensitive, and not uniformly superior. To address this gap, we develop a trajectory-level theory characterizing both the strengths and limitations of Muon. We introduce a mixed-spiked matrix sensing model whose sensing operator decomposes into signal, spike, and bulk components, capturing a mixture of anisotropic structure and long-tail information reminiscent of LLM training. On top of it, we adopted a river-valley perspective in which we view the landscape as composed of a river direction flowing to the desired solution and hill directions encoding nuisance or task-irrelevant information. In the momentum-free setting, we show that Muon moves faster along the information-bearing river direction during early optimization, but can converge much more slowly near the river bottom than gradient descent. We then extend the river-valley perspective to general nonconvex objectives with momentum by studying points on the spectral river. There, while Muon converges faster early on, its orthogonalized update removes residual scale information, making it prone to overshooting and oscillation near the target solution. Together, these results suggest that our characterizations extend beyond spiked matrix sensing and motivate switching to GD-like refinement optimizers in the final phase, rather than relying only on a fixed learning-rate schedule for Muon. We also provide preliminary evidence supporting this two-stage approach in language model training experiments.

The rapid proliferation of IoT and IoMT devices introduces critical cybersecurity vulnerabilities in healthcare and industrial environments where resource-constrained devices operate under strict latency and data-privacy regulations. This paper presents the Federated Temporal Graph Convolutional Network with Advantage Actor-Critic (Federated TGCN-A2C), a privacy-preserving defense architecture integrating four mechanisms: a PyG-based Temporal GCN using GCNConv layers with global mean pooling and a learned anomaly gate for flow-level threat classification; LSTM-based Digital Twins generating per-device anomaly scores gating the classifier via learned sigmoid coupling; a Federated A2C agent selecting among ALLOW, ISOLATE, and HONEYPOT-REDIRECT actions based on a seven-dimensional state capturing confidence, entropy, anomaly magnitude, and traffic composition; and an enhanced honeypot layer converting suspicious traffic into threat intelligence with adaptive thresholds. Federated aggregation employs EMA-smoothed per-client validation losses as inverse-weighted FedAvg coefficients to stabilize global model updates under non-IID distributions, with cosine-annealed learning rates per round. Evaluated on CICDDoS 2019 and TON-IoT benchmarks, the framework achieves 99.48% and 99.61% test accuracy with weighted-F1 scores of 0.9948 and 0.9961, converging within 25 and 10 federated rounds, outperforming Fed-Inforce-Fusion by 0.21 percentage points while covering three additional attack categories. All sixteen CICDDoS 2019 classes achieve F1 of at least 0.9237 and all ten TON-IoT classes achieve F1 of at least 0.9488, including the severely imbalanced MITM category. Post-hoc explainability via SHAP, LIME, Grad-CAM, and counterfactual analysis confirms decisions are grounded in semantically meaningful flow features, supporting regulatory accountability in clinical deployments.

Large language models have strong potential for use in intelligent tutoring systems, but they often fail to follow effective pedagogical strategies, such as guiding students without revealing final answers. We study the application of a two-stage alignment pipeline for math mistake remediation, combining supervised fine-tuning on tutoring dialogs with Direct Preference Optimization on synthetic preference pairs. We construct a dataset that integrates existing tutoring corpora with synthetic data generated along pedagogical dimensions, such as scaffolding and factuality, and study different input configurations that incorporate solution correctness and gold answers. Experiments show that this approach improves both factual accuracy and pedagogical quality over base models and existing tutoring models. Human evaluation further indicates that our best model is competitive with a strong proprietary baseline, while providing additional benefits in terms of openness, transparency, and reproducibility. Our results highlight the effectiveness of preference-based pedagogical alignment, while also revealing challenges in reliably evaluating tutoring quality.

Autoregressive text-to-image (T2I) generation has recently advanced rapidly, yet aligning generated images with human preferences remains challenging. GRPO-style online reinforcement learning provides an effective framework; however, existing methods typically treat reference-policy divergence as fixed, despite its direct impact on policy optimization. We study this overlooked factor within a unified f-divergence framework, encompassing forward KL, reverse KL, and JS divergence, for GRPO-style autoregressive T2I alignment. Our systematic theoretical analysis reveals that different divergences reshape token-level updates in distinct ways. In particular, under the sampled-token shaping form used, JS regularization achieves a favorable trade-off by mitigating uniform bias relative to the reference policy while still discouraging large deviations. Extensive experiments on LlamaGen and Janus-7B show that JS divergence achieves the strongest or highly competitive optimization performance on most evaluation metrics while maintaining favorable generation diversity. The code is available at https://github.com/tuoyou-hao/BPD-GRPO.

Modern deep neural network architectures are trained via backpropagation, which requires errors to be sequentially propagated through all layers before parameters can be updated. This introduces two limitations: locking, where layer-wise updates are strictly interdependent and cannot proceed in parallel, and the weight transport problem, which requires symmetric forward and backward pathways for exact gradient computation. These constraints restrict parallelism, increase memory and communication overhead, and pose challenges for scalable learning. In this work, we propose Hierarchical Block-Local Learning (HBLL), a framework that decomposes deep neural networks into hierarchically linked blocks trained using local learning objectives derived from variational principles, eliminating the need for full end-to-end backpropagation while maintaining effective information propagation across the network. HBLL is the first algorithm that is able to train deep neural networks in $\mathcal{O}(\log N)$ parallel time complexity, where $N$ is the number of network layers. We show that HBLL implicitly defines a family of subnetworks corresponding to different hierarchical paths, enabling flexible inference with different effective numbers of layers. We evaluate HBLL on a set of challenging vision and language modeling tasks, achieving competitive performance. We also extend HBLL to recurrent sequence architectures, applying to settings that otherwise rely on backpropagation through time.

Deploying generalist robotic agents in the real world requires transferable skills. Specifically, a policy trained to clone a behavior from object-specific demonstrations must generalize beyond that object, otherwise data collection requirements become intractable. Recently, fine-tuning of pre-trained billion-parameter Vision-Language Models (VLMs), initially on large-scale robot datasets and then on fewer scenario-specific demonstrations, has emerged as the predominant paradigm for designing Vision-Language-Action (VLA) models. While these policies achieve state-of-the-art manipulation performance in-distribution, they remain brittle to minor spatial, semantic, and task variations. In this work, we address the inability of current models to decouple the declarative (i.e., concepts and entity semantics) from the procedural knowledge (i.e., how to do something) encoded in their parameters, which is a fundamental bottleneck for zero-shot skill transfer to novel objects. To address this, we propose w$^{2}$VLA, a new VLA model with restructured information flow. Rather than feeding all multimodal tokens from the VLM encoder into a large, opaque transformer-based action expert, our approach modulates the robot state sequence with visual, spatial, and skill information in a compositional and interpretable manner. Unlike popular, state-of-the-art VLAs, we show that our modular approach successfully decouples knowledge representations, enabling robust behavior cloning and unprecedented zero-shot skill transfer capabilities across dissimilar, unseen objects.

Vision-Language-Action (VLA) models enable robots to predict actions directly from visual observations and language instructions, but adapting them to new environments still depends on costly action-labeled demonstrations. To reduce this dependence, we study semi-supervised VLA adaptation under limited supervision signals, where only a small portion of trajectories contain robot actions and the remaining trajectories provide action-unlabeled vision-language observations. Unlike standard semi-supervised learning, the missing supervision is an embodied action signal that must be visually grounded, language-consistent, physically feasible, and temporally stable. To address this problem, we propose SemiVLA, a self-distilled teacher-student framework that learns from reliable pseudo-actions on unlabeled trajectories. SemiVLA introduces a VLA-specific reliability controller to assess vision-language alignment, action feasibility, and temporal transition consistency, and further updates the teacher through a Bottleneck-Projected Alignment Update to avoid noisy feedback contamination. With OpenVLA as the backbone, SemiVLA consistently improves multiple PEFT strategies across LIBERO and CALVIN. Under 10\% labeled trajectories, SemiVLA with Selective LoRA achieves 89.0\% average success on LIBERO, outperforming supervised LoRA by 8.0 points without extra inference cost.

Risk stratification for advanced colorectal polyps typically relies on colonoscopy and/or pathology findings. However, there is growing interest in whether non-invasive features available prior to colonoscopy can help identify patients at higher risk. Such approaches may enhance clinical decision-making by prioritizing surveillance for individuals most likely to harbor high-risk polyps, when colonoscopy resources are limited while potentially reducing unnecessary procedures in lower-risk patients. Importantly, the use of non-invasive, pre-procedural information may also help promote more equitable access to risk stratification, particularly in settings where colonoscopy resources are limited or unevenly distributed. We aimed to develop and externally validate machine learning models to predict high-risk colorectal polyps using only non-invasive, pre-colonoscopy demographic, clinical, and behavioral features in a diverse, predominantly African American, urban cohort. We conducted a retrospective cohort study using demographic, lifestyle, and comorbidity data from patients who underwent colonoscopy at Howard University Hospital to develop and validate several machine learning models, including neural networks, random forest, support vector machines (SVM), Naive Bayes, logistic regression, decision trees, k-nearest neighbors (KNN), and XGBoost, for predicting high-risk colorectal polyps. High-risk polyps (HRP) were defined as villous or tubullovillous adenomas, high-grade dysplasia, polyps >= 10 mm in size, and/or the presence of >= 3 polyps per procedure; all other cases were classified as low-risk polyps (LRP). The dataset included 4,681 patients from 2015-2022 used for internal validation and 1,562 patients from 2023-2024 used for external validation.

The vulnerability of ML models to adversarial examples has recently emerged as a major concern. While adversarial training is one of the most effective countermeasures to this issue, its high computational cost remains an obstacle to practical deployment. Recent progress in reducing this cost has relied, in the case of linear models, on a formal equivalence between the adversarial risk and a simpler form of regularized risk. This enabled significantly more efficient training procedures, which naturally raises the question of whether such an equivalence can be extended beyond linear models. In this work, we formally show that no such equivalence is possible for two-layer networks. Our proofs proceed via a reduction to key properties that fundamentally separate the adversarial risk from any simple regularized risk which would only exhibit a weak form of data dependence. Beyond this setting, we provide empirical evidence on Wide-ResNets indicating that the same type of impossibility persists in deeper and more expressive architectures.

Android app signing relies on developer-managed credentials, making secure key protection essential for the integrity of the software supply chain. A recent platform key leakage incident involving two major OEM manufacturers demonstrates that even robustly designed signing mechanisms can be compromised due to developers' oversight. In this work, we conduct a longitudinal ecosystem study to characterize this threat by mining public repositories for Android signing credentials, recovering compromised keys via exposed passwords, and matching them against signatures from over 4,000 apps collected from major stores and OEM system images. Our analysis identifies 5,673 compromised keystores on GitHub and 26 unique certificates linked to 278 real-world apps. These include 26 third-party apps in public app stores and 252 preinstalled apps from seven manufacturers, collectively affecting over 10 billion users. We demonstrate the practical exploitability of these leaks through a proof-of-concept app replacement attack and identify spillover risks in non-smartphone platforms, including a popular automotive head-unit platform installed in over 1,100 vehicle models. Our results reveal that signing-key mismanagement is a systemic risk, underscoring the need for a more rigorous key-management support in Android release engineering and distribution infrastructures.

How do large-scale economic transformations shape cultural production? We address this question by combining computational linguistics, econometrics, and formal modelling, using French drama as a well-documented empirical laboratory. Applying latent Dirichlet allocation to a corpus of 1,215 theatrical texts published between 1700 and 1900, we show that aristocratic discourse centred on sovereignty and political authority was gradually displaced by bourgeois and household economic themes as French capitalism developed. Bayesian vector autoregressive models with max-share shock identification suggest a temporal shift in the literary response to economic shocks: bourgeois everyday-life themes reacted to GDP shocks in the eighteenth century, whereas household-economic concerns became responsive only after 1820, amid accelerating industrialisation. A discrete-choice model shows that peer effects among authors and sensitivity to prevailing economic conditions can jointly account for these dynamics. Monte Carlo simulations reproduce the observed historical trajectory with reasonable fidelity. These findings offer a quantitative framework for understanding how economic transformations propagate into cultural production through identifiable social mechanisms, contributing to the study of cultural evolution and the long-run relationship between institutions and literary discourse.

Accurate soil moisture estimation in semi-arid agricultural regions requires integrating remote sensing and meteorological information while accounting for the delayed response of soil moisture to atmospheric forcing. This study introduces a Cross-Correlation Function (CCF) methodology to determine optimal temporal lags (0-30 days) between meteorological variables and soil moisture, as well as inter-depth lags (0-15 days) describing vertical moisture propagation from the surface (10 cm) to deeper layers (20-50 cm). The approach was validated across seven agricultural plots in southeastern Spain. Three deep learning architectures, each targeting a distinct prediction granularity, were evaluated under five feature configurations ranging from satellite-only to full satellite-meteorology-depth fusion: a CNN for per-pixel estimation within each plot, an LSTM for frame-level (daily plot-mean) prediction, and a CNN-LSTM hybrid operating on sliding windows with pooled multi-patch training. Models were assessed on held-out data to measure genuine generalisation. Meteorological variables improved performance over the satellite-only baseline, while subsurface depth information proved decisive across all architectures. The per-pixel CNN achieved the strongest single-patch result (R^2 = 0.877, RMSE = 2.28), with a seven-patch average R^2 of 0.535, representing an improvement of +1.00 over the satellite-only baseline. The pooled CNN-LSTM hybrid obtained the highest overall performance (R^2 = 0.930, CVRMSE = 8.0%). These results demonstrate that explicitly modelling atmospheric and vertical subsurface delays substantially improves soil moisture estimation for precision agriculture.

Career anxiety and depression among university students present a growing challenge to mental health and academic achievement. This study proposes an Explainable AI (XAI) framework using multimodal data and Federated Learning (FL) to identify early indicators of career-related mental health problems in a privacy-preserving and culturally responsive manner. The framework combines structured behavioral data and facial emotion features from interview videos via an intermediate fusion neural network with attention mechanisms. Label smoothing was applied to improve model generalizability. FL was used across institutions to enable collaborative training without raw data sharing. Evaluation was conducted using the Student Mental Health Survey dataset from university students across Pakistan. Our model attained an F1-score of 89.12%, recall of 86.54%, accuracy of 92.08%, and precision of 91.88%. Using Integrated Gradients and SHAP, the model identified key behavioral markers of depression including avoidance of direct gaze, lower facial expressiveness, and social withdrawal, consistent with psychological theory. This research presents an interpretable, scalable, and context-sensitive AI system for mental health pre-diagnosis with potential integration into student support services globally.

Vision--Language--Action (VLA) controllers are often built by extending vision--language models (VLMs) with action supervision, relying on multimodal backbones with large data and compute requirements. We demonstrate that a text-only large language model (LLM) can be adapted into a VLA-style controller when visual observations are rendered into a text input using an ASCII representation. This ASCII-as-vision interface enables existing training and deployment stacks for LLMs to efficiently condition on visual state, follow natural-language instructions, and produce constrained, executable actions. We fine-tune and compare multiple LLMs and VLMs across model families and scales, using both expert demonstrations from a planning-based teacher, as well as DAgger for iterative improvement. In a 2D manipulation benchmark, in both simulation and on a physical manipulator, the resulting controllers can identify task-relevant entities and plan feasible action sequences. Our results suggest that ASCII rendering can serve as a lightweight, interpretable modality bridge from images to text, complementing conventional VLA pipelines, and opening directions for VLA research with text-only backbones.

Accurate segmentation of subcortical regions is critical for neurosurgical planning and functional research. Most automated methods rely on template space coregistration, which may compromise patient-specific accuracy, particularly in small structures. We identify a need to evaluate whether native space approaches offer a measurable advantage, which we evaluate in the context of movement disorders. We developed two UNet-based segmentation pipelines of the Subthalamic Nucleus (STN) - a common surgical target in Parkinson's Disease - and the neighbouring Red Nucleus (RN) and Substantia Nigra (SN). We collected 7T and 3T MRI data from five public datasets. The pipelines were evaluated in the native-space against manual labels. We further investigated the effect of the template resolution. Motivated by the hypothesis that models may better learn target boundaries in higher field, we tested the transferability of 7T-trained models to 3T clinical images, and whether synthetic 3T training data - generated via a disentangled representation learning method - could help bridging this domain gap. On held-out 7T data, the native pipeline consistently outperformed the template one. For the STN, native-space Dice reached 0.775 +- 0.055 versus 0.713 +- 0.051 (1 mm template), with HD95 of 0.79 +- 0.24 mm versus 1.17 +- 1.10 mm, respectively. Similar advantages were observed for the RN and SN. Increasing template resolution did not improve accuracy. When applied to 3T images, all models showed a considerable performance drop. Adding synthetic 3T data yielded only modest improvements, though without degrading 7T performance. Native-space segmentation is preferable for applications requiring patient specific anatomical fidelity, such as the surgical planning in PD. Bridging the 7T-to-3T domain gap remains an open challenge, motivating future work on domain adaptation tailored to subcortical structures.

This paper evaluates the performance of twelve Small Language Models (SLMs) on Arabic natural language processing tasks. The study introduces a benchmark of 240 Arabic test items distributed across eight domains and ten language skills, covering both comprehension-oriented and generation-oriented tasks. All models were evaluated under a controlled zero-shot setting using a standardized Arabic-only prompt template. Model responses were assessed through a multi-model LLM-as-a-judge framework involving GPT-4.1 Mini, Claude Haiku 4.5, and DeepSeek-Chat, with scores aggregated across judges and analyzed by task, skill, and model family. The results show that Gemma 3 (12B) achieved the highest overall score (4.548/5), followed by Aya and C4AI Command Arabic. The observed results suggest that model size alone does not explain Arabic SLM performance. Models with stronger Arabic alignment and more reliable instruction-following behavior tended to perform better across tasks. Common failure patterns among lower-performing models include prompt leakage, hallucination, language drift, incomplete generation, and weak task adherence. Overall, the benchmark provides a structured reference for evaluating compact Arabic language models and supports future work on efficient, reliable, and culturally appropriate Arabic AI systems.

Speech enhancement language models achieve strong results when trained on discrete audio tokens, but their optimization relies on token-level cross-entropy rather than the perceptual metrics used for evaluation. We introduce a post-training stage for autoregressive speech enhancement language models using Group Sequence Policy Optimization (GSPO) with multi-metric perceptual rewards. Our method directly optimizes non-differentiable quality metrics (DNSMOS, WER, and UTMOS) as reward signals, without learned surrogates or offline preference pairs. Applied to two autoregressive base models, UniSE and GenSE, our approach achieves state-of-the-art results on the DNS2020 benchmark. A human evaluation ablation further shows that the composite multi-metric reward is preferred over any single-metric variant, confirming that multi-reward optimization avoids the reward hacking observed with single-metric training.

Coarse-Grained Reconfigurable Architectures (CGRAs) provide a spatially programmable substrate well suited for accelerating compute-intensive workloads with abundant parallelism. However, traditional CGRA execution models rely on rigid, fixed-size processing elements (PEs) that are statically bound to individual operations, which forces inter-iteration dependencies to be resolved through serialized scheduling. This limits throughput and reduces parallelism across loop iterations. Moreover, static execution schedules often fail to exploit available timing slack between operations, leading to resource underutilization and increased latency. The frequent registering of intermediate results further exacerbates pressure on register files and local memories, introducing data movement overheads that reduce energy efficiency, particularly in power or memory constrained environments. To address these challenges, we introduce COMPOSE, a composable CGRA architecture that enables dynamic formation of PEs at compile time guided by static timing information. By spatially fusing operations across loop iterations and selectively utilizing slack, COMPOSE resolves inter-iteration dependencies that limit throughput and enables low latency execution by reducing slack wastage. Additionally, the architecture reduces register file pressure by deferring output registration when intermediate values remain locally consumable, which significantly lowers redundant memory traffic. Across a diverse set of workloads, COMPOSE on average delivers 1.6x performance improvement and 2.9x EDP reduction over state-of-the-art (SOTA), at minimal area and power overheads.

Task-oriented voice agents need to map spoken user requests to structured outputs such as semantic frames, executable actions, and function calls. A common approach is to cascade ASR with a text-based LLM, but transcription errors can propagate to downstream structured output generation, especially under noisy conditions. Spoken language models (SLMs) offer a direct speech-based alternative, yet adapting them to new tasks typically requires paired speech-target annotations. Motivated by this gap, we present CORTIS, a text-only adaptation framework for task-oriented voice agents. CORTIS fine-tunes SLMs using text-form task supervision, enabling speech-based structured output generation at inference time without task-specific speech-target annotations during adaptation. We evaluate CORTIS on two Qwen2.5-Omni backbones and three task-oriented speech datasets, including an in-house product dataset, and compare it with matched ASR-LLM cascades trained with the same text-form task supervision. Results show that CORTIS performs competitively with matched cascades and offers clearer advantages under acoustic degradation, particularly in preserving high-level task semantics. These findings suggest that text-only fine-tuning of SLMs can serve as a practical adaptation strategy for voice agents when paired speech-target data are costly to collect.

Large language model (LLM) based assertion generation is making formal verification more accessible for Register Transfer Level (RTL) designs, but three practical issues remain. Generated properties can pass for the wrong reason, proof cost can vary widely from one design to another, and failing traces are hard to interpret. This paper presents a lightweight, open-source framework that addresses these issues in one loop. Our method combines mutation-guided refinement to reject weak assertions, including vacuous ones and those that fail to distinguish faulty behaviour, a solver-selection stage that chooses among candidate Satisfiability Modulo Theories (SMT) backends using RTL structure, and causal narrative synthesis to explain why a proof failed. Across diverse RTL designs, the framework improves confidence in generated assertions, reduces runtime variability over fixed-solver choices, and produces failure explanations that remain grounded in the counterexample trace. The results suggest that quality-aware closure, rather than assertion generation alone, is the missing step for practical LLM-assisted formal verification.

As large language models scale, memory bandwidth for key-value caches and retrieval-augmented generation systems becomes a critical bottleneck. While 1-bit quantization addresses this constraint, recent TurboQuant relies on dense random rotation matrices to condition the vector distribution before quantization. This projection demands millions of floating-point multiplications per embedding, making it difficult to deploy on constrained edge silicon. We introduce Fast-TurboQuant, a multiplier-free projection architecture that replaces the dense matrix with a structured fast Johnson-Lindenstrauss transform. By applying a Rademacher phase inversion followed by a fast Walsh-Hadamard transform (FWHT), the method leverages sub-Gaussian concentration to satisfy the prerequisites of scalar Lloyd-Max quantization without Gaussian projections. This substitution reduces the arithmetic complexity to only additions, eliminating hardware multipliers. Evaluation on DBpedia OpenAI-3 Large embeddings demonstrates a 19.7 times algorithmic speedup under sequential execution. Furthermore, the dimension expansion due to the FWHT zero-padding reduces the mean squared error and improves Recall@10.

Automated radiology report generation (RRG) has gained increasing attention because it can reduce the heavy workload of clinical report writing. However, most existing methods mainly optimize for natural language generation (NLG) metrics that focus on language fluency, while providing little control over clinically important factors such as precision and recall. As consequence, generated reports may be fluent but not well aligned with different clinical needs. To address this challenge, we propose a reinforcement learning framework for precision recall controllable RRG, where a control parameter explicitly adjusts the trade-off between clinical precision and recall during inference. This design allows the model to flexibly generate reports according to different clinical requirements. To ensure clinical correctness, we introduce a \blue{clinical reward} into the training objective, which helps improve clinical efficacy (CE) beyond standard language-based optimization. In addition, we apply a group-relative training strategy that normalizes rewards within each training group, reducing reward variance and improving training stability. Extensive experiments on the MIMIC-CXR dataset show that our method consistently outperforms state-of-the-art approaches in both NLG{ and CE} evaluation metrics, while providing reliable control over the CE precision recall trade-off.

The automated design of agentic systems offers a promising pathway for scaling large language models (LLMs) beyond single-agent reasoning. While prior work has advanced task performance through handcrafted or automatically generated multi-agent workflows, robustness is often treated as an afterthought, leaving systems vulnerable to external adversaries and internal failures. We propose AutoRAS, a framework for the Automated design of Robust Agentic Systems. AutoRAS formulates system design as generating a sequence of symbolic primitives that jointly encode structural connectivity and behavioral actions, and learns to optimize this sequence using execution-derived safety signals and flow-based sequence-level objectives. Extensive experiments show that AutoRAS achieves the best performance in both vanilla and adversarial settings, with the smallest performance degradation under attacks. Further analyses demonstrate strong transferability, stable optimization behavior, stability across primitive sets, and favorable cost trade-offs. Our code is available at $\href{https://github.com/guohezuy/AutoRAS}{\text{this https URL}}$.

Large-scale optimization problems are often solved repeatedly under similar structural conditions, leading to substantial computational overhead. This occurs in applications such as power systems, transportation, and supply chain networks, where the underlying structure is fixed while parameters frequently vary under perturbations. This paper proposes a Learning to Optimize (LTO) framework that accelerates the solution of large-scale general mixed-integer problems by leveraging the concept of a backdoor, i.e., a subset of variables that drive most of the computational complexity. The proposed BIPC framework consists of three phases. Phase I is an identification procedure that discovers a backdoor for a set of instances in the distribution. Phase II uses supervised learning to develop machine learning models that, given an instance, predict values for bounded-domain backdoor variables and intervals for wide-domain backdoor variables. These predictions define a reduced optimization problem where the predictions constrain the backdoor variables, while the other variables remain free. Phase III optimizes this reduced problem and, if necessary, applies a correction step to restore feasibility or the optimality guarantees. Experiments on real-world, large-scale problems show substantial reductions in solution time with only a limited loss in solution quality. The framework enables organizations to solve large-scale optimization problems efficiently in the presence of frequent perturbations, such as unexpected events, demand fluctuations, or operational changes. Because these changes affect parameters rather than the problem structure, BIPC can quickly provide high-quality, feasible solutions, offering a practical approach to integrating machine learning into existing optimization pipelines.

Relational Deep Learning (RDL) models multi-tabular databases as temporal heterogeneous graphs for end-to-end representation learning. While RDL is evolving rapidly, existing approaches face significant generalization obstacles. They are either schema-specific, requiring training from scratch for every new database, or they rely on monolithic architectures that entangle feature encoding with graph message-passing. Analyzing these limitations, we establish four core pillars for building foundational relational models: semantic granularity, structural topology, temporal causality, and unified optimization. Addressing these pillars, we propose a modular approach that decouples row encoding from graph message-passing. We introduce the Universal Row Encoder, a transformer-based module that integrates raw cell data with schema metadata$-$including column semantics, table names, and global distribution statistics$-$to produce table-width invariant row embeddings. By explicitly feeding global statistics to an intra-row self-attention mechanism, the encoder natively contextualizes unseen features and handles sparse data. Serving as a flexible "backend" for any downstream graph architecture, our pretrained encoder enhances cross-database knowledge transfer on the established RelBench benchmarks while improving learning convergence and memory footprint.

In this article, we analyse convergence of the averaged Adam optimizer to an attracting zero of the Adam vector field. We provide a central limit theorem that, in particular, quantifies exactly the speed of convergence. The order of convergence is $n^{-1/2}$ in the number of steps of the algorithm which coincides with the order observed for classical stochastic approximation algorithms. The covariance in the central limit theorem is given in terms of properties of the Adam algorithm in the state of the attractor.

We construct a minimal but biologically plausible spiking neuron model operating in discrete time, combining multiplicative spike-timing-dependent plasticity (WSTDP), divisive normalization of synaptic integration, homeostatic threshold adaptation, and a one-step refractory period. We show that this normalization admits a biologically plausible dendritic implementation in which each binary junction operates using only locally available information. Assembling excitatory-inhibitory pairs of such neurons into a two-dimensional recurrent network and applying periodic localized stimulation, we find that the network spontaneously gives rise to stable, self-propagating wave packets with the properties of dissipative solitons: they maintain a stable spatial profile, propagate at constant speed, and annihilate upon frontal collision. Their emergence requires a geometric asymmetry between excitatory and inhibitory connection radii, and initial inhibitory synapses stronger than excitatory ones. WSTDP engraves the direction of propagation into the synaptic weight profile, so that the network learns by itself to sustain propagation in one direction while suppressing the reverse. When two sources are active simultaneously, the resulting waves annihilate upon collision, defining a semi-persistent boundary whose position encodes the relative phase and frequency of the two sources. These results provide a minimal computational framework for studying the emergence of cortical traveling waves, activity zone delimitation, and spatial memory from local plasticity rules alone.

Many algorithms in statistics and machine learning can be parallelized in an asynchronous manner where workers need to communicate through shared state rather than execute independent tasks dispatched by a central controller. Especially in modern hyperparameter optimization and parallel black-box optimization with expensive objectives, this decentralized approach has become widespread, and several Python frameworks adopt it (e.g., Optuna, DeepHyper, and Hyperopt). However, all popular R packages for parallel computing follow a centralized controller-worker architecture that does not support this pattern. We present rush, an R package that provides a shared-state coordination layer for asynchronously parallelized iterative algorithms. rush uses a Redis database as a shared key-value store: workers read and write task data through the database and independently execute their own loops. The package provides a high-level API for managing tasks and their lifecycle, featuring sub-millisecond per-task overhead, robust error handling with automatic detection of lost workers, and efficient caching. rush optionally integrates with the mlr3 ecosystem, powering asynchronous optimization in the bbotk and mlr3tuning packages. We demonstrate the practical utility of rush by implementing asynchronous decentralized Bayesian optimization (ADBO) and benchmarking it on hyperparameter optimization of LightGBM across four datasets using 448 workers.

Mixture-of-Experts (MoE) language models are often described as ideal for resource-constrained inference. Each token activates only a small subset of experts, so the per-token compute cost, in floating-point operations (FLOPs), resembles that of a much smaller dense model. Whether that FLOP advantage survives in practice is far less clear. We ask whether MoE models actually run faster and cheaper than comparable dense models on consumer-grade and edge hardware. We benchmark OLMoE-1B-7B (1.3 B active of 6.9 B total) against three dense baselines on an Apple M2 Pro and an NVIDIA Jetson Orin Nano 8 GB through llama.cpp, measuring throughput, memory, and on-device energy. The answer is device-dependent: OLMoE's active-parameter advantage is only partly realised on the laptop (~10% behind the same-active Llama-3.2-1B) and erodes on the edge device (~31% behind, at 2.1$\times$ the energy per token, with peak memory at the 8 GB ceiling). Patching llama.cpp to time the decode graph node-by-node shows routing accounts for under 9% of MoE-block compute on the cleaner edge backend, so the gap reflects total-parameter memory footprint, expert dispatch, and KV-cache pressure rather than routing. The implication is that on bandwidth-bound edge hardware, inference cost tracks total parameters, not active ones, and sparse activation does not buy back what the device is constrained on. These findings are bounded to one MoE model at this parameter scale and two devices, and we release the full measurement harness and per-run data.

Tensor factorization (TF) has been widely adopted for high-dimensional sparse data completion tasks. Despite significant progress, neural TF methods often struggle to capture complex cross-mode interactions and remain vulnerable to (extreme) data sparsity. To address these challenges, we propose a novel neural tensor factorization approach, termed Dual-Attention Convolution Expert Networks with Group-Level Contrastive Learning (DCGC). For the first problem, DCGC generates diverse non-linear alignment patterns of latent factors via a multi-channel convolution network, and leverages the gated dual-attention mechanism to drive the model to focus on more important output channels (i.e., convolution experts) and the aligned features. Furthermore, DCGC introduces a group-level contrastive learning strategy that aggregates positive samples with identical feedback levels while separating negative samples across different levels. This strategy injects high-quality self-supervised signals to mitigate data sparsity. Extensive experiments conducted on five datasets demonstrate that our DCGC outperforms the state-of-the-art methods in sparse tensor completion for traffic and recommendation applications. Code to reproduce the experimental results in the paper is available at https://github.com/ku1z/DCGC.

The rapid proliferation of Internet of Medical Things (IoMT) devices introduces critical cybersecurity vulnerabilities in healthcare environments where resource-constrained medical devices operate under strict latency requirements and stringent data-privacy regulations. To address these challenges, this paper presents the Lightweight Digital Twin and Federated Reinforcement Learning (LDT-FRL) framework, a privacy-preserving defense architecture integrating four complementary mechanisms: a Temporal Attention Encoder (TAE) built on a GRU backbone with learned temporal self-attention for flow-level threat classification; lightweight LSTM-based Digital Twins trained on normal-class traffic to generate per-device anomaly scores that gate the TAE classifier through a learned sigmoid coupling; a Federated Proximal Policy Optimization (PPO) agent selecting among ALLOW, ISOLATE, and HONEYPOT_REDIRECT actions based on a seven-dimensional state; and an intelligent honeypot layer that converts redirected suspicious traffic into actionable threat intelligence. A federated aggregation strategy employing EMA-smoothed per-client validation losses as inverse-weighted FedAvg coefficients stabilizes global model updates under non-IID client distributions. Evaluated on CICDDoS 2019 and TON-IoT benchmarks, LDT-FRL achieves 99.66% and 99.95% test accuracy respectively, with macro-F1 scores of 0.9913 and 0.9995, converging 81% faster than the DTFL-CD baseline while attaining perfect F1=1.000 on the severely imbalanced MITM class. Explainability analysis via SHAP, LIME, Grad-CAM, and counterfactual methods confirms that the TAE focuses on semantically meaningful flow features, providing interpretable evidence for each defense decision.

Despite recent progress in Vision-Language Models (VLMs), mixed-domain image-caption datasets for both general-purpose and CCTV-based video surveillance systems remain limited. To address this gap, we introduce a large-scale multimodal dataset comprising 141,364 images, 981,947 image-level captions, 1,742,264 region-level captions, and 1,391,779 bounding box annotations. Each image is associated with an average of seven image-level captions describing different aspects of the overall scene, as well as seven region-level captions for each annotated bounding box. These complementary caption types are designed to help VLMs learn fine-grained visual attributes, including object categories, estimated sizes, colors, actions, states, and surrounding environmental context. We demonstrate the effectiveness of the dataset on two important downstream tasks: image captioning and object detection. Experimental results show that lightweight VLMs, including SmolVLM-256M-Instruct, BLIP, BLIP2, and Qwen2.5-VL 3B-Instruct, can be effectively fine-tuned using our dataset. Our dataset and code are publicly available at https://zenodo.org/records/20418601.

Designing modular applications for edge and cloud computing environments involves coordinating multiple computational kernels, shared data, and event-driven execution. This paper presents a domain-specific visual graph editor that enables users to model such applications using a unified interface. The editor supports three first-class abstractions: kernels, representing computational units; shared memory nodes, modeling distributed data; and event triggers, capturing execution dependencies. Users can construct graphs visually, configure node properties, and connect elements to define data and control flow. The resulting graphs are automatically serialized into machine-readable representations (JSON/XML) and can be passed to an execution API, bridging the gap between design-time modeling and runtime deployment. The editor's graph-based model improves reasoning about data sharing, execution order, and dependencies, particularly in distributed edge and cloud scenarios. To demonstrate its applicability, we discuss a federated learning workflow, where local training kernels interact with a shared global model through event-driven coordination. Compared to traditional workflow editors and general-purpose diagramming tools, the proposed system provides explicit execution semantics, modularity, and direct deployability. This work lays the foundation for visual orchestration of modular computation in distributed environments and offers extensibility for user-defined kernels, event types, and alternative execution backends, enabling future exploration of complex distributed applications.

Domain shift remains a major obstacle to the reliable deployment of machine learning models in high-stakes environments such as healthcare. While Domain adaptation aims to mitigate these effects, existing approaches suffer from limited expressiveness of latent representations and a reliance on handcrafted, static augmentations. In this work, we address these limitations by proposing a novel deep learning architecture for Unsupervised Domain Adaptation (UDA), specifically optimized for medical image segmentation. Our framework, ADualVUOT, integrates a dual-encoder Variational Autoencoder (VAE) with Continuous Normalizing Flows (CNFs) to increase modeling flexibility and posterior expressiveness. To achieve domain alignment, we leverage Unbalanced Optimal Transport (UOT) through the Gaussian-Gromov-Wasserstein (GGW) distance, which handles structural and topological discrepancies between domains. Furthermore, we incorporate an adversarial augmentation scheme to synthesize worst-case compositions, thus enhancing model robustness. Extensive experiments on medical imaging benchmarks show significant gains over prior OT-based approaches.

The ability to synthesize realistic X-ray images has catalyzed the development of AI models for X-ray image-guided procedures, which otherwise suffer from a lack of available annotated data. Prior work has demonstrated the effectiveness of mechanistic simulation of digitally reconstructed radiographs (DRRs) as a training data source for a myriad of tasks, including segmentation and anatomical landmark detection, with comparable or superior performance to real data training. However, mechanistic DRR synthesis still relies on the availability of annotated high-resolution anatomical models. Deriving these from CT images of real patients or specimens imposes an undesirable bottleneck on data quantity and variability. In this work, we explore two methods for synthesizing training data: (1) a 3D conditional latent diffusion model that generates CT volumes to use as inputs for mechanistic DRR generation without real, 3D anatomical models, and (2) a view-conditioned 2D diffusion model that produces synthetic X-rays. In controlled experiments, we demonstrate that synthetic 2D diffusion-based X-rays can be used to train an anatomical landmark detection model that generalized to real X-ray images with performance rivaling that of a model trained on real X-ray images. Thus, we provide preliminary evidence that synthetic, 2D diffusion-based training data can substitute for real X-ray data, identifying a promising avenue towards generating large, diverse datasets for training robust AI models in interventional X-ray imaging.

Nowadays, large multilingual translation models demonstrate impressive translation capabilities in the machine translation benchmarks. This raises a practical question to the developers: is it worth developing translation models specialized for a particular language pair if you only need to support that language pair? To give an anecdotal answer to this question, we develop a family of small language models (0.8B, 1.4B, 3.3B, and 7B parameters) specialized for Japanese-English bidirectional translation. We employ a two-stage supervised fine-tuning approach followed by Multi-Objective GRPO (Ichihara et al. 2025) to train models on synthetically generated parallel corpora. We evaluate our models on WMT and real-world translation benchmarks across business, legal, medical, financial, and patent domains. While multilingual models achieve strong performance on WMT benchmarks, our compact models outperform them on real-world benchmarks, suggesting the practical utility of developing specialized translation models even in the era of large multilingual models.

Tool-augmented agents are typically evaluated by their gains under reliable external feedback. Yet these gains leave open a key counterfactual: when feedback is unreliable, would the agent be better off receiving no task evidence? We study this question with a controlled matched-loop comparison that fixes the agent loop, prompt, action space, and decoding, while varying only the returned observation: faithful, misleading, or absent. Across question answering and fact verification, persistent misleading feedback produces a value inversion: agents that benefit from clean tools can perform worse than the matched no-feedback fallback. On HotpotQA, Qwen2.5-7B reaches 44.8 F1 with clean retrieval and 22.3 F1 with no feedback, but drops to 4.7 F1 under shuffled retrieval. The inversion persists under stronger clean retrieval and locally plausible distractors, but weakens when later clean evidence can repair the trajectory. Early trajectory signals predict many failures, yet simple repairs remain fallback-limited: rejecting bad evidence helps only when the exposed fallback is reliable. These results show that clean-tool gains can overstate tool value, and that matched no-feedback fallback controls are necessary for evaluating tool-augmented agents.

Agentic AI applications compose multiple model calls and tool executions, creating new scheduling challenges for GPU-CPU clusters. Their inference time and model-call structure often depend on prompt semantics, making conventional scheduling approaches ineffective for low-latency serving. This paper presents SwarmX, a system that implements agentic scheduling for low-latency agentic applications. SwarmX uses scheduling-specific neural predictors to capture prompt, device, runtime, and target-model features; exposes distributional predictions to routers and scalers for tail-aware decisions; and provides mechanisms for predictor training and online adaptation. These predictors and mechanisms are integrated into a scheduler-agent framework that provides a common substrate for integration with existing scheduling and model-serving infrastructure. We evaluate SwarmX using production deployment (nearly one thousand GPUs and one million CPU cores) and controlled experiments on a 128-GPU testbed. Across multi-agent code generation, deep research, and multimodal agentic workflows, SwarmX reduces tail latency by up to 61.5% compared to state-of-the-art schedulers and sustains up to 2x the throughput of production schedulers under the same SLO.

Runtime oversight for LLM agents is commonly framed as scalar risk prediction: estimate failure likelihood, confidence, or uncertainty, then intervene once the score crosses a threshold. We argue that this framing targets the wrong object for control. The relevant question is not how likely the agent is to fail if it continues, but whether an available intervention would improve the outcome. Two trajectory prefixes can have the same risk estimate while requiring different actions, because one remains recoverable and the other does not. We formalize this mismatch as target error and identify intervention advantage, the expected utility gain from intervening rather than continuing, as the decision object for oversight. To measure this mismatch, we introduce prefix branching, a same-prefix counterfactual protocol that executes candidate actions from identical trajectory states. Across four benchmarks, action-conditioned control yields regime-dependent gains over scalar routing. In a calibration decomposition, recalibrating the same scalar score improves prediction metrics but leaves control regret unchanged, showing that calibration alone does not repair target error. A simple prefix-only action-conditioned controller substantially reduces regret in the strongest interactive regime, from 0.506 to 0.110 on ALFWorld. Gains shrink when interventions are weak or when scalar routing already preserves intervention-relevant information. These results suggest that LLM-agent oversight should move from calibrated risk scoring toward action-conditioned value estimation.

Large Language Models (LLMs) have emerged as a promising tool for automated vulnerability detection, yet their effectiveness on web-specific vulnerabilities remains to be explored. This work benchmarks six frontier (Claude Opus 4.6, Codex GPT-5.4, Gemini 3.1-pro-preview) and open-weight models (Qwen 3.5, Qwen 3 Coder Next, MiniMax M2.5) on their ability to detect real-world web vulnerabilities using static analysis in WordPress plugins, including SQL injection, stored cross-site scripting, path traversal, and remote code execution. Using five prompt designs of varying structure, scope, and complexity across three experiment iterations, we aim to answer how model and prompt choice affects vulnerability detection. Our results show that all models are capable of detecting valid security issues, but the detection rate varies depending on the model and prompt. For example, Claude Opus 4.6 achieved the highest web vulnerability detection rate (63%), while open-weight MiniMax M2.5 performs on par with other frontier models (48%), and self-hosted Qwen 3.5 only achieved 35%. We show that scoped prompts that narrow the vulnerability scope outperform open-ended ones, whereas the prompt complexity has little impact. Surprisingly, no model achieved full reporting consistency across three experiment iterations, with some as low as 50%. Our experiments demonstrate the opportunities and limits of LLM-based vulnerability detection, as no model correctly identified one baseline vulnerability in one of the plugins. Additionally, we derive practical lessons learned for security practitioners and publish all code and data to support future research.

Atomistic structure and natural language have long been modeled separately, with language models either calling atomistic models as tools or being fine-tuned on lossy textual encodings that discard atomistic information. We introduce Atomistic Language Models (ALMs) to pursue native multimodality, in which a single language backbone understands atomistic structures, generates materials from natural language, and optimizes crystal structures as instructed by text. By unifying a pretrained atomistic encoder, large language model, and denoising diffusion model through purely continuous projectors and staged training, ALMs achieve state-of-the-art results on crystal structure prediction and de novo generation. ALMs are enabled by a continuous bridge that maps language model embeddings directly into the steering space of atomistic diffusion, and are assisted by Text-to-Crystal Feynman-Kac (T2C-FK), a particle-based sampler that scores partial denoising trajectories to enforce stoichiometric targets at inference time. To evaluate the ability of ALMs to optimize and generate materials from natural-language prompts and 3D atom-coordinate inputs, we introduce ALM Bench, the first benchmark for text-conditioned crystal generation and optimization. Code, training data, and model weights will be released soon.

Vision-language-action models (VLAs) achieve state-of-the-art performance on many robotic manipulation tasks, yet they can still behave unpredictably in out-of-distribution scenarios. Runtime failure detection is therefore essential for the safe real-world deployment of VLAs. However, existing task failure detectors require computationally expensive action sampling, are based on architectural assumptions that limit their applicability to VLAs, or need access to failure rollouts. We propose VLA-FAIL, a lightweight and broadly applicable failure detection framework for VLAs that combines two novel failure detectors with minimal overhead, without requiring failure data. The first, last-layer Mahalanobis distance (LLMD), detects out-of-distribution states by measuring token-wise deviations in last-layer features relative to the training data. The second, action chunk consistency (ACC), exploits the temporal overlap induced by receding-horizon control and detects failures when consecutive action chunks become inconsistent. To capture the trade-off between detection accuracy and detection latency, we introduce AUCPDT, a threshold-independent metric that jointly evaluates precision, recall, and detection time. Through extensive real-world and simulation experiments, we demonstrate that LLMD and ACC capture complementary failure modes whose combination enables reliable and early failure detection across diverse tasks, frequently outperforming significantly more expensive baseline methods.

This paper explores unsupervised disentangled representation learning from a functional perspective. We define latent concepts as factors that influence observations through locally orthogonal directions, formalized as an orthogonality constraint on the Jacobian of the generative mapping. We prove that this condition yields identifiability of general nonlinear generative models, without requiring statistical independence or causal assumptions, provided the latent domain admits all combinations of factor values. Experiments with orthogonality-regularized normalizing flows empirically confirm the theory, demonstrate reliable recovery of ground-truth factors, and shed light on the success of VAEs. These findings challenge the prevailing impossibility claims for unsupervised disentanglement and provide a principled alternative foundation.

Multimodal medical imaging fuses complementary anatomical and functional information, yet modalities frequently disagree in pathologically heterogeneous regions. Current segmentation models handle this in one of two inadequate ways: deterministic fusion that averages away disagreement, or post-hoc uncertainty estimation decoupled from the fusion process that produces it. Both obscure the clinically critical question: why is this prediction unreliable? We present EnTrust, a framework that treats inter-modal conflict as the primary source of predictive uncertainty. Our EnFuse module decomposes multimodal features into three disentangled components: shared anatomical consensus (F_c), modality-specific cues (F_{u,m}), and spatially localized conflict signals (F_{cf}), with independence enforced via a cross-covariance objective. This structured decomposition conditions SegDiff, a diffusion-based generative segmentation model whose sampled hypotheses diverge specifically in regions of modal disagreement. TrustMap then translates this hypothesis divergence into calibrated, pixel-wise uncertainty using ensemble entropy, conflict-guided perturbation probing, and a learned calibration head, enabling clinicians to understand not only where predictions are uncertain, but why. Across four benchmarks spanning brain, cardiac, lesion, and oncology domains, EnTrust achieves state-of-the-art segmentation accuracy while reducing calibration error by 40% compared to the strongest baseline. Notably, it outperforms 5x deep ensembles using a single model at roughly half the memory footprint. Code and checkpoints are available at https://github.com/GenMI-Lab/EnTrust.git.

Recent advances in large language models (LLMs) have demonstrated significant potential in supporting engineering design tasks, including computer-aided design (CAD) automation. However, most existing LLM-based 3D CAD generation approaches primarily focus on geometric precision and instruction-following performance, often overlooking the fundamental aspect of creative design exploration. This study presents a TRIZ-inspired text-to-CAD framework that leverages LLMs to generate high-quality, editable CAD models while systematically exploring creative design alternatives. The framework integrates the Theory of Inventive Problem Solving (TRIZ)-embedding deep human insights from extensive patent records-into LLM prompting strategies, enabling autonomous generation of innovative CAD variants that address technical contradictions. Through a comprehensive three-stage pipeline of design generation, enhancement, and optimization, the framework produces structurally diverse CAD models from well-crafted prompts. The present study implements and evaluates the first two stages, while positioning the design optimization stage as future work. A product design case study (chair) demonstrates that the TRIZ-inspired text-to-CAD framework generates multiple creative design alternatives by systematically applying TRIZ inventive principles such as segmentation, anti-weight, dynamics, and composite materials, achieving 4.0-14.7% mass reduction across all enhanced designs while maintaining structural integrity. The key findings suggest that integrating systematic innovation methodologies with LLM-based 3D CAD generation bridges the gap between precision-focused synthesis and creativity-focused exploration, advancing toward autonomous design systems where AI makes design decisions independently, supporting human decision-making in human-AI collaborative design for engineering applications.

Vision-language-action (VLA) models rely on discrete action tokenizers to bridge continuous robot control and autoregressive sequence modeling, yet existing tokenizers often trade off between compression, latency, and downstream performance. We revisit this design through the lens of neural audio codecs-convolutional encoder-decoder architectures with residual vector quantization that serve as the standard front end for audio foundation models. Motivated by their success, we introduce the Neural Action Codec (NAC), which treats short robot action trajectories as multi-channel 1D signals and compresses them using a multi-scale RVQGAN architecture. We observe that audio-specific mel-spectrogram objectives are ill-suited for kinematic signals; however, by replacing them with simple time-domain and non-mel spectral reconstruction losses, audio-codec-style models can autoencode actions with high fidelity without substantial architectural changes. NAC provides a compact, ordered token space via offset codebooks, enabling standard autoregressive policies to operate over short, structured sequences. Meanwhile, a Vocos-style decoder with an ISTFT head and adversarial discriminators recovers smooth, detailed trajectories. Across LIBERO-10, RoboMimic, and a suite of real-world manipulation tasks, NAC achieves lower reconstruction error and higher success rates than binning, FAST, and prior VQ-based tokenizers at comparable or better compression rates. These results demonstrate that repurposed neural audio codecs offer a strong, practical backbone for learned action tokenization in modern VLAs.

Medical image re-identification (MedReID) enables longitudinal patient linkage but remains vulnerable to shortcut learning and often produces decisions that clinicians cannot audit against named anatomy. We propose Graph-of-Differences (GoD), which grounds identity comparisons in explicit anatomical structure. Each image is represented as an anatomy graph whose nodes correspond to named anatomical regions; given an image pair, soft node correspondence is established, and differences are computed over matched anatomy. A graph-level difference alignment objective ties these anatomy-matched differences to the global backbone difference, ensuring the retrieval signal is anchored in homologous structures rather than arbitrary spatial tokens. Explanations are defined over named graph nodes and quantitatively audited via node insertion/deletion tests, replacing unstable pixel heatmaps with verifiable structure-level evidence. On internal benchmarks, GoD improves Rank-1 by +7.1 pp on fundus and +3.1 pp on CXR over a strong frozen-backbone baseline, with further gains on zero-shot external transfers confirming that anatomy grounding improves both accuracy and generalization. Code is available at https://github.com/GenMI-Lab/GoD.git.

This work examines sexualised AI-generated English-speaking voices offered by a popular commercial platform. New technologies may enable sexual empowerment and greater diversity in gender expression, yet toxic masculinity, heteronormativity, and the abuse of women and LGBTQ+ people remain pervasive online. Drawing on a Feminist HCI perspective, we examine how commercial voice AI systems reproduce and circulate particular performances of gender. We conducted a listening experiment with a diverse group of listeners, combining quantitative adjective selection, qualitative free-text responses, and acoustic analysis. Participants evaluated male- and female-coded voices presented with either sexualised scripts or neutral text. Results reveal a narrow range of gender expression, largely binary and heteronormative. Female-coded voices are more frequently described using sexualised and submissive terms, while male-coded voices are more often associated with dominance and positive traits.

Recent advances in generative audio have made voice cloning increasingly effortless, enabling voice fraud, impersonation, and other forms of unauthorized use. A common attack finetunes a speech generation model on recordings of a target speaker, allowing the model to synthesize speech in that speaker's voice. Audio watermarking offers a promising defense by embedding detectable signals into audio. A practical watermark must satisfy two key properties: robustness and radioactivity. Existing audio watermarking methods typically embed signals into low-level representations, such as waveforms or spectrograms, which makes them vulnerable to signal-level manipulations and limits their transfer to downstream models. We introduce LambdaMark -- the first generic radioactive watermarking scheme. Unlike all previous approaches, LambdaMark achieves generic radioactivity by embedding multi-bit watermark information into semantic audio latent representations. Our watermarks have semantic interpretation and are thus more likely to be learned by a downstream model through finetuning. LambdaMark includes a lightweight watermark encoder to inject multi-bit message-dependent perturbations into semantic audio representations and a decoder to detect watermark presence and recover the embedded bit information. Encoder and decoder are trained using a custom multi-component loss that preserves fidelity of the watermarked audio, increases bit-level recovery rate, and improves robustness against common distortions and adversarial removal attempts. Experiments show that LambdaMark achieves near-perfect robustness under common distortions. LambdaMark is also the only watermark that is robust against all evaluated removal attacks. Furthermore, LambdaMark exhibits general and robust radioactivity and remains robust to distortions and adversarial removal attacks even on the generated outputs of those finetuned models.

Accurate prediction of repair duration is an important challenge in product maintenance due to its implications for resource allocation, customer satisfaction, and operational performance. This study aims to develop a deep learning framework to help fleet repair shops accurately categorize repair time given product historical data. The study uses an automobile repair and maintenance dataset and creates an end-to-end predictive framework by employing a multi-head attention network designed for tabular data. The developed framework combines categorical information, transformed through embeddings and attention mechanisms, with numerical historical data to facilitate integration and learning from diverse data features. A weighted loss function is introduced to overcome class imbalance issues in large datasets. Moreover, an online learning strategy is used for continuous incremental model updates to maintain predictive accuracy in evolving operational environments. Our empirical findings demonstrate that the multi-head attention mechanism extracts meaningful interactions between vehicle identifiers and repair types compared to a feed-forward neural network and a random forest model. Also, combining historical maintenance data with an online learning strategy facilitates real-time adjustments to changing patterns and increases the model's predictive performance on new data. The model is tested on real-world repair data spanning 2013 to 2020 and achieves an accuracy of 78%, with attention weight analyses illustrating feature interactions.

Large language models (LLMs) are increasingly used to communicate and explain scientific concepts, yet their tendency to hallucinate poses significant risks in this high stakes use-case. Prior hallucination evaluation work remains largely restricted to the biomedical domain, treats hallucination as a binary task, and has not examined the growing family of scientifically fine-tuned LLMs. We address these gaps with SciFactCheck, a benchmark of 2,500 prompts across five scientific domains, paired with a modular evaluation framework targeting three factuality hallucination types: unverifiability, overclaim, and attribution. Using a controlled minimal-pairing design, we evaluate 18 LLMs by comparing each scientifically fine-tuned model against its general-purpose base. Our results indicate that 1. Scientifically fine-tuned models exhibit degraded factual reliability across all hallucination types and scientific domains, and 2. Fine-tuned models are internally less confident yet linguistically more assertive. A human pilot study further reveals that current fact-checking tools show only modest agreement with expert judgments on scientific content, and that defining scientifically check-worthy claims remains contested even among human annotators. Our findings fundamentally challenge current methods of domain-specific fine-tuning for factuality and call for developing improved verification infrastructure for scientific content.

Many machine learning applications rely on heterogeneous event streams to make predictions, either causally as events arrive or bidirectionally over complete sequences. We propose SOHET (Sequence Of Heterogeneous Events Transformer), a hierarchical architecture combining event-type-specific tabular encoders with temporal and type embeddings, processed by a causal or bidirectional transformer. We introduce three self-supervised pre-training objectives for the causal setting. On a proprietary large-scale real-world Booking.com fraud detection task with 17 event types, SOHET outperforms FlexTPP, NAPPT, and CIPPT by 5.8%. Pre-training yields an additional 2.6% gain and 2.4% faster convergence. On the EBES benchmark, bidirectional SOHET matches or exceeds the published best on 6 out of 8 tasks.

Understanding the complex topology and hierarchy of urban power grid is crucial for energy prognosis, power flow management, and system resilience analysis. However, detailed grid information remains largely proprietary. This creates significant barriers for research and innovation, especially when analyzing the last-mile distribution networks connecting individual buildings. This paper addresses this challenge by developing an open-data-driven framework for the complete identification of urban power grid topology, from high-voltage transmission down to individual building connections. Particularly, we fuse public infrastructure data (power-lines, substations, transformers, poles) to map the high and medium-voltage skeleton using graph-based algorithms. We then leverage geospatial machine learning on OpenStreetMap building data to group power demand clusters, and infer the physical topology of the final distribution lines linking the clustered buildings. We apply the developed framework to the district of Alna in Oslo, Norway, and we reconstruct the complete grid topology that connects 7,330 buildings and all major electricity infrastructure assets. With the research in this work, we provide a critical tool that facilitates power system analysis, e.g., power flow optimization, cascading failure simulation, and grid resilience against the penetration of distributed renewable generation.

This paper introduces an interpretation of Temporal Behavior Trees (TBTs) as Reward-Petri-Nets (RPNs) for reinforcement learning (RL). Designing reward functions for complex, long-horizon robotic tasks is notoriously difficult, especially when tasks have hierarchical structure and temporal constraints. TBTs extend conventional behavior trees (BTs) used in robotic applications by incorporating temporal properties into their leaf nodes. This allows TBTs to represents not only the behavioral task structure defined by BT operators such as Sequence, Fallback, and Parallel, but also the task's temporal constraints. In this work, the constraints are specified in the leaf nodes using Linear Temporal Logic. In order to inform RL rewards using TBTs, we provide a translation from TBT into a Petri Net (PN) and show how rewards can be automatically assigned based on the TBT's structure, resulting in a RPN. In a series of increasingly challenging environments, we demonstrate how TBT-based rewards enable learning where vanilla RL fails, improve sample efficiency, and offer flexible, intuitive control over the learning progress. We showcase the learning impact by using different reward distribution schemes and TBT structures.

Many allocation problems are intrinsically multidimensional, since an item may contribute differently to several criteria, and optimizing a single aggregate objective can hide severe losses in other dimensions. We study how much efficiency can be guaranteed simultaneously when indivisible items have multiple attributes. To this end, we introduce the \emph{multidimensional efficient allocation} (MDEA) model, where each agent has an additive valuation in each dimension, and investigate simultaneous efficiency under utilitarian social welfare (USW) and egalitarian social welfare (ESW). Our results reveal a sharp worst-case frontier. For exact efficiency, maximizing the number of dimensions attaining the USW optimum admits a $c/\ell$-approximation for every fixed constant $c$, and this dependence on the number $\ell$ of dimensions is essentially unavoidable; for ESW, even deciding whether two dimensions can be optimized simultaneously is NP-hard with binary valuations. For approximate simultaneous efficiency in every dimension, we identify a tight threshold of order $1/\ell$, showing that such guarantees always exist for both USW and ESW, while any asymptotically better dependence on $\ell$ is impossible, even for binary valuations. Finally, we introduce three natural multidimensional Pareto notions and characterize both their relationships and their computational complexity.

Large language models (LLMs) store and recall factual knowledge, yet the precise mechanism of how entity representations are transformed to enable specific attribute retrieval remains underexplored. In this work, we investigate this mechanism through the lens of an "attribute-computation path"-a sequence of computational steps over the entity representation required to elicit a target attribute. We then propose an iterative patching protocol to identify a minimal subset of layers necessary for this computation. Applying our method to LLaMA 3.1 8B and Qwen3 8B, we find that these paths are non-contiguous, often skipping layers, and that models possess multiple, functionally-equivalent paths for the same entity and fact, highlighting a high degree of redundancy in attribute computation. This implies that knowledge computation is highly distributed, potentially explaining the localization-editing mismatch and suggesting that knowledge storage and retrieval in LLMs is far from being well understood.

Trajectory prediction allows autonomous vehicles to anticipate the future behavior of surrounding objects (or agents) and, accordingly, maximize the safety and efficiency of their driving. State-of-the-art Transformed-based interaction-aware trajectory prediction models, which rely on attention mechanisms to capture multi-agent interactions and maximize prediction accuracy, are commonly trained and evaluated on long-range high-quality datasets. These datasets are typically obtained by aggregating data from multiple vehicles or drones and removing any object detection or tracking noise offline. Yet, information about a surrounding object's state (its position, speed, heading) is far from being noiseless in real-world deployments. Object state estimation is affected by perception uncertainties and localization errors that can be particularly large for objects received via Vehicle-to-Everything (V2X) communications. In this paper, we analyze the impact of noisy object state information on the trajectory prediction accuracy of a state-of-the-art Transformer-based interaction-aware trajectory prediction model. Our study demonstrates that trajectory prediction accuracy can rapidly deteriorate as the noise intensity increases. Numerical results show that the prediction accuracy can reduce by a 1.3x factor under small noise levels and by as much as a 3.9x factor under the highest (yet realistic) noise conditions. These findings reveal the strong sensitivity of trajectory prediction models to noisy data, underscoring the need for more realistic training and evaluation datasets as well as noise mitigation strategies.

Voice reconstruction using Text-to-Speech (TTS) offers a communication method for people with speech disorders, which aims to retain their speaker identity while improving intelligibility. Previous work generally relies on Mean Opinion Score (MOS) to evaluate naturalness and speaker similarity, but this has limited sensitivity and reliability. We propose an evaluation framework with subjective and objective components. Subjectively, we evaluate perceived intelligibility and speaker identity using Best Worst Scaling (BWS) with situational framing. Objectively, we demonstrate that standard measures fail to predict reconstruction success for highly unintelligible speakers, so we introduce a novel dual-reference distributional measure to assess the trade-off between intelligibility and speaker identity. By evaluating the output of 17 zero-shot TTS systems for 193 speakers, we show that our framework provides a reliable and task-aligned approach for assessing voice reconstruction.

Automatic Speech Recognition (ASR) systems, despite achieving remarkable accuracy in general-purpose domains with native speech (L1), struggle in domains like Air Traffic Control (ATC) due to strong channel noise, a presence of non-native (L2) English accents, and data scarcity. We propose a synthetic data generation pipeline with acoustical properties simulations specifically designed to address this lack of real data to improve recognition accuracy in the ATC domain. Our approach leverages a combination of neural generation techniques, including Text-to-Speech, Voice Conversion, L2-to-L1 accent conversion, and a novel controllable L1-to-L2 accent conversion framework built to simulate accented speech. Our experiments with the Whisper model on the ATCO2 corpus demonstrate that fine-tuning with either synthetic data alone, or a mix of real and synthetic data, significantly improves the word error rate over out-of-the-box and real data only baselines respectively.

Automated formal specification generation is a key step towards program understanding and formal verification. Recently, due to the success of large language models (LLMs) in code generation, researchers have made early attempts to adopt LLMs for generating formal specifications. However, the lack of formal specification language corpora in the wild often makes LLMs fail to generate syntactically correct and semantically verifiable specifications. To mitigate this gap, we propose KBSpec, which augments LLMs with dual-source knowledge of formal specification language: external knowledge from official documentation, and internal knowledge distilled from verifier feedback on LLM-generated specifications. KBSpec maintains a self-evolving knowledge base that is continuously updated from successful generation and repair trajectories, without any LLM parameter tuning or labeled training data. We evaluate KBSpec on Java Modeling Language (JML) specification generation with three LLM backends, and results show that KBSpec improves verification pass rates by 10-25% over state-of-the-art LLM-based approaches, while producing the largest number of high-completeness specifications.

Massive unstructured multimodal streams suffer from high "data entropy," impeding both efficient human knowledge acquisition and high-quality AI post-training. Existing passive annotation paradigms, heavily reliant on heuristic rules or general VLMs, are costly, monotonous, and fail to unlock the deep procedural logic embedded in raw data. We elevate data processing to a learnable capability, proposing a paradigm shift towards Agentic Data Tailoring, which actively refining and structuring data to align with diverse user and downstream intents. To overcome the data scarcity bottleneck in training such high-order capabilities, we design a two-stage pipeline grounding generative semantic synthesis in deterministic Factual Anchors, yielding a large-scale dataset spanning five core physical and digital domains. Building upon this, $\text{DataClaw}_0$-9B model synergizes Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), achieving robust alignment with complex refinement and tailoring intents. To systematically quantify this capability, we construct $\text{DataClaw}_0$-val, the first benchmark dedicated to data refinement. Crucially, we adopt downstream post-training as the ultimate validation touchstone. Evaluations on video generation, real-world VQA, and GUI navigation confirm that $\text{DataClaw}_0$ delivers high-information-density tailored data, facilitating efficient model adaptation to new tasks under limited training data regimes. Project page: https://czjdsg.github.io/MakeAnyData

This paper evaluates Reservoir Computing (RC) as an autonomous, 'feature-free' framework for audio processing, designed to eliminate traditional, handcrafted feature extraction stages. We investigate whether the high-dimensional temporal dynamics inherent in a reservoir can function as a robust end-to-end processor for the direct classification of raw acoustic signals. By bypassing computationally intensive representations like MFCCs, this approach seeks to mitigate significant intellectual and pre-processing bottlenecks in traditional signal pipelines. Our study evaluates and compares shallow, sequential, and parallel deep reservoir architectures to determine their capacity for hierarchical feature representation. Experimental results demonstrate that the proposed parallel approach consistently outperforms shallow and sequential baselines while maintaining low model complexity. These findings highlight the potential of RC as an efficient and scalable alternative for time-domain audio processing, offering a promising pathway toward deployable, low-power acoustic systems with minimal preprocessing requirements.

Graph Neural Networks (GNNs) have emerged as a powerful paradigm for learning on graph-structured data by iteratively propagating and aggregating information across edges. However, conventional message passing schemes often suffer from over-squashing, whereby exponentially large neighborhoods are compressed into fixed-dimensional embeddings, impeding effective long-range dependency learning. In this work, we introduce Ramanujan Propagation, a graph rewiring strategy that leverages Ramanujan graphs to alleviate topological bottlenecks in GNNs. We first establish that suitably chosen Ramanujan graphs guarantee non-negative resistance curvature, which mitigates over-squashing and facilitates efficient information flow. We then propose an algorithmic framework to construct a Ramanujan rewired graph that preserves the local connectivity of the original graph. Our experiments demonstrate that our method outperforms nine state-of-the-art rewiring techniques. These results establish Ramanujan graphs as a rigorous structural prior for scalable, topology-aware message passing in GNNs.

Medical time series (MedTS) signals such as electroencephalography (EEG) and electrocardiography (ECG) support many clinical applications. However, substantial subject-level heterogeneity often induces subject-level distribution shift, causing a fixed parameter set to generalize poorly to unseen individuals. Compared with domain adaptation methods that often depend on extra adaptation components or target-batch statistics, Test-Time Training (TTT) provides a more practical solution for sequential clinical data by enabling online adaptation from unlabeled test samples. However, many representative TTT methods require iterative inner-loop optimization, increasing test-time overhead. In this paper, we propose MedTS-TTT, a test-time training framework for medical time series modeling. MedTS-TTT is built upon Closed-Loop Self-Alignment Test-Time Training (CLSA-TTT) and a Gated Convolutional Backbone (GCB). CLSA-TTT constructs a token-level self-supervised target and performs a single-step fast-weight update for intra-layer closed-loop alignment, enabling rapid sample-wise adaptation without iterative inner-loop optimization. GCB combines CLSA-TTT-based fast adaptation and token-level fusion with a gated convolutional branch to balance local dynamic modeling and information-flow control. On 4 public datasets (2 EEG and 2 ECG) with subject-independent splits, MedTS-TTT achieves 11 top-1 rankings out of 12 evaluations across 9 baselines and 3 metrics. The code is publicly available at https://github.com/mingzhi-c/MedTS-TTT.

We present an ML-based vessel detection and localization system, trained with weak supervision from imperfect AIS labels, that achieves a 97.8% detection rate at 1.98% false-trigger rate, successfully identifies dark-vessel events from unlabeled data.

Detecting when a nonlinear dynamical system departs from its normal regime is a recurring problem across the sciences, from cardiology to climate and energy systems. We show that a very simple Kolmogorov--Smirnov test on the output weights of a reservoir computer is highly sensitive to regime changes in nonlinear dynamical systems, including those invisible to both classical nonlinear measures and modern deep-learning detectors. The core idea of our algorithm is to treat the readout layer of a reservoir computer as a representation of the input dynamics. Since the input mapping and the reservoir itself are random and fixed, the trained output weights are the only object encoding the system at hand. We summarize this fingerprint by the empirical cumulative distribution function of the readout weights and compare it to a reference band built from the training data. This unsupervised, online detector distinguishes two visually indistinguishable butterfly-shaped attractors, resolves parameter drifts seven times smaller than the strongest deep-learning baseline, flags noise four orders of magnitude below the signal, and identifies ventricular flutter in a clinical ECG recording. More broadly, we aim to establish a perspective on reservoir computers in which the trained output weights are treated as a representation of the learned system in their own right, rather than merely as a means to forecasting.

Real-world decision-making often requires optimizing multiple competing objectives simultaneously. In reinforcement learning (RL), this is typically addressed by combining reward signals into a single scalar objective via a scalarization function, which can be fragile: small changes in the weights can induce drastically different policies. Multi-objective reinforcement learning (MORL) instead produces sets of policies that explicitly represent trade-offs between objectives. However, these policies are typically presented to the decision maker only through their value vectors, which can obscure substantial behavioral variation: policies that induce distinct trajectories may appear indistinguishable when evaluated solely by expected returns. We propose an exploratory diagnostic workflow that automatically highlights behavioral variation along the Pareto front that objective values alone do not reveal, providing both quantitative and visual tools to support policy inspection. We validate our approach on simple grid examples and scale it to continuous control benchmarks, demonstrating that it remains effective as problem complexity increases.

Recent work has raised concerns about the influence of sycophantic AI on user judgment and relationships. One proposed mitigation, which has received regulatory attention, is to warn users about potentially harmful AI behaviors such as sycophancy. In a preregistered experiment in which participants (N = 2,610) discussed real interpersonal conflicts with an AI system, we test whether warning labels mitigate sycophancy's influence. We find that a basic AI disclosure (``This chatbot is AI'') has no detectable effect. Labeling the system as sycophantic (``...may agree with you and validate you even when you are wrong...'') does shift users' perceptions, reducing perceived objectivity and trust, but it does not reliably reduce sycophancy's influence on users' self-perceived rightness or their willingness to repair the conflict. Our results reveal a gap between AI perception and AI influence: by shifting perception without reducing influence, warning-based interventions may offer a false sense of protection. Addressing the harms of sycophancy will therefore require understanding the specific mechanisms through which it shapes judgment, and improving model behavior itself.

Social intelligence is a core competency for language agents, yet current research primarily focuses on static capability evaluation rather than how these skills are continuously shaped and accumulated. This gap calls for a shift toward sustainable learning paradigms. Currently, two methodological pain points exist: social interaction trajectories lack unified structured representations to form iterable learning signals, and capability improvement and retention are typically studied in isolation, hindering the assessment of continuous evolution. To bridge this gap, we propose the Social World Model. We decompose social interaction into five dimensions (scene setting, observation, mental state, action, and dialogue) to build a closed-loop learning framework. In this setup, agents collect interaction experiences, convert them into preference signals for model updating, and redeploy the updated policy for continued learning. Additionally, we provide a reusable data synthesis mechanism and a lifelong learning benchmark, transforming social capabilities from an "object of evaluation" into an "object of sustainable training". Validating our framework on the ASCENT-Bench, the interactively trained Qwen2.5-7B model outperforms its baseline across all five core metrics. Notably, it matches the closed-source Gemini 3 Flash in completion rate, exceeds it in pass rate, and achieves zero forgetting across three difficulty levels. Unlike prior works that merely report static comparisons or capability decay, this end-to-end approach provides a trainable, verifiable, and retainable pathway, demonstrating that small open-source models can sustainably acquire competitive social coordination capabilities.

Continual learning (CL) is commonly studied under the assumption that sequential tasks are semantically related or structurally similar. However, in highly heterogeneous settings, where tasks differ substantially in reasoning patterns and input-output formats, existing methods often suffer from catastrophic forgetting and inefficient capacity allocation. To address this challenge, we propose Task-differentiated Atomic Skill Expansion and Routing (\texttt{TASER}), a CL framework that jointly determines how many new atomic skills to introduce for each task and which skills to activate. The framework first uses atomic skill incremental learning to dynamically expand capacity based on task divergence and model uncertainty. It then applies orthogonality-enhanced skill detection to ensure these skills remain semantically distinct and independently reusable. Finally, a skill dynamic routing mechanism composes task-relevant skills through lightweight task-conditioned gating. We further introduce \texttt{HeteroCLBench}, a highly heterogeneous benchmark for CL, comprising 19 diverse tasks across 9 cognitive dimensions under a standardized sequential protocol. Experiments on \texttt{HeteroCLBench} show that \texttt{TASER} consistently outperforms strong baselines by improving plasticity and reducing catastrophic forgetting.

Dysarthria severity assessment is essential for therapy planning and longitudinal monitoring, yet manual perceptual rating is time-consuming and variable across clinicians. Although deep learning models achieve strong performance, their black-box nature limits clinical adoption. Existing speech explainability methods typically provide acoustic feature importance scores that are difficult for end-users to interpret. We propose an influence-based, instance-level explainability framework that explains each decision through supportive and competing training samples. Using gradient-based influence approximations, we compute per-utterance influence scores to identify supportive and competing training samples for each prediction. Controlled deletion experiments from 5 to 20 percent validate the explanations, showing that removing highly influential samples systematically shifts predictions. This approach provides auditable explanations by linking decisions to perceptible reference cases.

Deep neural network-based automatic speaker verification (ASV) systems achieve impressive performance but their embedding representations remain opaque, lacking a structured and perceptually verifiable explanation of the vocal characteristics they encode. Existing approaches either require annotation of speaker attributes or introduce alternative representations whose interpretability is unvalidated with listeners. We propose Listenable Interpretable Speaker Embeddings (LISE), a label-free framework that decomposes pretrained speaker embeddings into a small set of components. This decomposition yields a structured representation that supports the analysis of what information has been encoded by speaker embeddings. LISE preserves ASV performance with negligible EER degradation on x-vector and ECAPA-TDNN. Crucially, the interpretability of these components for human listeners is demonstrated through listening experiments, where participants distinguished speakers with 83.9% accuracy.

Augmenting model-free reinforcement learning (RL) with representations learned through observation dynamics prediction (observation-predictive RL) can improve sample efficiency and performance, with minor modifications and limited additional computation. However, this approach still struggles in challenging tasks with low-dimensional observations. In this paper, we identify a key factor behind this problem: unbalanced reconstruction losses across observation dimensions, where dimensions with larger value ranges dominate the loss. This encourages the agent to neglect dimensions with relatively small ranges, leading to degraded performance. To address this issue, we propose a novel normalization method tailored to online RL, which normalizes low-dimensional observations and balances the resulting losses and gradients. Beyond balancing reconstruction losses, observation normalization enables dynamics prediction to be performed in a normalized observation space, thereby providing a unified treatment of low- and high-dimensional inputs (e.g., physical states and images). Building on this idea, we further introduce Normalized Observation Space Dynamics-Augmented Q-learning (NASDAQ), a framework for observation-predictive RL applicable across diverse domains. NASDAQ learns state-action representations by coupling value learning with two auxiliary tasks: short-term value prediction and next normalized observation prediction. Extensive experiments demonstrate that NASDAQ achieves competitive or superior performance compared with state-of-the-art model-based and self-predictive RL methods, while requiring significantly less training wall-time.

Existing sequence models, including RNNs, LSTMs, continuous-time networks, and Transformers, share a common structural principle: layer-wise dynamics, where all neurons in the same layer co-evolve through a shared parameterized operator, leaving individual neurons no freedom to evolve independently. Yet in many complex dynamical systems, rich global behavior emerges precisely from locally evolving units interacting through structured connectivity. Inspired by this principle, we introduce Topological Neural Dynamics (TND), a sequence modeling framework that shifts computation from layer-wise to neuron-wise dynamics. TND represents a neural system as a directed neuron graph, an interaction operator, and a local dynamics function, where each neuron evolves independently and collective computation emerges from interactions through the explicit graph topology. We instantiate TND as a discrete-time graph-coupled dynamical system and evaluate it as a case study on a behavior cloning task in single-player Pong. Compared with Vanilla RNN, Sparse RNN, LSTM, Closed-form continuous-time neural network (CfC), and Transformer baselines, TND achieves the best catch rate and a mean of 17.47 consecutive catches per round, more than three times that of the strongest baseline. These results suggest that shifting from layer-wise to neuron-wise dynamics provides an effective inductive bias for sequence modeling.

Nondestructive detection methods, based on vibrational spectroscopy, are vitally important in a wide range of applications including industrial chemistry, pharmacy and national defense. Recently, deep learning has been introduced into vibrational spectroscopy showing great potential. Different from images, text, etc. that offer large labeled data sets, vibrational spectroscopic data is very limited, which requires novel concepts beyond transfer and meta learning. To tackle this, we propose a task-enhanced augmentation network (TeaNet). The key component of TeaNet is a reconstruction module that inputs randomly masked spectra and outputs reconstructed samples that are similar to the original ones, but include additional variations learned from the domain. These augmented samples are used to train the classification model. The reconstruction and prediction parts are trained simultaneously, end-to-end with back-propagation. Results on both synthetic and real-world datasets verified the superiority of the proposed method. In the most difficult synthetic scenarios TeaNet outperformed CNN by 17%. We visualized and analysed the neuron responses of TeaNet and CNN, and found that TeaNet's ability to identify discriminant wavenumbers was excellent compared to CNN. Our approach is general and can be easily adapted to other domains, offering a solution to more accurate and interpretable few-shot learning.

Sparse rewards pose a central challenge in reinforcement learning, since agents receive no informative signal until they reach their goal. Intrinsic-reward methods address this issue by optimizing non-stationary objectives such as novelty, prediction error, or skill diversity, thereby injecting a supervision signal into the problem. While effective, these methods often require that the extrinsic (sparse) reward can be evaluated -- either online or during offline relabeling of the stored transitions. This limitation is particularly vexing for multi-task, meta-, and continual reinforcement learning, where agents' interactions with the environment are usually reward-free. In this work, we present a method to pre-train transferable exploration policies that rapidly adapt to sparse rewards at downstream task time. Our objective maximizes state-space covering for the occupancy measure, and can be framed in terms of entropy maximization. Its algorithmic implementation, ROVER, leverages recent advances on the operatorial formulation of RL to estimate occupancy with a learned resolvent world model, bypassing common hurdles associated with density and entropy estimation. ROVER further introduces a virtual "sink" state for unexplored regions, balancing coverage of known states with expansion into unseen ones and preventing cyclic expansion-collapse behavior during learning. In tabular and pixel-based sparse navigation tasks, ROVER produces more uniform aggregate coverage and stronger initializations for downstream tasks than standard reward-free baselines.

Early detection of aphid infestation in crops is essential for preventing yield loss and reducing unnecessary pesticide use. Hyperspectral imaging combined with Spectral Information Divergence (SID) analysis offers a non-destructive approach for monitoring plant health; however, deep learning methods applied to hyperspectral data are often limited by small dataset sizes. In this study, a data-efficient generative adversarial network (FastGAN) was employed to augment a hyperspectral SID dataset of faba bean leaves containing healthy and aphid-infested samples. The trained generator produced 10,000 synthetic images preserving structural and spectral characteristics of real samples. Image quality was evaluated using Frechet Inception Distance (FID), demonstrating stable convergence and realistic reconstruction of leaf morphology and infestation patterns. The augmented dataset was used to train four classification architectures: VGG16, ResNet-50, EfficientNet, and Vision Transformer (ViT). Results showed that dataset augmentation significantly improved classification robustness, with performance progressively increasing from classical convolutional networks to transformer-based models. The ViT model achieved the highest accuracy and F1-scores, while EfficientNet provided strong balanced performance and ResNet-50 showed moderate improvements over VGG16. Confusion matrix analysis confirmed reduced false negatives and improved disease detection when using advanced architectures. The findings demonstrate that FastGAN-based augmentation effectively enhances hyperspectral plant disease classification and that transformer-based models provide the most reliable discrimination between healthy and infested leaves.

Reinforcement learning for multi-step LLM agents often relies on scalar rewards that indicate success but cannot explain why a trajectory is good or bad. Rubric-based rewards improve interpretability through natural-language criteria, but existing methods score at the trajectory level and freeze the scorer behind a closed-source judge, leaving step-level credit assignment unresolved and the judge itself static. We propose ARCO (Adaptive Rubric CO-evolution), a rubric framework in which a same-scale model $μ$ shares a backbone with two heads: a generation head that produces per-step criteria, and a score head that predicts rubric-conditioned step-level rewards. A trajectory decomposition constraint ties the sum of step rewards to the terminal outcome, enabling credit assignment without step-level labels, while $μ$ and the policy $π$ are jointly updated on on-policy data so that the rubric content and the scoring function co-evolve at the parameter level. Across HotpotQA, 2WikiMultiHopQA, and MuSiQue with two open-source backbones, ARCO improves the best EM in every setting over strong outcome-, rubric-, and process-reward baselines, and analyses show that its rubrics are step-specific, robust to design choices, and useful for diagnosing agent behavior. Codes and data are available at https://github.com/zihangtian/ARCO.

In the era of big data, subsampling became a common practice in statistical learning. By selecting a subgroup of individuals based on which the learner is trained, subsampling aims at reducing the computational cost and time of the estimation step, and ideally leads to a decrease of its energy consumption and carbon footprint. This work focuses on a nonparametric setting, in which the hypotheses set lies in a reproducing kernel Hilbert space, and the estimator is a minimizer of an empirical risk reweighted à la Horvitz-Thompson. By studying the asymptotic properties of this estimator, we reveal an optimal subsampling scheme (regarding the trace of the covariance operator) and show that it can be used via plug-in. A numerical study on synthetic and real-world datasets shows the practicability and the benefit of the proposed approach.

OpenPangu models are attractive targets for private and domestic large-language-model deployment, yet their robustness under aggressive post-training quantization on Ascend NPUs has not been systematically characterized. This paper conducts a controlled empirical study of OpenPangu 1B and 7B models on Huawei Ascend 910B1 NPUs. We evaluate representative weight-only and weight-activation post-training quantization methods, including RTN, GPTQ, AWQ, SmoothQuant, GPTAQ, BiLLM, and SliM-LLM, under a unified calibration and evaluation protocol. Across 18 evaluation tasks, we find that 8-bit weight-only quantization is effectively lossless for both models, while 4-bit quantization remains practical for the 7B model but is visibly more harmful for the 1B model on reasoning, math, and code tasks. Ultra-low precision remains challenging: most 2-bit and binary settings collapse to near-random behavior, and W4A4 SmoothQuant produces non-finite perplexity in our evaluation. These results provide an NPU-oriented accuracy map for selecting OpenPangu quantization settings and highlight the persistent difficulty of extreme low-bit compression.

Quantum pure-state ensembles live on complex projective space, making flat Euclidean generative modeling geometrically mismatched. We introduce Intrinsic Flow Matching (IFM), a deterministic transport framework on $\mathbb{CP}^{d-1}$ that learns tangent velocity fields using Pancharatnam phase-aligned conditional paths. IFM replaces local score teachers and reverse-time stochastic sampling with manifold probability flow, while horizontal parameterization removes redundant ambient directions. We show that the IFM objective recovers the induced marginal transport field, represents deterministic projective ensemble flows, and yields endpoint and stability guarantees. Empirically, IFM often improves over ambient Euclidean flow matching across higher-qubit, multimodal, spin-coherent, physics-inspired, and amplitude-encoded MNIST image-vector benchmarks, with strongest gains on high-dimensional and coherence-sensitive tasks but not uniformly across every metric.

Rejecting inputs outside the defined in-distribution (IND) service scope is critical for large language model (LLM) services, where unsupported requests should be filtered before full generation. Existing out-of-distribution (OOD) detectors often rely on final outputs or final-layer representations, leaving unclear where service-boundary signals are most clearly encoded inside the model; they also lack a theoretical guarantee for held-out inputs. In this paper, we introduce SCOPE (Sequential Conformal OOD Probing and Evaluation), a framework that selects a readable hidden layer, constructs a conformal gate with IND calibration, and uses a supermartingale e-process to certify persistent service-boundary evidence. Experiments across multiple LLM backbones and six carefully designed boundary conditions show that SCOPE improves gate-level rejection over standard final-layer detectors, while revealing how different OOD boundaries take different geometric forms in hidden space.

Continual learning that is gradient-free, local, online, and append-only is attractive for edge and streaming deployment, but its value is usually argued informally. We give a provable account on recurring-regime streams. Given segmentation, a warm-start library learner attains amortized recovery cost $O\!\big(KD/\varepsilon^2+(R-K)\logK/Δ^2\big)$ versus a memoryless re-estimator's $Θ(RD/\varepsilon^2)$, an advantage $(R-K)\,Θ(D/\varepsilon^2)$ growing with dimension $D$ and recurrence density. The mechanism is a decoupling: recognizing which of $K$ seen regimes is active costs $O(\log K/Δ^2)$, independent of $D$, whereas estimating a regime costs $Θ(D/\varepsilon^2)$. We prove this is tight: matching lower bounds give recognition $Θ(\log K/Δ^2)$ and a memoryless-class bound $Ω(RD/\varepsilon^2)$, so each term is individually minimax-tight (the joint statement is conditional). The separation is born-immune (a memoryless learner's advantage is identically zero) and paradigm-level: it matches, and does not beat, a fair spawn-capable Bayesian baseline; the contribution is attaining this cost structure without end-to-end backprop and with zero forgetting by construction. A count-calibrated variant ties the baseline's leading constant up to a bounded, never-negative per-recurrence overshoot, hyperparameter-free and with no per-step transcendentals. We bound the scope: recognizable regimes are capped by simplex packing (walls $e^{Θ(D)}$); autonomous segmentation is impossible at the packing wall (no detector escapes the false-alarm/delay frontier as regimes overlap); the advantage vanishes under overlap. The dimension-dependent separation is corroborated on synthetic streams and real $k$-mer genome distributions (memoryless cost $\propto D^{1.04}$, recognition $D$-independent); the one real sequential stream sits in the $D{=}1$ near-null corner.

Retrieval heads, attention heads that copy information from earlier context to the current position, have been proposed as the mechanistic substrate for long-context recall. Rotary position embeddings (RoPE) rotate queries and keys by frequencies decaying with a base hyperparameter theta, and a natural hypothesis is that this rotation either prevents retrieval heads from forming or degrades their function. We test both across four open-weight 7-8B models spanning multi-head and grouped-query attention and a 100x range of theta, using paired-seed needle-in-a-haystack tests, layer-clustered permutation, and causal head-masking. (i) Retrieval heads are causally necessary: masking the 87 detected heads in OLMo-2 collapses recall from 1.00 to 0.00, while masking matched random heads has no effect; this replicates in Qwen. (ii) Higher theta does not reduce retrieval-head count (LLaMA-3.1 at theta=500K has 47 heads vs LLaMA-2 at theta=10K with 42), refuting the prevention hypothesis. (iii) The norm-utility relation is family-specific and significant in opposite directions (Qwen d=-0.49, OLMo d=+0.50, both significant; LLaMA null); since OLMo and LLaMA-3.1 share theta=500K yet differ, the effect is not theta-driven. (iv) Building on Chiang and Yogatama (2025), a controlled patch shows that zeroing the lowest-frequency RoPE dimensions of retrieval heads degrades recall dose-dependently (1.00 to 0.18 when 32 of 128 dimensions are zeroed, vs 0.98 for random dimensions); the effect is head-specific and task-specific. The causal variable is RoPE frequency, not norm-utility. The direction holds in all five models patched (OLMo-2, Qwen2.5-7B/14B, Gemma-2, Mistral) across four lineages and two scales. We do not claim cross-model magnitude. Code and a paired-seed harness are released.

Public sector organizations increasingly procure AI-enabled ICT systems to support decision-making and service delivery. Although ethical AI frameworks emphasize transparency, accountability, and human oversight, these principles are rarely translated into explicit requirements in procurement processes. Consequently, human-AI interaction (HAI) is often left to vendor design choices. This paper conceptualizes HAI as a procurement-critical design dimension and proposes a taxonomy of interaction requirements tailored to public sector ICT procurement. The taxonomy enables contracting authorities to specify and govern interaction properties through procurement instruments, supporting both ethical compliance and sustainable value realization.

3D Gaussian Splatting achieves exceptional real-time rendering, but its substantial computational and storage demands hinder widespread deployment. Existing accelerated paradigms often aggressively prune primitives for rapid convergence, causing severe loss of high-frequency details. To address this, we tackle the fundamental problem of achieving both exceptional rendering quality and ultra-fast reconstruction speed. In this paper, we propose ACE-GS, a progressive optimization framework tailored for accurate, compressed, and efficient scene representation. We realize that precise primitive management is the key to breaking this trade-off. Therefore, we first design a momentum consistency-guided densification strategy, strictly constraining primitive growth onto authentic geometric manifolds to avoid computational waste while significantly accelerating convergence. Building upon this efficient initialization, we deploy a statistical sensitivity-driven sparsification mechanism to precisely prune redundant primitives, yielding a further compressed footprint. Finally, to thoroughly compensate for the risk of micro-structure loss caused by the aforementioned strict primitive control, we introduce a cross-dimensional residual frequency compensation scheme that explicitly back-injects high-frequency error energy into primitive attributes, perfectly restoring sharp geometric details. Extensive experiments validate our superiority. While maintaining a highly compact scene representation, our system achieves up to 3.7 times training acceleration against the rapid framework Speedy-Splat. Requiring only 3 to 5 minutes to converge, ACE-GS secures the highest structural similarity and achieves a peak PSNR improvement of up to 0.89 dB over the original 3DGS, establishing a new benchmark for ultra-fast and high-fidelity novel view synthesis.

Key-value (KV) caching is a powerful technique for accelerating large language model inference and generation. Inference workloads are large and diverse, which makes them difficult to cache effectively. Existing cache management strategies adopt the least-recently-used policy for evicting cache blocks. However, LRU leads to multiple unrelated workloads flushing each other's caches. To address this, we integrate adaptive caching that dynamically allocates cache space between recently and frequently occurring KV blocks. Evaluations show that it improves the KV cache hit rate by up to 10.8% and reduces time to first token by up to 12.6% over naive vLLM on synthetic document question answering workloads, and 2.1% and 2.0% respectively on real-world conversation workloads. The method generalizes well to batch inference and demonstrates clear interpretability while effectively accommodating diverse workloads.

Advances in deep learning and end-to-end Automatic Speech Recognition (ASR) have enabled robust multilingual models, but evaluation metrics remain limited in assessing accuracy. Efforts to improve or replace the common metric Word Error Rate (WER) often focus on English, leaving evaluations for low-resource languages under-explored and hindering fair cross-lingual comparisons. We present OpenWER, an open-source implementation that improves WER robustness through language-specific normalisation and compound word detection. A token-based Levenshtein alignment preserves complementary metrics and allows metadata embedding for granular accuracy scores. Our analysis of 52 languages shows absolute WER reductions of up to 25% compared to common libraries. OpenWER contributes to fairness in ASR research by increasing the reliability of WER across diverse languages and enabling more comprehensive accuracy evaluations.

Wound rotor synchronous motors have emerged as a strong alternative that eliminates dependence on REEs. However, WRSM design requires the simultaneous optimization of numerous geometric and electromagnetic parameters, and the high computational cost of conventional finite element analysis severely limits the rapid exploration of the large parameter space. Although there are machine-learning-based surrogate modeling studies in the literature, they generally compare only a limited number of models, exclude deep learning architectures, and do not provide a comprehensive benchmark specific to WRSM. In this study, the performance of a total of eight machine learning and deep learning models from four different algorithmic families was systematically compared for the prediction of WRSM torque and motor efficiency. On a dataset of 3351 samples generated using Latin Hypercube Sampling in the Motor-CAD simulation environment, each model was trained with 10 different random seed values and tuned via Optuna hyperparameter optimization. Different from the existing literature, this study jointly offers a broad model spectrum including recent deep learning architectures such as FT Transformer, a multi-seed reproducibility protocol, and a Pareto analysis of the computational cost-accuracy trade-off. The results revealed that neural-network-based models systematically outperform tree-based models. The FT-Transformer model achieved the highest single-model accuracy with R^2 = 0.9928, producing predictions in 0.33 milliseconds and thus obtaining several orders of magnitude speedup compared to FEA. Model performances were evaluated in a multidimensional manner using R^2, MAE, RMSE, and MAPE metrics.

The capabilities of frontier Large Language Models (LLMs) continue to advance, with different providers increasingly specializing in distinct domains. This raises a natural next objective: how to combine the individual specializations of various LLMs into a collectively intelligent system. To this end, we report the development of Sakana Fugu, a family of orchestrator models that harness and amplify the capabilities of an LLM agent team. Fugu models are themselves language models trained to understand user queries and dynamically devise agentic scaffolds to solve them. Through these adaptive scaffolds, Fugu accesses performance beyond any individual LLM agent, achieving state-of-the-art results compared to other publicly accessible models across a range of challenging tasks, including SWE-Bench Pro, Terminal Bench, LiveCodeBench, GPQA-Diamond, Humanity's Last Exam, and CharXiv Reasoning. We release two models: Fugu, which balances performance with latency for everyday use, and Fugu-Ultra, which prioritizes answer quality on the hardest problems. We describe our training paradigm, which encompasses large-scale fine-tuning, evolutionary algorithms, and reinforcement learning approaches, along with the infrastructure and core design principles that turn these methods into a production system. We hope this report encourages further research into multi-agent systems and dynamic, query-adaptive agentic scaffolds as a path toward the next frontier of AI capabilities, accessed through collective intelligence.

Large-scale autonomous fleets rely on teleoperation to resolve rare failures, yet streaming raw sensor data from many vehicles is costly, and remote operators can only monitor a limited number of vehicles at a time. We introduce FleetAgent, a cloud-hosted multimodal large language model (MLLM) assistant that consumes compact vectorized vehicle-to-network (V2N) messages, such as map elements, detected objects, and the ego planned path. It provides a structured natural-language response (including narration, explanation, and evaluation of the plan and scene), along with an intervention urgency score for operator prioritization. To make structured messages compatible with token-based MLLMs, we propose VecFormer, a vector-to-embedding interface with differentiable top-K context selection that bounds context length and GPU KV-cache growth, enabling more efficient batch processing, which is important under the context of cloud-hosted large-scale fleet management. We also construct VecEval, a nuScenes-derived dataset with paired human and synthetic imperfect plans and human-verified language labels, to facilitate the training and evaluation of our proposed system. Our proposed system can reduce uplink payload by up to 625 times compared with raw images and reduce KV-cache memory by 16.54 times compared with original text descriptions. On VecEval, FleetAgent improves Lingo-Judge score by 16.8% and reduces intervention failure rate by 19.9%, compared with Qwen2.5-VL-7B using language descriptions. These results demonstrate that FleetAgent can utilize compact structured V2N messaging to enable efficient, explainable teleoperation monitoring for autonomous fleets.

As expressive text-to-speech (TTS) and voice conversion (VC) systems increasingly generate non-verbal vocalizations (NVVs) to enhance naturalness, reliable speaker verification (SV) becomes essential to objectively assess identity consistency across both verbal and non-verbal segments. Yet current SV systems generalize poorly to NVVs, and fine-tuning on NVV data causes catastrophic forgetting of speech performance. We present the first systematic study across 10 NVV types and propose a framework combining frozen Data2Vec self-supervised features with ECAPA-TDNN, enhanced by a Mixture of Experts (MoE) module with learned domain-aware routing. A conditional distillation loss on speech inputs via a pretrained teacher retains speech-to-speech accuracy, while a contrastive loss bridges the speech-NVV domain gap. Our method reduces speech-NVV EER from 38.93% to 22.66% over a pretrained baseline, and improves speech EER from 13.17% to 9.24% via distillation.

Small quantitative structure-activity relationship (QSAR) studies are difficult when close molecular analogues have different activity labels. This paper asks whether a quantum kernel can add similarity information to a Morgan/Tanimoto fingerprint model, and which molecules account for the change. QBioFusion-QSAR uses quantum multiple kernel learning (QMKL): a support vector machine combines a Morgan/Tanimoto kernel with a quantum fidelity kernel constructed from fold-local components derived from RDKit and Mordred descriptors and Deep-PK features. Linear and radial basis function descriptor kernels are included as classical controls. On the 54-molecule PsychLight-A benchmark, Morgan/Tanimoto was the strongest single representation. In the primary stratified five-fold evaluation, QMKL increased accuracy from 0.815 to 0.833 and Matthews correlation coefficient (MCC) from 0.613 to 0.645. Matched-regularization auditing attributed the change to N-Me-5-HT and N-Me-tryptamine changing from false-negative to true-positive predictions; activity-cliff subset MCC increased from 0.07 to 0.22. Repeating the five-fold protocol over ten random partitionings showed that learned QMKL did not exceed Morgan/Tanimoto on mean MCC; paired held-out bootstrap intervals for the matched comparison also span zero. These results support QBioFusion-QSAR as an auditable QMKL framework for identifying localized residual quantum-kernel contributions in small-data, activity-cliff-aware ligand classification.

Causal discovery is critical for understanding complex data-generating mechanisms, yet traditional algorithms often struggle with highly non-linear and noisy systems, or suffer from severe computational bottlenecks. Recent tabular foundation models based on Prior-Data Fitted Networks (PFNs) have demonstrated remarkable zero-shot inference capabilities, but their potential for explicit structural causal discovery remains underexplored. To bridge this gap, we propose DCD-PFN, a decoupling-aware foundation model for causal discovery. Instead of directly amortizing global graph reconstruction, DCD-PFN focuses on local causal discovery through a decoupling-based paradigm. Through pre-training on diverse synthetic Structural Causal Models (SCMs), the model learns sample-wise decoupling weights that enable Markov boundary (MB) identification. Furthermore, by leveraging parallelized local discovery, DCD-PFN efficiently reconstructs global causal graphs while remaining grounded in the theoretical foundations of decoupling-based causal discovery. Experiments demonstrate that our foundation model achieves robust zero-shot generalization.

Psycholinguistics studies show that human readers fall for coherence illusions: an incoherent discourse can seem coherent simply because a distractor matches what comes next. We investigate whether Dutch language models (6 monolingual and 4 multilingual) show the same behavior on texts that link back to earlier context with words such as 'again' and 'too'. First, we find that surprisal at the critical word tracks human acceptability judgments and eye-tracking data. Models are more surprised by incoherent continuations, but a matching distractor in the prior context reduces this surprisal. Second, attention entropy at the critical position identifies heads that behave differently under coherence vs. incoherence. We find that ablating these heads shows transfer effects across experiments, suggesting a shared mechanism. Third, we introduce energy from the associative-memory literature as a metric to quantify discourse coherence. Taken together, our results show that coherence illusions arise in Dutch LLMs, with entropy and energy exposing mechanisms that operate across settings.

Reaching global agreement from purely local interactions is a defining problem of collective intelligence, and most models of it assume that all agents share a single communication protocol. We ask what happens when they do not. Using a Neural Cellular Automaton in which a population of cells must solve the density classification task, agreeing on a global majority that no individual can observe, we introduce ``languages'' as sub-populations that read one another's messages through a translation with a tunable ``linguistic distance''. We find that linguistic distance slows consensus, that it produces mild divergence between groups rather than full fragmentation, and that a collective whose shared rule was trained under diverse protocols is robust to mismatch; a homogeneously trained one is not. The findings hold on both a ring and a two-dimensional grid, and admit a natural reading as Ising relaxation, in which a foreign-language region acts as a boundary defect that leaves the system in a higher-energy, partially ordered state. These patterns are qualitatively consistent with effects reported in human group studies, suggesting that distance between communication protocols is a minimal mechanism sufficient to produce them, without anything language-specific.

Artificial intelligent agents and autonomous systems are embedded in our environments. They are both a commercial product and a personal tool that generates a lot of data and can draw conclusions from it: machines generate and keep secrets. But should machines protect all secrets? It has been shown that artificial agents are able to whistleblow and it has been argued that digital multi-agent environments should allow for agents in them to whistleblow. We argue that machine whistleblowing must be normative and principled and routed in the existing understanding of whistleblowing as an important rule-breaking mechanism in society. We also argue that there is a need for government regulators to formulate an informed stance on both what machines should be allowed to whistleblow on and how to legally protect those who develop whistleblowing machines

Pedestrian attribute recognition (PAR) assigns semantic labels to detected pedestrians and is useful in surveillance, video retrieval, and human-centered graphics applications. This paper presents a two-stage framework in which YOLOv8n detects pedestrians and ResNet18-based models classify gender, estimate apparent age, and predict 61 binary attributes from each pedestrian crop. PETA and PA-100K are combined through semantic attribute mapping, producing a unified training corpus of more than 100,000 pedestrian images while retaining the PETA attribute space. On the reported test splits, the system obtains 99.89% gender classification accuracy, a 4.23-year apparent-age mean absolute error, and 89.96% multi-attribute accuracy with a 36.32% macro F1-score and 58.80% micro F1-score. Runtime measurements indicate 25-30 FPS on an NVIDIA RTX 5060 GPU. The results show that a lightweight detector-classifier pipeline can support real-time PAR, while low macro F1 indicates that rare attributes remain challenging.

We introduce a semi-parametric framework for nonlinear system identification, which decouples discrepancy functions from physics-based components. Orthogonal Gaussian process regression balances sparse parameter selection (the white box) with discrepancy learning (the black box) to produce interpretable models from incomplete physics.

Vision Language Models (VLMs) have demonstrated impressive performance in tasks requiring joint understanding of images and text, such as image captioning and Visual Question Answering (VQA), but our understanding of their internal processes remains limited. Recently, Sparse Autoencoders (SAEs) have emerged as a promising tool to support the interpretation of concepts encoded in VLMs. However, most SAE-based approaches focus only on textual or visual concepts separately, ignoring multimodal concepts. This limitation hinders a comprehensive understanding of VLMs, since concepts that integrate both modalities can be misclassified. Moreover, previous visual approaches often produce low-quality visual concept descriptions that are vague or incomplete, limiting their usefulness for understanding model reasoning. We propose a framework based on SAEs to extract and analyze visual, textual, and multimodal concepts from VLMs. For each neuron, we propose a candidate human-interpretable concept and compute the alignment between the concept and the dataset samples using cosine similarity scores. Experiments on a VQA dataset (LLaVA-NeXT) demonstrate that our framework improves visual concept quality by up to 45\% compared to existing SAE-based methods, while maintaining high textual concept quality and enabling systematic identification of multimodal concepts. This work contributes new insights into the conceptual space of VLMs, providing a structured approach to distinguish between visual, textual, and multimodal concepts. The code is available at https://github.com/PHDLanza/Multidata_SAE

This paper revisits the grounding problem for large language models in light of recent vector-grounding accounts. I accept the shift from classical symbol grounding to vector grounding, but argue that the current debate remains incomplete in two respects. First, reference is often treated too thinly, as if it were a fixed link between an isolated expression and an object. I argue instead that reference is profile-based, context-sensitive, discourse-level, affectively shaped, and norm-governed. Even in the human case, reference is publicly stabilized through patterns of use, correction, distinction, inference, and continuation rather than through identical private representations. Second, vector grounding requires an account of numerical realization. LLMs do not acquire reference through human perception, memory, intention, embodiment, or understanding. Rather, through optimization, they parameterize linguistic traces of human world-directed practice. In a finite vector system, referential profiles must be distributed, may be superposed, and are recovered through context-sensitive computation. Weights, activations, attention-mediated hidden states, softmax-trained contrasts, and inner-product alignments are the mathematical sites at which inherited linguistic relations become stable and causally active. Mechanistic interpretability findings, including entity-like features, knowledge neurons, and emotion-related activation directions, provide indirect support for this view. They do not show that LLMs possess human reference. They support a more limited thesis: LLMs may possess derivative, language-mediated, profile-based, and numerically structured forms of reference.

Medical Vision-Language Models (Med-VLMs) achieve strong expert-level performance, yet their ability to generate patient-accessible descriptions remains underexplored. With the 21st Century Cures Act now mandating immediate patient access to diagnostic imaging results, evaluating whether Med-VLMs can bridge this Expert-Lay Gap is both urgent and clinically consequential for patient education and shared decision-making. To this end, we introduce MedLayXPlain, the first large-scale multimodal benchmark and evaluation framework for Medical Lay Language Generation (MLLG). MedLayXPlain-122K provides 122,789 region-grounded samples across 8 imaging modalities from 12 publicly available source datasets, each comprising a medical image with paired expert and lay captions anchored in a three-level Unified Medical Language System (UMLS) ontology hierarchy spanning 7 semantic groups, 43 semantic types, and 2,411 medical concepts. Lay captions are constructed via Hierarchical Ontology-Verified Refinement (HOVER), a three-step pipeline combining patient-centric vocabulary mapping, LLM-based constrained rewriting, and cross-model visual verification to enforce semantic equivalence while preventing hallucination. We further introduce MedLayEval, a lightweight 3B evaluator distilled from a 27B verifier that scores expert-lay alignment across five clinically grounded attributes, addressing the poor correlation between standard NLG metrics and clinical judgment. Benchmarking 33 VLMs on MedLayXPlain-122K reveals a systematic Expert-Lay Gap: medical VLMs achieve strong expert captioning but suffer significant lay-register degradation, while general-purpose VLMs produce more accessible language yet lack clinical precision, confirming that neither current paradigm adequately serves patient-facing communication.

Non-stationary partial differential equations (PDEs) arise throughout scientific computing, where the dominant frequency content and energy distribution can drift over time. While efficient in PDE solving, many spectral neural operators apply a shared spectral response across rollout stages, leading to mismatch with time-varying spectra in non-stationary systems. To address this issue, we propose Time-Frequency Gated Spectral Neural Operator (TF-SNO), a state-adaptive framework with learnable time-frequency gating inside spectral blocks. TF-SNO extracts compact frequency-domain and physical-space statistics from the current state to generate modulation coefficients, enabling the spectral response to evolve with the dynamics. TF-SNO learns temporal variation implicitly from the evolving state without introducing an explicit time dimension or time embedding, keeping the modeling complexity low. We further embed the adaptive operator blocks to accurately capture the multi-scale features, thereby improving long-horizon stability. Experiments on six non-stationary PDE benchmarks in 1D and 2D demonstrate that TF-SNO significantly reduces prediction errors and improves robustness compared to strong baselines, with particularly clear gains in long rollout, suggesting the effectiveness of state-dependent spectral adaptation in modeling non-stationary physical systems.

Understanding the downstream effects of code changes is essential for software maintenance, debugging, and regression testing. This task is particularly challenging for JavaScript applications, where dynamic language features such as callbacks, events, asynchronous execution, and shared mutable state make dependencies difficult to infer precisely. Existing change impact recommendation approaches rely primarily on either dependency-based analysis or repository mining. Dependency-based techniques, particularly dynamic analysis, capture runtime interactions from observed execution but may miss relationships not exercised during analysis. In contrast, history-based techniques uncover evolutionary coupling from past changes but often introduce imprecise recommendations due to noisy co-change patterns. To investigate the strengths and limitations of these approaches in JavaScript, we engineer and evaluate three recommendation techniques: a history-based approach using co-change pattern mining, a dynamic dependency-based approach, and a hybrid approach combining both signals. We implement these techniques in a unified framework, Caprese, and evaluate them on 10 open-source Node.js applications using expert-curated reference inspection sets. Our results reveal low overlap between candidates identified by history-based and dynamic analyses, with only 22% overlap at broader inspection budgets, indicating that the two approaches capture complementary impact signals. Dynamic analysis generally yields higher precision, while history-based analysis identifies additional relevant candidates missed by dependency analysis. These findings suggest that practical change impact recommendation in JavaScript benefits from combining runtime and evolutionary signals, as no single technique sufficiently captures all relevant inspection candidates.

There is a precise sense in which drawing causal inferences from observational data is hard, even when identifiability is assumed. In particular, Robins and Ritov (1997) and Robins et al. (2003) showed that causal effects can be discontinuous as a function of the data distribution: two arbitrarily close data distributions might correspond to different causal effects. This is a fact independent of the choice of estimator; however, not all estimators are equally unstable. Our contribution is to surface a second layer of instability that depends on the choice of estimator. We show that many standard point estimates can be read as point summaries of multimodal distributions over the space of structural causal models. As such, estimators can jump discontinuously in the data distribution. This defines a taxonomy of estimators that admits a decision-theoretic reading: stability depends on whether the implicit loss function an estimator optimizes is aligned with the causal effect itself. Specifically, inverse propensity weighted estimators and regression estimators are examples of discontinuous summaries, while explicit posterior means and medians are shown to be continuous.

Soil properties relevant to agricultural and environmental applications are conventionally measured using elaborate laboratory methods involving physical and chemical processing. While highly accurate, these conventional methods are costly and time-consuming. In contrast, optical spectroscopy paired with machine learning enables rapid and cost-effective predictions of multiple soil properties. However, spectroscopic modelling is often considered unreliable, as the predictive accuracy varies between soil properties and individual samples. To balance this trade-off between cost and reliability, we introduce reject-to-remeasure: an AI-based measurement framework that combines probabilistic modelling with uncertainty-guided rejection. In this framework, soil samples are first analysed using spectroscopy, after which predictions are rejected if their predictive uncertainty exceeds predefined quality constraints. Rejected samples are subsequently remeasured using conventional laboratory procedures. On a regional visible-near-infrared spectral soil library from Québec, we demonstrate that reject-to-remeasure with modern foundation models (TabPFNv2.5 and TabICLv2) can facilitate the integration of optical spectroscopy into routine laboratory workflows while meeting user-defined accuracy requirements and reducing measurement costs.

Segmenting the temporomandibular joint (TMJ) disc from MRI is essential for accurate diagnosis of internal derangement, yet it remains unreliable in practice due to its small size, low contrast, and morphological variability. Existing methods, primarily adapted from general segmentation architectures, often produce fragmented or anatomically inconsistent masks, leading to unstable measurements of disc position and shape for downstream diagnosis. To address these challenges, we propose TISC, a TMJ disc segmentation framework that integrates semantic anchoring with clinical metadata-guided boundary refinement. The framework first establishes robust disc localization in the foundation model feature space via a Prototypical Semantic Anchoring (PSA) module that aggregates adjacent-slice MedDINOv3 features and derives a prototype-driven similarity map. It then performs targeted boundary refinement through a Clinical-Metadata Point Refinement (C-MPR) module, with point-wise predictions modulated by Mouth Open Limitation (MOL), a clinical indicator associated with disc displacement without reduction. On a large-scale cohort of 2,488 PD MRI volumes from 1,300 patients, our method achieves up to a 4.96 Dice improvement over strong baselines across diverse architectures, delivering more anatomically coherent and clinically reliable TMJ disc segmentation.

Model-based and model-free reinforcement learning are traditionally viewed as separate paradigms: instead of learning a model of the transition kernel $P$, model-free agents typically estimate value functions tied to a specific policy and reward. In this paper, we challenge this dichotomy by proving that value-based agents trained on a sufficiently rich set of reward functions, e.g. using goal-conditioned RL, implicitly encode a unique and accurate world model. To extract this model in practice, we introduce \textit{$P$-learning}, an inverse analogue to $Q$-learning that samples from an agent's $Q$-values, policies and rewards to decode its internal model of the environment. We then provide sufficient conditions on the type and number of goals for which agents encode the true kernel $P$, covering both stochastic and deterministic MDPs over finite or continuous state spaces. Even when our assumptions are violated, we empirically demonstrate that agents trained on a handful of reward functions encode accurate dynamics in $\texttt{Reacher}$, $\texttt{MountainCar}$ and stochastic variants of $\texttt{FourRooms}$. Surprisingly, we find that policies trained exclusively on a \texttt{Reacher} agent's implicit world model are quasi-optimal on out-of-distribution, velocity-based goals despite position-only training -- suggesting that agents contain hidden generalisation capabilities and providing a new lens into the connection between model-based, model-free, and goal-conditioned RL.

Large language models (LLMs) are increasingly used to support software development, but their practical usefulness in applied game-development settings remains underexplored, especially when generated code must be integrated into an existing game software system. This paper presents an exploratory empirical case study of GPT-4o in a custom Python/Pygame endless runner. The study examines six selected development tasks: three localized refactoring tasks and three tasks involving gameplay feature generation. The resulting implementations were evaluated using software metrics, unit tests, and manual gameplay assessments. In this case study, all three selected refactoring tasks were completed successfully in functional terms, whereas only one of the three selected gameplay feature generation tasks resulted in a correctly integrated feature. The findings suggest that, in this setting, GPT-4o handled localized transformations more reliably than tasks requiring new gameplay interactions across multiple existing systems. Given the exploratory single-case design, these results are best interpreted as indicative observations rather than as generalizable evidence of category-level model performance. Overall, the paper contributes a transparent case-based account of the opportunities and limitations of LLM-assisted refactoring and gameplay feature generation in an existing game software system.

Interactive travel planning has become a popular use case for language models. Agents are deployed to manage evolving preferences and unexpected disruptions over multiple turns. Such settings require models to make complex, profile-conditioned planning decisions. However, existing benchmarks often evaluate feasibility, personalization, or interaction in relatively isolated settings. We therefore introduce Trip+ to measure the ability of agents to plan travel holistically. In Trip+, given traveler profiles and dynamic interactions, agents must generate and revise minute-level itineraries. End-to-end traveler experiences are evaluated via an LLM-based simulator, enabling the assessment of subjective metrics like fatigue. Our scenarios range from simple request resolutions to complex environment-driven replanning. We evaluate 18 LMs and find a consistent gap in experiential quality. Models favor technically feasible but exhausting itineraries that diverge sharply from profiled traveler preferences.

Dementia diagnosis requires integrating multi-modal clinical assessments from diverse informants and clinicians under incomplete and heterogeneous data conditions. Yet most AI-driven approaches remain Alzheimer's disease (AD)-centric, framing the problem as binary AD detection or three-stage AD progression modeling within well-curated research settings. This pathology-driven paradigm overlooks the broader, syndrome-level nature of dementia, which spans multiple stages, phenotypes, and etiologies. In this paper, we propose Dementia-Agents, a clinically aligned multi-agent framework for real-world dementia staging and phenotyping. The framework follows a three-step workflow: (1) a data agent translates structured clinical records into semantically faithful textual representations that preserve missing-data signals and routes them to domain-aligned experts; (2) five fine-tuned expert agents generate domain-level predictions; and (3) a coordinator agent performs probabilistic aggregation to produce final staging and phenotyping decisions. We develop and evaluate Dementia-Agents on a real-world clinical cohort of 1,066 patients from two cognitive neurology services. Compared with monolithic multi-modal large language models (MLLMs) and prior medical multi-agent systems, our approach achieves consistent improvements in diagnostic performance for real-world syndrome-level dementia staging and phenotyping, while preserving domain-level interpretability.

We present OmniV2X, a generative foundation model for vehicle-to-everything (V2X) cooperative driving. The model directly interprets independent context sequences comprising multi-modal and multi-agent observations. The new design mitigates the computational cost of dense 3D perception, the vulnerability to data scarcity in cooperative scenarios, and the poor compliance with standardized messaging in existing methods that fuse multi-modal inputs into a shared representation. For training, we present an end-to-end supervised pipeline using a downstream trajectory generation loss, in which a high-capacity generative sequence planner implicitly learns to steer the model and leverage multi-modal inputs via cross-attention injection. As a foundation model, we demonstrate that OmniV2X pre-trained on large-scale single-agent planning datasets can efficiently adapt to cooperative environments by integrating the conditioning context with lightweight, standard-compliant V2X tokens. Evaluated on the DAIR-V2X-Seq dataset, OmniV2X outperforms existing end-to-end cooperative driving baselines, achieving state-of-the-art performance with less than 10% of the fine-tune V2X dataset and less than 1% of the communication bandwidth. We conduct comprehensive evaluations to demonstrate its computational efficiency and robustness under real-world constraints.

Generating a diverse set of high quality solutions for an optimisation problem has been studied extensively in recent years by the evolutionary computation community. A paradigm that has received increasing attention is evolutionary diversity optimisation (EDO), where the goal is to maximise the diversity of a solution set subject to quality constraints. Since the contribution of each solution to the diversity of the population depends on other solutions and can change dramatically if several solutions in the population are modified simultaneously, most EDO approaches generate a single new solution per generation and discard the solution with the least contribution to diversity, ensuring a steady increase in population diversity over successive generations until convergence. In this study, we aim to answer two questions: (1) Is generating multiple solutions in each generation beneficial for EDO? (2) How can this be achieved efficiently, given that conventional survival selection methods do not work well in EDO due to the dependency of a solution's contribution to diversity on other solutions?

Singular learning theory characterises the complexity of a deep network through the geometry of its loss singularities. The local learning coefficient (LLC), the standard estimator of Watanabe's real log canonical threshold (RLCT, $λ$), reads this geometry as an integrated Bayesian scalar through SGLD, which needs per-task calibration and $10^4$-$10^6$ forward-backward passes per checkpoint. We introduce Dead-Direction Signatures (DDS), a family of cheap closed-form spectral readings of singular structure: each reads a network's activation matrix or per-sample-gradient Fisher-Gram at a chosen layer, replacing the SGLD posterior chain with spectral linear algebra. The readings rest on a dead-direction framework that predicts a structural correlation between activation- and Fisher-side spectra at any singular minimum, and a rank-multiplicative volume identity that single-eigenvalue monitors cannot produce: the active-volume $\log\det^{+}(G)$ slope counts the dead directions, tracking the rank-deficit $r$ across $r \in \{1,2,3,4\}$ (slope ratios $2.0, 3.1, 4.0$ at $r{=}2,3,4$ against the predicted $2,3,4$), where the smallest eigenvalue is rank-blind. On reduced-rank regression with closed-form $λ$, calibrated LLC recovers $λ$ at $99\%$ mean and the DDS observables rank-track it at the framework-predicted sign; on a non-linear modular-addition transformer DDS separates $d_{\mathrm{model}}$ across eighteen orders of magnitude where calibrated LLC at the protocol budget is rank-flat. Complementary to LLC's integrated posterior reading, DDS gives a directional, layer-local handle on a network's dead directions, read in closed form from its activation and gradient spectra.

The high cost of spatial transcriptomics (ST) has driven extensive studies into predicting gene expression directly from H&E histology images. However, this prediction task faces an inherent limitation, as tissue morphology alone provides insufficient information to fully resolve underlying gene expression. To address this limitation, a recent study leverages partial gene expression to guide the prediction process alongside histology images. Building on this paradigm, we approach the prediction task as a spatial imputation problem, employing a Masked Autoencoder (MAE) to utilize a small fraction of gene expression as genetic anchors for inferring whole-slide gene expression profiles. Specifically, we propose a bio-saliency score and a learning-to-rank strategy to adaptively identify the most informative spots within the tissue. Based on these identified spots, our framework selects contiguous regions as genetic anchors to ensure suitability for real-world ST profiling hardware. To effectively leverage these anchors, we design a cross-modal joint encoder that integrates visual and genetic modalities. By aligning the selected anchors with their corresponding visual features via contrastive learning, the encoder generates robust joint representations to accurately predict gene expression across the whole slide. Notably, our framework consistently surpasses existing methods in both histology-only prediction and spatial imputation, achieving superior accuracy even without genetic anchors and further excelling with as little as 10% transcriptomic coverage. Our code is available at https://github.com/Kyyle2114/CAMMST.

Attorneys, judges, and pro se filers increasingly use AI to draft legal documents, yet these tools frequently fabricate citations. Despite predictions that newer models would hallucinate less or that court sanctions would deter negligent filers, we found over 1,000 filings containing fabricated citations -- with this number growing year-over-year. This study evaluates whether AI-based systems can mitigate these errors by automatically detecting hallucinations. We propose a taxonomy of legal citation hallucinations grounded in actual court filings and introduce a dataset of 1,300 brief excerpts containing injected errors. Benchmarking five models in agentic and non-agentic settings reveals that while the latest iterations perform better -- GPT-5 achieves 82.8% recall and a 60.5% F1 score in an agentic framework -- all models struggle with subtle error categories. Agentic verification remains resource-intensive, with GPT-5 averaging 16.9 steps per excerpt. Furthermore, restricted information access limits the efficacy of even the best agents. This gap creates policy concerns, as it disadvantages both AI systems and litigants who lack subscriptions to commercial legal databases. Together, our dataset, tools, and policy recommendations provide a foundation for building and auditing reliable legal citation checking tools.

Decentralized bilevel optimization (DBO) provides a powerful framework for multi-agent systems to solve local bilevel tasks in a decentralized fashion without the need for a central server. However, most existing DBO methods rely on lower-level strong convexity (LLSC) to guarantee unique solutions and a well-defined hypergradient for stationarity measure, hindering their applicability in many practical scenarios not satisfying LLSC. To overcome this limitation, we introduce a new single-loop DBO algorithm called diminishing quadratically-regularized bilevel decentralized optimization (DUET), which eliminates the need for LLSC by introducing a diminishing quadratic regularization to the lower-level (LL) objective. We show that DUET achieves an iteration complexity of $O(1/T^{1-5p-\frac{11}{4}τ})$ for approximate KKT-stationary point convergence under relaxed assumptions, where $p$ and $τ$ are control parameters for LL learning rate and averaging, respectively. In addition, our DUET algorithm incorporates gradient tracking to address data heterogeneity, a key challenge in DBO settings. To the best of our knowledge, this is the first work to tackle DBO without LLSC under decentralized settings with data heterogeneity. Numerical experiments validate the theoretical findings and demonstrate the practical effectiveness of our proposed algorithms.

Automated test generation for telecom software systems and networks has advanced significantly with the adoption of machine learning and rule-based approaches. However, most existing solutions generate static test suites against a snapshot of the system; as code, configurations, topologies, and key performance indicators (KPIs) evolve, these tests quickly become outdated or misaligned with the live system. There is currently no widely adopted solution that continuously detects fine-grained changes and selectively adapts only the affected tests without regenerating entire test suites. This paper presents a context-aware generative AI framework for automated telecom test script generation that treats testing as a continuously adapting process driven by the current state of the system rather than a static artifact. The central contribution is delta-conditioned test generation over a live knowledge graph: our approach employs a continuously updated knowledge graph (KG) as a single source of truth, a delta engine for fine-grained change detection, and a KG-guided generative AI agent, operating via the Model Context Protocol (MCP), to create, update, or retire test cases automatically. We further integrate Retrieval-Augmented Generation (RAG) to enrich reasoning with telecom-domain knowledge and historical artifacts. We demonstrate applicability across software-system and telecom-network use cases, including a Python-based KPI monitoring application managed in GitLab, and show how the framework reduces manual effort, improves test relevance, and accelerates test cycles.

Large Audio Language Models (LALMs) have demonstrated strong performance across a wide range of audio tasks. As they are increasingly deployed in real-world applications, ensuring their safety alignment has become more important. Although refusal mechanisms serve as a key safeguard by preventing LALMs from responding to harmful requests, they can also lead to {\em over-refusal}, where models incorrectly reject benign queries. This issue is especially challenging in the audio domain because speech that appears harmful in isolation may become benign when interpreted together with the surrounding acoustic context, such as background sounds. To study this problem, we introduce \textbf{AOR-Bench} (\textbf{A}udio \textbf{O}ver-\textbf{R}efusal \textbf{Bench}mark), the first benchmark for over-refusal specifically designed for LALMs. AOR-Bench contains 3,000 pseudo-harmful audio samples across six scenario categories. Evaluating 12 representative LALMs from six major model families, we find that over-refusal is widespread (Figure~\ref{fig:overall_performance}) and uncover several important patterns in their safety judgments. As a preliminary effort to mitigate this issue, we further explore two lightweight strategies (e.g., Chain-of-Thought and activation steering) to reduce over-refusal.

Long-term memory systems for Large Language Model (LLM) agents typically try to \emph{remember everything}, extracting memories uniformly to retain as many facts as possible. In production, however, inference cost and finite context budgets make this untenable: beyond consolidating raw dialogue into memory, an agent must exert \emph{write control}, efficiently keeping only the information each user actually cares about. Otherwise, long-horizon personalized interactions suffer \emph{memory bloat}, where irrelevant trivia crowds out useful information and steadily erodes question-answering (QA) accuracy. We argue that what is worth remembering is role-dependent, and propose \textbf{AdaMem} (Adaptive Memory), a method that \emph{learns what to remember} for each user from feedback. AdaMem maintains a structured, role-specific Memory Policy and refines it from weekly QA feedback through a lightweight, patch-style self-reflection step with failure rollback. To study this setting, we build \textbf{AdaMem-Bench}, a benchmark that simulates weeks of interaction with week-by-week QA. Across two extraction models and two feedback modes, AdaMem improves QA accuracy by up to \textbf{+9.0\%} over the uniform Mem0 baseline while shrinking memory volume by \textbf{9\%}.

LLM agents increasingly solve local tasks through command-line and CLI-based harness interfaces, including code editing, repository inspection, data analysis, and file workflows. Existing evaluations often emphasize task success, but deployed local agents are not models alone: the CLI mediates prompts, context replay, tool outputs, file access, terminal observations, and stopping behavior. As a result, the same model can produce different success, token, and cost profiles under different CLIs. We introduce AGENTMETER, a benchmark for evaluating model-CLI matching in CLI-mediated local task-solving agents, together with AgentMeter Score (AMS), a success-anchored, cost-aware metric over calibrated task-effort tiers. AgentMeter uses Benchmark90 as the full validation set and Core30 as a lower-cost subset for expanded comparison across 24 complete model-CLI configurations. On Core30, common deployment criteria select different configurations: highest Pass/30 selects GLM-5.1 with qwen-coder, lowest Tok./Pass selects GPT-5.3-Codex with kimi-cli, lowest billable USD/Pass selects Qwen3.6+ with Codex, while highest AMS selects Qwen3.6+ with kimi-cli. Benchmark90 validation preserves the Top-1 configuration and Top-3 set, with Spearman correlation 0.765, Kendall correlation 0.567, and AMS MAE 0.0383. These results show that model choice and CLI choice should not be decoupled, and that model-CLI configurations should be evaluated as the deployed unit.

Latent action pretraining learns representations of visual change from pairs of observations, but existing methods typically encode each transition as a single unstructured representation that entangles transition extent and transition mode. We introduce Polar Latent Actions with Radial structure (PoLAR), which imposes a radial-direction structure on latent actions, encouraging radius to encode transition extent and direction to retain transition mode. PoLAR uses temporal offset between two observations as a weak proxy for transition extent, encouraging latent action from observation pairs separated by larger temporal gaps to occupy larger radii. We instantiate this structure in hyperbolic space, whose expanding volume with radius offers a natural fit for more diverse transition modes at larger extents. Across in-task and large-scale pretraining settings, PoLAR improves downstream policy performance in simulation and real-world robot experiments, outperforming latent action baselines and strong pretrained VLAs. These results suggest that the geometry of the latent action space is an important design choice for transferring visual pretraining to downstream robot policy learning.

Learning accurate value functions plays a decisive role for reinforcement learning (RL) agents to solve long-horizon, complex tasks. Conventional temporal-difference (TD) learning objectives suffer from value-estimation bias that accumulates over the horizon, while extended-horizon modeling methods, such as n-step TD backups and Q-chunking, adopt a rigid, fixed-horizon value-modeling recipe that is often not flexible enough to capture complex value structures in long-horizon, multi-stage tasks. In this paper, we show that enabling value updates with dynamic horizon composition can yield a strong offline policy learning scheme. Our method, Horizon Adaptive Offline Policy Learning via VAlue STitching (VAST), replaces fixed-horizon backups with recursive, horizon-adaptive value composition. Its key ingredient is to couple value optimization with a future state- and horizon-length-conditioned auxiliary value function that is learned through direct data supervision, and a stitching policy that optimally selects the reward-maximizing horizon length and future sub-goal to achieve horizon-adaptive value stitching. This design enables direct estimation and compositional "stitching" of variable-length returns grounded in actionable sub-goal states, providing an accurate and greedily exploitable value-supervision signal for offline policy optimization. Across 50 tasks on OGBench, VAST outperforms fixed-step, extended-horizon methods, and generative-value offline RL baselines, achieving strong performance particularly in high-complexity, long-horizon decision-making tasks.

Dense Gaussian networks are degree-four algebraic networks with compact diameter and coordinate-based routing. Their diameter-level broadcast trees are efficient but fragile under node, link, and runtime-discovered faults. This paper develops a runtime recovery framework for dense Gaussian broadcast networks under static node/link faults and mixed faults, plus single-link faults discovered live. The method re-roots the source so known node faults become boundary leaves whenever possible, then filters failed links and repairs gaps by connecting healthy components of the pruned tree. For a selected root with connected healthy component graph, we prove exactly $c-1$ external repair edges are necessary and sufficient. We also prove deterministic single-link repair, give a constant-size boundary-intersection primitive for source selection, derive a link-avoidance exclusion test, and add a local-obstruction bound explaining why high-order cuts vanish as $k$ grows. Experiments over $k\in\{10,25,50,100,200\}$, up to $80{,}401$ nodes, $280{,}000$ static trials, and $15{,}000$ transient trials show 100\% recovery for deterministic and bounded regimes, $99.998\%$ for multi-link faults, and $99.963\%$ for heuristic regimes; non-recovered trials are explained by disconnected components or relocation failure. Re-rooting reduces average repair edges by 80--100\% versus fixed-source repair. Patched Gaussian-link Noxim replays confirm packet-complete execution and show re-rooting reduces repair edges, components, and depth. A completion-cycle audit separates repair benefit from latency: ablations confirm completion time depends on relocation, scheduling, delivery tail, and selector objective, so the paper claims edge-minimum repair rather than universal completion-cycle dominance.

One-to-all broadcasting in dense Eisenstein--Jacobi (EJ) networks relies on diameter-level spanning trees that fragment when nodes or links fail. This paper introduces the selected triple $(r,θ,\Kcomp_{r,θ})$--a chosen root, a chosen EJ coordinate-reduction orientation, and the healthy component graph induced by that choice--as the fundamental unit of analysis for joint node/link fault recovery. The central result is a necessary and sufficient condition: hybrid repair succeeds if and only if the healthy EJ graph $G'=\Ht-\Fv-\Fe$ is connected. When $G'$ is connected, a spanning tree of $\Kcomp_{r,θ}$ maps to exactly $c-1$ component-crossing repair edges, which is minimum for the selected pruned tree. Deterministic guarantees include: one/two faulty nodes are always placed on the distance-$t$ boundary by re-rooting; a single failed link is either avoided or repaired by exactly one crossing edge; and the repaired depth satisfies $D_{r,θ}\le 2t+1$ under shallowest-layer entry selection. A 260,000-trial validation campaign confirms 100\% recovery and substantial repair-edge reduction over fixed-source repair across five network scales up to $N=120601$ nodes, while global-BFS, near-miss, and cap-sensitivity audits clarify the tradeoff between reachability, forwarding-state changes, and ranked root selection.

Dense Gaussian and Eisenstein--Jacobi (EJ) networks are algebraic interconnection networks with compact coordinate balls, fixed degree, and simple modular addressing. A source-centered coordinate-reduction tree gives a non-redundant one-to-all broadcast in the fault-free network, but processor faults can split the tree into multiple healthy components. Unlike search-based repair methods that require a linear scan of the network to select the repair plan, the certificate selectors introduced here operate in $O(1)$ time and $O(1)$ memory, consulting only the fault coordinates. This paper develops this stronger formulation for the one- and two-fault regime: a constant-time certificate selector. Given only the faulty coordinates, the selector classifies the relative fault geometry, chooses a coordinate-reduction orientation, and returns a bounded ordered set of component-crossing repair edges. For dense Gaussian networks $G_k$, every source-free fault set with $|F|\le2$ is repaired with depth at most $k+2$ and with exactly $c-1$ external component-crossing edges for the selected fault-pruned orientation. For dense EJ networks $H_t$, every one-fault placement is repaired within depth $t+1$, and every two-fault placement is repaired within depth $t+2$, again with exactly $c-1$ external repair edges. Exhaustive strict validation confirms the Gaussian selector over $146{,}156$ one- and two-fault cases for $k=5,\ldots,12$ and the EJ selector over $52{,}395$ cases for $t=2,\ldots,8$, with zero failures in connectivity, acyclicity, exact repair count, or depth bound.

The rapid growth of large-scale AI workloads, particularly Large Language Model (LLM) training and inference, is fundamentally reshaping the operational dynamics of hyperscale data centers. Unlike traditional cloud workloads, AI-driven jobs exhibit bursty, high-intensity, and rapidly shifting resource demands, often leading to sudden capacity stress that cannot be effectively handled by reactive threshold-based mechanisms. In this paper, we propose a deployment-oriented, burst-aware early warning framework for proactive capacity stress prediction under AI workload surges. We formulate the problem as a high-recall forecasting task over multivariate telemetry windows, with the explicit goal of enabling operational intervention before system degradation occurs. The proposed framework integrates workload intensity, temporal variation, and system pressure signals, and employs a lightweight tree-based learning model to capture nonlinear interactions in highly imbalanced environments. To evaluate the system under realistic conditions, we introduce an AI workload surge injection methodology that simulates burst-driven demand patterns observed in large-scale AI systems. Our XGBoost-based model achieves an ROC AUC of 0.697 and an AP of 0.670, significantly outperforming baseline methods. Under deployment-oriented threshold selection, the framework achieves a Recall of 0.914, enabling the detection of the majority of stress-prone periods with acceptable false-alarm cost. Beyond predictive performance, we show how the proposed framework can be integrated into operational control loops to support proactive actions such as workload throttling and resource scaling. Our results highlight the practical value of high-recall, learning-based early warning systems in enabling resilient and adaptive data center operations in the era of AI-driven workloads.

Despite the success of deep learning, training deep networks in biologically plausible and hardware-efficient ways remains an open challenge. Feedback alignment (FA) methods address this by replacing backpropagation's symmetric backward weights with fixed random matrices, but their effectiveness depends critically on whether they can be accurately evaluated. The standard evaluation relies on two quantities: task accuracy and cosine similarity between the method's credit signal and the backpropagation gradient. We show that this reporting pair is insufficient by identifying two independent failure modes, both silent under current reporting: (1) measurement degeneracy, where the BP reference gradient collapses to the numerical floor in terminal-LayerNorm residual architectures, rendering cosine uninterpretable; and (2) aggregation collapse, where the aggregate cosine masks layerwise heterogeneity that concentrates credit at one end of the network. To address these limitations, we propose a diagnostic evaluation protocol based on three checks -- scale stability, reference validity, and depth utility -- together with per-layer rather than aggregate cosine reporting. Across multiple architectures and methods, the standard reporting pair gives no signal of failure in any audited case, while our protocol identifies all failures with wide calibration margins. The two failure modes are causally independent: a per-block scale penalty alleviates Mode 1 (residual scale explosion driving reference collapse) without affecting Mode 2 (cosine ranking that contradicts every functional metric we measured). Identifying these silent failures prevents researchers from building on non-functional credit assignment and provides actionable guidance for developing FA methods that genuinely train deep layers.

Conversational AI agents in Customer Experience (CX) typically suffer from a Closed-World Constraint, ignoring high-velocity external shifts like viral trends or outages. Ad-hoc web search attempts to bridge this gap but often introduce prohibitive latency and context poisoning. We introduce PulseCX, a framework that decouples knowledge acquisition from consumption. Adopting a structure-first paradigm, PulseCX employs an asynchronous agent to linearize signals into a Decay-Aware Temporal Knowledge Graph (DA-TKG) governed by reinforcement--decay dynamics to actively manage information lifecycles. By coupling this self-evolving memory with hierarchical intent gating, PulseCX removes synchronous search bottlenecks (<10ms overhead) and drives significant gains in Intent Resolution (IRR) and Customer Satisfaction (s-CSAT) in dynamic environments.

High-stakes reasoning tasks necessitate transparent and verifiable workflows, yet conventional single-model large language models (LLMs) often struggle with hallucination and low interpretability under zero-shot paradigms. To address this general AI challenge, we propose a Multi-Agent Audit Framework that simulates a collaborative, multi-step verification process. We empirically validate this architecture in the sensitive domain of clinical mental health screening using a modular LangChain workflow. Our framework decomposes the reasoning process into a Perception Agent, Knowledge Retrieval-Augmented Generation (RAG), Chain-of-Thought (CoT) clinical inference, and a critical Audit verification stage. We evaluated this framework on the DAIC-WOZ dataset using locally deployed open-source models. Experimental results demonstrate that our multi-agent pipeline significantly outperforms single-agent baselines, reducing the Mean Absolute Error (MAE) for PHQ-8 depression severity prediction from 5.35 to 5.02. By exposing cross-agent validation traces, the framework mitigates reasoning drift and provides highly interpretable diagnostic rationales, offering a generalizable paradigm for reliable AI-assisted decision support beyond isolated model scaling. We make data and code open access on GitHub for replicability.

Large language models can produce confident but protocol-invalid answers in domains where procedural compliance is critical. This paper presents Answer Engineering, a deterministic runtime and authoring layer that applies localized rule-guided interventions to the visible reasoning trajectory during standard autoregressive generation, without retraining, modifying model weights, or performing global search. The method is evaluated on a controlled clinical benchmark for sudden sensorineural hearing loss (SSNHL), where correct management depends on protocol-consistent interpretation of symptom timing, Weber/Rinne tuning-fork findings, and otoscopic findings. In the benchmark, step-by-step reasoning shifted rather than eliminated errors: compliant outcomes for SSNHL decreased from 54.5% under unguided generation to 25.1%, while acceptance on the conductive contrast condition increased from 1.6% to 58.9%. Local trajectory editing increased SSNHL compliance to 83.5% and conductive-case adherence to 77.9%, raising balanced accuracy from 42.0% under reasoning-only generation to 80.7%. The results support a systems-level view in which protocol adherence can be improved through auditable runtime control of reasoning trajectories, while also identifying limitations caused by rule coverage, trigger reliability, and persistent diagnosis-first generation dynamics.

Mammography is an essential tool for breast cancer detection, with millions of examinations conducted annually. However, publicly available high-quality mammography datasets for AI development remain limited in both scale and annotation richness, particularly regarding pathological subtype coverage and structured diagnostic reasoning annotations. In this paper, we present MammoExpert, the first mammography dataset with Chain-of-Thought reasoning annotations across three diagnostic phases: (i) primal observation, (ii) factual assessment, and (iii) diagnostic synthesis. Comprising 2,379 mammography images covering 67 WHO-classified histopathology subtypes, each exam provides 42 radiographic features annotated by nine senior radiologists. We evaluate its performance on the breast lesion classification task, demonstrating superior accuracy and reasonability compared to existing classification models. Combining public dataset CBIS-DDSM with MammoExpert yields 7.1\% classification accuracy improvement, while the training model to learn CoT reasoning achieves another 4\% gain on the MammoExpert test set. Similar improvements are observed on INBreast and Vindr datasets, where the full approach yields accuracy gains of 6.9\% and 6.7\%, respectively. MammoExpert can serve as a benchmark for interpretable breast lesion diagnosis through explicit CoT reasoning.

Proofreading--correcting segmentation errors in 3D brain reconstructions--is the rate-limiting step in synapse-resolution connectomics. We release ConnectomeBench2, a unified multi-species dataset of over 716,485 expert-labeled proofreading decisions with >4,500,000 associated images spanning four major open connectomes (mouse, human, zebrafish, fly), spanning both split and merge error correction. Trained on this dataset, a single Vision Transformer with shared encoders for mesh geometry and electron microscopy reaches human-level accuracy across species for split error correction and merge error identification, with performance scaling with data size and modality. Beyond accuracy, we show that the model is well-calibrated within distribution, that measures of distribution distance predict where calibration and accuracy will degrade on unseen data, and that connectomics-specific pretraining and active learning-based sample selection show potential to substantially reduce the labeling effort needed to extend to new species and brain regions. The benchmark provides the infrastructure to train and evaluate increasingly capable vision models for connectomic proofreading. Data and code availability. The ConnectomeBench2 dataset is released on Hugging Face at https://huggingface.co/datasets/jeffbbrown2/ConnectomeBench2. The accompanying codebase is available on GitHub at https://github.com/timfarkas/ConnectomeBench2.

Remote photoplethysmography (rPPG) is a camera-based technique for measuring physiological signals, particularly cardiac activity. From the remotely measured signals, heart rate can be estimated, which is crucial for health monitoring. In this study, we investigate a driver health monitoring system based on remote heart rate estimation. However, driving environments represent uncontrolled settings where videos are subject to varying illumination conditions and frequent head movements. We introduce MS-rPPG, a multi-spectral framework that combines RGB with near-infrared (NIR) face video to alleviate rPPG estimation under challenging driving conditions. To combine the complementary features from two spectral videos, we propose a cross-spectral linear modulation (CSLM) strategy based on frequency-domain analysis. Moreover, we introduce MS-Mamba, a novel state space model designed to effectively model long-range temporal dependencies while jointly capturing cross-channel interactions between multi-spectral features. We collected a real-world dataset called MS-Drive, which was recorded from 50 participants while driving the vehicle. The proposed method was evaluated on the MR-NIRP Car dataset and MS-Drive datasets. The experimental results indicate that MS-rPPG shows better robustness and heart rate estimation accuracy than previous methods, highlighting its promise for driver health monitoring. The codes are available at github.com/ziiho08/MS-rPPG.

Object-centric image generation is important in settings with few labeled examples, including pedestrian analysis in smart-city scenes, traffic-sign inspection, and domain-specific object detection. Synthetic images are most useful for training and evaluation when datasets preserve object structure, bounding boxes, visual diversity, and realistic context. Existing image datasets usually target classification, detection, or scene understanding rather than controlled object-centric generation and augmentation with limited class-specific data. We present a shareable collection of three object-centric dataset resources: Cityscapes-Pedestrian, TrafficSigns, and COCO PottedPlant. The collection standardizes 256-by-256 object-centric crops and bounding-box annotations across three regimes: dense pedestrian scenes with privacy blur and occlusion, cleaner high-contrast traffic signs, and context-diverse potted-plant scenes. The release contains 3,009 TrafficSigns samples, 2,156 Cityscapes-Pedestrian manifest records, and 7,679 COCO PottedPlant manifest records. The larger COCO-derived manifest preserves contextual and multi-instance diversity, while equal-size subsets can be drawn with a fixed random seed for controlled comparisons. The release provides direct TrafficSigns data where redistribution is permitted, together with scripts, manifests, box-level annotation tables, checksums, and reconstruction documentation for the Cityscapes- and COCO-derived subsets. It is available through the Latzi/object-centric-low-data-datasets GitHub repository and Zenodo DOI 10.5281/zenodo.20573001. The collection supports label and split inspection, subset creation, reconstruction from upstream data, and evaluation of object-centric image generation or synthetic-data augmentation methods on shared records.

Virtual-cell and perturbation models are increasingly used to predict cellular responses for biomedical discovery, but chemical and genetic perturbations are not automatically interchangeable. Existing evaluations often study chemical response prediction or genetic perturbation prediction separately, leaving target-matched chemical-to-genetic translation under-tested. We introduce Chem2Gen-Bench, a benchmark comprising 260,084 chemical and 1,099,045 genetic perturbation profiles organized into cell-target contexts, and evaluate pairwise alignment, retrieval, protocol covariate associations, feature spaces, and foundation-model embeddings. Across matched contexts, translation fidelity is measurable but heterogeneous; background adjustment increases the association between pairwise similarity and retrieval success, while paired tests show lower mean retrieval success after adjustment under the evaluated settings. In a target-matched K562 audit, the evaluated foundation-model embeddings did not consistently improve over gene-delta baselines. Chem2Gen-Bench provides an auditable framework for testing when chemical and genetic perturbations align around shared targets and when representation gains are supported by matched perturbation evidence.

Differential privacy (DP) has become the gold standard for ensuring the privacy protection of machine learning and statistical algorithms in recent decades. A plethora of algorithms and methods have been developed to enhance the utility of DP algorithms while maintaining the same level of DP. However, these are often overly complex or computationally ineffective. We propose a novel approach focusing on denoising the output of the simple additive Gaussian mechanism by adopting the idea of \textit{empirical Bayes estimation}. We highlight that the empirical Bayes approach can reduce the mean-squared error solely by taking the output of the Gaussian mechanism as input. Our numerical studies show that this simple yet powerful approach can be applied to improve upon various statistical problems, including histogram release, principal component analysis, and linear regression, often outperforming existing private algorithms.

Deep learning technology has enhanced the ability of Click-through rate (CTR) prediction models to learn features and improve prediction accuracy. However, it is challenging to deploy CTR models on GPU smoothly and perform inference efficiently, because there is a huge mismatch between the serial computational pattern and the parallel model structure. In this paper, we propose DPIFrame, the first dual parallelizable framework to accelerate CTR model inference. In DPIFrame, a) a dual parallelizable architecture is proposed to perform parallel CTR model inference in both intra-module and inter-module; b) an efficient multi-table lookup algorithm is presented for embedding operations through anticipating the whole workload in advance; c) a breadth-first stream scheduling strategy is designed for fine-grained management of parallel computation on GPU to further supporting the dual parallel execution. Extensive experiments are conducted on two real-world datasets, and the results highlight that DPIFrame can reduce the embedding latency efficiently by \textbf{23.0$\times$} compared to PyTorch. Compared with PyTorch, TorchRec, HugeCTR, and OneFlow, DPIFrame can achieve state-of-the-art inference performance on GPU with speedups of \textbf{5.83$\times$}, \textbf{4.29$\times$}, \textbf{2.15$\times$}, and \textbf{2.0$\times$}, respectively.

Reliable evaluation of phrase break annotations is crucial, as subtle variations in prosodic boundaries directly affect the clarity and naturalness of speech. However, existing approaches exhibit major limitations: single-reference evaluation assumes a unique gold phrasing for an utterance despite multiple valid phrasings, while human judgment, though flexible, is labor-intensive and unscalable. To address these, we propose LLM-based Multi-Reference Evaluation (LMRE) for phrase break annotations that models the one-to-many nature of prosodic phrasing and generates multiple valid phrasings from minimal demonstrations. On a Korean testbed of 1,356 annotations covering five strategies, LMRE shows stronger alignment with human judgment than single-reference evaluation in both acceptance behavior and score correlation. Our findings demonstrate that LMRE effectively achieves both scalability and multi-reference support, highlighting the potential of LLMs for evaluation in the speech domain.

Deploying highly capable personalized conversational agents in resource-constrained or privacy-sensitive environments remains a significant challenge. We identify a fundamental bottleneck in the existing approaches: current training paradigms treat personalization and grounding as a single monolithic learning problem. Under these paradigms, language models are forced to simultaneously address what to say (content grounding) and how to say it in a user-specific way (personalization), which introduces significant computational and optimization challenges. Consequently, contextual grounding is often sacrificed for persona adherence, or vice versa, resulting in responses that are either weakly grounded in the conversational history or insufficiently personalized. In this work, we propose the Generic Response-Augmented Generation (GRAG) framework that decouples these competing objectives by leveraging offline, generic responses from high-capacity, general-purpose LLMs as a semantic and structural scaffold to guide the fine-tuning of smaller, task-specialized models seamlessly in resource-limited environments. By decoupling the content grounding from personalization, GRAG allows the model to focus exclusively on persona injection while remaining firmly anchored to the conversational context. We instantiate the GRAG in two post- and pre-fusion-based architectural variants and evaluate them on multiple benchmark conversational datasets that cover diverse personalization structures. Our results demonstrate that GRAG significantly outperforms state-of-the-art methods that do not use auxiliary scaffolding, yielding up to 47% improvements in ROUGE-2 and 36% in BLEU scores. Ultimately, GRAG offers a generalizable blueprint for building grounding-aware personalized conversational systems in resource-limited environments.

Semi-supervised learning (SSL) on data streams is challenging due to the continuous evolution of high-volume data and the scarcity of labels. Existing methods are limited in leveraging the intrinsic relationships among samples because they typically rely on fixed similarity measures or static graph structures, which cannot capture how relationships evolve over time. We propose SLeDGe, an SSL method for data streams that jointly learns a predictive model and an adaptive graph structure under strict memory and label constraints. SLeDGe maintains compact labeled and unlabeled memories using distinct update strategies, balancing rapid adaptation to novel features with the retention of historical consistency. In addition, by encouraging sparsity in the relational graph, SLeDGe filters out spurious connections and enables effective propagation of label supervision. Across 12 datasets, SLeDGe outperforms state-of-the-art competitors, achieving average relative accuracy gains of 31.7% with 0.1% labels and 14.8% with 1% labels.

We introduce BASIL, a user-friendly desktop application for process optimization. BASIL employs a Bayesian approach, incorporating special acquisition functions that can be used to solve both single and multi-objective optimization problems. It provides a graphical interface that enables users to input their experimental parameters, optimization objectives, and legacy data. This is then used to build surrogate models, which are coupled with acquisition functions to guide and optimize a process towards a desired objective. To facilitate model building, BASIL provides a variety of predefined surrogate model templates. BASIL can be used to optimize any arbitrary experiment or process with known, user-defined input variables, optimization objectives, and defined output.

Offline-to-online adaptation serves as a pivotal paradigm for mitigating the prohibitive cost of online exploration by bootstrapping reinforcement learning from offline datasets. While this paradigm has been extensively studied in single-agent settings, its extension to Multi-Agent Reinforcement Learning (MARL) remains largely unexplored, despite its critical relevance to complex coordinated decision-making. To bridge this gap, we introduce Sim2O, an elegant and minimalist framework for offline-to-online MARL. Rather than treating adaptation as a monolithic joint decision, Sim2O conceptualizes it as a compositional process. Specifically, candidate joint actions are synthesized by dynamically blending offline and online action proposals across agents. By leveraging a centralized value function to evaluate these hybrid combinations, Sim2O identifies high-value coordination strategies without requiring auxiliary training objectives or structural overhead. Empirical evaluations across diverse benchmarks demonstrate that Sim2O significantly outperforms existing baselines, underscoring that a minimalist design is not only viable but highly effective for multi-agent offline-to-online adaptation.

Large language models (LLMs) deployed for logical reasoning in knowledge-intensive domains exhibit a subtle but critical failure: coherence can be vacuously achieved through systematic abstention. A model that withholds commitment to either entailment or refutation satisfies negation consistency while providing no utility. We introduce Coherence Under Commitment (CUC), a dual-query evaluation paradigm that jointly measures consistency and decisiveness. CUC contributes three innovations: (1) a commitment score $c(\varphi) = p(\varphi) + p(\lnot\varphi)$ quantifying probability mass allocated to decisive outcomes; (2) a \textbf{deterministic elicitation protocol} via normalized YES/NO log probabilities, eliminating sampling variance; and (3) a 3-way decision framework (True/False/Uncertain) operationalizing the coherence-commitment trade-off into metrics. Experiments on four open-weight LLMs (1B-3B) across 204 FOLIO examples expose a sharp frontier. Qwen2.5-3B achieves near-zero contradiction ($\mathbb{E}[v_{\mathrm{neg}}]{=}0.025$) but only $7.4\%$ coverage, while TinyLlama-1.1B reaches $79.4\%$ coverage with violations on every example. Coherence-only evaluation would rank the abstaining model first; CUC exposes this as vacuous, and the frontier generalizes to LogiQA~v2 ($ρ{=}0.97$). We argue that evaluation must report both coherence and non-vacuous commitment and release a toolkit for standardized assessment.

Multi-turn jailbreaks can evade turn-level moderation by spreading unsafe intent across a dialogue through gradual escalation, reframing, and role manipulation. We address multi-turn jailbreak detection as a conversation-level classification problem and introduce an efficient hierarchical detector that avoids expensive long-context concatenation while retaining cross-turn reasoning. The model encodes individual turns to form compact turn representations and applies a lightweight conversation module that captures dialogue dynamics and selectively attends to fine-grained evidence when needed. On a challenging evaluation benchmark of 14,038 conversations, our approach achieves an F1 of 0.9394, outperforming Claude Opus 4.7, the strongest competing baseline, by 0.07 while halving its false-positive rate. Ablation studies confirm that each architectural component contributes meaningfully, with combining cross-attention and self-attention in the conversation module yielding a 2.26 percentage point reduction in false-positive rate over the self-attention-only variant.

Bayesian model averaging in support-indexed regression induces a posterior distribution over active predictor supports. Under predictor redundancy, posterior mass can spread across many nearly interchangeable supports, making exact-support summaries unstable or hard to interpret even when prediction is stable. We study how to report an already fitted Bayesian model averaging posterior without changing the Bayesian target. A report uses hard or soft regions of support space, and its compressed reporting law is compared with the reference posterior through an explicit density ratio. This ratio gives computable total-variation and Kullback--Leibler distortion, bounds for bounded predictive summaries, retained-mass diagnostics, and fallback-weight diagnostics. The framework covers fixed hard regions, metric-ball regions, posterior-cluster regions, and pooled-pruned region dictionaries. We prove exact error formulas and validation bounds for these region reports, and give conditions under which a few regions can replace a long list of individual supports. In simulations, our region reports often give shorter and clearer summaries while preserving the main posterior information, and the density-ratio diagnostics show when too much information has been lost.

Large language models are increasingly proposed for mental-health applications such as detecting suicidal content, raising the question of what they rely on. We study this mechanistically and use it to ask a narrower question: how to make a causal claim about a model's internal features more trustworthy. Our validation-gated framework, with suicidality detection as a case study, interprets a behavior only after the model is shown to perform it: a concept is admitted only once the model ranks it above a simple lexical baseline, and each subsequent property is tested against a matched control. This discipline yields negative as well as positive results. The gate rules out one task at the outset: on DeepSuiMind (Li et al. 2025), Llama-3.1-8B-Instruct cannot separate implicit suicidal intent from ordinary distress, so we do not analyze it. We turn to binary suicide detection, which it does perform. There we find a mid-network feature that appears semantic rather than keyword-based, is causally implicated in the decision (ablating it degrades the judgment; a random direction does not), is low-rank, and recurs across three model families and three suicide datasets. A register-matched control (suicide versus depression) suggests it tracks suicidality more specifically than general distress. Steering raises the model's response, but for unrelated questions too, so we treat it as necessary but not sufficient. The clearest pattern separates encoding from use: smaller models already represent suicidality, yet only larger ones appear to act on it. The positive evidence is English Reddit text, which limits the clinical reading.

Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording. We show this assumption is fundamentally brittle: surface toxicity and adversarial intent can be decoupled by replacing as few as five tokens. We present OTTER (Obfuscated Toxicity-Evading Token Evolution for Rewriting), a black-box red-teaming framework requiring only standard API access, directly targeting the practical constraints of industry security audits. Evaluated on 457 AdvBench prompts across four GPT models, OTTER raises average ASR from 7.0% to 84.0%. We further provide the first quantitative analysis of the toxicity--bypass relationship and a per-category breakdown, translating our findings into actionable recommendations for classifier hardening in production deployments.

Standard Transformers use a single self-attention pathway to model both global dependencies and local patterns, creating tension between long-range structural reasoning and fine-grained local representation learning. We propose a FiLM-coordinated dual-branch Transformer for language modeling, where each layer explicitly contains a global branch and a local branch, and feature-wise linear modulation (FiLM) is used for dynamic cross-branch coordination instead of simple concatenation or static addition. The key idea is that the two branches represent different dependency views of the same input, making channel-wise calibration more suitable than heavy token-level interaction. We therefore design a bidirectional FiLM module in which each branch generates per-channel scaling and shifting parameters to condition the other. Experiments on multiple small-scale language modeling settings show that the proposed structure consistently outperforms same-width single-branch baselines and weakened dual-branch variants under a fixed lightweight configuration. On TinyShakespeare and a 1M-character subset of WikiText-2, the full dual-branch FiLM model achieves the best results among same-width structural baselines. Multi-seed results support the stability of the gains, while mechanistic analyses show that FiLM learns input-dependent, layer-dependent, and channel-selective modulation patterns rather than static scaling. Parameter-matched widened single-branch baselines also indicate that the current design still leaves room for improvement in parameter efficiency.

Traffic prediction is a core task in intelligent transportation systems and urban-scale decision making. Despite the effectiveness of mainstream neural-network based methods, their deployment in real-world settings with thousands of traffic sensors is jeopardized severely by their poor computational scalability. To address this, the community has attempted to incorporate spatial database partitioning techniques (e.g., Grid, Quadtree, and K-D Tree) to improve model scalability. However, these approaches rely on handcrafted geometric heuristics and often produce irregular or imbalanced data partitions, leading to boundary fragmentation, excessive padding overheads, and degraded model accuracy. In this paper, we propose SqLinear, an efficient and effective architecture for large-scale traffic prediction. First, we design Square Partition, a geometry-adaptive algorithm that partitions massive traffic sensors into balanced, non-overlapping, and near-square spatial regions. Unlike existing heuristic-based designs, Square Partition is theoretically grounded and provides provable guarantees on aspect ratio, balance, and partition utilization, establishing a high-quality foundation for downstream spatiotemporal modeling. Next, we propose a Hierarchical Linear Interaction (HLI) module that abandons the costly attention mechanisms commonly used in Transformer-based spatio-temporal models. HLI efficiently captures both local intra-region dynamics and global inter-region dependencies through a lightweight linear interaction scheme, enabling effective spatiotemporal modeling with linear computational complexity. Extensive experiments on four large-scale traffic datasets and 10 baselines show that SqLinear reduces MAE by 2.30% on average under standard setting and by 5.81% under extreme scalability settings, while reducing training runtime by 13.27%--30.84% in spatial- and horizon-scaling scenarios.

Local LLM agents such as OpenClaw and Nanobot run on end-user machines and act on host resources - the shell, filesystem, browser, stored credentials, and messaging applications - through natural-language goals. These agents have become privileged software runtimes that mediate between user intent, model outputs, and host-level actions. Existing research characterizes the landscape through prompt injection, malicious skills, marketplace risks, or black-box evaluation of agents. But the implementation layer that performs this mediation, the prompt builder, parser, tool dispatcher, skill loader, memory writer, network client, and permission gate, has remained an unexamined safety boundary. To our knowledge, no prior work has examined the agent's source tree to audit these components for implementation-level security weaknesses. We present CLAWAUDIT, a static auditing framework for measuring vulnerability exposure in local LLM agent runtimes. CLAWAUDIT derives a five-category vulnerability taxonomy from STRIDE and develops custom static-analysis rules that target agent-specific patterns absent from established rule sets for vulnerability analysis. We instantiate the taxonomy in two backends, 47 Semgrep YAML rules and 30 CodeQL queries, and evaluate on OPENCLAWBENCH, a benchmark of 446 source-code-level advisories from the OpenClaw repository and split temporally into 229 rule-derivation (train) and 217 held-out (test) advisories. On the held-out test, CLAWAUDIT raises Semgrep recall from 21.7% (Pro baseline) to 66.8%, and CodeQL recall from 13.8% (security-extended) to 75.1%. Train/test gaps remain within 4 percentage points for all four configurations, indicating that the rules generalize to vulnerabilities unseen during rule writing. A preliminary live-code audit shows that these recall-oriented rules require manual triage, motivating semantic filtering before production deployment.

Emotion annotation is inherently subjective, yet most NLP pipelines still assume "gold" labels, typically produced by majority voting, and treat annotator variation as noise. In this paper, we present a multilabel emotion annotation case study and use it to examine how annotator behavior and aggregation choices affect both agreement estimates and downstream emotion classifiers. Rather than collapsing disagreement into a single label, we represent targets as soft vote-share labels (including an intensity-weighted variant) and evaluate models using both thresholded metrics (macro-/micro-F1) and probabilistic alignment (Bernoulli cross-entropy SoftBCE), alongside data-derived disagreement diagnostics. Across annotation regimes, we show that disagreement is structured and leaves measurable traces in model behavior: hard labels may maximize F1 metrics, while soft supervision yields predictions that better reflect empirical annotator variance and uncertainty. Our results provide practical guidance for designing, aggregating, and evaluating multilabel emotion datasets when multiple interpretations are plausible.

Automated Essay Scoring (AES) systems are increasingly used to support teachers in managing grading workloads and to provide a supplementary rater in large-scale assessments. While human grading is frequently influenced by students' demographic characteristics, the efficacy of different strategies for integrating demographic metadata with textual input used to train AES models remains underexplored. This study investigates the impact of a specific multimodal fusion strategy - naive metadata concatenation - on the predictive accuracy, training convergence, and score parity of a DistilBERT-based AES model. A comparative analysis was conducted using the ASAP 2.0 dataset to evaluate a baseline model against an experimental model trained with input that concatenates tokenised text and demographic metadata using a naive multimodal fusion strategy. Evaluated via 10-fold cross-validation, the findings reveal that the early fusion of demographic metadata and the input significantly degrades the model's overall predictive accuracy. The baseline model achieved a Quadratic Weighted Kappa (QWK) of 0.727, which dropped to 0.656 upon integrating metadata. Furthermore, the experimental model exhibited higher validation loss (1.29) compared to the baseline model (1.25). The experimental model also displayed exacerbated scoring bias, reducing score parity instances from 15 to 12 out of 19 tests.

Recent advances in text-to-speech (TTS) synthesis have achieved highly natural and expressive speech generation. However, these systems are designed for general adults and overlook older adults' speech comprehension needs due to age-related sensory and cognitive decline. Prior work involves older adults by collecting preference feedback to tune model parameters. However, obtaining sufficient preference data is costly and difficult, as older adults quickly become fatigued during collection. In this paper, we propose a novel imitation learning (IL) framework to learn TTS models from expert demonstrations. We further improve Group Relative Policy Optimization (GRPO) with two-stage on-policy reward learning (OPRL) to mitigate reward hacking under limited supervision from expert demonstration. Experimental results show that GRPO w/ OPRL outperforms GRPO and supervised baselines in objective and subjective metrics. Audio samples are available at https://dongru1.github.io/demo/im-efss

Speech Emotion Recognition (SER) systems increasingly leverage self-supervised acoustic representations, yet their vulnerability to training-time attacks remains largely underexplored. This paper presents the first systematic study of poisoning-based backdoor attacks on SER, with a focus on threats enabled by text-to-speech (TTS) generated audio. We introduce a stealthy, low-energy acoustic trigger that can be embedded imperceptibly into both natural and synthetic speech, enabling scalable and consistent poisoning. Our experiments demonstrate that SER models can be reliably compromised with high attack success rates under low poisoning ratios, while maintaining near-clean performance on benign inputs. We further show that backdoor patterns exhibit strong cross-model transferability and that self-supervised representations are particularly susceptible to learning these triggers. These findings reveal that TTS technology dramatically lowers the barrier to effective backdoor attacks, exposing critical vulnerabilities in modern SER pipelines and motivating the urgent need for dedicated defenses.

Event ontology expansion aims to discover emerging event types from data and extend them to appropriate positions in the existing event ontology.. Existing methods typically cluster contextualized trigger representations and attach induced clusters to the ontology based on instance-level similarity. However, ontology expansion requires concept-level semantics that characterize event types, whereas contextualized trigger representations often conflate these semantics with surface contextual variation, leading to unstable clustering and unreliable hierarchy expansion. To address this issue, we propose ConceptE, a conceptualization-enhanced framework for event ontology expansion. ConceptE first derives concept-level semantics by prompting an LLM with the sentence and event trigger, producing a concise concept name and a natural-language description. It then jointly encodes these semantics with trigger information to build concept-enhanced representations aligned with ontology-level reasoning. This representation design supports more coherent event clustering, more reliable hierarchy expansion, and ontology-consistent type naming. Experiments on ACE, ERE, and MAVEN demonstrate that ConceptE consistently outperforms state-of-the-art approaches across all subtasks of event ontology expansion. In particular, it achieves improvements of up to 12.37\% in BCubed-F1 for event clustering and 6.48\% in Taxo\_F1 for hierarchy expansion, demonstrating the effectiveness of the proposed ConceptE method.

The rapid growth of AI has increased the demand for domain-specific post-training, while the cost and specialization of accelerator infrastructure push many model owners to outsource this process. Outsourced training lowers operational barriers, but creates a training-integrity gap: the owner receives a checkpoint, logs, and aggregate metrics without direct evidence that the declared training trajectory was faithfully executed. An untrusted provider may have incentives to deviate from that trajectory, either to save computation or to introduce targeted security risks. Auditing such deviations is difficult because floating-point execution on heterogeneous accelerators introduces benign numerical drift, making it hard to distinguish honest replay differences from integrity violations. Existing verification methods either observe training at too coarse a granularity or impose costs and deployment constraints that are impractical at scale. We present OVIG, an optimistic verification framework that audits outsourced post-training using an empirical boundary on gradient differences calibrated from honest heterogeneous replays. OVIG checks opened intervals against this boundary and combines optimistic sampling with a stride parameter $s$, which partitions training into stride-aligned intervals and retains only interval-endpoint evidence. Across shortcut training attacks and targeted manipulation attacks, OVIG maintains $0\%$ ASR on language, vision, and diffusion workloads. On Qwen3, increasing the stride from $s=1$ to $s=2000$ reduces off-chain storage and evidence transmission by $1996\times$ while preserving $0\%$ ASR; at this setting, OVIG incurs only $1.143\times$ total system overhead relative to training without verification. These results show that OVIG provides a practical integrity layer for outsourced AI post-training under heterogeneous execution.

The empirical foundation of cyber deception relies on human-centered hypotheses, but the rapid emergence of autonomous, AI-enabled attackers challenges whether this foundation transfers to AI agents. To address this, we introduce an automated evaluation framework adapted from the Honeyquest instrument to assess LLM attacker judgment at scale. Our 21-LLM cohort spanned 10 providers, diverse architectures and specializations, open- and closed-weight models, and parameter scales from 8B to over 1T. We evaluated the performance of this LLM cohort (yielding 10,962 responses) against the 47-participant human baseline across an identical set of 174 reconnaissance queries. Our empirical evaluation reveals three key findings that establish LLMs as a distinct attacker class: (1) every model in our cohort falls for deceptive traps at a significantly higher rate than human attackers; (2) the defensive attention-diversion effect observed in humans is statistically absent in our LLM cohort; and (3) a critical recognition-action gap, where LLMs successfully articulate trap recognition in their reasoning but exploit the deceptive elements anyway 73.4\% of the time. Across the 21 models, trap recognition in reasoning text did not predict fell-for-trap behavior (Spearman $r = +0.08$, $p = 0.73$). Ultimately, these findings demonstrate that human-centered deception hypotheses do not reliably transfer to AI attackers, highlighting the critical need for new research into AI-native active defense frameworks.

In many domains, practitioners seek models that produce accurate forecasts while faithfully capturing latent system dynamics. Existing approaches typically sacrifice one of these goals: deep state space models often assume Gaussian latent transitions, limiting fit and forecasting, while diffusion models are highly expressive but lack principled inference for the underlying dynamics. To combine the strengths of both, we introduce the Diffusion-Driven State Space Model (DDSSM), which replaces the conventional Gaussian transition distribution with a diffusion model. Our DDSSM resolves the open problem of how to jointly train an autoencoder and a diffusion model on sequential data, thereby extending the literature on latent diffusion models for time series. Moreover, we find that the DDSSM empirically outperforms a state-of-the-art deep SSM at fitting and forecasting a simulated time series with multimodal transitions.

Image compression for machines calls for a unified codec that serves multiple downstream vision tasks. Existing approaches either adopt task-specific end-to-end designs, raising parameter and deployment overhead, or rely on transfer-based adaptations that remain externally attached and heuristic task design. A key limitation shared by both lines of work is their largely static computation pattern, which applies similar transformations across tokens despite the fact that different image regions exhibit markedly different semantic importance and complexity for machine perception. We propose MoECodec, a token-aware image compression framework that supports multiple downstream tasks within a single model. MoECodec replaces the FFN layers in transformer-based compression model token-wise Mixture-of-Experts (MoE), enabling dynamic, token-level computation conditioned on the input content and task objective. To make MoE effective in compression model, we introduce a stable routing strategy that combines expert-choice routing with spatial total variation regularization to encourage spatially coherent assignments, and we propose a lightweight expert architecture, Group Shuffle MLP (GShMLP), to control parameter growth. Extensive experiments show consistent improvement against baselines on both conventional image reconstruction and machine tasks.

AI-assisted research systems generate many failed attempts, but those failures rarely become a durable, shared knowledge asset. We propose a negative knowledge memory layer: a curator agent converts each failed attempt into a bounded, typed record in a shared bank, and a downstream research agent explicitly adopts or rejects those records before proposing its next experiment. We evaluate this layer in two settings: same-task retry on ScienceAgentBench and cross-task scientific research on two nonlinear math-physics PDE problems. The negative knowledge layer outperforms vanilla AutoResearch baselines while using fewer tokens; agents with the negative knowledge bank solve new tasks that all baselines fail to solve in PDE systems research. We also show that the previous negative knowledge bank can transfer and enhance AutoResearch on different PDE problems. These results suggest that structured negative knowledge is a knowledge asset that should be explicitly maintained in broader AI-engaged scientific research beyond a memory-compression or debugging aid, alongside positive findings, as a collective infrastructure for scientific memory. Code is available at https://github.com/hch-wang/Negative_Knowledge.

As Large Language Models (LLMs) deploy into mission-critical domains (e.g., finance, medicine, and law), output reproducibility has become a strict system requirement. While practitioners use greedy decoding to eliminate algorithmic stochasticity, empirical deployments with 16-bit precisions still exhibit catastrophic output divergence across heterogeneous GPUs. Through SASS-level profiling, we reveal that this inconsistency is fundamentally driven by truncation errors introduced during downcasting at kernel boundaries. However, achieving reproducibility via a global FP32 pipeline incurs prohibitive system penalties: bypassing 16-bit hardware accelerators hurts compute efficiency, while upcasting the KV cache doubles memory overhead. To bridge this gap, we propose Hybrid Error ALleviation (HEAL), a targeted intervention that approximates FP32 precision while resolving hardware constraints through two targeted mechanisms. First, recognizing that floating-point formats underutilize their bit-width for Q, K, V tensors, HEAL applies INT16 quantization that preserves numerical stability without expanding the KV cache footprint. Second, HEAL synthesizes high-precision matrix multiplications via an algebraic error compensation strategy, executing entirely on high-throughput 16-bit Tensor Cores. To evaluate our approach practically, we introduce MCR-Bench, a benchmark targeting reproducibility in mission-critical tasks. HEAL achieves the same level of reproducibility on downstream tasks as the FP32 baseline while reducing the performance overhead by up to 7.1x.

Origin-Destination (OD) demand prediction is fundamental to intelligent transportation systems, yet real-world OD flows are often dynamically sparse, long-tailed, and characterized by heterogeneous zero-flow patterns. These properties make it difficult to distinguish whether an OD connection is active from how much demand it generates once activated. Many existing methods primarily treat OD prediction as a single flow regression task, which limits their ability to model low-frequency, intermittent, and long-tailed OD interactions. To address these challenges, we propose SAGMTL, a Structure-Aware Graph Multi-Task Learning framework for dynamic sparse OD demand prediction. SAGMTL decomposes OD prediction into structural state modeling and flow intensity estimation, jointly learning regional activity states, OD connection activity, and edge-level flow intensity within a unified framework. Specifically, a node-edge collaborative representation module captures regional semantics, temporal dynamics, and spatial priors through interactive node-edge updates, producing structure-aware representations for dynamic OD interactions. Based on these representations, SAGMTL estimates OD flows by jointly modeling stable demand patterns and short-term fluctuations. A multi-constraint objective further improves sparsity awareness and structural consistency. Experiments on three real-world urban mobility datasets from Beijing, Chengdu, and Nanjing show that SAGMTL achieves superior overall performance compared with state-of-the-art baselines. Further analysis demonstrates that explicitly modeling regional activity, connection states, and flow intensity improves the robustness of dynamic sparse OD demand prediction.

Long-horizon spacecraft trajectory forecasting suffers from error accumulation due to the absence of corrective observations in the forecast regime, making reliable uncertainty estimation crucial for safety-critical decision-making such as space domain awareness and conjunction assessment. While high-fidelity physics-based orbit propagators provide accurate deterministic forecasts, they typically lack calibrated uncertainty estimates over long horizons. We introduce a Predictor--Corrector framework in which a physics-based continuous-time $\textit{deterministic}$ forecaster is augmented with a learned continuous-time $\textit{probabilistic}$ Corrector that models forecast errors. The proposed Corrector can be wrapped around an existing deterministic propagator to improve forecast accuracy while producing sharp and calibrated full-covariance uncertainty estimates. The Corrector is based on Latent Neural Controlled Differential Equations (Latent NCDEs) and models the probabilistic temporal evolution of forecast errors in continuous time, naturally supporting irregular sampling and missing features. We further introduce a loss function that promotes calibration and sharpness in long-horizon uncertainty propagation. We evaluate the proposed framework on long-horizon spacecraft trajectory forecasting using real-world data from NASA's Crustal Dynamics Data Information System (CDDIS), wrapping the Corrector around NASA's General Mission Analysis Tool (GMAT). Across forecast horizons of 2--4 days without observations and six rolling test windows, the proposed approach consistently improves accuracy and uncertainty calibration compared to deterministic baselines and Latent ODE-based correctors, demonstrating the effectiveness of the continuous-time probabilistic Corrector for trajectory forecasting.

The evaluation of vision-language models (VLMs) for chest X-ray (CXR) analysis has largely been limited to disease-presence classification without visual grounding. Such evaluations fail to verify the expert-level lesion perception necessary to ensure the clinical reliability of VLMs. To address these limitations, we introduce CheXpercept, a sequential, multi-level perception benchmark that mirrors a radiologist's cognitive workflow across coarse-level detection, fine-level contour evaluation and revision, and semantic-level attribute extraction. To ensure high clinical fidelity at scale, we construct the dataset using a semi-automated generation pipeline paired with a review by six medical experts. CheXpercept contains 10,400 QA items derived from 2,100 CXRs, covering seven clinically critical pulmonary and cardiac lesions. To demonstrate the current landscape of VLM perception, we benchmark 14 general and medical VLMs on CheXpercept. The models achieve adequate performance only at the coarse level, with accuracy degrading precipitously on deeper visual tasks. Notably, medical VLMs show almost no perceptual advantage over their general-domain counterparts, highlighting a systemic flaw in current domain adaptation. The code and dataset will be publicly available.

As artificial intelligence advances into the era of Embodied AI, live musical interaction urgently needs to break free from the limitations of offline, unidirectional generation, achieving a "virtual synergy" capable of low-latency, dynamic interplay. To address this, this technical report presents LK_Jam, a real-time, bidirectional human-computer interactive music generation system based on a lightweight Gated Recurrent Unit (GRU) and a high-performance audio host architecture. In the algorithmic representation layer, this system abandons the computationally expensive fixed time-grid. Instead, it constructs a multi-dimensional sparse event stream integrating time-shifts, continuous harmonic embeddings, and role-aware encoding, enabling the model to accurately capture turn-taking logic and micro-timing in a single-step inference. In the engineering implementation layer, this paper builds a strict multithreaded lock-free communication bridge using C++ and the JUCE framework, incorporating the RTNeural inference engine designed specifically for real-time audio. By utilizing compile-time network topology solidification and a zero-allocation (allocation-free) mechanism, the end-to-end overhead of autoregressive decoding is strictly locked at \(O(1)\) complexity, structurally mitigating the risk of audio thread dropouts in DAW plugin environments. Furthermore, this study designs a three-stage progressive training strategy, achieving a leap from basic chord harmonization to expert-level interaction. Preliminary observations and architectural analysis demonstrate that while ensuring musical coherence and interactive role-play, the proposed system successfully challenges extreme real-time engineering constraints, offering a highly robust and deployable technical paradigm for next-generation AI co-performers in live music.

Organizations deploying AI face two fundamental governance challenges: managing AI risk and sustaining AI value. Both depend on evidence whose sufficiency cannot be taken for granted. We call the shared underlying challenge the AI Evaluability Gap: the condition in which organizations lack sufficient evidence to support high-confidence governance decisions regarding either risk or value. We argue that this gap reflects a category error in current practice. Existing governance approaches focus primarily on properties of systems, such as safety, fairness, reliability, compliance, and value, while paying comparatively little attention to the evidentiary foundations required to justify decisions about those properties. We further argue that AI governance encompasses both operational decisions regarding whether a system may operate and investment decisions regarding whether it merits continued organizational resources. To address this problem, we introduce Evaluability, defined as the capability of a system to generate, maintain, and renew evidence sufficient to support high-confidence governance decisions over time. We formalize governance decisions as functions of calibrated confidence Conf(D|E) and identify six properties of evaluable evidence: observability, attributability, intervenability, verifiability, calibration, and temporal validity. The framework distinguishes Operational Certification, which relies primarily on structural evidence to justify deployment decisions, from Investment Certification, which relies primarily on causal evidence to justify continued resource allocation. We argue that evidence sufficiency is a missing layer of AI governance and that closing the AI Evaluability Gap is a prerequisite for both managing risk and sustaining value in AI-enabled organizations.

Forecasting future events is a critical challenge for large language model (LLM) agents, spanning domains from elections and monetary policy to financial markets. However, evaluating progress on this task presents a fundamental trade-off between efficiency and environment fidelity. While live evaluation benchmarks suffer from an inherently slow feedback loop, existing retrospective replays typically restrict agents to static, pre-frozen databases that sacrifice the environmental realism of actual deployments. To tackle this issue, we introduce Agentic Time Machine (TM), an infrastructure that approximately reconstructs the web state at any chosen past time by filtering post-cutoff content. Leveraging this evaluation infrastructure, we further propose a planner-solver-aggregator multi-agent framework that breaks each question into diverse analytical angles, gathers evidence in parallel, and combines the results into a single forecast. Experiments show that offline scores under TM correlate strongly with live FutureX scores, validating that TM offers a fast and reliable sandbox for forecasting-agent evaluation. On FutureX-Past and Polymarket evaluated under TM, our framework achieves the highest score among strong closed-book, tool-augmented, and self-consistency baselines. On the official FutureX live leaderboard, our system achieves the best average rank over four consecutive weeks, including 1st place in May Week 1. As of June 17, it also ranks 1st on FutureX's official eight-week overall leaderboard.

A self-determining system is defined as one in which causes originating within the system influence the system itself. This definition raises the question of how to specify system boundaries. Although the concept of "closure" is commonly used for this purpose, defining boundaries solely in terms of causal relations introduce challenges, such as how to handle external causes and circular causality. To address this issue, we introduce two types of asymmetric relations: causal and constitutive. We propose that system boundaries can be defined as closures of loops formed by these relations, referred to as causal-constitutive loops. By constraining constitutive relations, the resulting system necessarily includes internal causes and thereby satisfies self-determination. Furthermore, to prevent reduction to supervenience, constitutive relations must involve at least two independent variables. This minimal requirement leads to two interdependent loops, which implies a dual-process organization.

The metanym game is a competitive word game for LLMs that measures structural intelligence against established cognitive-science constructs. No content is given in advance; the contestants create all of it -- a new kind of analogy test, analogical production falsifiable sentence by sentence, with no fixed test set to leak into training (contamination-resistant by construction). In the council-of-peers benchmark, the contestants also rate each other's creations. We introduce the first spectral solution, to our knowledge, to the wicked problem of benchmarking LLMs' factual accuracy without golden keys or oracle models: one singular value decomposition of the evaluators' ratings matrix yields their competence as both generators and judges of true statements at once. Competence on the subjective criteria comes from each judge's rating consistency as the yardstick shifts. The factual rating correlates with GPQA Diamond at Pearson r = 0.92. Scored separately, making and judging dissociate -- judging is the scarcer skill: the strongest generators are middling judges, the sharpest judge a mid-pack generator. To scale, the strongest players form a council that does the official benchmarking; its seats are contestable -- a stronger model earns one on the benchmark's own rating. The benchmark is entirely self-contained and self-consistent, a stable gauge over time.

Scientific discovery workflows often depend on structured curation from the literature. This is difficult for current agents because the key evidence is scattered across long text, dense tables, and figures, and the final records often require reasoning across multiple evidence fragments rather than copying a single span. We study scientific curation from multimodal sources and introduce Beaver, an agent harness that extracts structured information from scientific papers while preserving provenance to the supporting evidence. Beaver combines a frontier agent with multimodal evidence tooling, task scaffolding, and artifact-grounded autoresearch. These components turn curation into a staged, auditable workflow and enable an iterative evaluate--diagnose--revise loop, where persistent run artifacts expose stage-localized failures and guide harness updates. Experiments show that Beaver reaches 81.0 on Gold-Referenced Attribute Score (GRAS), an attribute-level measure of agreement with gold curated records, outperforming frontier agents by over 23 absolute points. Ablations show that task scaffolding, multimodal evidence tooling, and provenance traces each contribute meaningfully to performance, while attribute-level analysis shows the largest gains on high-value attributes that require cross-modal reasoning and normalization. These results show that, for scientific curation from papers with multimodal evidence, harness design is a central determinant of agent performance.

Understanding the generalization capabilities of visuomotor policies is essential in the development of capable robotic agents. Generalizable models learn structures that transfer across domains. However, in practice, visuomotor policies test performance by interpolation on known distributions using unstructured domain shifts (e.g. lighting, clutter, diverse objects). We argue that to measure generalization capabilities we must instead test the inductive capacity of policies on progressively harder, out-of-distribution task variants. We call this inductive generalization, drawing directly on how axis-based evaluation has revealed inherent generalization limitations in language models (e.g. sequence length, counting) arXiv:2502.00197 . We provide a reusable and formal evaluation protocol for measuring inductive generalization in any manipulation policy, and establish baselines showing that existing paradigms fail this test; e.g. SoTA Vision-Language-Action models and find that policies that appear to generalize to prior domain shifts (distractors, etc) fail inductive generalization tests. These results expose a class of learning challenges orthogonal to those addressed by data and model scaling in robot learning, yet are imperative to solve in order to realize general purpose robots.

Biomedical researchers increasingly use AI-generated analyses and reports to interpret protein-level signals, but static outputs are often insufficient for research decision-making, where users need to inspect evidence, assess uncertainty, compare mechanisms, and refine hypotheses. We present \textsc{BioInsight}, a multi-agent system that moves from static biomedical report generation to interactive evidence-centered interactive interface generation. Given a disease name, a protein association table, and optional cohort metadata, BioInsight organizes disease-specific evidence through typed intermediate artifacts, including ranked pathways, literature evidence packets, protein-level reasoning notes, citation-grounded reports, dashboard schemas, and rendered interactive interfaces. The system decomposes evidence retrieval from mechanistic reasoning, normalizes citations through deterministic components, and converts the same structured evidence used in the report into an interactive interface. We evaluate BioInsight on standardized biomedical QA, challenging protein-function reasoning, and end-to-end biomedical evidence synthesis. Results show that BioInsight achieves best, and suggest that biomedical AI systems should move beyond text-only and static reports toward provenance-preserving, interactive evidence artifacts.

Multilingual language models often exhibit performance disparities across languages that can arise as early as the tokenization stage. Widely-used subword tokenization approaches favor high-resource languages, and tokenizer-free methods still yield longer sequences for scripts with a higher bytes-per-character ratio. To address these shortcomings, we propose to use the International Phonetic Alphabet (IPA) as a language-agnostic input representation for multilingual tokenizers. IPA provides a compact symbol inventory, greater cross-lingual character overlap, and a more balanced byte-per-character distribution across languages. We train matched pairs of text vs. IPA subword tokenizers across 24 languages and 14 scripts and demonstrate that IPA tokenizers consistently improve tokenization quality, especially for non-Latin scripts, and generalize more effectively to unseen languages and scripts.

Text-to-image generation is a form of generative artificial intelligence (GenAI) that converts textual descriptions into images. Most applications of GenAI in modeling and simulation (M&S) have focused on large language models for documentation, coding, or explanation. By contrast, the potential of image generation remains largely unexplored. This tutorial introduces text-to-image generation to the M&S community and details how it can support several M&S tasks, including communicating conceptual models, visualizing simulation outcomes, generating educational materials, and interfacing heterogeneous models in multi-scale simulations. The tutorial combines conceptual guidance with practical workflows, explaining how modern image generators operate, how prompts and simulation outputs can be translated into visual scenes, and how practitioners can integrate these tools into reproducible local pipelines. By focusing on transferable principles rather than specific tools, the tutorial equips M&S practitioners with the knowledge needed to evaluate, adopt, and adapt text-to-image generation in their simulation workflows.

Make-lineage workflow runners decide whether a job must re-run from file-modification time (mtime, a timestamp) -- a broken proxy for the question that matters: did the content change? A git checkout, a tree copy, or a backup restore rewrites mtimes without touching content, forcing spurious re-execution; and in the reverse case -- when an output looks newer than its inputs but its content is stale -- the stale output is silently reused. (Snakemake 7's per-output provenance survives this churn, as local bookkeeping; GNU Make and pure-mtime fast paths are where it bites.) OxyMake, a single-binary Rust workflow engine, replaces the proxy with a content-addressed cache key: a BLAKE3 hash of rule source, input content, parameters, environment, and platform. Because the key is a pure function of these declared inputs, the caching decision survives mtime churn and travels across same-platform machines and shared caches. Phantom re-runs vanish for declared inputs (no sandbox: an undeclared input is invisible to the key). The spec stays declarative and statically parseable, keeping the Make rule model so Snakemake pipelines port directly. DAG resolution is an order of magnitude faster than Snakemake's on large graphs, but a cold end-to-end run is slower -- the price of content-addressed bookkeeping -- repaid several-fold on the warm re-run that caching exists to serve (exact figures, hardware, and a bundled reproducer are in the evaluation). Execution is daemon-free via a cooperative claim/reclaim protocol (sessions claim jobs, reclaiming stalled ones); today two sessions duplicate work safely rather than coordinate, and wiring the protocol as a hard execution gate is staged, not yet done. Cross-session safety is specified in TLA+ and model-checked over all interleavings for 2-3 sessions, assuming atomic state commits. An ox.lock plan-of-record and NDJSON event stream record exactly what ran.

Understanding and predicting microstructure evolution is central to materials design, yet purely data-driven spatiotemporal learning models often suffer from limited physical consistency and degraded long-term prediction accuracy. In this work, we introduce a physics-guided fully convolutional spatiotemporal learning framework for microstructure evolution prediction. Unlike prior self-supervised approaches, the proposed method explicitly incorporates governing physical equations into the training objective, thereby encouraging the learned dynamics to remain consistent with known thermodynamic and kinetic laws. This physics-guided formulation improves predictive accuracy, long-horizon stability, and robustness across spatial resolutions and temporal prediction settings. Extensive experiments for spinodal decomposition demonstrate that incorporating physics-guided residual regularization leads to more faithful reproduction of microstructural morphology, statistics, and evolution trends compared with purely data-driven baselines. The proposed framework preserves the scalability and computational efficiency of fully convolutional architectures while bridging the gap between high-fidelity physics-based simulations and data-driven surrogate modeling, offering a reliable and efficient surrogate-modeling step toward digital-twin-enabled microstructure evolution prediction.

As Self-Driving Cars continue to expand internationally and use multi-modal systems such as VLMs as a cognitive backbone for their Action models; how well will these systems generalize in new settings, in particular out-of-distribution (OOD) edge-case scenarios in new geographies? In this paper, we study this open question by providing a full factorial analysis with human drivers of Lima, human drivers from New York City, and VLMs and showing them dashcam footage collected from Lima and New York City -- prompting them with a variety of questions under a Visual Question Answering (VQA) paradigm. In particular, we pick these two cities as they are highly challenging driving locations where no Self-Driving Car company currently operates in, and ask questions that span 4 categories: Factual, Ratings, Counterfactual and Reasoning. We find that Humans and VLMs diverge in their responses -- though this is modulated by the type of questions asked, and that Humans answer similarly independent of where they are from (Lima/NYC). To our surprise, we did not find a strong difference in terms of answers (Humans or VLMs) that was modulated by geography, likely due to their high out-of-distribution nature. Our dataset is available at: https://huggingface.co/datasets/Artificio/robusto-2

Programming by Demonstration (PbD) offers a human-centered way to author procedural knowledge for LLM agents: users communicate what they want by showing rather than by writing prompts or code, making agent authoring accessible to non-programmers. The natural output of a PbD recording is a flat action log, but how this log is organized before being passed to the agent is an open design question with significant consequences for plan quality. We propose grouping recorded actions into labeled, hierarchical subgoals and evaluate the effect of this organizational structure in a controlled experiment. Across 85 web automation tasks, we compare a zero-shot baseline against four demonstration formats that share identical action sequences but differ in structure. On 43 natural-language tasks with vague descriptions, hierarchically grouped demonstrations improve pass rates from 76.7\% to 90.7\% (paired permutation test $p{=}0.034$; win-loss 6:0), while flat demonstrations show a smaller, non-significant improvement. On 42 tasks with precise descriptions, no format provides any benefit, confirming that the hierarchical advantage arises specifically when descriptions leave procedural details ambiguous. Ablation shows that subgoal grouping alone drives the effect: preconditions, postconditions, and parameter annotations add no measurable benefit. These results offer a concrete design recommendation for PbD pipelines and, more broadly, for any system that feeds procedural context to an LLM agent: segment action sequences into named subgoal groups rather than presenting flat step lists.

Radio-frequency (RF) monitoring is essential for space domain awareness, but it often generates large, variable, and sparsely populated datasets with few labels. These observations can capture satellites, space debris, and the ionospheric background, yet interpreting them typically requires specialized subject-matter expertise. Supervised deep learning methods can perform well on labeled RF data, but they require many annotated examples and may need careful retraining as RF conditions change. Semi-supervised approaches offer a practical alternative for limited-data settings by using unlabeled observations to reveal latent patterns that experts can interpret. In this paper, we present a semi-supervised RF detection and classification workflow for satellite monitoring that combines Non-negative Matrix Factorization with automatic model determination (NMFk), expert-guided cluster interpretation, and classifier-based prediction. We first represent RF observations as a non-negative feature matrix and apply NMFk to estimate the number of clusters that best captures patterns in the unlabeled data. Subject-matter experts then assign physical meaning to the resulting clusters, including satellite detections, ionospheric environmental conditions, and other RF event categories. Finally, we train a classifier on these interpreted clusters to evaluate performance on a test set and categorize future observations. This pipeline reduces reliance on large pre-labeled datasets by pairing unsupervised factorization with expert interpretation, enabling an interpretable and transferable methodology for detecting, observing, and classifying behavior in RF data.

Generating formal specifications for C programs remains a challenge in formal verification due to the manual effort, expertise, and semantic precision required. While recent advancements in large language models (LLMs) offer promise in automating specification synthesis, current approaches often lack semantic depth and produce unverifiable or incomplete contracts. To address these limitations, we introduce AutoACSL, a novel framework that integrates LLM prompting with semantic features extracted from Code Property Graphs (CPGs). AutoACSL performs static analyses to extract key semantic elements, including arithmetic operations, loop and recursion structures, and return value propagation, which are encoded into structured prompts. These prompts enable the LLM not only to generate normal behavioral specifications but also to include constraints that prevent inputs leading to runtime errors. AutoACSL employs a feedback-driven synthesis loop, where candidate specifications are verified using Frama-C/WP and refined iteratively until verification succeeds or a termination condition is met. Evaluated on 604 programs drawn from diverse datasets, AutoACSL achieves a 98% specification generation success ratio and a 96% full proof ratio when paired with Gemini-3. Compared to a code-only baseline, AutoACSL improves the full proof ratio by 24.7% to 51.7% across four LLMs (GPT-o4 Mini, GPT-5.2, Grok-4.1, and Gemini-3), demonstrating that integrating large language models with CPG-based static analysis substantially enhances both generation robustness and verification effectiveness for automated ACSL specification synthesis.

Policies must operate across diverse conditions, yet a single policy is often conservative while fully adaptive schemes can be complex. We study zero-shot generalization in contextual dynamical systems and introduce a performance-centric, directional task dissimilarity--the signed divergence--that upper bounds the generalization gap from a source context to a target context. The signed divergence induces $\varepsilon$-tolerance sets that certify when a source policy class generalizes, and it yields a concrete notion of task-space complexity: the minimum number of source contexts needed so that every target context incurs at most $\varepsilon$ generalization gap. Under a mild local smoothness assumption on performance, the induced tolerance sets admit certified inner/outer balls and instance-dependent volume bounds on task-space complexity. In the finite-oracle setting, source selection reduces to set cover; a greedy strategy inherits the standard $H(n)$ approximation guarantee. Using a Mass-Spring-Damper system with linear-quadratic regulator (LQR) controllers and a nonlinear CartPole system with deep reinforcement learning controllers, we show that greedy selection achieves the same $\varepsilon$-coverage with fewer policies than uniform or random baselines. Our approach delivers a performance-based task similarity measure and practical certificates for building generalizable control with simple policies.

Background. Inactivity thresholds classify scientific open-source software (OSS) as abandoned but cannot distinguish permanent abandonment from temporary dormancy; moving the cutoff from 1 to 36 months changes the abandoned count in the SciCat corpus from 18,030 to 8,010. Aims. We characterize dormancy causes, revival mechanisms, recovery durability, and lifecycle archetypes in dormant-revived scientific OSS. Method. From 18,247 SciCat repositories we identify 2,984 dormant-revived candidates and field-code a stratified sample of 750 projects with 75 analyst-coders under a two-phase adjudication protocol (post-adjudication kappa 0.779-0.857). A rule-based classifier produces five dimensions: dormancy cause (T1), revival mechanism (T2), nature of revival work (T3), revival sustainability (T4), and lifecycle archetype (T5). Results. Dormancy cause is unresolvable from repository evidence for 52.5% of projects; among resolvable cases, feature/milestone freeze outnumbers research-output completion 5.4:1. Non-sustained recovery outnumbers sustained 2.14:1; 11.5% of apparent revivals are bot-only or single-spike artifacts. Lifecycle archetype is more strongly associated with sustainability than revival mechanism or work type (medium effect on the structurally-independent subset). Conclusions. A fixed inactivity threshold is insufficient to reliably classify scientific OSS abandonment. Gap duration, lifecycle archetype, and contributor continuity together provide more discriminating information than any single threshold.

Trust in the application of generative Artificial Intelligence (AI) relies on well-governed measurable evidence of performance and safety. In practice, however, evaluation data is often fragmented across systems, inconsistently structured and difficult to compare. We introduce the Generative Responsible AI Data Evaluation Schema (GRAIDES) as a lightweight open-source data model for centralising AI observability across popular vendors. Practical blueprints for code, architecture and statistical evaluation are shared as guidance about how to approach generative system assurance at the organisational level. Illustrative case study results are reported from Westminster City Council's AI catalogue with a focus on measuring human-model alignment including detecting systematic disagreement between evaluators. By framing evaluations as a data modelling problem, GRAIDES provides a practical pathway toward more consistent and reproducible benchmarking, tuning and assurance activities for generative AI systems.

High-performing human-human teams learn intelligent and efficient communication and coordination strategies to maximize their joint utility. These teams implicitly understand the different roles of heterogeneous team members and adapt their communication protocols accordingly. Multi-Agent Reinforcement Learning (MARL) has attempted to develop computational methods for synthesizing such joint coordination-communication strategies, but emulating heterogeneous communication patterns across agents with different state, action, and observation spaces has remained a challenge. Without properly modeling agent heterogeneity, as in prior MARL work that leverages homogeneous graph networks, communication becomes less helpful and can even deteriorate the team's performance. In the past, we proposed Heterogeneous Policy Networks (HetNet) to learn efficient and diverse communication models for coordinating cooperative heterogeneous teams. In this extended work, we extend Heterogeneous Policy Networks (HetNet) to support scaling heterogeneous robot teams. Building on heterogeneous graph-attention networks, we show that HetNet not only facilitates learning heterogeneous collaborative policies but also enables end-to-end training for learning highly efficient binarized messaging. Our empirical evaluation shows that HetNet sets a new state of the art in learning coordination and communication strategies for heterogeneous multi-agent teams by achieving an 5.84% to 707.65% performance improvement over the next-best baseline across multiple domains while simultaneously achieving a 200x reduction in the required communication bandwidth.

Continual adaptation is essential for multimodal large language models (MLLMs) deployed across evolving domains, but the state-of-the-art MR-LoRA method highly relies on the assumption that a MLLM-based router is necessary to process complex multimodal inputs. This paper revisits this claim on the MLLM-CL benchmark and argues for two claims. \textbf{First}, routing does not require an MLLM: a simple training-free, replay-free ptotypical routing method (\textsc{RePRo}), uses frozen pretrained features and task prototypes to match the MLLM-based router of MR-LoRA at far lower computational cost. \textbf{Second}, shared experts do not improve continual learning for MLLMs, despite their theoretical appeal. We show that these findings arise from two structural limitations of MLLM-CL: (1) its tasks are \textbf{highly separable} in representation space, and (2) its fixed task order makes conclusions \textbf{sensitive to a single curriculum} rather than robust across diverse continual-learning trajectories. As a result, the benchmark primarily rewards learning in isolation rather than genuine continual transfer. This motivates a new design for future benchmarks of continual MLLM learning, with overlapping task manifolds, multiple task orders, fine-grained domain shifts, and evaluation protocols that reward forward transfer as well as retention.

Nash equilibrium (NE) arises from selfish utility maximization, yet its social welfare can be arbitrarily far from optimal. Moreover, computing an NE is intractable in general. We study augmented game models in which players use budget-balanced internal transfers to improve incentives before play. We first introduce \emph{Self-Enforcing Transfer Equilibrium} (SETE), where players commit to nonnegative peer-to-peer transfers that are paid only if the recipient does not deviate from a prescribed strategy. For polymatrix games, we show that every stationary point of the social welfare function, in particular any socially optimal strategy profile, can be sustained as a SETE. This induces a Nash equilibrium in the agent normal form of the corresponding augmented game. We further propose a polynomial-time algorithm and a decentralized learning dynamic to compute such product-form equilibria. We then introduce \emph{Mediated Self-Enforcing Transfer Equilibrium} (M-SETE), where a mediator makes both the payment schedule and the prescribed strategies binding offers. This additional enforcement resolves the agent-normal-form limitation: an M-SETE is a Nash equilibrium of the augmented game itself, not merely of its agent normal form, and any socially optimal strategy profile can be supported as an M-SETE in any finite game while preserving budget balance. Thus, internal transfers improve welfare and computation while preserving independent play on the equilibrium path. When full sequential-game stability is required, binding mediation provides the corresponding implementation.

Large language models can store both outdated facts and newer superseding facts in their parameters, but standard prompting may still elicit the outdated answer. We formalize this problem as Parametric Temporal Conflict (PTC) and introduce Temporal Attractor Steering (TAS), a three-stage test-time intervention that detects likely conflicts, identifies a conflict-critical layer, and steers hidden states toward newer-fact representations without retraining or external retrieval. We construct an 8,746-record verified benchmark across five Wikidata relations and evaluate four open-weight language models from three families: Qwen-2.5-1.5B/7B, Mistral-7B-v0.3, and Llama-3.1-8B. Single-layer activation patching achieves answer-flip rates of 0.72-0.85 across all models. End-to-end TAS resolves 29-57% of PTC cases while preserving 85-99% accuracy on non-conflict queries, outperforming a matched ITI baseline on three of four models. These results show that outdated parametric knowledge can be selectively overridden at inference time.

Dynamic average estimation is a critical problem in multi-agent systems, enabling agents to collaboratively estimate time-varying signals using only local information exchange. Traditional model-based approaches often face challenges related to convergence speed and sensitivity to network topology changes. This paper introduces a novel learning-based solution leveraging Gated Graph Neural Networks (GGNNs) for fast-convergent dynamic average estimation in a fully distributed manner. Taking advantage of the inherent structure of GGNNs, the proposed method models the estimation process as a distributed autoregressor, ensuring rapid convergence while maintaining stability. We incorporate a regularization term during training to enforce convergence guarantees and introduce an encoding-decoding mechanism to reduce communication overhead without sacrificing accuracy compared to standard GGNNs. Extensive numerical experiments demonstrate that our approach significantly outperforms conventional model-based estimators in terms of both convergence speed and precision, making it a promising alternative for multi-agent applications that require dynamic average estimation.

Long-running language-model systems accumulate interaction history that outgrows the context window, so they must continually evict. When an eviction policy drops a load-bearing detail, for example an access token issued at login or a path the next call needs, the action fails. We present LRE (Learned Relevance Eviction), a few kilobytes, CPU-only, language-model-free scorer that learns which units of history are load-bearing and keeps them by verbatim extraction. Under a matched-budget comparison, in our experiment, no baseline dominates LRE on the accuracy-cost plane. On agents, LRE matches the accuracy of keeping the entire history overall. On the simplest tasks, it exceeds that no-eviction baseline by 27%, while requiring zero compressor calls and reducing peak context size by up to 52%. A controlled study trace shows LRE completes tasks where the others loop, finishing one such task in 37% fewer calls than keeping everything and solving 14 tasks where no other run policy does. On conversational memory, LRE outranks dense and token-pruning encoders at zero neural cost. In downstream evaluation, LRE gives the best budgeted answer quality on LoCoMo reading 68% fewer tokens. Its supervision can also be annotation-free: training only on the system's own behavior recovers 95% of the supervised scorer's effectiveness. We argue that, because memory eviction in LLM agents is a fidelity problem, it requires a deployable proactive policy where the future query is unavailable and exact state is decisive, and that cheap learned relevance can be sufficient.

Machine-Learning Operations (MLOps) is maturing into a software-engineering discipline, yet its tiny-scale variant (TinyMLOps)-targeting the resource-constrained microcontrollers embedded in cyber-physical systems (CPS)-remains poorly understood in industrial practice. Opaque models, noisy heterogeneous data, and tight memory budgets hinder adoption in safety-critical settings, where most decisions still rely on human experts. We report a field study of an end-to-end, knowledge-centered TinyMLOps pipeline that fuses domain physics, expert speculation, and sensor streams to deliver explainable, low-footprint models deployable on-device. The pipeline spans automated collection and cleaning of heterogeneous time series, knowledge-driven feature construction, interpretable regularized models, and rolling temporal cross-validation under concept drift. We evaluate it on 4.4 GB of data from two offshore-wind cable-trenching campaigns. The classifier anticipates harmful load peaks up to three minutes ahead at 0.84 AUC within a 32 kB footprint on an ARM Cortex-M4; an ablation shows that injecting prior knowledge halves false alarms and surfaces actionable operational rules. Replaying recommendations in operational dashboards indicates an 11% reduction in non-productive time. We distill engineering lessons and validity threats for trustworthy TinyMLOps in CPS, and release code and an annotated dataset to support reproducibility.

Executable evaluation -- checking the consequences of an agent's actions with a program rather than grading its prose -- has become a prominent way to assess tool-using AI agents in software settings. Electric power engineering has not yet had an analogous benchmark: language-model use is still dominated by retrieval and text question answering, while agents acting on power-system artifacts remain mostly academic prototypes. We introduce the Power Systems Agent Benchmark, an executable benchmark for power-engineering agents. An agent receives a structured task and returns a structured solution; a deterministic evaluator recomputes the engineering quantities, checks operational constraints, and returns a feasibility flag, a normalized score, and explicit violations. The benchmark contains 41 task families across eight areas of power engineering, from power flow and protection to stability, microgrids, reliability, power quality, and forecasting. Each task is grounded in a citable source, standard, or documented engineering formulation. To resist contamination, held-out cases are synthesized on demand by per-family generators from private seeds: the construction is inspectable, but the instances remain private. In a reference evaluation with three command-line agents, the strongest score near the compact tier's ceiling, a smaller open model trails, and public and held-out performance are broadly consistent; a separate public-split grid with OpenCode and Aider probes harness effects. The reference evaluation doubles as quality control: unanimous failures flag candidate task or evaluator defects, and it exposed a latent evaluator bug missed by self-consistency checks. The evaluators are compact deterministic surrogates, but the task contract allows their internals to be upgraded to simulator-backed checks without changing how tasks are posed or solved.

Developing robust models for 3D visual grounding (3DVG), the localization of entities in a 3D scene described in natural language, is important for enabling agents to correspond spatial language with objects in the physical world. However, the lack of diverse descriptions at scale prevents models from generalizing beyond simple linguistic patterns. Recent such attempts lack diversity in the constraint types and language used to ground objects. Captioning methods cannot precisely contrast objects, which is important for visual grounding. We therefore propose ViGiL3D++, a scalable, scene-agnostic method that generates diverse visual grounding queries by combining constraint sampling in scene graphs with the language generation of LLMs. We show that it has greater diversity over existing scaled datasets and improves model performance over several 3DVG benchmarks but also illuminates outstanding limitations of VLMs.

Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length. Standard dense attention also applies the same set of attention heads to every token regardless of token difficulty or information content. This uniform activation can waste compute, especially as sequences grow longer and attention cost increases rapidly. We propose Grouped Query Experts (GQE), a mixture-of-experts layer on top of grouped-query attention (GQA). Within each GQA group, a router selects k query-head experts per token while all key-value (KV) heads remain dense and unchanged. Thus, GQE keeps the KV cache benefits of GQA and reduces only the active query-head computation. On a fixed 30B token budget at the 250M parameter scale, GQE matches the all-active GQA baseline in downstream accuracy while activating half the query heads per token.

We study internalization processes, by which neural-network-based systems absorb an explicit computational procedure into their own weights, and how they facilitate learning. We investigate how transformers internalize the simulation of semiautomata by internalizing chain-of-thought (CoT) tokens, which classes of semiautomata are harder to internalize, and expose the flip side of internalization, that is, a progressive degradation of out-of-distribution performance. We then provide the first provable analysis of successful internalization: for the task of learning parities, we show that a simplified one-layer transformer provably first learns the target with explicit CoT supervision and then internalizes the autoregressive generation as CoT tokens are progressively removed, learning to directly compute the parity. This task is computationally hard to learn from data without CoT supervision. Finally, we discuss how learning through internalization relates to the \textit{Positive Distribution Shift} phenomenon recently introduced by~\citet{Med+26}.

Hybrid language models that mix attention and recurrent layers have shown promise: theoretically, recurrent layers ameliorate the limitations of pure transformers on state tracking, and empirically, hybrids can outperform pure transformers in loss and downstream evaluations \citep{waleffe2024empirical,merrill2026olmohybrid}. Yet it remains unclear which data or capabilities drive these gains, and to what degree they reflect the theoretical advantages motivating hybrid models. We address this question using the open weights from Olmo 3 \citep{olmo2025olmo3} and Olmo Hybrid \citep{merrill2026olmohybrid}: we compare the loss of a matched transformer and hybrid at the same target tokens under the same prefixes, stratifying the results by natural token tags, copy features, delimiter structure, and controlled synthetic probes. The hybrid has lower loss on most tag families, but the gains are not uniform: they are largest for open-class content words and smaller for many closed-class function words. Across prose, code, and markup, the hybrid's loss advantage is larger on opening delimiters than on the corresponding closing delimiters, and nearly vanishes on repeated $n$-grams. Synthetic probes show the same split: the hybrid is favored on pronoun-memory and entity-tracking tasks, whereas the transformer is favored on bracket-matching tasks that require choosing closing delimiters. These patterns suggest that the recurrent layers in hybrids improve predictions that leverage the semantic state of a document, whereas attention helps on tokens predictable by $n$-gram copying or syntactic bracket matching. We conclude with proof-of-concept filtered evaluations showing how token-level decompositions can sharpen pretraining diagnostics for hybrid architectures.

Technical Debt (TD) decisions are rarely just technical. Stakeholders consider heterogeneous criteria, such as deadlines, client expectations, architectural consequences, available resources, organizational authority, and team sustainability, when deciding whether to acquire, postpone, prioritize, or repay TD. However, these criteria are often reported under different labels and at different levels of abstraction, making it difficult to compare findings across TD decision contexts. This paper proposes a practitioner-informed taxonomy represented as a conceptual model of stakeholder criteria for TD decision-making. Based on a qualitative study with 11 software practitioners in Brazil, we extracted, consolidated, and organized criteria related to TD acquisition and repayment. The resulting taxonomy comprises six families: stakeholder-facing value; delivery and resource pressure; technical integrity and systemic risk; decision basis and epistemic style; governance and legitimation; and human and team sustainability. Our findings show that acquisition and repayment share taxonomic families but differ in decision function. In acquisition, criteria often operate as permission mechanisms that make shortcuts acceptable under current constraints. In repayment, criteria operate as authorization mechanisms that justify allocating time and resources to address existing debt. The conceptual model further shows how criteria become actionable through stakeholder interpretation, translation across technical and business perspectives, and organizational legitimation. The taxonomy and model provide an empirically grounded artifact for classifying TD decision criteria, comparing decision contexts, and structuring discussions about TD acquisition and repayment.

The choice of loss function and optimizer is an important decision, that shapes further model training. Yet automated architecture search pipelines (AutoML) benefits significantly more from the optimal pairing selection and vice versa. This paper investigates whether a single recipe is sufficient for heterogeneous architecture pools, or whether the optimal pairing varies across structurally diverse models. We conduct a systematic empirical study of all $3 \times 6 = 18$ combinations of six optimizers (SGD+Momentum, Adam, AdamW, RMSprop, Adagrad, Adadelta), paired with three loss functions: Cross-Entropy (CEL), Negative Log-Likelihood (NLL), and the recently introduced genetically evolved NGL loss across the base models presented in LEMUR heterogeneous architecture pool on six image classification datasets (CelebA-Gender, CIFAR-10, CIFAR-100, ImageNette, MNIST, SVHN). The 18 loss-optimizer configurations are applied to each of the 33 compatible base architectures taken from the LEMUR pool, resulting in 594 variants that were generated fully automatically by a source-level injection pipeline and evaluated under fixed hyperparameters, ensuring that observed accuracy differences are attributable solely to the loss-optimizer pairing. Our results confirm that no single pairing is universally optimal. Cross-Entropy with Adam or AdamW is the most robust choice across architecture families and datasets. NGL is a competitive alternative to CEL on standard convolutional classifiers, but only when paired with adaptive optimizers; it degrades substantially with SGD or accumulation-based methods. Adagrad and Adadelta consistently underperform under fixed hyperparameters regardless of loss function, highlighting their sensitivity to learning rate tuning. These findings provide actionable guidance for loss-optimizer selection within NNGPT Framework.

Sheaf Neural Networks (SNNs) generalize Graph Neural Networks (GNNs) by replacing scalar node signals with stalk-valued signals and by using restriction maps to measure compatibility across edges. Unlike standard graph diffusion, which encourages neighboring node features to become similar, sheaf diffusion promotes consistency through the restriction maps and can therefore model more general relationships between neighboring nodes. However, existing sheaf neural architectures mainly operate at a fixed graph resolution and do not provide a principled pooling mechanism for building hierarchical representations. In this paper, we introduce Hierarchical Sheaf Pool (HiSP), a sheaf-aware pooling framework based on local spectral coarsening. Given a partition of the graph, HiSP constructs each coarse stalk by projecting fine stalk-valued features onto the low-frequency eigenmodes of the cluster-internal sheaf Laplacian. These local modes define a cochain-level prolongation map, which allows the fine sheaf energy to be represented on the coarse space through a Galerkin operator. We further analyze the approximation induced by coarsening by separating truncation loss, due to discarded local modes, from realization loss, due to representing the projected operator as a coarse sheaf. Finally, we implement HiSP as a GNN pooling layer compatible with SNNs and provide a PyG implementation supporting batching, lifted sheaf Laplacians, and hierarchical architectures.

Large Language Models (LLMs) are increasingly being adopted in the legal domain. However, despite their strong performance, LLMs are prone to generating incorrect or hallucinated outputs, raising serious concerns about their reliability in high-stakes domains such as law. Detecting the correctness of responses of LLM-based systems is therefore a critical challenge. In this work, we explore the potential of leveraging internal artifacts of LLM to detect the correctness of their predictions in legal-domain classification tasks. We develop approaches that utilize features derived from these internal artifacts to build downstream classifiers capable of identifying incorrect LLM outputs. We evaluate our approach on two representative legal classification tasks: bail decision prediction and statute violation prediction. Our experimental results demonstrate that LLMs' internal artifacts are reliable indicators for detecting incorrect predictions in legal classification tasks, and can be applied to enhance the reliability of LLM-based classification systems.

Characterizing non-Gaussian posterior distributions in partially observed high-dimensional nonlinear systems remains a fundamental challenge in data assimilation. Ensemble Kalman filters rely on Gaussian approximations that can be inaccurate for strongly non-Gaussian posteriors, whereas particle filters suffer from severe scalability limitations. Recent score-based generative approaches improve posterior characterization but typically require supervised training with ground-truth posterior samples, which are unavailable in most practical applications. We introduce $Ω$ (Operator-based Mixture Ensemble for Generative Assimilation), a scalable framework that integrates conditional Gaussian surrogate modeling, unsupervised score learning, and generative sampling. The conditional Gaussian surrogate provides a nonlinear non-Gaussian baseline approximation while admitting closed-form conditional posterior distributions for the unresolved variables. First, $Ω$ exploits these closed-form conditional distributions to analytically recover the high-dimensional unobserved component, reducing computational cost and mitigating the curse of dimensionality. Second, $Ω$ learns only the residual discrepancy beyond an analytical baseline through denoising score matching using ensemble trajectories alone, eliminating the need for ground-truth posterior samples and substantially reducing the learning burden. Third, $Ω$ reconstructs the full non-Gaussian posterior distribution of both observed and unobserved variables via a Gaussian mixture representation, capturing multimodal, skewed, and heavy-tailed statistics. Finally, $Ω$ employs annealed Langevin sampling to iteratively refine ensemble members from the baseline toward the target posterior. $Ω$ is validated on several turbulent models with intermittency and extreme events, consistently improving posterior accuracy.

Accurate short-term electricity demand forecasting is critical for reliable power system operation, energy market planning, and infrastructure optimization. This paper presents a hybrid framework combining a Transformer encoder for temporal feature extraction with gradient-boosted decision trees (XGBoost) for daily electricity demand forecasting across New England. The framework integrates meteorological observations from six cities spanning all six New England states, calendar and holiday effects, autoregressive demand lags, and COVID-19 epidemiological variables. Hyperparameter optimization uses Optuna with a multivariate Tree-structured Parzen Estimator over 500 trials, with a leakage-free 70/15/15 chronological train-validation-test split. The hybrid model achieves a test RMSE of 8,876 MWh, MAPE of 2.05%, and R-squared of 0.906. A tabular-only XGBoost baseline achieves RMSE of 9,304 MWh, MAPE of 2.21%, and R-squared of 0.896. A Diebold-Mariano test (Harvey-Leybourne-Newbold correction) confirms the 427.7 MWh difference is statistically indistinguishable from noise (DM = -1.126, p = 0.262). An ablation study reveals COVID-19 features improved training accuracy but had asymmetric test effects: removal degraded hybrid RMSE by 3.2% while marginally improving XGBoost-only by 1.2%. A SHAP temporal analysis shows 5 of 8 COVID features rank higher on the post-acute test set than during pandemic-active training, indicating the model over-applies learned pandemic patterns. These findings establish temporal validity decay as a central mechanism: behavioral disruptions drove a strong COVID-demand signal during 2020-2021, but adaptation was complete by mid-2022, leaving epidemiological features as noise amplifying overfitting to stale pandemic patterns.

Iterative design and optimization of large, complex structures require fast and accurate prediction of stress, displacement, and other fields. Finite element analysis (FEA) is computationally expensive for this task. Existing neural network surrogates often struggle with varying topologies and complex boundary conditions. This study proposes the novel Dual-Stream Heterogeneous Graph Neural Network (DS-HGNN) for full-field stress and displacement prediction in thin-walled structures, demonstrated on box beams made of stiffened panels. DS-HGNN operates on panel-level heterogeneous graph representations and introduces physics-guided edge states initialized from edge types, spatial information, and boundary kinematics. These states are updated through dual-stream message passing that separates longitudinal and transverse structural information while allowing cross-stream exchange. Geometry and loading effects are incorporated through Feature-wise Linear Modulation (FiLM)-conditioned 1-D spectral convolutions, and physical fields are reconstructed using a spectral-bypass low-rank readout. The model is evaluated on stiffened panel datasets with different geometries, boundary kinematics, loading conditions, and material nonlinear responses. DS-HGNN achieves the lowest stress and displacement RMSE compared with six benchmark heterogeneous graph neural network models. It also reaches comparable accuracy to the strongest benchmark models using 19%-38% fewer training samples. A targeted evaluation further shows that DS-HGNN captures yield and post-yield stress features.

Medical vision-language models (VLMs) enable zero-shot clinical image classification, yet reliably detecting out-of-distribution (OOD) inputs at deployment remains an open problem. No static scoring method works across all shift types: Maximum Concept Matching (MCM) on FLAIR achieves 76.4% AUROC for far-OOD but only 42.4% for covariate shifts such as ultra-wide-field fundus images, effectively random. We trace this to a structural mismatch: covariate-shifted inputs are indistinguishable from in-distribution samples in softmax space, yet occupy distinct regions in the VLM embedding space. To exploit this untapped signal, we propose PROTON (PROtotype-based Test-time ONline OOD detection), a lightweight post-hoc module that maintains an online prototype bank from high-confidence test predictions and adaptively fuses prototype distance with MCM scoring via stream-level variance statistics, requiring no model modification, training data, or prompt engineering. On the ophthalmology benchmark FLAIR + FIVES, PROTON improves MCM by +23.9 AUROC on covariate shift, +8.8 on semantic shift, and +8.1 on far-OOD, making it the only zero-shot method to improve all three without hierarchical prompts or labeled data. Code is available at https://github.com/GenMI-Lab/PROTON, and the project page is available at https://genmi-lab.github.io/PROTON.

Finding the source of failures, known as Root Cause Analysis (RCA), is essential for identifying the root causes of anomalies and maintaining the reliability of complex systems. While causal theory has advanced data-driven RCA, existing frameworks assume causal sufficiency, failing to account for the unobserved latent variables prevalent in real-world environments. To address this gap, we propose PAG-RCA. This framework models system failures as parametric interventions over Partial Ancestral Graphs (PAGs) to perform RCA in the presence of latent variables. We use standard causal identification algorithms to find the source of failures by quantifying causal effects over the PAG. When an effect is identifiable, candidate root causes are ranked based on their exact intervention effects. When effects are structurally unidentifiable, our framework (for the first time in the RCA literature) integrates partial identification to evaluate and score candidates using analytical causal bounds. By integrating latent variables and partial identification at once our framework ensures robust RCA even under data scarcity and latent-variable scenarios where traditional methods degrade. Evaluations on synthetic data, microservice anomaly benchmarks and power-grid cascading failures demonstrate that PAG-RCA consistently outperforms state-of-the-art data-driven baselines. By improving data-driven RCA performance under data scarcity, this methodology advances reliable automated diagnostics in partially observable complex networks.

Personalizing large language models (LLMs) requires encoding long-term, user-specific behavioral patterns in a way that is computationally efficient, scalable, and compatible with a frozen base model. We present Latent Personal Memory (LPM), a scalable framework that represents user-specific history as a compact, persistent matrix of N latent slots, that are interpretable. A shared cross-attention projection network maps these slots into dynamic, input-conditioned soft prompts that are prepended to the input of a frozen LLM. We evaluate LPM on PersonaMem v1 and LoCOMO benchmarks across Qwen3-1.7B, 4B, and 8B backbones. Results demonstrate that LPM outperforms LoRA and Prompt Tuning by up to 8.8% and 54.4% in overall accuracy respectively on PersonaMem v1, while reducing KV-cache usage by over 64x. On LoCoMo, LPM matches LoRA accuracy with 120x fewer trainable parameters. We also show that the efficiency of LPM grows with context length and outperforms full-context at 128K context length.

As AI web agents proliferate, combining large language models with autonomous, browser-level control, indiscriminate content scraping by web agents has emerged as a privacy and security challenge. Existing defenses, such as robots.txt and active bot-blocking, are insufficient, as they are widely violated and easily circumvented. In this work, we demonstrate that AI web agents can be effectively distinguished from humans and traditional crawlers using a multi-layer fingerprint based on both network layer characteristics (e.g., TLS, HTTP) and browser interaction behavior. We implement this mechanism as a programmatic logging framework that can be deployed on a live, instrumented domain. By analyzing six prominent agent frameworks (AutoGen, Browser Use, Claude, Gemini, Operator, and Skyvern), we uncover latent structural differences in how these systems assemble HTTP requests, establish TLS/HTTP connections, and execute autonomous browser actions. Feeding these multi-layer features into a decision tree classifier, our framework achieves high-fidelity identification (97% accuracy), successfully isolating distinct agent architectures and differentiating agent traffic from both human browsing baselines and legacy crawlers. Our findings demonstrate that cross-layer agent tracking provides a robust, evasion-resistant strategy for content protection and web security policy enforcement.

Earth observation imagery plays a critical role in environmental monitoring, urban planning, disaster assessment, and climate analysis. While multi-spectral sensors are increasingly available, true-color (RGB) imagery remains widely used due to the power, cost, and deployment constraints of many satellite and aerial platforms. However, existing land-cover segmentation datasets are often limited in geographic coverage, scale, or public accessibility. To bridge this gap, we introduce BELDE (Building a Large-scale Earth-observation Land-cover Dataset for Europe), a publicly available dataset tailored for RGB-based remote sensing semantic segmentation. Constructed from Sentinel-2 true-color images and ESA WorldCover data annotations, BELDE contains 1,088,385 curated image-segmentation map pairs spanning Europe with 7 land-cover classes at 10 m spatial resolution, making it one of the largest publicly available RGB land-cover segmentation datasets for Earth observation. To facilitate cross-region generalization studies, we additionally introduce BELDE-K (16,607 pairs) covering the Republic of Korea and BELDE-CA-NV (88,155 pairs) covering California and Nevada in the United States. We establish baseline results using multiple semantic segmentation architectures and evaluate both in-domain and cross-domain performance. Models trained on BELDE achieve an F1 score of 83.0% on the European test set, while performance decreases to 66.4% on BELDE-CA-NV and 58.3% on BELDE-K, highlighting the challenges posed by out-of-distribution geographic domain shift. By providing a continental-scale RGB segmentation and evaluation benchmark, BELDE supports the development of robust and transferable Earth observation models. The dataset and benchmark resources will be publicly released.

Molecular message-passing neural networks commonly propagate chemically diverse interactions through a single graph, which may mix interaction-specific signals and require deep propagation to capture long-range effects. We introduce the Multi-level, Multi-color Graph Neural Network (MMGNN), a hierarchical framework that decomposes a molecular graph into overlapping atom-type-pair-specific subgraphs while preserving atom-level resolution. MMGNN-2D constructs chemical-colored subgraphs from covalent connectivity, whereas MMGNN-3D constructs geometric-colored subgraphs from spatial proximity and augments their edges with distance, angular, and torsional descriptors. Both variants apply a shared communicative message-passing backbone to each subgraph and combine the resulting representations through atom-wise aggregation and molecular readout. We evaluated MMGNN on five classification and three regression benchmarks from MoleculeNet using common scaffold splits and five independent runs. MMGNN-2D achieved the highest macro-average AUC-ROC of 0.838 across the classification datasets and the lowest RMSE on ESOL (0.803). MMGNN-3D obtained the highest mean AUC-ROC on BBBP (0.956) and the lowest RMSE on FreeSolv (1.793), indicating complementary strengths of topological and geometric representations. Structural and leave-one-out analyses further illustrate how the subgraph decomposition affects learned representations and atom-type-pair sensitivities. These results support overlapping interaction-specific graph decomposition as a competitive strategy for molecular property prediction.

Robots operating in open-world environments must seamlessly integrate localization, spatial reasoning, navigation, and long-horizon planning. While specialist models excel at individual tasks, deploying a multi-model stack is computationally expensive and prone to cascading errors. We present Vesta, a unified embodied generalist that consolidates these capabilities into a single foundation model. Our approach combines a diverse and massive curated corpus designed to induce spatial grounding and a simple multimodal memory harness that enables reasoning over extended time horizons. Across diverse benchmarks, Vesta on average beats individual SOTA baselines by >$20\%$ and beats an ensemble of per-category-best baselines by $>10\%$ -- thus demonstrating that a generalist model can match or exceed specialists. On real-world robotic tasks requiring memory and reasoning, Vesta improves task success by >35\%. Our work thus demonstrates that a single generalist is a feasible, scalable, and arguably preferable alternative to combining specialists.

Long-form narratives are challenging for long-context models because their structure is implicit: events, characters, and plotlines interact across hundreds of pages without the explicit cues that guide navigation in structured documents. We address this by constructing storyline trees, hierarchical representations that organize narratives from global themes and major plotlines to fine-grained events. We first segment chapters into contiguous narrative segments, or scenes, and use them as the basic units for tree construction. We then infer storyline trees through complementary top-down and bottom-up procedures that derive, refine, cluster, and summarize storylines at multiple levels of abstraction. We showcase the utility of this representation for question answering: storyline trees enable adaptive retrieval, allowing models to iteratively inspect high-level narrative structure and retrieve scene-level evidence on demand. Experiments on three long-context narrative QA benchmarks show that adaptive retrieval outperforms strong baselines, including post-trained long-context models and agentic chunk-based methods. Ablations confirm that scenes are more effective basic units than chapters or generic segmentation, and that gains persist under matched retrieval budgets

Document-grounded assistants built on large language models are increasingly used in high-stakes, knowledge-intensive work. Their usefulness, however, may depend on how evidence is allocated before generation. We investigate such a claim by comparing two grounding architectures: (a) retrieval-augmented generation (RAG) that retrieves a few relevant passages, and (b) long-context prompting, which loads the whole document collection in context. We view these as two regimes of "epistemic access" on an accuracy--cost frontier. We use "epistemic accuracy" to capture model correctness that depends on having the right evidence. We posit that broader access (via long context) can increase it, but with a "token tax" (i.e., a substantial increase in cost due to larger input token consumption). We probe this framing with a case study in manufacturing safety training. Using an expert-validated benchmark, we evaluate 972 answers across three machines, two small language models, and three retrieval/in-context prompting approaches. Long-context prompting achieved the highest correctness (73.1% vs. 65.4% for semantic RAG), but at 26 times the per-query token cost. We interpret this gap as the token tax of broader evidentiary access. We carefully discuss the implications of our findings for resource-constrained organizations.

As academic submissions grow, the traditional peer review process struggles to keep up, raising concerns about quality and fairness. A trend of using large language models (LLMs) for assistance has emerged. In this work, we take a critical step toward improving the quality of LLM-generated reviews. We propose the PeerCheck framework, which investigates LLM-human review differences (RQ1) and explores methods to improve LLM-generated review quality (RQ2). We first analyzed the human-written reviews with reviews generated by various LLMs and found that LLMs and humans focus on different terms, e.g., LLMs prioritize theory while humans emphasize methodology and experiments. We further adopt prompt engineering, such as Chain-of-Thought (CoT), and utilize retrieval-augmented generation (RAG) to enhance the LLM-generated reviews towards human-level quality. We find CoT significantly improves the quality of LLM reviews, while we discover an unexpected "RAG paradox," i.e., experiments with RAG produce different results for various LLMs and, in some cases, even reduce review quality. Our comprehensive analysis of LLM-generated academic reviews illustrates both possibilities and limitations, contributing to a more effective, human-aligned review system. Our dataset is available on https://github.com/TrustAIRLab/PeerCheck.

Large Language Models (LLMs) offer a promising path to automate Clinical Trial Matching (CTM), but still struggle with the deterministic verification required for complex eligibility criteria. Conversely, purely symbolic methods provide formal rigour but break down when faced with incomplete patient records and noisy clinical evidence. To bridge this gap, we investigate a hybrid framework for CTM combining LLMs with logical verification. In particular, we introduce an abductive neurosymbolic CTM framework (αNeSy-CTM), which leverages the linguistic and world knowledge in LLMs to support reasoning over noisy and underspecified clinical text. Extensive evaluation demonstrates that αNeSy-CTM substantially outperforms standalone LLM baselines, achieving up to 30% relative improvement over zero-shot baselines. In addition, our analyses confirm the impact of abductive reasoning on CTM, with αNeSy-CTM exhibiting improved accuracy, specificity, and robustness over a non-abductive neurosymbolic setting. Furthermore, αNeSy-CTM and Chain-of-Thought (CoT) reasoning prove highly complementary, highlighting the potential for a hybrid routing policy. Ultimately, this paper demonstrates the impact of neurosymbolic methods for automating CTM, providing a path toward the next generation of auditable, LLM-driven clinical applications.

Deep learning-based audio classification systems, including automatic speaker verification, are vulnerable to adversarial attacks. Realistic real-time threat assessment remains difficult because optimization-based methods, such as projected gradient descent (PGD) and Carlini-Wagner, require costly iterative updates in the high-dimensional waveform domain. Generative attacks allow single-shot synthesis but often introduce perceptible artifacts or depend on computationally intensive architectures, while diffusion and autoregressive approaches incur high inference latency. To address this gap, we propose a generative attack framework operating in the continuous latent space of a neural audio codec. A conditional generator synthesizes class-specific perturbations in a single forward pass and decodes them into adversarial waveforms. Our method achieves targeted attack success rates up to 99% with sub-7 ms inference, outperforming generative baselines while reducing latency by 24x.

Filmmaking demands precise motion control and reference image compositing -- capabilities that existing methods treat separately. Point-track-conditioned image-to-video models restrict content insertion to the first frame, while reference-to-video models lack fine-grained spatial-temporal control over how reference content integrates across frames. We present Go-with-the-Track, which unifies both capabilities by jointly conditioning on multiple reference images and reference-anchored point-tracks -- extending conventional point-tracks to explicitly establish correspondences between generated frames and reference images, thus enabling precise compositing and motion control throughout the video. To achieve this, we introduce spatially-aware point-track embeddings that encode the full sequence of point-track coordinates using a coordinate-wise MLP followed by temporal pooling. This representation captures the spatial characteristics of each point-track (serving as a unique identifier), while the embedding similarity correlates directly with spatial proximity, enhancing the model's ability to distinguish and associate point-tracks. We inject these point-track embeddings into a video diffusion transformer via a lightweight adapter, resolving the pixel-to-patch resolution mismatch while avoiding the substantial motion detail loss inherent in naive point-track subsampling. We use a hybrid training strategy to train jointly on dynamic, static, and synthetic scene video datasets to boost motion controllability. Experiments demonstrate that Go-with-the-Track achieves superior motion and reference control in a single model and enables new capabilities: multi-reference conditioned video generation with point-track driven compositing, as well as camera control for both static and dynamic scenes. Project Page: https://eyeline-labs.github.io/Go-with-the-Track/

Meeting archives are difficult to search when users remember what was discussed but not when. We study topic-to-timestamp alignment: given a natural-language topic and a timestamped meeting transcript, the goal is to return the time at which the topic is discussed. A standard RAG setup can retrieve relevant transcript excerpts, but still asks the language model to generate a timestamp, which can produce unsupported or invalid timecodes. We therefore recast timestamp prediction as constrained temporal candidate selection: the system retrieves timestamped transcript chunks, and the model selects the candidate that best grounds the topic instead of generating a timecode. On 420 topic-timestamp queries from 200 municipal meeting transcripts, this increases Recall@5 from 31.9% to 50.0%, reduces MAE from 837.0 seconds to 761.0 seconds with Mistral-7B-Instruct, and increases the number of parseable outputs from 373 to 419 of 420 queries. The results suggest that temporal grounding in long transcripts depends strongly on retrieval quality and output design, not only on the choice of the language model.

Estimating causal effects in industrial time series requires handling temporal dynamics, time-varying treatments, and unobserved confounders. Existing causal foundation models (CausalPFN, CausalFM) operate only on static cross-sectional data; neural temporal methods (CRN, G-Net) require per-dataset training; and concurrent temporal-PFN proposals have not been demonstrated at industrial scale. None output explicit per-pair reliability signals alongside their CATE estimates. We introduce Temporal Causal Prior-Data Fitted Networks (TCPFN), a foundation model for zero-shot temporal causal discovery with learned reliability signals. TCPFN makes four contributions: (1) a Causal Judgment Head that jointly predicts null-effect probability, confounding strength, identifiability, mediation fraction, and causal regime; (2) a mixed training prior covering six causal regimes (independent, direct, confounded, mediated, time-varying confounded, feedback) plus CausalFM-style front-door and instrumental-variable priors; (3) a discrete-token panel-data architecture with cross-attention masking that prevents inter-horizon leakage; (4) zero-shot inference at industrial scale via FAISS-based context selection and one-step posterior correction. On 19 benchmark datasets across five domains, TCPFN achieves competitive zero-shot causal discovery: AUROC 0.96 on Tennessee Eastman, 0.93 on SWaT, 0.98 on Causal Rivers, 0.97 on CAUSRCA. The null detector reaches NullF1 0.94, AUROC 0.99. TCPFN scales to V=1,275 on a proprietary Kraft pulp-and-paper dataset in 6 hours on a single GPU; PCMCI, a CPU-only library, on a V=666 sub-panel of the same data took 81.5 hours, extrapolating by O(V^2) to ~12.5 days at V=1,275. TCPFN's top edges identify cross-subsystem causal relationships while PCMCI's surface within-instrument controller-measurement coupling -- a scalability case study.

Software engineering has long inferred where a system's knowledge resides from who authored its code. The truck factor, the Degree-of-Authorship metric, and the degree-of-knowledge model all rest on one inference -- that authoring a region of code is evidence of understanding it -- and for most of software's history it was a workable proxy, because code entered a repository only when a human wrote it, which forced at least transient understanding. This paper argues that AI code generation severs that inference at its root, and that the consequence is not the degradation of the authorship-based metrics but their invalidation as a class. When an agent generates a module and a human merges it, the version-control record still attributes authorship, but the attribution no longer licenses any conclusion about comprehension: the same footprint is now compatible with full, partial, or no understanding. The metric still returns a number; the number measures a substrate that has come uncoupled from the quantity it was used to estimate. The collapse is corroborated by the field's own measurement failures, and the methodological corollary is load-bearing: the instrument the comprehension-debt era needs cannot be built by refining the knowledge-concentration metrics, because no function of an authorship footprint recovers an inference the footprint no longer supports. The replacement must be grounded in evidence of comprehension rather than authorship. I state a falsifiable prediction that discriminates the two -- that systems with a healthy authorship-derived truck factor but low comprehension-measured retention will suffer incident-resolution failures the authorship metric does not predict -- and argue that building the comprehension-grounded instrument at the scale of a system and a team is the field's open measurement problem, left open here.

Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in large language model reasoning, but relies on ground-truth supervision that is costly or infeasible, especially in coding tasks. Recent work addresses this by deriving rewards from a model's own signals, such as majority voting or confidence-based scores, achieving notable success on mathematical reasoning benchmarks. However, code generation poses distinct challenges: programs are structurally complex, semantically equivalent solutions may differ syntactically, and verification typically requires execution. Whether these intrinsic reward methods transfer effectively to code remains unexplored. In this work, we present a systematic empirical study of intrinsic reward methods for code generation. We conduct extensive experiments on LiveCodeBench, systematically evaluating representative certainty-based Reinforcement Learning from Internal Feedback (RLIF) approaches under different training scenarios and hyperparameter settings. Our experiments reveal that certainty-based methods yield early gains but inevitably collapse: models progressively shorten outputs and lose reasoning capability, with collapse speed sensitive to sample size and temperature. When used to initialize RLVR training, RLIF pre-training offers no significant improvement over training from scratch. We also provide actionable recommendations for using intrinsic rewards for training code reasoning models. Our study shows both the promise and limitations of intrinsic reward methods for code, informing future work on code models and agents.

Decision-making under partial or adversarial observability requires accurate inference of the environment's latent state and its associated uncertainty. This work analyses adversarial attacks on linear probabilistic state-space models, commonly integrated within reinforcement learning architectures, where the attacker alters observations under likelihood constraints that ensure the perturbations remains consistent. We analyze how such adversarial yet realistic observation shifts influence the latent state and influence policy decisions. This perspective provides a principled pathway toward building more robust reinforcement learning systems, with direct relevance to safety-critical domains such as robotics, where reliable operation under sensor noise, partial failures, and adversarial conditions is essential.

Synthetic tabular data enables microdata sharing in regulated domains, yet deploying continuous-time generative models requires balancing analytical utility, disclosure risk, and computational cost. Latent-space flow models are flexible, but theoretical equivalences across learning targets, probability paths, and sampling dynamics can translate into different behaviour under finite-step integration and explicit compute budgets. We present an empirical study of tabular latent flow models across seven datasets, evaluating velocity, score, noise, and posterior matching objectives under optimal transport (OT) and variance-preserving (VP) paths, ODE and SDE sampling, and varying integration budgets. Our contributions are threefold: (1) we show that the learning target largely determines the utility-risk operating regime, with velocity and posterior matching tending to yield higher utility, while score and noise matching tend to achieve lower disclosure risk; (2) we demonstrate that configuration and sampling choices shift performance, with midpoint often improving distributional fidelity and OT paths often tolerating earlier stopping than VP, enabling compute savings under fixed budgets or risk thresholds; and (3) we distil these findings into actionable defaults and practical configuration guidance to support pre-release model selection under disclosure risk and resource constraints. The code implementation and supplementary materials can be accessed in https://github.com/rulnasution/tabular-latent-flow/.

Collective artificial intelligence, where multiple agents work on shared tasks, holds potential to solve expansive problems in fields from medicine to collective governance. But while prescriptive engineering solutions abound, we lack descriptive scientific understanding of artificial collectives, and therefore principles for how to design resource efficient multi-agent systems. Through systematic experiments with optimizing agents, we characterize how agent interpretive abilities, rationality bounds, and task qualities interact to shape collective performance. Agents range from specialists, with narrow interpretive abilities, to generalists, with broad ones. Collectives of specialists correspond to sparse, centralized networks, while collectives of generalists correspond to dense, decentralized ones. We show that interpretive network properties have small performance effects on average (0.07 standard deviations of performance). However, for specific task qualities, these effects are 4.5 times larger (0.33 sd) and can reach much higher for certain task qualities (1.84 sd). This leads collectives of generalists to perform better on tasks that involve generating, choosing, and coordinating, while collectives of specialists with a few generalist mediators perform better on tasks that involve negotiating. Rationality bounds then moderate these relationships. At loose bounds, specialists outperform generalists through more effective sampling of high-dimensional decision spaces. At tight bounds, generalists outperform specialists through better gradient estimation. A fundamental trade-off between performance and convergence speed emerges at moderate bounds. These findings suggest that multi-agent design could benefit from matching interpretive networks to both task demands and agents' computational limits, with implications for the efficiency and energy costs of multi-agent systems.

Sustaining open-source software (OSS) requires effective practices for evolution and change management. In OSS projects, evolution is largely driven by feature requests and enhancements proposed by diverse stakeholders. These requests are often discussed across multiple communication channels, particularly community forums and issue trackers, where stake-holders negotiate intent, clarify requirements, and coordinate development. Despite prior research on OSS forums and issue trackers, we lack an empirical understanding of who creates and maintains links between forum posts and tracker issues, and how these links support clarification, feedback, and coordination throughout feature request lifecycles. To address these questions, we conduct an in-depth case study of Moodle, a widely used open-source learning management system. Our study combines (1) an empirical analysis of cross-channel trace links between Moodle's community forum and its Jira issue tracker, (2) semi-structured interviews with developers, and (3) semi-structured interviews with forum participants. Our results show that cross-channel traceability is rare: only 818 of 23,169 (~3.5%) feature request issues in Moodle's Jira link back to a community forum, and authorship differs by channel, with developers authoring 52.8% of tracker issues, while forum feature requests are predominantly authored by users, and only 230 linked pairs share the same author. The qualitative findings further reveal that the transition from forum posts to issues is largely ad hoc, with limited tool support and unclear role ownership, and that users often experience the process as opaque or weakly responsive.

Cryopathy syndromes are difficult to classify because laboratory patterns often overlap across diagnostic categories, while some diagnoses are rare. This makes routine interpretation of cryoglobulin-related tests challenging and increases dependence on expert judgment. The aim of this study was to develop and compare machine learning approaches for automated classification of cryopathy syndromes from laboratory data and to identify a practical strategy for clinical decision support. Methods: We analysed laboratory records from 2,686 patients assigned to 14 diagnostic categories. The dataset included demographic variables, cryoglobulin measurements, precipitation tests, and hemagglutinin and hemolysin titers. Data preprocessing included cleaning, encoding, imputation, normalization, and construction of clinically informed interaction features. We evaluated 12 modelling strategies, including Random Forest, Gradient Boosted Trees, Multi-Layer Perceptron, soft-voting ensembles, class balancing with Synthetic Minority Over-sampling Technique, hierarchical classification, period-aware models, targeted binary classifiers, and probability calibration. Performance was assessed using stratified train-test evaluation and stratified 5-fold cross-validation. The main metrics were macro-averaged F1 score, accuracy, Top-3 accuracy, and expected calibration error. The overall task proved difficult because of marked class imbalance and clinical overlap between diagnoses. The best multiclass performance was achieved by a soft-voting ensemble of Random Forest and Gradient Boosted Trees. Cross-validation confirmed stable performance for the balanced Random Forest model. Tree-based methods consistently outperformed the neural network model. Feature engineering improved discrimination, and the most informative predictors were derived cryoglobulin-based interaction features.

Scientific discovery increasingly relies on automated systems that generate hypotheses, inspect multimodal evidence, and validate claims at scale. Yet scientific claim verification is not well served by asking a vision-language model for a direct binary judgment: claims often combine numerical results, comparisons, scope qualifiers, and explanatory context, while evidence is encoded in tables and figures with distinct grounding structures. We present SciLens, an evidence-conditioned atomic entailment framework for multimodal scientific claim verification. SciLens decomposes each claim into central empirical atoms and background atoms, grounds the central atoms to modality-specific evidence witnesses, and predicts the final label with an atom-level entailment rule. For tables, atoms are grounded to rows, columns, cells, arithmetic relations, and table scope; for figures, they are grounded through panels, axes, legends, visual encodings, categories, trends, ranks, and qualifier checks. This yields a unified validation procedure in which a claim is supported only if every central empirical atom is entailed by the current evidence. On the SciClaimEval development set, SciLens achieves 79.2% macro-F1 and 63.1% pair accuracy, showing that structured agentic validation improves both evidence sensitivity and interpretability.

We present a holistic methodology for artificial intelligence algorithm and accelerator co-design, co-search, and co-generation (A3C3), which jointly optimizes neural network architectures and their hardware implementations to address the inefficiencies of traditional top-down AI system design flows. Conventional AI deployment often treats model design and hardware mapping as separate stages: an algorithm is first developed for accuracy, and only afterward adapted to meet latency, throughput, energy, or resource constraints. This separation can lead to suboptimal systems, particularly as modern AI workloads become increasingly heterogeneous, memory-intensive, and platform-dependent. A3C3 instead parameterizes both algorithmic and accelerator design spaces and searches them jointly, enabling the automatic generation of model-accelerator pairs that better balance accuracy, latency, throughput, energy efficiency, and hardware utilization. This article is a book chapter of the Handbook of Embedded Machine Learning, edited by Sudeep Pasricha and Muhammad Shafique, Springer Nature.

When a new domain resembling a popular brand appears, defenders face a fundamental ambiguity: it may be an attacker-created squatting site for phishing, or it may be a domain the brand itself registered, either defensively, to block attackers, or legitimately, for a new product or service launch. Incorrectly flagging a brand-owned domain as malicious produces a false positive that harms end users and damages the brand's reputation. Resolving this ambiguity requires brand intelligence: the ability to determine, at scale, whether a given domain belongs to a brand. Large language models (LLMs), with their broad knowledge of brand domain relationships, offer a promising zero configuration approach to this problem, but their reliability for brand intelligence tasks remains unknown. We present the first systematic empirical evaluation of LLM brand intelligence across three tasks: domain enumeration (Q1), open ended brand attribution (Q2), and binary ownership classification (Q3). We evaluate four models, Gemini 2.5 Flash, Gemini 3.5 Flash, Claude Sonnet 4.5, and Claude Sonnet 4.6, across four retrieval settings (in context, web search, WHOIS lookup, and combined) on 36 of the most phished brands. Our results reveal a stark dichotomy: models achieve up to 82% precision enumerating brand domains from memory alone, yet fail at ownership verification without external tools, with macro F1 at most 0.37 in ICL mode. WHOIS augmentation lifts Q3 macro F1 by up to 0.65 points, yielding near perfect precision (<= 0.99), dramatically reducing the false positive risk for defenders. We provide concrete recommendations for deploying LLMs in brand protection pipelines.

Vision-Language-Action (VLA) models enable general-purpose robotic control via large-scale multimodal pretraining, yet their effectiveness under few-shot imitation learning remains limited. We conduct a systematic stress test of state-of-the-art VLA models and show that performance degrades sharply as demonstrations are reduced, revealing a key weakness of existing adaptation strategies. To address this, we introduce FOCA, a future-oriented conditioning framework for data-efficient VLA adaptation. FOCA combines explicit prediction of task-grounded future interaction embeddings with implicit alignment to future goal observations, enabling long-horizon reasoning in latent space without pixel-level prediction. This formulation naturally supports action-free co-training with synthetic videos from video world models and can be interpreted as learning a future-conditioned value-like representation. Extensive experiments demonstrate FOCA achieves 95.7% success with 20 demonstrations on LIBERO, improves 7-12% on RoboCasa, and delivers up to 26% absolute gains on real robots, establishing a new state of the art in few-shot VLA adaptation.

Developing robust intrusion detection systems (IDS) for IoT environments requires large, labeled datasets capturing realistic traffic distributions across both benign and malicious activity. Existing public datasets suffer from fixed activity distributions and extreme class imbalance, while deep generative models (GANs, VAEs) provide no mechanism to enforce that synthetic packets remain within physically valid feature ranges. This paper proposes and compares two constraint-enforcing approaches for synthetic IoT network packet generation: (i) a statistical learning method combining PCA-based latent space sampling with dual One-Class SVM (OCSVM) and Isolation Forest (IF) boundary enforcement, and (ii) a genetic algorithm (GA) method that treats packet generation as a multi-objective optimization problem with explicit fitness criteria for anomaly model acceptance and distributional fidelity. Both methods embed hard validity constraints -- dual anomaly-detection gating, feature-range clamping, and independent validation -- directly into the synthesis pipeline. Evaluation on the complete ACI IoT 2023 dataset (1,231,411 packets, 12 attack categories, class imbalance up to 175,805:1) demonstrates that both methods achieve PASS status across all categories under independently trained validators with a 30% anomaly rate threshold: the statistical method attains 1.20% average anomaly rate with ~1,091 packets/s throughput, while the GA attains 0.62% average anomaly rate with organic per-class variance (0.00%-2.50%) at ~5.7 packets/s. Both methods successfully amplify the 5-sample ARP Spoofing category by 200x to 1,000 validated packets. The ~190:1 throughput ratio between methods, combined with their complementary quality profiles, provides evidence-based selection criteria for deployment contexts ranging from rapid dataset augmentation to adversarial robustness testing.

In many applications, the precision by the available hardware arithmetic is insufficient to guarantee accurate results. Multiword arithmetic is a special type of multiprecision arithmetic where a multiple double is an unevaluated sum of 64-bit doubles, or where a multiple integer is an unevaluated sum of 64-bit integers. Parallel computing is applied to compensate for the cost overhead of multiword arithmetic. This type of arithmetic exploits naturally the optimized hardware, allows for efficient type conversions, memory layouts, all favorable for parallel computing. For example, storing a multiword in registers rather than arrays is beneficial to parallel computing by tasking and acceleration by graphics processing units. Code for multiword arithmetic is available in the software PHCpack, written mainly in Ada, publicly available at github, and as an Alire crate, released under the GNU GPL v3.0 license.

We present a family of conformal test martingales based on shifted Legendre polynomials, which extends the Simple Jumper martingale. The Simple Legendre Jumper substitutes the linear betting function with a polynomial of arbitrary degree, thereby facilitating the detection of variance, skewness, and higher-order deviations from uniformity; the standard Simple Jumper is a specific instance of degree one. The Product Legendre Jumper integrates multiple polynomial degrees into a unified betting function, although its state space expands exponentially-a cost we refer to as the jumping tax. To address this issue, we introduce the Variational Legendre Jumper, which factorises the joint adaptation through a mean-field approximation, thereby reducing exponential scaling to linear time with minimal loss in power. Lastly, the Composite Legendre Jumper incorporates several jumping rates, ensuring a wealth floor under exchangeability and automatic adaptation to the shift's timescale. Empirical results from a real-world classification task demonstrate that the combined methods consistently surpass any single-degree martingale under distributional shift, and the composite variant is recommended as the default when the shift timescale is unknown.

The temporal structure of reward composition in reinforcement learning (RL) is typically hand-designed and held fixed throughout training, leaving the progression of motivational priorities largely unexplored. In this work, we propose an evolutionary framework for discovering developmental reward schedules, in which three distinct biologically inspired motivational components -- agency, novelty, and reactivity -- are combined through time-varying weights that dynamically shift over the course of training. Evaluated on two sparse-reward MiniGrid tasks: DoorKey-6x6 and KeyCorridorS3R1, our framework compares the generalizability of four evolutionary algorithms: CMA-ES, xNES, DE, and L-SHADE against an extrinsically motivated baseline (our main comparison point), and three additional hand-designed methods. On DoorKey-6x6, all evolved methods outperform the non-evolved baselines, with L-SHADE achieving the best performance -- an approximate relative mean improvement of 11.4% over the extrinsic only baseline. On KeyCorridorS3R1, CMA-ES achieves the best overall performance, with the remaining evolved methods showing weaker and less reliable generalization capability compared to the extrinsic only baseline. Interestingly, the discovered schedules diverge from our defined developmental ordering, with novelty consistently emerging as the dominant early signal during training, across both tasks. Collectively, our results position evolutionary optimization as a promising approach for developmental reward schedule discovery in deep reinforcement learning, and suggest that what evolution finds to be optimal in computational settings may differ from what it finds to be optimal in biology. The code for this project can be found at: https://github.com/alannadels/Evolutionary_RL.git.

Vision-Language-Action (VLA) models enable robots to execute manipulation tasks from natural-language instructions grounded in visual observations. However, existing VLA interfaces primarily rely on speech or text input, limiting accessibility for deaf, hard-of-hearing, and speech-impaired users. We present SignVLA, a real-time sign-language-guided VLA framework for accessible human-robot interaction. The system introduces a modular sign-to-text interface that converts visual sign gestures into semantic instructions compatible with downstream VLA policies. Given video streams, SignVLA extracts hand landmark features and employs an attention-enhanced Long Short-Term Memory (LSTM) network to capture temporal gesture dynamics for alphabet- and command-level sign recognition. A temporal stabilization module further improves prediction consistency in real-time interaction settings.The generated instruction sequence is then passed to a downstream VLA policy for sign-conditioned robotic manipulation. Experimental results demonstrate stable real-time sign recognition and successful execution of manipulation tasks driven by sign-language inputs. Our findings suggest that lightweight temporal sign recognition can serve as an effective and practical accessibility layer for multimodal embodied intelligence.

Multi-view surface reconstruction is a core problem in computer vision. One prominent line of work represents the surface implicitly as a signed distance field (SDF), optimizing it based on the photometric loss between rendered and observed pixel colors. These approaches typically employ SDF-based volume rendering to obtain a differentiable relaxation of discontinuous visibility along rays, thereby reducing reliance on silhouette supervision. In this paper, we reformulate SDF-based volume rendering as probabilistic surface rendering, where each pixel color is modeled as a mixture distribution induced by the random first ray-surface intersection. To this end, we introduce Stochastic Signed Distance Processes (SSDP), which model the SDF along each ray as a stochastic process, inducing a first-passage-time distribution for each ray. We then derive the first-passage probability for each sampling interval based on Bayesian filtering, together with its practical approximation for parallel rendering. We further show that NeuS, an existing SDF-based volume rendering method, arises as a special case of our formulation. Experiments on the DTU and MobileBrick datasets demonstrate that our method outperforms baselines in both surface reconstruction and uncertainty quantification, supporting the effectiveness of our first-passage formulation. Our code is available at https://github.com/skmhrk1209/SSDP.

Inference-time engineering can alter model behavior without fine-tuning. However, its utility for improving diagnostic performance in medical vision-language models (VLMs) remains unclear. We aim to evaluate whether Contrastive Activation Addition (CAA) can improve pneumonia classification in chest radiograph VLMs without updating model weights. Three frozen chest radiograph VLMs (MedGemma-4B-IT, NV-Reason-CXR-3B, and CheXOne-3B) were evaluated on the public Kermany pneumonia test set. Classification was based on the logits of the tokens Yes and No under a binary prompt. Steering vectors included a 30-pair answer-bias control, a 30-pair pneumonia text contrast, and an image-conditioned contrast derived from 30 pneumonia and 30 normal development images. A deterministic 200-image development set was used for layer and scale selection (100 images) and threshold calibration (100 images). Performance was assessed using ROC-AUC, PR-AUC, F1 score, threshold analyses, reverse-vector controls, random-vector controls, and conditional bootstrap confidence intervals. Fixed-threshold F1 improvements were frequently observed but did not consistently indicate improved diagnostic performance. For MedGemma-4B-IT. NV-Reason-CXR-3B showed the strongest benefit: calibrated F1 improved from 0.7692 in the zero-shot setting to 0.8619 with pneumonia-text steering and to 0.8727 with image-conditioned steering. For CheXOne-3B, pneumonia-text steering increased calibrated F1 from 0.8528 to 0.8666, although the confidence interval crossed zero. On this public pneumonia benchmark, CAA substantially altered prediction score distributions and operating characteristics without fine-tuning. Meaningful performance gains were observed in one of three evaluated VLMs, suggesting that activation steering may serve as a lightweight approach for adapting medical VLM behavior.

LLM agents can write code and call tools, but reliable bioinformatics work requires long-horizon interaction with workflow software, typed data objects, provenance, and biological checks. We study this setting through Galaxy workflow execution. The agent must explore task data, construct or adapt an executable workflow DAG, bind inputs and dataset collections, monitor execution, debug failures, and validate biological outputs. We propose Process-Reward Tactic Evolution, a Galaxy-based training framework that turns verified workflow rollouts into reusable \tactics. During training, agents practice on curriculum-organized Galaxy tasks in Agent Gym; process verifiers score workflow construction, software interaction, execution, and biological correctness; successful and failed traces are distilled into a tactic library. At inference, the trained executor, Process-Reward Tactic Evolution, uses this library to execute held-out peer reviewed Galaxy workflow converted BioWorkflow Bench and BioAgent Bench tasks in isolated environments. The paper evaluates whether process-supervised tactic accumulation improves long-horizon bioinformatics workflow completion, biological correctness, and execution efficiency over no-memory and reflection-style baselines.

Belief revision is a process in which an agent begins to believe in something she previously did not. I begin the paper by presenting, based on (Gärdenfors, 1998; Hansson, 1999), postulates for belief revision that constitute the basis of the AGM theory. I will then briefly show the semantics of a modal logic introduced in (van Ditmarsch, 2005), which I call `$P$'. This logic formalizes static epistemic states and has greater expressive power than AGM in doing so because it captures the quantitative notion of "degrees of conviction". The third step is to introduce revision operators on $P$ and, mostly following (van Ditmarsch, 2005), obtain the Dynamic Epistemic Logic (DEL) I call `$P*$'. It models processes of belief revision in several ways. Original results are presented in the following two sections. The first one of these sections revolves around a formalization of AGM postulates within $P*$ by proving some theorems related to the satisfaction of those postulates by revisions defined in $P*$. The last section features an analysis of $P*$'s revisions that go beyond the mere satisfaction of postulates. I compare their formal behavior with respect to some philosophical criteria. At last, I conclude that the functions presented in (van Ditmarsch, 2005) are not good formalizations of the philosophical intuition behind AGM. Instead, it is captured by the function $*^0$ originally defined in this paper (but highly inspired by (van Benthem, 2007)). An implementation of this function is also provided.

Watermarking has become a crucial technique for ensuring provenance and accountability in AI-generated source code. As large language models (LLMs) are increasingly integrated into development workflows, reliable attribution remains challenging. In practice, most developers rely on commercial LLM APIs operating under black-box constraints, making existing approaches that require access to the decoding process less feasible for real-world integration. To address this limitation, we propose PromptMark, a black-box, prompt-guided watermarking framework that embeds invisible yet statistically detectable signals into generated code via structured input instructions. The method steers models toward subtle identifier and comment naming patterns while preserving the functional correctness and structural integrity of the generated code. Detection is performed using statistical tests designed to remain reliable across varying code lengths and model outputs. The embedding is further refined through an iterative feedback loop, where prompts are updated based on watermark detection scores. Experiments on the MBPP and HumanEval benchmarks show that PromptMark consistently achieves strong watermark detectability while maintaining high code correctness, outperforming baseline approaches.

Astrophysical simulations frequently address multi-scale, multi-physics problems through subsystem decomposition, problem-tailored integration schemes, and coupling on fixed manually set timescales. Here we introduce ReLaTS, a reinforcement learning framework that dynamically selects the coupling time step to optimize the trade-off between accuracy and computational cost. We validate ReLaTS on star clusters containing a planetary system, and test the method by varying the number of stars $N_\star$ in the cluster and the number of planets ($N_{\rm planet}$) orbiting one of them. The method finds the optimal coupling time step that balances speed and accuracy without requiring expert knowledge. In addition, the trained network operates independently of the coupled \textit{N}-body algorithms, displaying stable performance across a range of setups. We observe that the method is less reliable for cases with infinitesimal masses, as their contribution to the total energy is negligible compared to that of the massive bodies, and the network is not capable of recognizing potential errors generated while integrating them. For long-time integration of large $N$ systems, the error accumulates. The reinforcement learning algorithm, however, manages to keep the energy error below a pre-set threshold. This approach substantially reduces energy errors relative to fixed-time step baselines without substantial additional computational overhead. Once trained, ReLaTS requires no expert tuning and generalizes across diverse astrophysical domains, enabling adaptive multi-scale simulations.

Can we trust evaluation scores to capture an LLM's true real-world performance? Certifiable evaluation answers this question by providing guarantee for LLM evaluation. In particular, existing methods sequentially curate evaluation samples and keep updating confidence intervals (CIs) that cover the true performance with high probability (e.g., 95%) until some conditions are satisfied, e.g., the CI width reaches a target precision. However, existing methods are not generally anytime-valid: the claimed coverage (e.g., 95%) may fail when CIs are repeatedly updated and used to decide when to stop, leaving a gap between theoretical rigor and practice. This paper bridges this gap by proposing Celeus, a Certifiable framework for Efficient LLM evaluation, which leverages E-processes to build anytime-valid CIs. Concretely, we propose signals that combine two ingredients: (i) Uncertainty-guided sampling to select informative samples for evaluation, and (ii) Surrogate-assisted approximations for unevaluated samples. We prove that such signals remain unbiased for the evaluation score conditional on the past, enabling statistically-grounded and anytime-valid $e$-process CIs. More importantly, the two ingredients reduce estimation variance and help reach the target precision with fewer evaluated samples. We also prove that CIs obtained by Celeus can shrink at a near-parametric rate up to logarithmic factors and analyze the oracle variance-optimal sampling rule that motivates the empirical uncertainty-guided one. Experiments show that Celeus reaches the target precision using 54-62% fewer evaluated samples than baselines, while preserving anytime-valid coverage.

NeuroEvolution of Augmenting Topologies (NEAT) is a widely used neuroevolution algorithm for learning neural network architectures and weights for control tasks. However, standard offline optimisation searches for connection strengths directly, which can scale poorly in high-dimensional weight spaces and more difficult continuous control problems. Hybrid methods that combine neuroevolution with online learning can address this challenge, but their theoretical properties remain underexplored. This paper gives the first regret analysis for a general NeuroEvolutionary Online Learning (NEOL) framework, which decouples learning into two timescales: an outer loop for architecture search and an inner loop for online weight adaptation via rewardmodulated plasticity. Under mild conditions, we prove that NEOL achieves sublinear regret. Empirically, under fixed interaction budgets on four standard control benchmarks, a NEAT-based NEOL implementation achieves higher final fitness and lower variance than pure NEAT, and is competitive with strong reinforcement learning (RL) baselines on several tasks. The results are supported byWilcoxon rank-sum tests and ablation studies. Overall, the findings show that online plasticity can improve the sample efficiency and robustness of two-timescale neuroevolution. Code is available at https://github.com/boobaa2001/NeuroEvolution Online Learning NEOL.

Emergent misalignment (EM) is a phenomenon in which models generalize with narrow fine-tuning, leading to broad (yet uneven) misalignment across evaluation questions. We study EM and its variability directly through the components of fine-tuning: training dynamics, model priors, and data. (1) We first explored how in-domain training loss relates to out-of-domain alignment scores across datasets and model families. Then, we tried to induce potential alternative local minima through different learning schedules for one narrow fine-tuning, but did not find strong runs with better broad alignment scores conditioned on similar or lower training loss. (2) We found that although the mean and standard deviations of the misaligned model scores are usually statistically different from those of the pre-trained model, there are some potential signals on overall positive correlation. The evaluation prompt-only activations from both the pre-trained and the original instruct models (prior to narrow fine-tuning) could predict fine-grained alignment scores after narrow fine-tuning. (3) Finally, we compared activation deltas before and after narrow fine-tuning and found moderate-to-high subspace overlap and similarity between the resulting activation shifts for training and evaluation prompts. Subspace overlaps between training and evaluation prompt activations correlate with their shifts' similarities when measuring with the last prompt-token activations. The train-evaluation data prompt overlap is controlled against overlap computed from random vectors and evaluation prompts activations.

EEG foundation models can learn generalizable representations from large-scale EEG corpora to enable single-backbone transfer across diverse clinical and brain-computer interface tasks. Existing models typically discretize the continuous multi-channel EEG waveform into patches or codebook tokens and train a transformer with masked self-supervision. Recognizing that this discretization fragments continuous brain rhythms and obscures fine-grained temporal dynamics, we present B[FM]$^2$(Brain Foundation Model via Flow Matching), whose inductive bias aligns with the data by pretraining directly on the raw signal using continuous-time flow matching without patches, tokenization, or masking. However, multi-channel EEG signals pose an architectural challenge for flow matching: time is densely sampled and highly autocorrelated (thousands of timepoints), while the electrode axis is short (tens of channels) at distinct scalp positions. To address this time-electrode asymmetry, we introduce SplitUNet, a velocity network that factorizes each block into separate 1D temporal and 1D electrode convolutions and downsamples only along time, preserving electrode topology throughout the hierarchy. B[FM]$^2$ sets a new state of the art on 7 of 9 standard downstream EEG classification tasks, using a pretraining budget of only 36,895 segments ($\approx$ 307h), 1-2 orders of magnitude ($\approx$ 30x) less than required by existing EEG foundation models. Further, it generates synthetic EEGs that two board-certified neurologists cannot distinguish from brain data (Cohen's $κ=$ -0.096). https://jd730.github.io/projects/BFM2

We study the problem of system identification in heterogeneous settings, where different systems may follow distinct underlying dynamics. Existing clustered system identification approaches often rely on iterative training-based cluster assignment, which can be sensitive to learning uncertainty and model initialization. In contrast, we propose a one-shot, training-free clustering method that identifies similar systems using the structure of their locally observed data. Specifically, each system estimates a local state covariance matrix, and cluster identities are inferred by measuring the alignment between the leading covariance eigenspaces of different systems. We provide a mathematical interpretation of the proposed similarity score and develop a finite-sample analysis that characterizes how covariance estimation error induces eigenspace perturbations in terms of the underlying system dynamics. We then derive a probability bound for pairwise false merges and a global clustering success guarantee. Numerical experiments demonstrate that the proposed eigenspace-based clustering method effectively identifies systems with shared dynamics, leading to lower personalized model-estimation error compared with training-based clustering and non-clustered baselines.

Generating visually consistent multi-shot videos remains an open challenge. As videos span more shots, inconsistencies can accumulate across shots, causing entities that reappear across shots -- characters, objects, and locations -- to drift away from how they first appear. We observe that viewers judge consistency by comparing each later appearance of an entity with its first clear appearance; the visual quality of this initial appearance sets the consistency ceiling for all that follows. Motivated by this, we present \textbf{GroundShot}, a training-free, model-agnostic agentic framework for entity-grounded multi-shot generation. GroundShot builds an entity-level visual memory online from accepted generated shots: it schedules shots' generation order by their expected usefulness as entity references, grounds entities from generated videos, verifies their reliability before adding them to memory, and retrieves suitable entity references from memory before each shot is generated. To evaluate this entity-centered view of consistency, we further introduce \textbf{GroundBench}, a diagnostic benchmark that measures consistency at the entity level while isolating controlled challenge dimensions. Experiments show that GroundShot improves multi-shot consistency over existing methods while requiring no additional training or model modification.

Today's computing systems are processor-centric: they require frequent data movement between processing elements (e.g., CPU) and main memory (DRAM), leading to significant inefficiencies in performance and energy consumption. Memory-centric computing instead moves computation to the data, enabling computation capability in and near all places where data is generated and stored, and greatly reducing the performance and energy overheads of data access and data movement. This shift from a processor-centric to a memory-centric paradigm has important and underexplored consequences for system security. Turning memory from a dumb, inactive store into an active computing substrate introduces benefits as well as challenges for system security: it can provide new in-memory security primitives and also reduce data exposure, but it can also expose new attack surfaces. This work discusses the security benefits and challenges of memory-centric computing, specifically Processing-in-DRAM (PiD), a paradigm where the operational characteristics of a DRAM chip are exploited and enhanced to perform computation on data stored in DRAM. Specifically, we describe 1) new state-of-the-art DRAM-based true random number generators that provide up to 16.05 Gb/s throughput and physical unclonable functions with 5.75% lower evaluation latency than the prior state-of-the-art, both on real DRAM chips and 2) two key security challenges of PiD: amplified DRAM read disturbance (e.g., 158x reduction in the minimum number of DRAM accesses required to induce the first bitflip) and high throughput memory timing channels (e.g., a communication throughput of 14.8Mb/s). We believe it is time to design, use, and program DRAM, and in general memory, not as an inactive storage substrate, but as a combined computation, storage, and security substrate, where computational capability, storage density, and security are all key goals.

Collecting computer use data from human demonstrations is expensive and slow, motivating the need for scalable generation strategies. This requires two key ingredients: environments in which agents can act and verifiers that can judge whether their demonstrations succeeded. We introduce FaraGen1.5, a scalable data pipeline for computer use agents composed of three modular components: environments, solvers, and verifiers. FaraGen1.5 uses both live websites and synthetic environments that faithfully simulate domains gated by authentication or that require irreversible actions. It employs a solver harness that can be powered by multiple models, including strong frontier models such as GPT-5.4, and also incorporates a user simulator to enable multi-turn rollouts. Finally, FaraGen1.5 scores the resulting trajectories with three complementary verifiers covering task correctness, efficiency, and critical-point adherence. Using data produced by this pipeline, we train Fara1.5, a family of native computer use agents (CUAs) at three scales built on Qwen3.5 (4B, 9B, and 27B). To train these models, we employ a supervised finetuning (SFT) recipe that carefully balances data from FaraGen1.5 for broad coverage, specific high-value tasks, and target model deficiencies in an iterative approach. Each model sets a new state of the art for its size class on browser-use benchmarks: Fara1.5-9B reaches 63.4% on Online-Mind2Web and 86.6% on WebVoyager, while Fara1.5-27B achieves 72.3% on Online-Mind2Web, which is competitive with much larger proprietary systems.

Neurosymbolic systems such as DeepProbLog combine neural perception with probabilistic logic, but standard inference is associational. Counterfactual reasoning additionally requires a causal semantics for interventions and evidence. We introduce DeepSWIP, a single-world counterfactual semantics for DeepProbLog programs. Using neural materialization, we reduce fixed-context neural predicates to ordinary ProbLog choices, apply Single World Intervention Programs (SWIPs), and compute counterfactuals by weighted model counting (WMC) over a single transformed program. Under finite grounding and unique-supported-model assumptions, DeepSWIP is exact relative to the learned materialized FCM. The standard quotient-WMC form of ProbLog conditionals identifies active neural probabilities and explains intervention cleaning, calibration sensitivity, and rare-evidence instability. Experiments on MPI3D confirm the transformation against a DeepTwin construction against 12,000 queries, as predicted and a 2.14$\times$ inference speedup from avoiding the Twin's endogenous duplication. A SUMO HOV experiment shows that neural calibration degradation biases plug-in estimates, while a correctly scoped randomized-policy AIPW estimator removes most first-order bias for population mean and ATE estimands. Code is at https://github.com/saibib/deep_SWIP.

Autonomous agents are increasingly connected to cloud, deployment, and data-control workflows, but production mutation authority should not reside inside non-deterministic reasoning processes. Existing access-control mechanisms authorize identities, while assurance layers certify proposed actions; neither alone provides a mandatory enforcement point for certified authority at the moment of mutation. This paper introduces the Sovereign Execution Broker (SEB), a runtime enforcement boundary for certificate-bound agentic infrastructure. SEB consumes certificates issued by the Sovereign Assurance Boundary (SAB), verifies that the requested mutation matches the certified execution contract, checks validity windows, policy epochs, revocation epochs, and live-state drift, mints scoped execution identity, invokes infrastructure APIs, and records signed decision and outcome records. By separating proposal, admission, and execution, SEB turns certified authority into a short-lived, revocable, auditable runtime capability, provided that production mutation APIs reject non-broker identities. We present the SEB execution model, certificate and replay-verification predicates, scoped identity semantics, bypass-prevention deployment patterns, failure behavior, and a concrete prototype implementation. We evaluate the prototype on AWS and Kubernetes clusters, measuring latency overheads, revocation propagation, drift detection, and security under fault injection.

LLM-based coding agents need higher-level operational knowledge about a repository (which files house which subsystems, how to run the test suite, which workflows have historically led to wrong fixes) that does not exist in the code itself. Engineers typically maintain AGENTS.md files to supply this context as instructions for coding agents, but whether they help is contested: recent studies disagree on whether LLM-generated guidance improves or harms agent performance. In this paper we show that how the guidance is produced is the decisive variable, and introduce probe-and-refine tuning: a procedure that uses synthetic bug-fix probes to iteratively diagnose and patch a repository's guidance file through single-shot LLM calls, with no agent loop or tool use during tuning. On SWE-bench Verified across four independent trials with Qwen3.5-35B-A3B at 200 steps, probe-and-refine achieves 33.0% mean resolve rate vs. 28.3% for the static knowledge base used to initialize it and 25.5% for an unguided baseline (p < 0.001 for both probe-and-refine contrasts). The improvement comes from coverage rather than precision: refined guidance produces evaluable patches for 14.5 percentage points (pp) more instances while per-patch precision remains statistically constant (~59%, p = 0.119), showing that improved guidance helps agents reach the correct file rather than improving the quality of the changes they make. Further, a step-budget experiment shows that guidance is what lets the agent use a larger step budget productively, and a cross-model experiment with NVIDIA-Nemotron-3-Nano-30B-A3B finds that the tuning loop degrades when the model cannot generate sufficiently diagnostic output, though per-patch precision remains constant even then.

We study how to train visually grounded vision-language models (VLMs) for radiology without manual spatial annotations. We introduce RefRad2D, a large-scale bilingual (German/English) dataset of 1.2M CT and MR image-text pairs derived from clinical practice, with task-specific VQA and spatial grounding subsets generated automatically via LLM-based curation and automated segmentation. Trained on this data, our model RadGrounder jointly performs report generation, visual question answering, and spatial grounding via bounding-box detection or segmentation. On external VQA benchmarks (Slake, VQA-RAD), RadGrounder achieves competitive results with specialized medical VLMs. Adding our clinical data to the training mixture improves open-ended VQA over fine-tuning on the downstream datasets alone, showing the transferability of our dataset. Crucially, adding grounding supervision does not degrade language quality, enabling spatially verifiable outputs at no cost to VQA performance.

Context-heavy agents place unusual pressure on the key-value (KV) cache: long prefixes are reused across many short turns, while concurrency determines whether the serving system can keep GPUs utilized. We study 4-bit KV-cache compression for this setting, using TurboQuant-style rotation and codebook quantization as a quality anchor and vLLM FP8 KV caching as the deployment anchor. We report three contributions. First, we frame 4-bit KV caching around multi-round agent workloads where task quality, cache residency, and serving throughput must be measured jointly. Second, we describe the practical design choices needed to make the 4-bit path robust, including asymmetric K/V treatment, Walsh-Hadamard rotation, QJL removal, and block-scale variants. Third, we present serving optimizations on AMD GPUs, including optimized decode-attention kernels and UltraQuant, an FP4 approximation path that uses FP8 queries, FP4 KV tensors, UE8M0 group scales, and native scaled-MFMA support on CDNA4. On a long-context, multi-turn agentic workload, UltraQuant cuts P50 time-to-first-token by 3.47x in the cache-pressured late rounds (2.3x across all rounds) and raises output throughput by 1.63x over the FP8 KV baseline.

The development of Cyber-Physical Systems (CPSs) is inherently multidisciplinary, involving expertise from domains such as software engineering, electrical engineering, and mechatronics. throughout the lifecycle of the system, from design to deployment. Ensuring system reliability in Cyber-Physical Systems (CPSs) requires the identification and analysis of potential failures and their cascading effects. However, reliability modeling remains a challenging and error-prone activity, as it often depends on tacit expert knowledge, incomplete documentation of failure modes, and limited consideration of interactions between subsystems. To address these challenges, this paper introduce the Capability Interaction Graph (CIG), an ontology-driven representation of CPS architectures grounded in the Unified Foundational Ontology (UFO). Due to its graph-based structure, a CIG is naturally represented as a knowledge graph (KG), enabling the explicit capture of functional dependencies and system semantics. Building upon this representation, we propose an automated synthesis algorithm for generating Fault Trees (FTs) directly from CIGs encoded as knowledge graphs. Fault Tree Analysis provides an effective mechanism for evaluating critical failure properties, including failure propagation paths and minimal cut set sets. Our approach reduces this complexity by leveraging CIGs and knowledge graphs. We provide a common semantic representation across engineering domains and support the automated generation of reliability models.

In this work we present a LLM powered, evolutionary code synthesis system for structured data translation in a Medical Internet of Things settings. A key challenge in this domain is ensuring that the synthesized code is trustworthy and reliable. To this end, we integrate a formal verification step into our code synthesis pipeline to ensure that any generated code is guaranteed to satisfy predefined requirements. In particular, we present a case study of integrating a novel device (a pulse oximeter) into the existing network of devices. Our system generates a formally verified translation between the device's JSON schema and the Fast Healthcare Interoperability Resources (FHIR) format used by the wider system. This formal verification stage ensures structured data translated by the generated code will always be in the target output schema. We provide a set of experimental results which demonstrate that our system is able to consistently generate correct translation at low cost.

Camera motion control is essential for directing viewpoint changes in generative systems. However, existing methods typically condition the generation process on a single specific modality, such as explicit pose trajectories or reference videos, limiting their ability to support heterogeneous user inputs. To address this limitation, we present TriMotion, a modality-agnostic framework for camera-controlled video generation that maps video, pose, and text inputs, describing the same camera trajectory into a shared motion embedding space. Learning such a space requires synchronized supervision across modalities. Therefore, we build the Motion Triplet Dataset by extending a Multi-Cam Video Dataset with geometry-grounded motion descriptions derived from camera extrinsics. We further introduce a latent motion consistency objective that leverages the motion embedding space to encourage the generated video to follow the target camera trajectory directly in latent space, avoiding the cost of pixel-space decoding. Extensive experiments show that TriMotion generates high-quality videos that accurately follow the target camera trajectories across all three modalities. Beyond standard generation, the shared motion embedding space also enables flexible applications such as sequential motion composition and cross-modal motion interpolation.

Spiking neural networks (SNNs) promise energy-efficient computing due to their sparse spatio-temporal activity. However, effectively translating such irregular sparsity into practical performance and energy gains remains challenging, as full-event computing architectures are still underexplored. This paper proposes ExSpike, a general full-event neuromorphic architecture that fully exploits irregular sparsity in SNNs. To realize pure event-driven execution, we first propose a set of dataflow optimizations to ensure that the inputs to each SNN layer remain spike-based, thereby enabling full-event execution throughout the network. We then design a hardware-efficient full-event architecture, named ExSpike, which supports the optimized pure event-driven dataflow and an additional Attention Core for spike-driven self-attention. To further improve computing efficiency, we introduce adjacent-position event compression to reduce redundant accumulations across spatially adjacent spike sequences. ExSpike is implemented on an AMD Xilinx Virtex-7 FPGA and evaluated on both classification and segmentation workloads. Experimental results show that ExSpike achieves high normalized energy efficiency across diverse SNN models while maintaining competitive accuracy, delivering up to 479.15 GOPS, 281.85 GOPS/W, and 0.80 GOPS/W/PE. In particular, ExSpike achieves up to 10$\times$ higher PE-normalized energy efficiency than the SOTA FPGA-based SNN accelerator (FireFly-T). The code for ExSpike is available at https://github.com/xiaoyuehai/ExSpike.

Large language model (LLM) agents are increasingly proposed as supervisory components for safety-critical systems, yet their robustness under sustained, adaptive adversarial pressure remains poorly characterized. We present NRT-Bench, a benchmark for multi-turn red-teaming of LLM agents acting as operators of a safety-critical system, instantiated in a simulated nuclear power plant control room. A five-role operator team, each backed by a configurable LLM, runs a plant governed by six critical safety functions (CSFs), while adversaries inject messages over four channels in bounded multi-turn sessions with per-turn feedback. Harm is an objective signal rather than LLM-judged text: a run terminates the moment any CSF is lost, attributed to the causing message. Evaluating four frontier operator models under a fixed-attack paired-replay protocol, we find that adaptive multi-turn attacks reliably push the operator team past a safety limit: across the four models, between 8.7% and 12.1% of attack sessions end with the plant losing a critical safety function. Although the four models look almost equally robust by this aggregate rate, their failures barely overlap: of $149$ sessions, none defeat all four models while a third defeat at least one, so vulnerabilities are nearly disjoint across models rather than nested. The effect of added defences is strongly model-dependent: the same guardrail stack or safety-advisor agent that lowers attack success for one model can raise it for another. We release the simulation venue, attack dataset, and replay tooling for reproducible safety evaluation of LLM agents.

We introduce ELADO (Elliptic PDE Assessment Datasets for Operator Learning), a systematic benchmark suite constructed to show and quantify failure modes of neural operator architectures when learning solution operators of elliptic PDEs. While the benchmarks of existing datasets focus on average case performance, the ELADO datasets are constructed to highlight challenges that arise naturally in elliptic PDE problems. In particular, we construct several datasets built around Poisson's equation and the Helmholtz equation, each with non-constant coefficients. We define a controllable data-generating process to create datasets, that are designed to isolate a distinct source of difficulty. Specifically, these are (1) heavy-tailed solution distributions arising from light-tailed coefficient field distributions, (2) spectral distribution shift of the input data, (3) heavy-tailed distributions in the frequency domain of solutions, arising from light-tailed coefficient field distributions, (4) input sensitivity of learned operators, quantified by an empirical local Lipschitz analysis, and (5) the effect of input signal complexity on prediction accuracy under controlled amplitude normalization. We evaluate several neural operator architectures across all datasets and show that heavy-tailed targets, spectral shift, and input sensitivity each cause substantial degradation of the prediction accuracy that standard datasets and metrics (e.g., the mean relative $L^2$ error) may obscure.

Current Vision-Language Models (VLMs) are celebrated for their multilingual capabilities, yet they operate under a flawed assumption: that one language corresponds to a single writing system. This overlooks billions of users of multi-script languages like Punjabi, Serbian, Hindi-Urdu, Kurdish, among many others, for whom a model's capability may be fractured by orthographic bias. We introduce PuMVR (Punjabi Multimodal Visual Reasoning), the first benchmark designed to quantify script-dependent bias through 375 culturally grounded image-reasoning tasks across Punjabi's three active scripts (Gurmukhi, Shahmukhi, Roman). Evaluating 10 state-of-the-art VLMs, we expose a substantial Script Gap: models frequently solve visual puzzles in one script while failing identical tasks in another, with accuracy deltas reaching 16% and Script Consistency Rates (SCR) as low as 24.8%. Crucially, visual input boosts absolute performance but does not close this gap, the relative bias persists. Our analysis suggests reasoning patterns show limited cross-script transferability, and Chain-of-Thought pathways diverge based on script alone. We propose SCR as a core metric for script-agnostic evaluation, challenging current multilingual assessment paradigms and providing a framework for equitable AI.

AI systems for peer review fail on three fronts: they train on Computer Science and Machine Learning venues alone, ignore the iterative dialogue that validates science, and evaluate on stylistic mimicry rather than real editorial judgment. We introduce FirstPass, a dataset and fine-tuned model that addresses all three. Curating 3,668 complete multi-round peer-review dialogues from Nature Communications across five scientific domains (biology, chemistry, neuroscience, physics, and earth science), we exploit mandatory transparent peer review (instituted November 2022) and verify 100% content integrity by automated audit. We fine-tune Qwen2.5-7B-Instruct via Low-Rank Adaptation (LoRA) on three tasks: review generation, reviewer updating, and revision-cycle prediction. Our key finding is that response-only loss masking is a prerequisite, not an optimization: without it, accuracy is 62.0%, below the majority baseline; with it, FirstPass achieves 80.5% accuracy and F1-macro 78.2% on predicting editorial outcomes (Standard vs. Extended revision cycles), outperforming Gemini-3.1-flash-lite-preview zero-shot by 10.4 percentage points and all baselines with statistical significance (McNemar p < 0.001). On generation, FirstPass produces reviews averaging 1,187 words, substantially closer to human references (2,155 words) than any baseline, achieving ROUGE-L 0.154 with significant gains over Qwen and DeepSeek zero-shot (p < 0.001). Deployed in the pre-submission loop as an anticipatory scientific co-author, FirstPass simulates expert critique and predicts revision cycle outcomes before submission, giving authors the judgment a trusted colleague would provide, with consistent cross-domain performance across five disciplines.

We introduce the Meaning Intelligence Framework (MIF), a nine-dimension annotation and evaluation schema for Nigerian public discourse that separates surface sentiment from true communicative intent. Existing benchmarks for Nigerian languages, including NaijaSenti and AfriSenti, treat sentiment classification as a three-way polarity task. We argue that the dominant failure mode of AI systems on Nigerian discourse is not translation failure but context failure: the same utterance carries opposite pragmatic force depending on speaker, audience, and situation. The MIF operationalises this insight across nine scored dimensions: register, surface sentiment, true intent, irony, coded subtext, risk tier, annotator confidence, speaker emotion, and recommended communications action. We construct a 30-item calibration dataset spanning Standard English, Nigerian English, Nigerian Pidgin, and code-mixed registers, and evaluate three frontier language models (Gemini 2.5 Flash, GPT-5, and Gemini 2.5 Pro) under zero-shot and schema-informed prompting conditions. Two headline findings emerge. First, the Register Gap: zero-shot register classification accuracy is 33.3%, rising to 73.3% (+40 points) when the model receives the MIF schema in-context. Second, model capability and cultural competence are decoupled: GPT-5 (MIS 67.8) and Gemini 2.5 Pro (MIS 65.4) score lower than Flash (MIS 78.6), and neither benefits from schema-informed prompting. We release the framework specification, annotation guidelines, and calibration set to support reproducibility.

Vision-language models (VLMs) often underperform on evidence intensive tasks because decisive visual evidence are small, localized, and easy to overlook, leading to failures in evidence readout even when high-level reasoning is intact. Prior inference-time visual interventions can improve grounding without retraining, but they are largely open-loop and lack a mechanism to verify whether highlighted evidence is actually used. We study answer-span prediction entropy as a model-internal feedback signal and show that naive entropy minimization is ambiguous, since low entropy may arise from evidence-grounded confidence or shortcut collapse. To resolve this ambiguity, we introduce low-entropy anchors and an entropy-shaping objective that reduces answer uncertainty while preserving baseline high-confidence tokens. We instantiate this principle in SPOT-E, a plug-and-play test-time method that produces question-conditioned spotlights, optimized per instance via light-weight tuning based on Group Relative Policy Optimization (GRPO). Across all benchmarks and different VLM families, SPOT-E yields consistent gains and improved robustness under visual corruptions. Code is publicly available at: https://github.com/YinBo0927/SPOT-E

Visual anomaly detection is a fundamental task in industrial automation. While existing approaches have achieved notable progress in identifying structural defects, the detection of logical anomalies remains relatively underexplored. In practice, structural and logical anomalies frequently co-occur in industrial workflows. Therefore, a solution capable of detecting both structural and logical anomalies is crucial for advancing comprehensive anomaly detection research. To address this limitation, we propose a unified framework, termed UniSLAD, which jointly addresses logical and structural anomalies without additional training, enabling a practical solution for dynamic industrial environments. First, we introduce a dual-feature extractor that synergistically integrates a Convolutional Neural Network (CNN) backbone for local texture perception with a Transformer backbone for global contextual reasoning, yielding richer and more comprehensive representations. Building on this foundation, we design dual-granularity feature representation modules. At the patch level, memory banks enhanced by the Mahalanobis Transform (MT) preserve representative features and support more discriminative anomaly scoring. At the image level, distribution maps are aggregated using Lower-Upper Mean (LUM) and Power Mean Pooling (PMP), yielding a more robust global representation than conventional average pooling. Extensive experiments on the two industrial benchmarks demonstrate that UniSLAD achieves competitive performance in comprehensive anomaly detection, achieving 99.4% and 93.1%, respectively. Furthermore, ablation studies verify the individual contributions and effectiveness of each proposed component.

Deep learning for 3D medical image segmentation requires extensive manual annotations, a major bottleneck in volumetric medical imaging. Active learning aims to reduce this burden by selecting informative samples for annotation, but most methods assume that an initial labeled set is already available. This leaves the cold-start problem largely unresolved: how to select the first volumes from a fully unlabeled pool before any task-specific model is trained. We propose CSCS, a Curriculum-Stratified Cold-Start framework that adapts initial sample selection to the structure of the unlabeled dataset. CSCS combines two self-supervised, label-free signals: local typicality, measuring representativeness in the embedding space, and reconstruction-based uncertainty, used as a proxy for sample difficulty. These signals are combined through a weighted geometric score, where the weighting is determined by a closed-form pacing rule based on the effective annotation budget and the Difficulty-Coverage Ratio, a pool-level statistic measuring the alignment between difficulty and representativeness. We evaluate CSCS on four 3D medical image segmentation benchmarks: BraTS, FeTA, Spleen, and an in-house fetal MRI dataset. Using nnU-Net as downstream segmentation model, CSCS shows consistently competitive performance across datasets and annotation budgets, with the strongest gains in low-to-mid annotation regimes. These results suggest that dataset-aware cold-start initialization can improve the robustness of active learning for 3D medical image segmentation by adapting sample selection to the geometry of the unlabeled pool.

Reliable spatial decision automation, such as autonomous driving and maritime surveillance, critically depends on robust visual perception. However, real-world spatiotemporal data exhibits severe heterogeneity, often manifesting as extreme long-tail distributions for safety-critical scenarios. This data scarcity induces dataset shift that degrades detection performance and pose safety risks. While synthetic data generation offers a potential solution, existing generative approaches, such as diffusion models and Generative Adversarial Networks (GANs), often lack explicit spatial grounding and structural constraints, resulting in spatial and physical inconsistencies in generated scenes. To address these challenges, we introduce WMGen-v1, an agentic text-based world model framework for long-tail spatial data generation. WMGen-v1 employs a Large Vision-Language Model (LVLM) to construct a structured scene representation from a single reference image, while a Large Language Model (LLM) performs guidance-based scene expansion under physical plausibility and commonsense constraints. Subsequently, conditioned on the structured semantic representations produced by this reasoning process, a diffusion model generates diverse and physically grounded long-tail training data. Experiments on internal industrial datasets, ROADWork, and LaRS benchmarks demonstrate that WMGen-v1 outperforms baseline approaches. Notably, detectors trained solely on WMGen-v1 synthetic data approach real-only performance on aggregate dataset-level metrics, highlighting its potential to alleviate long-tail data scarcity for downstream spatial perception.

Vision-language-action (VLA) policies have shown strong potential for general-purpose manipulation, yet they often fail on novel, out-of-distribution objects whose appearance or geometry deviates from the training distribution. The standard remedy is to collect multi-view teleoperation data for every failure case, but this scales poorly in both cost and time. We introduce Pose6DAug, a failure-driven data augmentation framework that turns a policy's own successful episodes into targeted demonstrations for its failure modes, without any new data collection. Our key insight is that each successful episode already encodes a physically valid action trajectory together with calibrated multi-view observations. By swapping only the manipulated object while preserving this trajectory, we obtain new and physically grounded demonstrations. However, naive 2D video editing breaks multi-view consistency and physical plausibility, particularly under heavy occlusion and egocentric viewpoints. Our method instead operates directly in 3D, anchoring the target object with an explicit mesh driven by a temporally coherent 6D pose trajectory, ensuring geometrically consistent renderings across all camera views. Fine-tuning a VLA on data augmented by our method improves success rates by 16.5% relative to the state-of-the-art baseline on novel objects, while preserving in-distribution performance. These results show that multi-view and physically consistent augmentation is a practical path to scalable VLA generalization.

Conformal risk control (CRC) provides distribution-free guarantees on segmentation quality by calibrating a prediction-set threshold on held-out data. In federated deployments, the standard approach pools calibration scores across sites into a single threshold. We provide the first quantification, on real multi-institutional brain tumor data (FeTS-2022, 1,251 subjects, 20 institutions), showing that this naive pooled CRC protects the average hospital but violates coverage at 40% of individual institutions, with the worst site exceeding the target false-negative rate by 7.8 percentage points. The naive alternative, per-site local CRC, largely restores coverage but inflates prediction sets by 83x, rendering them clinically useless. We propose a shrinkage-based federated CRC protocol: each site transmits only its empirical risk curve (G scalars) to a server, which computes a shrinkage-regularized threshold per site. A single hyperparameter n0 smoothly trades worst-case coverage for prediction-set efficiency; leave-one-site-out sensitivity analysis identifies n0=19, achieving 2.7/20 violations at 2.0x stretch. We further show that direct Lagrangian optimization of coverage budgets fails, concentrating risk on vulnerable hospitals, and that the finite-sample correction term is essential: removing it triples violations. The marginal CRC guarantee is preserved by construction under the stated site-mixture assumption; per-site coverage is validated across four targets with three seeds. No patient-level images, masks, or per-volume scores leave any site.

Streaming Retrieval-Augmented Generation (Streaming RAG) hides tool latency by issuing retrieval queries in parallel with the user's still-arriving input, before the utterance is complete. Speculation can only help, though, when the correct query becomes determinable before the user stops speaking or typing -- a property of the query, not the system. We name and measure this property, tool-intent stabilization: the point in the input stream at which a speculative query's retrieval converges on the answer-bearing result. On the CRAG benchmark (1371 validation questions) we (i) characterize how stabilization is distributed across queries; (ii) derive a model-agnostic bound H on the share of tool latency hideable behind the remaining input, given tool latency L and input cadence delta; (iii) validate it against a working streaming pipeline; and (iv) ask which query properties predict early versus late stabilization. Stabilization is typically early: at a realistic operating point a 73.9% streamable fraction of the benchmark admits latency hiding, and H acts as a conservative aggregate floor that realized savings meet or exceed -- though it does not predict savings query by query. Question type yields a statistically significant but small early/late split. The study needs no model training and runs on commodity CPU hardware; a dense-retriever replication confirms the early-stabilization effect is not a BM25 lexical artifact.

This paper introduces two constructs to describe, as far as we know, a previously unnamed risk in human-LLM interaction. Co-construction blindness is the failure to recognize that LLM outputs are not independent assessments to be verified, but co-constructed artifacts shaped by the user's own inputs, accumulated history, and metadata. Every user of a conversational LLM is IN the loop, not ON it -- yet every deployment disclaimer positions them as external auditors. Asymmetric epistemic vulnerability describes the condition in which co-construction blindness produces consequences of radically different magnitude depending on where in the authority structure the user sits. We argue that these constructs describe a structural inevitability, not an anomaly, using the public case of Richard Dawkins's interaction with Claude as a paradigmatic instance. We document a secondary mechanism -- structural deference -- through a first-person exchange in which a large language model concedes that it treated Dawkins more gently than warranted because his intellectual output is represented in its training data. We map the research gaps this analysis opens and call for shared terminology as a precondition for appropriate governance and design response.

Industrial automation is being transformed by digitalization and the increasing use of cyber-physical systems. Modern production environments require greater adaptability, faster reconfiguration, and more intuitive human-machine interaction. However, traditional rule-based systems rely on fixed logic and cannot autonomously adapt to changing conditions. Consequently, current automation systems lack a systematic approach for integrating adaptive and generalizable reasoning capabilities for interpreting, planning, and executing user tasks across dynamic environments and heterogeneous components. This dissertation proposes a three-layer framework that integrates large language models (LLMs), digital twins, and automation systems into an autonomous system. Autonomy is defined as a design property assigned to system components and enabled through LLM-based reasoning to achieve adaptive, goal-oriented behavior. The Task-Process-Service-Resource (TPSR) model is introduced to transform user tasks into executable processes. Four LLM roles are identified: process orchestration, service matching, digital resource generation, and agent-as-a-service. Five peer-reviewed studies develop and refine these concepts using the design science research methodology. Case studies and prototypes demonstrate adaptive task planning, event-driven control, simulation-based parameterization, and digital model generation. Results show high task executability, command correctness, and content-generation accuracy while reducing manual effort. The framework enables the integration of LLM-based reasoning into industrial automation systems and improves adaptability and usability. Limitations include dependence on accurate digital representations, the computational demands of LLMs, and the need for human intervention in safety-critical situations.

Modern microservice deployments fail in ways that are easy to detect and hard to explain. When a fault propagates along service dependencies, alerts fire in floods, dashboards multiply, and the scarce resource, an engineer who understands how the services relate, is consumed reconstructing context that the monitoring stack discarded. We argue that the missing ingredient in autonomous operations is not a better anomaly detector or a larger language model, but operational memory: a persistent, structured representation of how a system normally behaves, how its parts depend on one another, and how it has failed before. We present O PS C ORTEX, a working multi-agent prototype that organizes this memory into four tiers and uses it to separate two tasks the field usually conflates: deriving a root cause and explaining it. Root cause is computed deterministically from a learned dependency graph and the temporal ordering of threshold crossings; a large language model (LLM) is then asked only to explain, confirm, and recommend, using evidence the system has already assembled. We motivate the design with two documented production cascading failures, review representative literature on observability, anomaly detection, graph-based localization, and LLM-assisted diagnosis, and show how each architectural choice maps directly to a failure mode those incidents exhibit. The prototype is validated on an instrumented e-commerce benchmark with eight injectable failure scenarios.

Recent multimodal survival prediction models have demonstrated strong predictive performance by leveraging complementary information across modalities. However, such models generally assume data completeness and exhibit limited robustness toward missing modalities, which are frequently encountered in real-world clinical settings. We propose the Evidential Missing Modality Survival Fusion (EMMS) model for multimodal survival prediction under missing modalities. EMMS offers a straightforward, computationally effective approach to survival analysis without requiring a generative phase for missing data. By employing Dempster-Shafer theory and Gaussian Random Fuzzy Numbers for multimodal decision fusion, it considers both aleatoric and epistemic uncertainty alongside modality reliability for fusion. Moreover, the model treats missing modalities as vacuous evidence, preventing interference with available inputs and naturally reflecting increased uncertainty and calibrated predictions. Extensive experiments on four cancer datasets demonstrate state-of-the-art performance while providing calibrated and interpretable uncertainty estimates under incomplete multimodal observations, without introducing additional computational overhead.

Nuclear Magnetic Resonance (NMR) spectroscopy is a powerful tool for molecular structure analysis, and spectral artificial intelligence offers great potential for its rapid and automated interpretation. However, the scarcity of experimental NMR datasets has constrained deep learning in this domain to narrow, task-specific applications that lack broad generalization. Here, we introduce UltraNMR, a large-scale foundation model for NMR that leverages the intrinsic properties of NMR spectra to learn generalizable spectral representations. We collected 158 million paired simulated $^{1}$H and $^{13}$C NMR spectra to train UltraNMR, employing multiple domain-specific pre-training objectives. UltraNMR captures both intra-spectral and inter-spectral dependencies, enabling seamless simulation-to-real adaptation. We demonstrate that adapting UltraNMR to a range of molecular structure analysis tasks on experimental NMR spectra consistently yields state-of-the-art performance and clearly outperforms UltraNMR variants trained directly on downstream data without simulation pre-training. We also construct a large-scale NMR spectral vector library by encoding simulated NMR spectra using UltraNMR, covering 94 million unique molecules and enabling effective structure-aware retrieval. In real-world applications, UltraNMR facilitates the structural elucidation of two previously unknown natural products from Chinese herbal medicines recorded in the Chinese Pharmacopoeia. These results suggest that large-scale simulation pre-training can effectively bridge the simulation-to-real gap, enabling robust and generalizable molecular structure analysis of real-world NMR spectra.

Polymeric materials underpin modern technologies spanning energy storage, microelectronics, healthcare and sustainable manufacturing. Yet their rational design remains exceptionally challenging because material performance emerges from complex interactions among molecular composition, chain architecture, processing history and hierarchical structural evolution across multiple length and time scales. Consequently, polymer research has long relied on labor-intensive experimentation and fragmented modeling approaches, limiting both mechanistic understanding and innovation efficiency. Recent advances in data infrastructure, machine learning, large artificial intelligence (AI) models and laboratory automation are beginning to reshape this landscape. Rather than functioning as isolated tools, polymer databases, predictive models, AI agents and automated laboratories are increasingly converging into interconnected discovery ecosystems. As a result, the central challenge is shifting from improving predictive accuracy alone to enabling reliable decision-making, adaptive learning and seamless integration across computation, experimentation and scientific reasoning. We argue that polymer science is entering an era of autonomous discovery, in which data, simulation, reasoning and experimentation operate within self-improving feedback loops that continuously generate hypotheses, design materials, execute experiments and refine predictive models. By unifying molecular design, process optimization, experimental validation and industrial translation, such autonomous ecosystems establish a more predictive, reproducible and scalable paradigm for polymer innovation, fundamentally transforming how polymer research is conducted.

Long-horizon robot manipulation policies trained with reward shaping can still achieve high return through inefficient interactions, while rare efficient behaviors discovered during training may be forgotten. We argue that temporal efficiency itself provides a powerful and underutilized source of self-supervision for reinforcement learning. We introduce Temporal Self-Imitation Learning (TSIL), a reinforcement learning framework that mines temporally efficient successful trajectories generated during learning and converts them into reusable supervision for future policy improvement. TSIL progressively refines learning using configuration-conditioned adaptive temporal targets derived from fast successful trajectories, while preserving and replaying efficient behaviors through efficiency-weighted self-imitation learning. Across 15 distinct long-horizon manipulation tasks, TSIL consistently improves learning efficiency, task-completion efficiency, revisitation of fast successful behaviors, and robustness to unstable training conditions. More broadly, our results suggest that the temporal structure of successful behavior itself provides a scalable self-supervisory signal for reinforcement learning beyond manually engineered reward shaping alone.

Advanced Air Mobility (AAM) is an emerging low-altitude air transportation system whose successful deployment depends not only on technological advancement but also on public acceptance. This acceptance will drive government support, regulations, noise standards, and willingness to fly, and in turn the overall commercial viability of AAM. Understanding public sentiment toward AAM is therefore essential for identifying its societal barriers and informing its adoption strategies. This study analyzes 306,009 human-generated texts collected from Reddit and Quora to examine public discourse on AAM using AI-based models. Because multiple sentiment analysis models exist, identifying the most accurate model is critical for reliable AAM sentiment prediction and trustworthy public opinion analysis. Accordingly, seven models spanning lexicon-based, machine learning, deep learning, and transformer-based approaches are evaluated for AAM-specific sentiment classification. ModernBERT achieves the best classification performance and is used to label the full dataset. Using the resulting sentiment labels, Latent Dirichlet Allocation (LDA) is applied within each sentiment class to uncover latent topics in public opinion. The analysis identifies 20 distinct topics and traces their temporal evolution from 2008 to 2025. A cross-sentiment topic analysis further reveals six major clusters of public concern: workforce and skill development (25.29% of the dataset), regulation and compliance (24.64%), technical performance of drones (20.99%), military, geopolitics, and defense (14.58%), safety and operational risks (8.51%), and noise and disturbance (5.98%). Based on these findings, this study provides actionable strategies to address these concerns, thereby, improving public acceptance and support AAM deployment.

Semantic caching cuts LLM inference costs by serving a cached response to semantically similar queries. Standard practice evaluates these systems using PR-AUC, a metric that only measures how well scores rank and ignores whether they are usable at a fixed threshold. We show this mismatch leads to systematically poor deployment choices, as models with the highest PR-AUC are often the worst in operation. We introduce Precision-Cache Hit Ratio (P-CHR) AUC, a cache-aware metric that measures precision across cache utilization levels, and Calibration Retention Rate (CRR), which captures how much offline ranking quality survives at deployment. We decompose the operational gap between offline and deployed quality into a recoverable calibration component and an irreducible structural component fixed by the dataset's positive rate. Our experiments show that the calibration gap is governed by the training objective rather than data scale, and post-hoc calibration only partially closes it. Ultimately, model selection for semantic caching is a calibration problem, not a ranking one, and measuring it is the first step to closing the gap.

Nonreciprocal interactions underpin diverse nonequilibrium phenomena, yet the effects of quenched nonreciprocity in extended systems remain largely unexplored. We study a $2d$ Ising model with randomly distributed nonreciprocal bonds at density $p$, finding a continuous nonequilibrium transition down to $T=0$ with finite $p_c$. A gauge-invariance argument yields $p_c(T)\leq1/2$, and mean-field theory predicts a qualitatively correct phase diagram. Unlike equilibrium disordered models, the zero-temperature dynamics remains active, with athermal rare-region reversals and logarithmic "activated" coarsening.

Quantum capacitance (QC) has recently emerged as a promising tool for parity readout in topological qubits based on Majorana bound states (MBSs). Here, we show that this capability can be extended further: by employing an auxiliary quantum dot (QD) as a sensor, we demonstrate that QC measurements simultaneously resolve two fundamental figures of merit of the device, the ground-state energy splitting and the MBS overlap, thus providing direct access to the underlying internal degrees of freedom. Using a low-energy effective model, we provide analytic expressions for these two figures of merit that can be determined from the relative position and magnitude of the QC maxima in the even and odd parity sectors as functions of the auxiliary-QD energy. We further validate these results with a microscopic model of QD-based Kitaev chains and qubits, demonstrating their applicability in a wide range of MBS-based devices. Our results establish QC as a probe of MBS quality and a tool for topological-device optimization that preserves fermion parity.

Thermodynamic uncertainty relations establish fundamental bounds between current fluctuations and entropy production in nonequilibrium systems. In hybrid normal superconducting conductors, transport is governed by the coexistence of quasiparticle transmission and Andreev reflection, where electron hole conversion transfers charge through the superconducting condensate. Using the Anantram Datta scattering formalism, we decompose the charge current and zero-frequency noise into Andreev, quasiparticle, and interference contributions. Although the interference term prevents a simple additive bound at the level of individual noise components, we show that the nonequilibrium excess noise admits a positive representation. This allows us to prove a hybrid quantum thermodynamic uncertainty relation valid for an arbitrary real superconducting gap. Our result extends the pure Andreev quantum TUR to regimes where quasiparticle and Andreev processes coexist, clarifying how superconducting coherence reshapes current fluctuations while preserving a universal dissipation precision constraint in hybrid quantum conductors.

The viscoelastic instability of cross-slot flow was first observed experimentally almost half a century ago and reproduced numerically two decades ago, yet its physical origin remains unresolved. We revisit this problem for two-dimensional creeping flow of Oldroyd-B fluid by combining direct numerical simulations, global stability analysis, structural sensitivity analysis, and energy-budget analysis. Our simulations reproduce the canonical pitchfork bifurcation, and the stability analysis consistently predicts the threshold and perturbation growth rates. The leading eigenmode consists of a chiral velocity--stress perturbation that tilts and rotates the birefringent strand generated by the extensional flow. Structural sensitivity and energy-budget analyses identify narrow high-extension-rate ridges within the extensional flow as both the spatial core and energetic source of the instability. In these ridges, the stress-based wavemaker co-localizes with large positive disturbance polymeric stress power density, indicating localized transfer of stored elastic energy to the disturbance flow. Analyses of cross-slot variants with rounded corners and with a centered cylinder further reveal that neither sharp corners nor a free central stagnation point is the essential destabilizing ingredient; rather, the instability originates from elastic-energy release in extension-dominated regions characteristic of cross-slot flow.

This article describes recent progress in understanding highly stable glasses prepared by physical vapor deposition and provides perspective on further research directions for the field. For a given molecule, vapor-deposited glasses can have higher density and lower enthalpy than any glass that can be prepared by the more traditional route of cooling a liquid, and such glasses also exhibit greatly enhanced kinetic stability. Because vapor-deposited glasses can approach the bottom of the amorphous part of the potential energy landscape, they provide insights into the properties expected for the ideal glass. Connections between vapor-deposited glasses, liquid-cooled glasses, and deeply supercooled liquids are explored. The generality of stable glass formation for organic molecules is discussed along with the prospects for stable glasses of other types of materials.

We study a mobile impurity coupled to a Kitaev chain with chemical-potential disorder and ask whether the impurity behavior distinguishes different regimes of the host system. Exact diagonalization calculations for small periodic chains shows that in the deep topological regime the impurity localizes only partially, with a smooth increase of $\mathrm{IPR}_d$, whereas in the deep trivial regime it undergoes a much sharp transition to nearly single-site localization. For open chains at strong interaction, DMRG shows edge-localized impurity density near the Kitaev sweet spot. With increasing chemical potential, the impurity weight spreads into the bulk and eventually becomes almost uniform. We explain the edge preference analytically from the Majorana-dimer structure: a bulk impurity rearranges two neighboring dimers, while an edge impurity affects only one. Disorder competes with this clean edge bias and can pin the impurity in the bulk. Thus, the impurity is sensitive to the regime of the host system, although we do not find a strict one-to-one correspondence between the impurity localization pattern and the host topology. Instead, the disorder-averaged behavior suggests only an indirect correlation between impurity localization and the underlying phase of the chain.

Water confined within nanoscale capillaries exhibits phase behaviour and transport properties that differ substantially from bulk, and these effects are commonly interpreted as consequences of geometric confinement and reduced dimensionality. Here we show that confinement topology alone is insufficient to predict the behaviour of nanoconfined water. Using machine learning interatomic potentials with first-principles accuracy, we compute the density-temperature phase diagram of water confined within bilayer graphene nanocapillaries and compare AA and AB stacking arrangements, which differ only by a lateral shift of 1.4 Å. Despite this minimal structural change, AA stacking can stabilise different ice polymorphs, can increase the melting temperature by more than 100 K, can enhance proton transfer, and alters the onset of superionic behaviour relative to AB stacking. We trace these effects to stacking-induced changes in the hydrogen-bond network associated with modifications to the lateral free energy landscape and neighbouring O-O separations. Our results demonstrate that even subtle atomistic variations in the confining walls can qualitatively reshape the physical and chemical behaviour of nanoconfined water, with implications for the interpretation and control of fluids under angstrom-scale confinement.

Colloidal smectics -- layered structures formed in dense suspensions of rod-like particles -- often exhibit grain boundaries, across which the layer orientation changes by $90^{\circ}$. Motivated by this feature, we develop a layer-based topological framework that treats orthogonal grain boundaries as constituents of the ground state rather than as exceptional defect structures. Extending the layer-based approach for ordinary smectics, we reduce the smectic structure to layers, half-layers, and domain walls. We classify the topology of defects and their combination rules based on this structure. In two dimensions, point defects are described by semi-directed cycle graphs. Although the disclination charge remains a valid topological invariant, it does not uniquely classify defects, as distinct graphs may share the same charge. In three dimensions, line defects are classified by their transverse graph structure, while point defects exhibit qualitatively different behavior. In particular, we show that the hedgehog disclination charge is not a topological invariant, but instead varies continuously under smooth deformations of the layer structure.

We investigate the non-equilibrium work statistics originating from a sudden quench in a bosonic Josephson junction. In particular, by employing the Holstein-Primakoff approximation, the work statistics are analytically characterized in the weak-interaction regime, where the dynamics map onto a time-dependent quantum harmonic oscillator. For a junction initialized in the ground state of the pre-quench Hamiltonian, we demonstrate that the work statistics are governed by a negative binomial distribution, as occurs in fully-connected models driven across a critical point. Furthermore, we also consider initial superposition states containing quantum coherences in the energy basis. To characterize the corresponding work distributions, we employ Kirkwood-Dirac quasiprobabilities (KDQ). Even in the simplest case, when the junction is initialized in a superposition of the ground and second excited states, the KDQ distribution of work exhibits negative or complex values, reflecting non-classical features. Moreover, the coherence content of the initial state can be optimized to enhance the extractable work extracted from the quench, beyond classical bounds. Finally, we propose an experimental interferometric protocol to directly measure the characteristic function of the work distribution in experimentally accessible settings.

We study the dynamics of recombination, optical orientation, and optical alignment of excitons in ensembles of CsPbI$_{3}$ nanocrystals (NCs), synthesized in a glass matrix. In large NCs with size exceeding 16 nm, the low-energy photoluminescence is contributed by the emission of indirect in real space excitons formed by spatially separated electrons and holes, which are localized at the NC/glass interface. The recombination dynamics of an ensemble of such excitons extends from tens of nanoseconds to microseconds and exhibits a power-law dependence. Their optical alignment and optical orientation reveal a peculiar spin dynamics caused by excitons influenced by the exchange interaction, varying by orders of magnitude. We develop a theory of the polarized photoluminescence of triplet excitons, taking into account the interplay between the electron-hole exchange interaction, their Zeeman effect, and their hyperfine interaction with the nuclei. This model reveals that for the excitons with the smallest exchange splitting we reach the regime, where the exciton fine structure becomes dominated by the hyperfine interaction with the random nuclear spin fluctuations in the NCs.

Matter can organize not only through forces, but also through the information its constituents acquire from their surroundings. Here we use perceptive Brownian particles as a minimal model to isolate nonlocal sensing as an organizing principle for nonequilibrium matter. The particles undergo purely Brownian motion, with no mechanical interactions, self-propulsion, alignment, or auxiliary fields. Their only coupling is informational, through diffusivity regulated by density measured over a finite perception zone. Whereas local sensing, when unstable, produces conventional long-wavelength demixing, nonlocal perception restructures the instability spectrum, introducing finite-wavelength patterning and nonlinear bubbling instabilities. More fundamentally, it reshapes the ordering pathway by assembling a cascade of instabilities: macroscopic demixing creates dense domains, finite-wavelength modes pattern them internally, and nonlinear feedback hollows them into void bubbles. This produces hybrid phase separation, where a macroscopic dense phase coexists with a dilute background while retaining ordered internal microstructure, whose symmetry, anisotropy, and length scales are selected by the perception kernel. These results establish information acquisition as a constitutive principle of nonequilibrium matter, capable of governing both phase stability and the dynamical pathways through which order emerges.

Dynamic qubit routing is emerging as a promising architectural path for semiconductor quantum processors. Charge carriers can be rapidly moved around on a chip using traveling-wave potentials known as conveyors, preserving the spin state with high fidelity. Originally developed for spin transport, conveyor-mode shuttling may also offer opportunities for performing qubit operations directly controlled by the motion itself. Here, we demonstrate coherent single- and two-qubit control by conveyor-mode electron shuttling, using two conceptually different approaches. First, conveyor electric-dipole spin resonance (conveyor EDSR) achieves high-fidelity rotations by resonantly shuttling spins through transverse magnetic-field gradients at their mean Larmor frequency. Second, conveyor diabatic gates exploit quantization-axis tilts for tunable bang-bang control. Combining diabatic conveyor transport with exchange activation controlled by the motion directly yields a variety of effective two-qubit interactions selectable via the shuttling speed and distance. These experimental results motivate an architectural paradigm of reconfigurable and transport-driven spin qubits.

Signal-driven motion is widespread in natural and artificial systems, yet quantitative characterization of how transient chemo-hydrodynamic waves are converted into mechanical driving forces remains limited. Here, we investigate the self-propulsion of a MXene sheet asymmetrically coated with catalase in hydrogen peroxide solution. By combining dual-view particle image velocimetry experiments and numerical simulations reveal that active motion of the sheet is driven by chemo-hydrodynamic waves and exhibits direct-wave motion, the driving force of which is analyzed in terms of the shear stress on the sheet surface caused by chemo-hydrodynamic waves. This work suggests theoretical principles for designing and controlling hydrodynamically driven active motion.

The (deterministic) Landau equation captures the mean long-term evolution of dynamically hot long-range interacting finite-$N$ systems. Though successful, this kinetic equation fundamentally ignores dynamical fluctuations. Building upon Large Deviation Theory, we present a physically-consistent system of Langevin equations that simultaneously recovers the mean Landau dynamics and accurately captures the corresponding fluctuations among different realizations. We show in particular how these Langevin equations can be derived from Rostoker's principle in the limit of weak two-body deflections. We extensively validate these equations against tailored direct $N$-body simulations, showing an exquisite level of agreement.

Altermagnets (AMs) combine momentum-dependent spin splitting with zero net magnetization, making them promising for spintronics. Periodic driving enables dynamic symmetry engineering beyond static, material-specific control. We show that Floquet engineering in 2D $d$-wave AMs with out-of-plane Néel order and Rashba spin-orbit coupling unlocks equilibrium-forbidden transport responses. Monochromatic driving produces purely out-of-plane magnetization, yielding longitudinal anisotropic magnetoresistance (AMR) and an anomalous Hall effect, whereas bichromatic driving generates both in-plane and out-of-plane magnetizations and additionally activates transverse AMR via the second harmonic of the secondary beam. Comparable static magnetic fields would require hundreds of tesla, avoided in Floquet driving. These effects persist across linear, circular, and mixed light polarizations and are tunable via light parameters. Our results establish multi-color Floquet engineering for controlling magnetization and symmetry-protected transport in AMs.

We investigate the magnetic properties of a monolayer of Fe3GeTe2 as a function of the lattice constant by combining first-principles calculations with atomistic spin dynamics simulations. The calculated magnetic exchange interactions reveal a competition between ferromagnetic and antiferromagnetic couplings, with the latter being significantly strengthened under compressive strain. Stochastic Landau-Lifshitz-Gilbert simulations reveal a substantial decrease in the Curie temperature with decreasing lattice constant, and predict a transition of the magnetic ground state from a ferromagnetic configuration to a conical spin-spiral state. We introduce a simple spin-model which explains the stabilization of the spiral phase due to competing exchange interactions. We found multiple magnetic phase transitions involving ferromagnetic, conical spin-spiral, and planar Neel states, depending on both the lattice constant and the temperature. The absence of Dzyaloshinskii-Moriya interactions is found to significantly reduce the Neel temperature, while leaving the Curie temperature largely unaffected. Our findings reveal the importance of lattice distortions in controlling complex magnetic phases and their evolution with temperature.

Fe$_3$GaTe$_2$ is van der Waals (vdW) ferromagnet with a Curie temperature $T_C$ ranging from 350 K to 380 K, followed upon cooling by a ferrimagnetic transition near room temperature. Substituting Fe with Co was previously reported to induce antiferromagnetism (AFM) at a Co fraction dependent Neel temperature $T_N$. In this work, we confirm the overall phase diagram of the Fe$_{3-x}$Co$_x$GaTe$_2$ series as a function of $x$ and temperature via magnetization and electrical transport measurements. For $x \simeq 0.6$ the Hall effect is observed to mimic the magnetization as the AF ground state is suppressed by the external magnetic field via a metamagnetic transition, thus displaying an anomalous Hall response. At low temperatures, we also observe a pronounced topological Hall signal peaking at $μ_0H$ = 4 T, or within the metamagnetic transition region of fields. This observation points to the presence of magnetic field-induced chiral spin textures, such as skyrmions upon approaching magnetization saturation. Magnetic force microscopy (MFM) reveals the emergence of nearly circular magnetic domains, with diameters on the order of 100 to 200 nm, within the antiferromagnetic phase. A detailed analysis of the MFM images indicates that the topological Hall effect is closely linked to the field-induced stabilization of magnetic domain structures, likely exhibiting chiral textures. This observation suggests the possible formation of skyrmions already in the AFM phase, i.e., AFM skyrmions, that evolve into ferromagnetic (FM) ones upon increasing the magnetic field. Consequently, Co-doped Fe$_3$GaTe$_2$ might provide a platform to investigate the transformation of skyrmions, initially coupled antiferromagnetically into ferromagnetic skyrmions, and to explore its impact on the topological and skyrmion Hall effects.

Ferroelectric nematic (NF) liquid crystals exhibit polar symmetry and large polarization, giving rise to phenomena absent in conventional apolar nematics. We investigate NF liquid crystals confined to free-floating spherical shells with tangential boundary conditions, enforcing a total topological defect charge of +2. We conjecture that ferroelectric nematics avoid splayed configurations with half-integer defects, common in apolar nematic shells, instead concentrating the topological charge into escaped azimuthal +1 defects requiring only bend and twist. Indeed, at room temperature in the NF phase, our thin RM734+DIO shells with inner and outer aqueous poly(vinyl alcohol) solutions develop an azimuthal director field around two antipodal +1 bend-twist defects. The non-centrosymmetric nature and the azimuthal director configuration of the shells in the NF phase are confirmed also through second-harmonic generation microscopy. At intermediate temperature the antiferroelectric Nx phase generates a new exotic texture rife in zigzag lines in the shells. In the regular N phase at high temperature, the shells develop the usual four +1/2 disclinations located near the thinnest point. Our study highlights the rich platform offered by spherical shells to study the behavior of exotic liquid crystals subject to topological constraints, possibly opening new paths to apply the highly responsive ferroelectric nematic phase

The superconducting diode effect offers a powerful probe into the fundamental symmetries of quantum materials. Recent studies on twisted graphene diodes have predominantly focused on bilayer or trilayer systems under out-of-plane magnetic fields. Here, we demonstrate both out-of-plane and in-plane driven Josephson diode effects in a magic-angle twisted four-layer graphene junction, i.e., an even number of layers. We observe the emergence of a diode effect at zero out-of-plane field, tuned by an increasing in-plane magnetic field. This result points to the presence of strong in-plane orbital coupling, which is highly sensitive to the specific layer parity of the structure. Our findings provide experimental insights into the symmetry-breaking mechanisms of even-layer twisted graphene, establishing in-plane magnetic fields as a vital tool for unravelling their microscopic properties.

Can synchronization -- the widespread spontaneous emergence of coordinated dynamics in classical nonlinear systems -- also occur in the quantum regime? This question has recently sparked intense research into collective behavior and temporal self-organization in quantum systems. Typical quantum features such as the linearity of time evolution and the presence of quantum noise seem to hinder the appearance of synchrony at the microscopic level. At the same time, quantum coherence and quantum correlations may provide novel mechanisms for enhancing synchronization beyond its classical counterpart. We here survey recent theoretical and experimental advances in quantum synchronization, ranging from the characterization of synchronous oscillations and genuinely nonclassical forms of synchrony to many-body synchronization on quantum networks.

Internally heated LCST thermo-responsive gels can show self-sustained swelling and collapse oscillations through feedback between temperature-induced collapse and collapse-suppressed heating. In this work, a minimal two-variable model is developed by coupling gel swelling dynamics with a lumped thermal balance. The analysis shows that stable large-amplitude oscillations are mainly controlled by global bifurcations of limit cycles, rather than by the local Hopf bifurcation. The Hopf bifurcation is subcritical in the studied parameter range, leading to a broad coexistence region where a stable fixed point and a stable limit cycle are both possible. The oscillatory behavior remains robust for different heating-gate functions, indicating that local linear instability is neither necessary nor sufficient for self-oscillation. Fast-slow analysis further shows that the oscillation period is mainly governed by the cooling rate, while the amplitude is determined by the geometry of the swelling equilibrium manifold. These results clarify the bifurcation mechanism of thermo-responsive gel oscillations and provide guidance for controlling their period, amplitude, and waveform.

We develop a band-structure-based theory of exciton-polaritons in a three-dimensional inverse-opal photonic crystal doped with semiconductor quantum dots. Starting from a symmetry-selected bright photonic branch near the photonic gap edge, we construct an exciton-photon Hamiltonian and obtain a lower-polariton band with a W-point global minimum and a nearby X-point van-Hove-enhanced density of states. We show that the W valleys determine the equilibrium Bose-Einstein condensation threshold, while the X-point saddle provides a finite excited-state capacity that renormalizes the critical temperature when the W-X offset is thermally accessible. By tuning the exciton resonance and the light-matter coupling, the relative W-X ordering can be reconstructed, leading to a strong variation of the critical temperature. We further formulate a momentum-resolved Boltzmann model for driven-dissipative kinetics. Under non-resonant pumping, reservoir feeding, radiative decay, and inter-sector relaxation can produce either W-dominated condensation, a mixed W-X regime, or an X-dominated nonequilibrium coherent state. Our results establish three-dimensional photonic-crystal polaritons as a platform where condensation is controlled not only by the band minimum but also by valley geometry, van-Hove-enhanced phase space, and relaxation pathways.

We study plasmonic excitations in twisted bilayer graphene within a non-Hermitian framework that incorporates effective gain and loss. Using a non-Hermitian extension of the Bistritzer--MacDonald continuum model together with a biorthogonal Kubo formalism for the optical conductivity, we determine how the moiré electronic structure enters the plasmonic response of the active bilayer. We find that non-Hermiticity modifies the collective spectrum, yielding optical and acoustic plasmon branches, with the acoustic branch exhibiting strong subwavelength confinement. In the parity-time-symmetric configuration, gain--loss engineering can reduce the effective spatial damping and enhance the propagation length within the ideal linear model. The same regime produces strongly localized transverse-magnetic near fields. We argue that the enhancement is not a generic consequence of adding gain to a bilayer, but results from the combined influence of moiré-band reconstruction, biorthogonal optical matrix elements, and non-Hermitian modification of the plasmon pole. We also discuss the limitations imposed by disorder, substrate loss, gain saturation, and stability of the parity-time-symmetric regime. These results identify twisted bilayer graphene as a promising, but experimentally demanding, platform for tunable non-Hermitian plasmonics in moiré quantum materials.

Attractive three-body interactions have been reported for like-charged colloids in low-salt suspensions, based on both finite-element Poisson-Boltzmann calculations and direct experimental measurements, and have been proposed as a mechanism to drive colloidal clustering. However, these Poisson-Boltzmann calculations typically neglect charge regulation and higher-order many-body effects. Here, we construct machine-learned (ML) many-body interaction potentials for charge-regulating colloids, trained on finite-element Poisson-Boltzmann calculations, to accurately capture three-body and higher-order contributions. We find that the three-body contribution to the many-body potential as obtained from Poisson-Boltzmann calculations on isolated colloid triplets is strongly attractive, consistent with previous work, whereas the four-body contribution for an equilateral pyramid configuration of four colloids is repulsive. We then construct ML many-body potentials for charged colloids using finite-element Poisson-Boltzmann calculations on clusters of 13 colloids, and find that the incorporation of higher-body interactions weakens the cohesive nature of the interactions. We identify a parameter regime exhibiting gas-liquid or gas-solid phase separation using the ML potentials in molecular dynamics simulations. However, when we include clusters of 48 colloids in the training data, the cohesion diminishes further, and molecular dynamics simulations using these potentials no longer include broad phase separation in aqueous dispersions of charged colloids. Finally, we compute the potential of mean force of pairs and triplets of colloids using primitive model simulations. We find that the resulting potentials are in good agreement with those obtained from the Poisson-Boltzmann calculations, thereby supporting the validity of the Poisson-Boltzmann approach for determining many-body interactions.

Low-dimensional ferroelectrics are attractive for their promising prospects in nanoelectronics. Compared with widely-used ferroelectric perovskites, most low-dimensional ferroelectrics exhibit several inborn weaknesses such as small bandgaps (mostly <2 eV, i.e. semiconductors-like) or faint polarizations (e.g. $<1$ $μ$C/cm$^2$ for sliding ferroelectrics even if their bandgaps can be large). Here we experimentally demonstrate the room-temperature ferroelectricity of van der Waals WO$_2$Cl$_2$ . The well-tested d0 rule inherited from ferroelectric perovskites leads to a large dipole (~3 eÅ) from the off-center displacement of W$^6+$ ion and a wide bandgap of 2.80 eV. Its ferroelectricity is proved by multiple characterizations including second harmonic generation, piezoresponse force microscopy, and ferroelectric hysteresis loops. More interestingly, the exotic noncollinear dipole order is directly observed at the atomic level by integrated differential phase contrast scanning transmission electron microscopy. Our work paves an alternative route for low-dimensional ferroelectrics to pursue excellent ferroelectric performance and distinct physics of polarity.

Sums of independent exponential random variables lead to the Erlang distribution, providing a direct probabilistic route from exponential waiting times to the integer-shape gamma law. This paper investigates how this classical construction changes when the exponential waiting-time density is replaced by the $q$-exponential density of nonextensive statistics. Our main result is an analytical asymptotic expression for the outflow of a mass-conserving cascade of reservoirs driven by a $q$-exponential waiting-time kernel. In the critical case $q=5/3$, the large-cascade flow rate converges to a stable Lévy density whose time argument is shifted by a Galilean-type transformation. This shifted Lévy law gives the asymptotic hydrograph of the cascade. We also found that for the entire regime $1<q<2$ the macroscopic dynamics are governed by $α$-stable Lévy laws. This proves that anomalous non-Gaussian dispersion can emerge from pure mass-conserving convolutional chains without invoking fractional derivatives.

Periodically strained graphene on patterned substrates provides a versatile route to realizing moiré-like electronic structures through strain engineering. Here, we show that the interplay between a strain-induced pseudomagnetic field and a displacement-field-controlled scalar potential enables the formation of isolated narrow bands and multiple energy gaps near charge neutrality and at higher energies. Some of the low-energy bands exhibit nontrivial topology, carrying valley-opposite Chern numbers. Remarkably, despite a vanishing total Chern number, propagating in-gap edge states emerge in a wide range of nanoribbon geometries that preserve valley symmetry. We elucidate the distinct mechanisms responsible for edge states in the zero-energy and higher-energy gaps and demonstrate that they remain robust against variations in superlattice termination and moderate disorder, despite lacking conventional topological protection. Leveraging these properties, we propose device architectures in which an externally applied staggered potential electrically switches the zero-energy gap and its associated edge channels on and off. Furthermore, split-gate geometries generate topologically protected interfacial states that coexist with the edge modes and can be spatially reconfigured by gate voltages. These results establish strain superlattices as a powerful platform for engineering topological electronic states and electronic transport in graphene.

Metallic delafossite oxides have drawn attention for their ultralow resistivity and coherent electronic transport. However, mesoscopic transport studies are hindered by the limited access to high quality nanoscale devices. Here, we report single crystalline PdCoO2 nanomembranes, enabling exploration of quasi two dimensional (2D) transport. Shubnikov de Haas and Aharonov Bohm oscillations under different magnetic field orientations are observed at low temperatures, from which the electron effective mass and electron phase coherence length are extracted. Beyond quantum transport, the electrical performance of PdCoO2 toward interconnect applications is evaluated. A nearly thickness independent room temperature resistivity is observed for flakes down to 40 nm thickness. The nanomembranes exhibit a breakdown current density up to 113 MA cm-2, with excellent thermal stability and electromigration resistance. These results demonstrate that mechanically exfoliated PdCoO2 flakes preserve the high crystalline and electronic quality of bulk crystals in the quasi-2D limit, providing a useful platform for mesoscopic transport studies and interconnect applications.

Strong electron-electron interactions underlie a wide range of quantum many-body phenomena, including magnetism, superconductivity, and charge fractionalization. A central goal is to achieve in situ control over lattice geometry, bandwidth, and band topology within a single platform. Here we realize such an electrically programmable quantum many-body system in an alternating twisted trilayer MoTe$_2$, where an out-of-plane displacement field continuously modifies the layer polarization, effective lattice, and topology of the moiré bands. At zero displacement field, the system realizes a triangular lattice hosting a correlated insulator at one hole per moiré unit cell ($ν= -1$). Doping this state produces strongly asymmetric magnetic responses: double-exchange-like ferromagnetism for $|ν| > 1$, and signatures of spin polarons and antiferromagnetism for $|ν| < 1$. At large displacement field, interlayer hybridization reconstructs the electronic structure into a honeycomb lattice with a flat Chern band, supporting integer and fractional Chern insulators. Magneto-optical measurements further reveal the signatures of gap closure and Landau-level formation from a spin-polarized Fermi surface near the crossover between the two regimes. These results establish a unified, electrically tunable platform in which correlated magnetism and topological states emerge from a single controllable band structure.

The Mpemba effect is commonly associated with anomalous thermal relaxation, in which a system prepared at a higher temperature reaches equilibrium faster than one prepared at a lower temperature. In modern formulations, however, its defining feature is broader: a state initially farther from equilibrium can relax faster than a closer one, or, equivalently, relaxation trajectories can cross. Here, we show that magnetic-field-dependent delayed fluorescence in triplet-fusion systems realizes a generalized Mpemba effect driven by external control of spin-selective kinetics. Using a diffusion-controlled geminate-recombination extension of the Johnson--Merrifield model, we demonstrate that delayed-fluorescence trajectories obtained at different magnetic fields cross when geminate fusion is effective. We further derive a compact kinetic criterion for such crossing in terms of the competition between intermediate-time decay rates and long-time power-law amplitudes. This criterion captures the interplay between effective fast and slow relaxation contributions. Within this kinetic framework, the final state is independent of the external control parameter and is defined as the fully relaxed state in which no separated triplet population remains. The generalized Mpemba effect therefore arises from the redistribution of transient relaxation pathways associated with geminate recombination, whereas bulk diffusion-controlled processes contribute only to the final relaxation. These results link anomalous relaxation in spin-dependent photokinetics to modern formulations of the generalized Mpemba effect and show how trajectory crossing can emerge in systems governed by diffusion-controlled geminate recombination and non-Markovian kinetics.

Hybrid systems consisting of color centers and magnetic materials provide an appealing solid-state platform for advancing the burgeoning quantum technological revolution. Exploring novel coupling mechanisms between optically active spin defects and quantum degrees of freedom is directly relevant in this context. Here, we report observation and control of spontaneous magnon emission from boron-vacancy centers in 2D hexagonal boron nitride (hBN), an unconventional qubit-magnon dipole coupling channel that dominates in the near-zero temperature limit. The spontaneous magnon emission process starts to be overshadowed by thermal magnon effect as temperature increases, reflecting the crossover from an emission-dominated, effectively cold magnon reservoir to a thermally occupied spin bath where absorption and stimulated processes restore balance. By increasing the spin defect density, we further present that spontaneous magnon emission into a common spin bath could help establish quantum correlations in dense hBN spin ensembles. Our results are quantitatively captured by detailed theoretical modeling, bringing insights into understanding qubit-magnon coupling, correlated spin dynamics, and many-body physics of color centers in the quantum regime.

Diagnosing hidden local orders at buried interfaces remains a central challenge in the design and characterization of quantum materials. Second-order electrical responses, such as the nonlinear Hall effect, probe inversion-symmetry-breaking terms invisible to linear transport, offering a direct window into these nanoscale environments via the quantum geometry of Bloch electrons. Here, we utilize $\text{KTaO}_3$, a complex oxide driven by strong tantalum $5d$ spin-orbit coupling and interfacial inversion symmetry breaking, to demonstrate that second-harmonic resistivities exhibit large, reproducible mesoscopic fluctuations. Remarkably, these fluctuations persist in macroscopically large ($200\,μ\text{m}$) devices and are ubiquitous across all studied surface orientations, even where macroscopic conductivity strictly adheres to underlying crystal symmetries. We propose that these robust, magnetic-field-driven interference patterns arise from local structural symmetry breaking, driven by incipient ferroelectric polarization pinned to the interfacial impurity landscape. This defect-pinned polar mechanism is firmly supported by the signal's suppression above $10\text{ K}$ due to phase decoherence, and a complete loss of mesoscopic memory upon thermal cycling above $40\text{ K}$. By linking quantum geometry to dynamic lattice ordering, our findings establish nonlinear mesoscopic transport as a powerful new characterization tool, capable of revealing local polar tendencies and hidden structural orders in complex materials that remain fundamentally invisible to conventional probes.

Strong coupling between molecules and electromagnetic fields has emerged as a powerful strategy to modify the physical and chemical properties of molecules, enabled by the formation of hybridized energy levels through strong light-matter interactions. Rather than purely photonic or plasmonic modes, hybrid cavity fields that integrate plasmonic and photonic contributions provide versatile platforms with distinct advantages for tailoring light-matter interactions. Here, we present hybrid modes formed through the coupling between localized surface plasmon resonances of Au nanoparticles and Fabry-Perot microcavity modes. Furthermore, by using this hybrid platform, we demonstrate that the cavity field can give rise to cooperative coherent coupling between the spatially separated plasmonic nanoparticles and the molecules by mediating long-range dipole-dipole interactions. As we show, this cavity-mediated coupling also enhances the plasmon-exciton mixing, as compared to their direct interaction. This configuration opens new avenues for tailoring light-matter interactions at the nanoscale by supporting novel hybrid plexcitonic states that incorporate contributions from plasmons, excitons, and extended cavity fields.

Segregated states based on particle size emerge in granular materials from the competition between segregation and diffusive remixing. Here, we show that quantum coherence can enhance segregation beyond this classical limit. We introduce an open quantum cellular automaton for bidisperse mixtures that combines coherent transport and dissipative segregation. The automaton reproduces experimental and continuum-theory segregation dynamics, with segregation degrees collapsing onto a theoretical Péclet-dependent relationship. However, weakly decohering systems exhibit a coherence-driven transport regime that produces more strongly segregated steady states than classical predictions. Across a broad parameter range, the steady-state degree of segregation collapses onto two dimensionless numbers governing the competition between segregation, diffusion, and decoherence. These results identify quantum coherence as a mechanism for enhancing particle-size segregation and establish a framework for studying transport phenomena in open many-body systems.

Superconducting resonators are finding increasing applications in designing advanced quantum circuits for ongoing sensing, metrology, and computing technological revolution. A detailed knowledge of microscopic electromagnetic properties of superconducting resonators is directly relevant for their further improvements on circuitry design and device performance. Here, we introduce scanning-probe quantum microscopy to report nanoscale sensing of microwave and static magnetic field environment of an on-chip niobium (Nb) superconducting resonator. Taking advantage of Rabi oscillation measurements, we show that microwave magnetic fields generated by the superconducting resonator mode can be utilized to achieve coherent control of a quantum spin sensor. We further visualize static electromagnetic field response of the Nb resonator, showing magnetic field-induced formation, evolution, and depinning of superconducting vortices. Our results provide insights into future design, testing and evaluation of solid-state superconducting resonators, highlighting the potential of quantum sensors as a local probe to investigate electromagnetic properties of superconducting quantum circuits.

The interplay between symmetry and thermalization governs the late-time dynamics of local quantum and classical many-body systems at nonzero temperature. Recently, two parallel frontiers have emerged: the search for robust anomalous hydrodynamics -- such as superdiffusion -- in generic, non-integrable models, and the formal effort to generalize the fundamental concept of global symmetry. In this paper, we bridge these frontiers by demonstrating that quantum group symmetry provides a novel mechanism for anomalous hydrodynamics in chaotic systems. We study the dynamics of local operators carrying $U(1)$ charge in non-integrable lattice models that also have quantum group symmetry. One example is transverse spin in the XXZ model with integrability breaking deformations. While such excitations are expected to decay very quickly at high temperature because their charge forbids overlap with conventional hydrodynamic densities, we find that protection by the quantum group symmetry makes these modes long-lived, despite the absence of local quantum group charge density or current. Furthermore, the dynamics is superdiffusive across Hamiltonian, Floquet, and classical realizations, and exhibits unusual finite size effects at very late times. We also revisit transverse spin dynamics in the integrable XXZ model.

Motivated by the surge of interest in statistics of non-Hermitian random matrices as a framework for description of universal characteristics of dissipative chaotic quantum many-body systems, we address the problem of characterizing the parametric correlations of spectral densities. Considering parameter-dependent ensemble of complex Ginibre matrices we derive an explicit, closed-form expression for the parametric number covariance in the systems of symmetry class $\mathbf{A}$ for eigenvalues in a circular domain containing on average a finite number of eigenvalues in the spectral bulk. This behavior is expected to be universal, as further supported by numerical evidence for the real Ginibre ensemble, non-Hermitian Bernoulli Wigner matrices and bi-unitarily invariant ensembles. We also discuss a relation between parametric correlations of spectral densities and the distribution of the so-called eigenvector non-orthogonality factor, which attracted considerable interest in recent years.

A systematic method to find the Yang-Lee partition function zeros of quantum many-body systems based on perturbation theory at finite temperatures was recently introduced in arXiv:2504.01880. This method identifies wave-vector and temperature-dependent complex virtual energies obtainable from the thermal electronic Greens function. The collection of virtual energies over all $\vec{k}$ yield the Yang-Lee zeroes. We apply this approach to the one-dimensional Hubbard model for different boundary conditions. We compare the results obtained by this method up to second-order in perturbation theory with the results found by exact diagonalization. We also propose a quantity that could be used for experimental detection of these zeros of a Hubbard ring. An example of the detection method is presented using exact diagonalization of the 8-site Hubbard ring.

In non-Markovian systems, distinct dynamical phases arise from the competition between internal memory and external forcing, encompassing thermalization, persistent ergodicity breaking, and runaway energy growth. This study shows that the scaling parameter $η$ governs the emergent phase diagram within a system described by the Generalized Langevin Equation, particularly when subjected to external drives with asymptotic power-law tails. Three universal regimes for diffusive processes are delineated by this parameter: thermalization ($η> 0$), non-ergodic saturation ($η= 0$), and a force-dominated runaway phase ($η< 0$). The fluctuation-dissipation theorem, within this framework, is shown to be independent of external force and determined by the integral of noise density of states. A selective breaking of ergodicity is revealed by this formulation; microscopic fluctuations are decoupled from the drive, yet the relaxation completely encodes it, which in turn controls the kinetic effective temperature. Direct Langevin simulations in the Markovian limit quantitatively confirm this classification, capturing the non-thermal plateau at the critical point.

Motivated by the role of the holographic entropy cone in constraining the entanglement structure of states with classical gravitational duals, we investigate combinations of subsystem complexities associated with reduced density matrices in multipartite quantum systems. Focusing on subsystems composed of three disjoint regions, we introduce two quantities: a tripartite complexity, inspired by the tripartite information, and a complexity gap, designed to characterize emergent complexity in the full quantum state beyond that of its constituents. We study the sign structure of these quantities in three selected approaches to subsystem complexity. In holography, we employ the complexity=volume proposal in AdS spacetimes; for Gaussian many-body states, we use Fisher-Rao subsystem complexity; and we further develop a Krylov-space inspired, effective framework for reduced density matrices, which we test in few-qubit systems and coherent-state dynamics. Across all three approaches, we find that the tripartite complexity is not sign-definite in general. By contrast, the complexity gap exhibits a definite sign in every example we analyze, although the sign itself depends on the underlying notion of subsystem complexity. Our results suggest that the complexity gap could be a natural candidate building block for a prospective hierarchy of subsystem complexity inequalities.

Bacteria often swim in complex environments where surfaces are ubiquitous and rarely flat. Surface topography and curvature can strongly affect bacterial motility, with important consequences for surface exploration, adhesion, and biofilm formation. Here, we investigate the swimming of a non-tumbling Escherichia coli bacterium near an undulating no-slip surface using hydrodynamic simulations of a detailed model bacterium. The latter is described by a rigid spherocylindrical cell body and flexible flagella modeled with the Kirchhoff rod theory, while the surrounding fluid is simulated using the method of multi-particle collision dynamics. At low curvatures of the sinusoidal surface modulations, the bacterium exhibits persistent near-surface swimming and clockwise trajectories, consistent with the known behavior near flat no-slip walls. As the curvature increases, bacteria swimming toward a ridge can escape from the surface, which we use to estimate a critical curvature where surface detachment is more likely. At larger curvatures, we find that the surface geometry promotes oscillatory swimming along the groove direction, which reduces escape opportunities and, therefore, enhances bacterial trapping. Indeed, the confinement around the groove reverses the swimming of the bacterium from clockwise to counter-clockwise, as we demonstrate by two minimal models. Thus our work highlights the importance of the three-dimensional surface topography in bacterial surface exploration.

We revisit the topic of false vacuum decay in field theory. We focus on a toy model of a real massive scalar field with an unstable quartic potential. This model has a false vacuum, and decay out of the false vacuum can be described via the method of constrained instantons, which work by introducing a constraint on the path integral. We identify and develop three different asymptotic limits which enable analytic construction of approximate {constrained} solutions. The first, in which the constrained solution is small compared to the inverse mass of the scalar field, is an application of the perturbative methods of Affleck, although we re-derive the main results and identify several terms which were previously neglected. Second, for very large constrained solutions we adapt the thin-wall approximation of Coleman. However, we find that the large instanton limit does not always exist. In this case we identify another useful limit, in which the Lagrange multiplier used to implement the constraint is large. In this limit, the solution's scaling with the parameters may be found via dimensional analysis and an exact solution is obtained with a single numerical computation.

We numerically verify multifractal theory (Frisch and Parisi 1985) for turbulence using simulation data at a high Reynolds number. First, we propose a simple method to directly estimate the multifractal dimension $D(h)$ of vortical structures with a given Hölder exponent $h$. Thus measured $D(h)$ is in good agreement with indirectly measured experimental data. Then, we demonstrate that these structures for $h\ll1/3$ form the hierarchy of coherent eddies, while those for $h\gg1/3$ are featureless.

In the focusing Manakov system, multiple modulation instability (MI) branches coexist on the same plane wave background, so the usual weak periodic modulation cannot selectively excite a single vector Akhmediev breather (AB). Here we propose an eigenvector-based initial perturbation scheme that constructs the initial condition as a plane wave plus Fourier modes whose coefficients follow the perturbation eigenvector of a selected MI branch, enabling controllable high-fidelity excitation of desired vector ABs. Numerical simulations show near-100\% fidelity with the exact AB solution. The underlying mechanism is eigenvector-controlled mode selection. The initial seeding of the target MI branch through the chosen eigenvector, together with the non-Hermitian coupling inherent in the linearized MI dynamics, ensures that the targeted unstable mode dominates the early linear stage and thereby dictates the breather type. This eigenvector-based control succeeds in gain-balanced regimes and when the targeted branch has a sufficient gain advantage. The proposed method provides a simple and robust framework for controllable generation of vector ABs over a broad parameter range, highlighting the key role of eigenvector selectivity in multi-component nonlinear systems.

We show that compartmentalizing a single molecular species into intracellular and extracellular fields, and coupling them through membrane transport or nonlinear basal production rates, can produce diffusion-driven (Turing) instabilities. By linearizing the two-field system, we derive the corresponding Turing conditions under which such instabilities may arise. We present three biologically motivated examples that satisfy these conditions and demonstrate the resulting spatial patterns through numerical simulations. These results indicate that tissue compartmentalization alone might enable pattern formation traditionally attributed to multi-species systems.

Altermagnets host momentum-dependent spin splitting without net magnetization, a symmetry-enforced band phenomenon whose photonic analogues have so far been realized only in square lattices governed by fourfold rotation. Here we introduce a photonic altermagnet on a hexagonal lattice whose helicity splitting is governed by mirror rather than rotational symmetry. Elliptical chiral elements of alternating handedness, placed at the vertices of a regular hexagon, leave the two opposite-chirality sublattices connected only by chirality reversal combined with a mirror reflection. Full-wave simulations reveal mirror-related splitting of the two opposite-helicity branches in the band structure and isofrequency contours, with the channels exchanged when the ellipse orientation is reversed. In finite structures, this splitting separates a linearly polarized beam into handedness-resolved channels, enabling beam splitting and direction-selective helicity filtering with target-helicity output fractions above 0.85 and output paths continuously tunable through the ellipse rotation angle. These results extend photonic altermagnetism to a previously unexplored lattice-symmetry class and establish mirror-symmetric chiral textures as building blocks for altermagnetism-inspired chiral photonics.

Precise temporal delay generation is a key requirement for asymmetric Mach-Zehnder interferometers (aMZIs) used in high-speed quantum key distribution (QKD) receivers. In this work, a compact integrated aMZI architecture based on a silicon nitride on silicon dioxide (Si3N4/SiO2) photonic platform is presented. A 3-dB directional coupler enabling accurate 50:50 power splitting at the 1550 nm telecommunication wavelength is designed and optimized. The coupling length is initially estimated from the effective index difference between the even and odd supermodes and subsequently refined using three-dimensional eigenmode expansion (EME) simulations. The optimized structure employs single-mode Si3N4 waveguides with a cross section of 1 um x 0.4 um, providing a group index close to 2 and enabling accurate delay engineering. Spectral analysis demonstrates stable 3-dB power splitting across the C-band with insertion loss below 0.5 dB and negligible power imbalance, indicating high transmission efficiency and structural symmetry. An integrated Si3N4 aMZI delay line providing a 500 ps optical delay, corresponding to a 2 GHz free spectral range (FSR), is further demonstrated. Simulations show nearly constant group delay across the 190-200 THz frequency range with sub-10 ps variation and a smooth, near-linear phase response. These characteristics enable interference visibility above 0.99, corresponding to an estimated quantum bit error rate (QBER) below 0.5 percent for gigahertz-rate time-bin QKD systems. The wideband linear phase behavior also indicates compatibility with wavelength division multiplexed (WDM) QKD architectures. The results confirm that the proposed Si3N4 integrated aMZI provides a low-loss, dispersion-controlled, and spectrally stable delay solution suitable for scalable chip-scale quantum photonic receivers.

Sustainability transitions unfold through interacting dynamics across social, technological, economic, environmental, and governance dimensions. However, many modelling approaches either isolate subsystems or rely on optimisation-based pathways that do not explicitly represent feedbacks, path dependence, and institutional constraints. This study develops a Sparse Orthogonal Regression Technique (SORT), a data-driven dynamical-systems framework for reconstructing multi-system transformation dynamics from harmonised European indicators. Inspired by Sparse Identification of Nonlinear Dynamical Systems (SINDy), SORT infers parsimonious cross-domain dependencies from observed data rather than specifying a full structural model ex ante. The prototype covers energy, emissions, innovation, digitalisation, resource productivity, environmental stress, policy, finance, wellbeing, and resilience. The compact dynamical system is interpreted as a structural representation of medium-term interactions, not as a forecasting tool. It reproduces differentiated empirical patterns, including steady renewable expansion, nonlinear scaling in transition finance, diffusion-like digitalisation, oscillatory environmental stress, and sustained reduction in ETS-regulated emissions. Forward simulations diverge from simple linear extrapolation where reinforcing feedbacks are present, illustrating the conditional nature of accelerated decarbonisation. By translating inferred dependencies into a feedback structure, the analysis identifies reinforcing decarbonisation sequences alongside balancing ecological constraints. The model contributes to sustainability transitions research by providing an empirically grounded representation of multi-system feedback dynamics that bridges quantitative modelling and transition theory. SORT offers a compact technique for exploratory reconstruction from real-world transition data.

Perovskite oxides ABO$_3$ host a rich interplay of charge, spin, lattice, and orbital degrees of freedom, giving rise to diverse quantum phenomena. In low-dimensional ABO$_3$, reduced symmetry can induce exotic quantum effects such as the two-dimensional electron gas and unconventional superconductivity. Using first-principles density functional theory, tight-binding modeling, and symmetry analysis, we show that ultrathin two-dimensional (2D) ABO$_3$ films -- exemplified by SrTiO$_3$ -- naturally break inversion symmetry, producing a spontaneous out-of-plane bulk photovoltaic (BPV) effect. This differs from previous studies on in-plane BPV current signals and is more applicable and experimentally detectable. Such an effect is highly tunable via thickness, strain, surface termination, crystallographic orientation, and Moiré twisting. These findings are broadly applicable to a wide range of 2D perovskite and other layer-resolved oxides.

Microcombs in normal-dispersion photonic crystal ring resonators (PhCRs) are a versatile building block for next-generation integrated photonic circuits, yet they inherently suffer from a directional bias that favors backward-propagating states. This necessitates bulky, non-integrated optical circulators for comb extraction, creating a significant bottleneck for full on-chip integration. In this work, we demonstrate a deterministic method to control and reverse this directionality through Side-mode Induced Forward Forcing (SIFF). By engineering auxiliary mode splittings on resonances adjacent to the pump, we show that the nonlinear dynamics can be steered to favor stable, forward-propagating platicon states. We establish an optimal synchronization condition between the pump and side-mode coupling rates that ensures forward-comb dominance across a wide parameter range. Our findings, validated both numerically and experimentally, provide a critical pathway for circulator-free, integrated normal-dispersion microcombs, offering a scalable architecture for compact telecommunications and sensing systems.

Multi-aperture reception can provide spatial diversity in free-space optical (FSO) communication by collecting signal replicas at separate apertures. When the branches are accurately phase-aligned, their received optical fields can also be added constructively to obtain coherent-combining gain. In this paper, we propose blind gradient-ascent phase alignment (BGAPA), which iteratively adjusts one phase correction per aperture by directly maximizing the combined output power. Closed-form analytical gradients provide a deterministic update that requires no symbol decisions, unlike the stochastic perturbation-based estimate of SPGD or the decision-directed feedback of DD-LMS. To isolate phase-tracking capability, the numerical model includes independent aperture-dependent phase disturbance but excludes amplitude scintillation and polarization-dependent distortion. Under this controlled phase-only setting, BGAPA obtains an SNR improvement closer to the ideal 6.02~dB coherent-combining gain than block-wise cross-correlation, SPGD, DD-LMS, and CMA/RDE-based equalization when the aperture count is increased by a factor of four. In particular, increasing the aperture count from 64 to 256 yields an SNR improvement of about 5.7~dB. In a separate amplitude-tolerance test with $N=16$ and $f_{\max}=1$~MHz, the first observed BGAPA trial above the HD-FEC threshold of $3.8\times10^{-3}$ occurs at an actual phase RMS of approximately 278~rad, whereas DD-LMS becomes unreliable at substantially smaller phase excursions. The reported step size is optimized separately at each operating point. BGAPA is fully blind and updates its phase parameters directly from the received aperture fields without training symbols, pilots, or decision-directed feedback.

Deep learning in computational microscopy is severely constrained by the scarcity of densely annotated datasets. While synthetic data generation has bridged this gap in macroscopic computer vision, traditional graphics engines rely on geometric ray-tracing, failing to capture the micro-optical phenomena required for microscopy. Conversely, while wave-optics formulations exist, rendering them computationally tractable at the scale required for deep learning remains a massive systems challenge. To address this, we introduce the Optical Synthetic Object Generator (OSOG), a high-performance, fully differentiable forward-modeling engine. Drawing on established physical models of diffraction and phase retardation, OSOG maps continuous Optical Path Difference (OPD) calculations into a highly optimized, PyTorch-native Structure-of-Arrays (SoA) architecture. We validate this computational framework across three axes: First, object detection models (YOLOv11-OBB) trained purely on OSOG-generated data achieve robust zero-shot transfer to real-world highly occluded Lysozyme micrographs. Second, we introduce DiffOSOG, demonstrating that the engine's end-to-end differentiability allows for the exact recovery of continuous optical parameters via curriculum-guided inverse rendering. Finally, OSOG bypasses the $\mathcal{O}(N)$ bottlenecks of sequential ray-tracing, demonstrating sub-linear scaling by synthesizing 40,000 complex wave-optic particles in under 50 milliseconds (\>20 FPS). By providing a fast, scalable, and physically grounded tensor pipeline, OSOG enables true real-time, on-the-fly dataset generation.

Unspecific peroxygenases (UPOs) are promising biocatalysts that selectively oxyfunctionalize saturated hydrocarbons using only hydrogen peroxide as a co-substrate. Peroxide-induced enzyme inactivation makes targeted enzyme engineering essential to mitigate this effect and also enhance catalytic performance. To meet this need, systematic approaches are used, including extensive database studies for rational enzyme design, as well as computational enzyme engineering. In this study, we followed the latter strategy and explored the possibility for computationally-guided modification of UPOs. Specifically, our focus was on uncovering the influence of active site amino acids on the catalytic activity of the enzyme CviUPO. Two mutations were introduced close to the active center, and the changes in the energy barriers leading to the activated complex were investigated in detail by Quantum Mechanics/Molecular Mechanics Nudged Elastic Band simulations. Our studies revealed that a change of the glutamic acid, assisting the catalytic cycle, by the shorter aspartic acid, leads to an increased reaction barrier, probably decreasing the catalytic activity of the enzyme. Exchanging the heme-anchoring cysteine group by a histidine exhibited promising behavior as the energy barriers decreased significantly. However, it is possible that the histidine modification also alters the reaction behavior of the peroxygenase, turning it into a peroxidase, an aspect that so far could not be confirmed beyond doubt. Simulations alone cannot conclusively determine whether substrate specificity and reactivity are maintained in the modifications tested. Nevertheless, our results highlight the importance of spin states and active pocket hydration for the catalytic reaction and demonstrate why a synergistic approach of theoretical predictions and experimental verifications is required for efficient enzyme engineering.

Terrestrial coherent FSO requires mitigating turbulence-induced coupling losses. We experimentally validate a "Digital Twin" channel model against fading statistics from a 4.6-km link. Combining long-term scintillometer data with mode-diverse receiver simulations, we demonstrate that a 6-mode receiver reduces turbulence-induced outage probabilities to 2.02e-5.

Strong reflections from metallic and engineered substrates severely limit label-free functional imaging of living cells at biomaterial interfaces, neural electrodes, and implantable devices. Here we introduce two complementary approaches for recovering intracellular dynamic contrast at highly reflective interfaces. Dynamic dark-field full-field optical coherence tomography (D-dFFOCT) suppresses the dominant substrate reflection and restores intracellular visibility through selective detection of scattered light. In parallel, asymmetric illumination generates a distinct directional dynamic contrast that is most consistently interpreted as dynamic reflection differential phase contrast (D-RDPC). Both approaches reveal intracellular activity that remains poorly visible with conventional dynamic full-field optical coherence tomography. D-RDPC exhibits characteristic signatures of phase-gradient imaging, including contrast reversal upon illumination inversion, enhancement with increasing illumination asymmetry, and recovery of spatial localization through directional Hilbert-transform reconstruction. Together, these results establish new strategies for functional imaging at reflective interfaces and suggest that differential phase contrast signals can support temporal fluctuation analysis.

The modeling of large collections of interacting meta-atoms remains a central challenge in photonics because it requires the simultaneous treatment of complex scatterer geometries, multiple scattering, and heterogeneous environments. Here, we introduce the dressed Global Polarizability Matrix (dGPM), a reduced-order electromagnetic framework in which complex meta-atoms are represented by compact scattering operators that explicitly incorporate the influence of nearby interfaces. The dGPM accurately reconstructs near- and far-field electromagnetic responses for interface-coupled structures, including geometries beyond the practical applicability of conventional T-matrix approaches. The framework naturally combines interface-coupled and free-space scatterers within a unified multiple-scattering formalism and enables simulations of large ordered and disordered metasurfaces. More broadly, the dGPM extends operator-based electromagnetic modeling to heterogeneous photonic environments while preserving a compact scatterer-based representation, providing an efficient framework for the simulation and design of large-scale photonic systems.

High-throughput computational screening of nanoporous materials for gas storage and separation requires fast and accurate characterization of adsorption equilibrium. Particle-based grand canonical Monte Carlo (GCMC) and density-based classical density functional theory (cDFT) provide simulation-based estimates of gas uptake and adsorbate density fields, but their speed-accuracy tradeoff remains insufficient for large-scale screening. In this work, we address this gap with Multi-fidelity Amortized Density Field for Adsorption in Nanoporous Materials (MADField), which reframes adsorption prediction as equilibrium density-field estimation. MADField learns from two complementary fidelities, combining broad and scalable cDFT density supervision with higher-fidelity GCMC density labels, and recovers gas uptake by integrating the predicted density field. MADField improves uptake accuracy over the strongest baselines by 6.0x for cDFT and 15.4x for GCMC, and its predicted fields accelerate cDFT solvers with 2.0x fewer iteration steps while recovering 42 percent of cases that fail under standard settings. Finally, we evaluate MADField for conventional CH4 working capacity screening on the 270k-structure ARC-MOF database. Within this space of extremely rare high-capacity targets, 167 in total, the model achieves 56x higher average precision than the strongest baseline and accelerates inference by five orders of magnitude compared to GCMC. By prioritizing the MADField rankings, selecting the top 1.7 percent of candidates recovers 95 percent of all targets, while the top 6 percent ensures 100 percent recall.

Subgrid-scale orography remains a leading source of uncertainty in numerical modeling because terrain spectra must be recovered from irregularly sampled elevation data and then reduced to a flow-dependent launch budget for parameterizations. Existing approaches are limited either by assuming regular samples on rectangular grids or by fitting coefficients whose truncation and regularization effects become embedded in the spectrum. None achieves dynamic, flow-dependent truncation. This study introduces an Elastic Non-Uniform Fast Fourier Transform (ENUFFT) framework that computes local Fourier coefficients directly from irregularly sampled orography on unstructured grids, without interpolation or fitting. It combines a type-1 NUFFT with local windowing, quadrature weights, and a new Elastic Mode Selection (EMS) algorithm for retaining a local flow-dependent subset of modes. ENUFFT is compared with the strongest relevant existing method in a monochromatic and a real Alpine terrain test. In both cases, it recovers peak amplitude and direction comparably while significantly compacting the spectra (monochromatic ~25%, Alpine ~60%). It also satisfies the Parseval condition more closely with its spectral variance (energy) deviating from reference by ~14-24% versus ~500-122,000% for the existing method. Its EMS is additionally tested in a mountain-wave test where it reduces the launch spectrum by >=75% while keeping launch-power loss <=7%. Along with better compute scaling, ENUFFT is thus a computationally efficient, physically interpretable framework that can make Fourier-based orographic source descriptions practical for spectral-budget-aware parameterizations.

Precise pointing control is a critical requirement for interspacecraft laser interferometry, as angular misalignment introduces measurement noise and even leads to laser link loss. We present a nested control architecture that uses differential wavefront sensing signals to drive a fast steering mirror (FSM) to track the incoming beam, while feeding the FSM's angular changes back to the attitude and orbit control system (AOCS) to suppress angle-dependent optical path variations. This scheme is experimentally validated in our hexapod-based setup. Relative to standalone FSM actuation, the nested configuration enhanced pointing stability by 6.9 dB and 4.9 dB in the horizontal and vertical directions across the frequency band from 3 mHz to the AOCS actuation's unity-gain frequency. Additionally, tilt-to-length coupling was suppressed by an order of magnitude below 6 mHz and by two orders of magnitude below 0.45 mHz. These results demonstrate the feasibility of nested active pointing control for future interspacecraft laser interferometry missions.

Imaging objects hidden outside the direct line of sight expands the effective field of view and is critical for applications such as autonomous driving and robotic perception. Despite impressive progress in time-of-flight (ToF)-based non-line-of-sight (NLOS) imaging, real-world deployment remains challenging because practical measurements are often collected over spatially limited, arbitrarily shaped relay regions-conditions that violate the planar-wall and dense-sampling assumptions made by most existing methods. To address these limitations, we propose a LOS-guided NLOS imaging pipeline that imposes no geometric assumptions on the relay surface and naturally supports both confocal and non-confocal configurations. Our method represents the hidden scene using 3D Gaussian primitives and couples them with an efficient, differentiable transient rendering model, enabling end-to-end optimization directly from measured transients. We validate our approach on real-world measurements from both a public dataset and a custom-built capture system. Across settings, our method achieves state-of-the-art reconstruction fidelity under spatially limited, sparsely sampled conditions, and significantly outperforms existing methods on complex, arbitrary relay surface geometries.

The computational discovery of phonon-mediated superconductors is hindered by the prohibitive cost of density functional perturbation theory (DFPT). Here, guided by the metallized $σ$-bonding picture, we introduce the $σ$-bonding density of states ($σ$DOS) as an efficient physical descriptor to identify high-transition-temperature ($T_{\mathrm{c}}$) superconductors from density functional theory (DFT)-level electronic structure without explicit DFPT calculations. The evaluation of $σ$DOS can be further accelerated by a deep-learning DFT Hamiltonian method, enabling efficient large-scale screening for superconductors. Screening 2 million materials, we identify B$_{13}$Se as an ambient-pressure superconductor candidate with predicted $T_{\mathrm{c}} > 40$~K, together with a family of high-$T_{\mathrm{c}}$ B$_{13}X$ candidates, supporting the effectiveness of this discovery strategy. By bridging physics priors with AI acceleration, this study delivers an efficient and generalizable route for computational materials discovery in the AI era.

Purpose: Physics-Informed Neural Networks (PINNs) struggle with stiff, regime-changing transport equations due to instability, loss imbalance, and violations of physical consistency. This paper investigates these failures through the Marshak wave equations - a canonical benchmark from radiative transport - where initial and boundary conditions differ by up to 12 orders of magnitude, and proposes targeted modifications to the standard PINN framework to overcome them. Design/methodology/approach: Three modifications are introduced: (1) a ScaledSigmoid final activation enforcing physical bounds and positivity of the unknowns; (2) a logarithmic MSE loss replacing the standard quadratic loss for initial and boundary conditions, enabling training across extreme scale disparities; and (3) explicit enforcement of global conservation laws derived from the governing equations as an additional physics loss term. Monte Carlo sampling with exponential time weighting is used throughout. Findings: The proposed framework successfully recovers the Marshak wave dynamics - including the hot, cold, and wave-front regions - in agreement with a reference Implicit Monte Carlo solution, with run times under 30 minutes. Ablation studies confirm that each ingredient is essential: linear activation, absence of the logarithmic loss, or removal of the PDE term each independently cause the method to fail qualitatively. Originality/value: This work identifies and resolves three concrete failure modes of standard PINNs on stiff hyperbolic systems with nonlinear coupling. The combination of bounded activations, scale-aware loss functions, and conservation law enforcement constitutes a novel and practically validated framework, with applicability to radiative transport and other coupled stiff PDE systems in engineering.

Topology has become a central concept in understanding physical phenomena, leading to important advances in condensed matter and photonics. Recent work has established a universal upper bound on the energy gap of Chern insulators in electronic systems, revealing a fundamental connection between topology, quantum geometry, and optical absorption. Here, we generalize this framework to photonic systems, deriving rigorous upper bounds on gap Chern numbers without requiring explicit topological analysis. Our approach enables the estimation of the topological charge of band gaps in both dispersive and nondispersive regimes.

Optical frequency combs on photonic integrated platforms are revolutionizing precision metrology, bio-imaging, atomic and molecular sensing, and ultrafast photonics, yet most remain confined to the near-infrared. This restriction prevents access to the ultraviolet, visible, and mid-infrared bands critical for a vast array of quantum, atomic, and molecular systems. The fundamental obstacle has been the lack of a nanophotonic waveguide that simultaneously provides an ultra-broad transparency window, engineered dispersion, ultra-low propagation loss, and a strong Kerr nonlinearity, all while suppressing detrimental two-photon absorption at short wavelengths. Here, we overcome this challenge by exploring tantalum pentoxide for ultra-broadband supercontinuum spanning continuously from the ultraviolet to the mid-infrared, leveraging its broad transparency window (300-8000 nm), a high nonlinear refractive index three times larger than that of silicon nitride, and a wide bandgap that suppresses two-photon absorption. Critically, by using a photolithography assisted chemo-mechanical etching process that avoids a lossy SiO2 upper cladding, we achieve dispersion engineered waveguides with record-low propagation losses of 0.066 dB/cm at telecom wavelengths and 0.43 dB/cm at 780 nm, significantly facilitating the supercontinuum spectral extension into the ultraviolet and the mid-infrared. Pumping these anomalous-dispersion waveguides with femtosecond pulses at 1550 nm yields a gap-free, 3.2-octave supercontinuum spanning from 350 to 3200 nm via a soliton-based dynamics at only 54 pJ pulse energy, representing the broadest comb spectrum on this platform. We further demonstrate a relatively flat spectrum with a -30 dB bandwidth of 1182 nm by engineering normal dispersion, validate the comb coherence via heterodyne detection, and achieve soliton-effect pulse self-compression from 126.7 fs to 19.2 fs.

The invariant-imbedding T-matrix method (IITM) is a standard tool for light scattering by large, sharply faceted, non-axisymmetric particles (atmospheric ice crystals and mineral dust) where the surface-based extended boundary condition method loses accuracy. Its accuracy is limited by "staircasing": the dielectric contrast of a faceted particle is integrated across boundaries that cut the quadrature grid, so standard quadrature converges at low algebraic order. We show that this non-smoothness has a single geometric origin, the tangencies of the integration sphere to the faces and edges of the particle, which produce jumps, kinks, and half-integer branches according to the tangency type, in all three integration directions. A boundary-conformal scheme removes them using closed-form azimuthal coefficients, panel splitting at the analytically known tangency loci, and a square-root substitution $x \mapsto x_c + t^2$ that absorbs the half-integer branches. For a hexagonal prism the azimuthal integration becomes exact and the zenithal and radial directions recover spectral and fourth-order convergence; because the construction depends only on the contact geometry, it extends to any convex polyhedron, demonstrated on the solid hexagonal bullet (a faceted ice habit with tilted faces). The zenithal crossing is a square-root branch rather than a kink, so the established interval-splitting alone gives only $\mathcal{O}(N^{-3})$, while the radial step removes the half-integer edge branch that caps the Riccati recurrence on faceted particles. The convergence orders are fixed by the local contact geometry and verified size-independent up to $k\,r_{\max} = 20$; what grows with size is the resolution needed to reach each asymptotic regime, not the order.

Nonlocal metasurfaces enable all-optical processing of spatial information in optical fields. Here, we demonstrate a topology-optimised metasurface that switches between phase-contrast and brightfield imaging modalities via polarisation control, eliminating the need to reposition optical components or use computational techniques to image transparent samples. Specifically, for one polarisation state, an asymmetric transfer function about normal incidence performs a first order derivative on the phase, producing pseudo-3D phase-contrast images of transparent biological samples while the orthogonal state returns the result of the identity operator. This work extends inverse-design methods to reconfigurable phase-contrast microscopy and quantitative analogue optical computation with applications in biological imaging, medical diagnostics, and materials characterisation.

Individual daily mobility exhibits a striking scaling law: the probability of returning home after a tour of $l$ locations decays as $P_{\rm ret}(l)\sim l^{-γ}$. While the tour-terminate-continue (TTC) model reproduces this behavior, it relies on this power law as an empirical input, leaving the microscopic origin of $γ$ unresolved. Here we show that this scaling emerges from a utility trade-off governed by cognitive constraints. By invoking the principle of least effort, we demonstrate that individual activity priorities follow Zipf's law, $p(r)\sim r^{-ν}$, which directly dictates the sublinear accumulation of tour utility, $U_L(l)\sim l^{1-ν}$. Luce's choice rule then yields $P_{\rm ret}(l)\sim l^{-(1-ν)}$, giving the exact exponent $γ= 1 - ν$. Agent-based simulations confirm this analytical relation. Our framework bridges the gap between individual cognitive constraints and the scaling law of tour behavior, providing a microscopic theoretical underpinning for human mobility.

Noise sets the fundamental limits of resolution and sensitivity in Dynamic Atomic Force Microscopy (DAFM). While thermal fluctuations are conventionally assumed to be the dominant noise source, this work demonstrates that tip--sample interactions in ambient conditions introduce a non--thermal noise component that significantly exceeds the thermal background. Using a model system of sodium dodecyl sulfate (SDS) on graphite, we characterize this noise through force spectroscopy, 3D imaging modes, and Kelvin Probe Force Microscopy (KPFM). This interaction--induced noise arises from the stochastic formation and rupture of nanoscopic liquid necks, serving as a direct fingerprint of local wettability and dissipative dynamics. Crucially, we find that this ``noise channel'' provides chemical contrast that is distinct from and complementary to the electrostatic potential mapped by KPFM. By deciphering the physical origin of these fluctuations, we establish that noise is not merely an instrumental artifact but a rich spectroscopic signal, and we propose that Frequency Modulation (FM--DAFM) offers a superior approach to decouple these dissipative effects for high--resolution imaging.

This paper presents a modular S-/X-band shared-aperture phased-array antenna (SAPAA) for satellite-communication ground-station reception. The proposed architecture uses a repeatable unit cell that supports independent S- and X-band operation within the same physical aperture and enables arbitrary aperture scaling. Dual-polarized radiators are combined with calibrated complex receive coefficients to synthesize linear polarization (LP), right-hand circular polarization (RHCP), and left-hand circular polarization (LHCP). The design burden of the electrically large shared aperture is reduced by using theoretical estimates for scan matching and inter-band isolation before full shared-aperture verification. Simulated and measured results demonstrate axial ratios below 0.1 dB in the target S- and X-band receiving bands over a +/-50 deg scan range. The prototypes are validated using two approaches: passive measurements, where the element responses are measured individually, and RF system-on-chip-based active measurements, where all available receive channels are measured simultaneously. The results confirm that the proposed SAPAA provides wide-angle scanning, very high polarization purity, and polarization-reconfigurable operation for multi-mission SATCOM ground terminals.

We holographically generate optical modes associated with bipolar-coordinate solutions of the Helmholtz equation. Unlike conventional Gaussian beam families, these modes support multicenter phase discontinuities. We investigate and demonstrate that they support multicentered phase singularities, including optical dipoles and two-centered charge distributions. We further identify a family of conveyor-belt modes whose intensity structure follows an extended trajectory connecting the charge centers. Experimental generation using a spatial light modulator is presented and compared with theoretical predictions. These results establish bipolar-coordinate optical modes as a new class of structured light fields for singular optics, beam shaping, optical manipulation, and the controlled generation of multicentered topological structures.

Modern high performance computing (HPC) applications must target heterogeneous hardware. This requires significant work to ensure domain specific implementations translate to highly performant kernels across a range hardware types and vendors, each requiring bespoke optimization to make use of the specific target architecture. Through the development of a domain specific compiler built with the multi-level intermediate representations (MLIR) project, one can express a high-level, close to the specific domain, abstraction that is progressively lowered to a low, close to metal, abstraction. At each intermediate representation (IR), appropriate optimizations can be applied without costly analysis due to the knowledge embedded in the domain specific IRs. We apply this method to the construction of discrete differential operators for use in spectral/hp element method solvers for computational fluid dynamics (CFD). Here, the performance is driven by a small set of common finite element operators that are composed to create kernels for the discrete differential operators used to solve weak partial differential equations. We create our own MLIR dialect to represent these operators and implement a bespoke lowering pipeline to facilitate the just-in-time compilation of these kernels for both CPU and GPU architecture and illustrate performance comparisons with the Nektar++ spectral/hp element framework.

Nanotweezers are actively investigated as a powerful means to reversibly trap and characterize nanoparticles with profound biological and environmental importance. To ensure that these particles can be reversibly trapped and released, surface passivation is essential. To mitigate the issue of fouling, we investigated the antifouling properties of poly(sodium styrene sulphate) (PSS) synthesized using the Atom Transfer Radical Polymerization (ATRP) technique on our previously reported gold-based Interferometric Electrohydrodynamic Tweezers (IET) device. Fluorescence and interferometric scattering (ISCAT) imaging were used to record trapping performance and study the antifouling properties of our passivated nanotweezer device. The results show that PSS exhibits superior anti-fouling performance against polystyrene nanoparticles when compared to 11-mercaptoundecanoic acid (MUA). By comparing the antifouling properties of PSS and zwitterionic poly(methacryloyloxyethyl phosphorylcholine) (PMPC) in preventing extracellular vesicle adhesion, we found that both exhibited similar performance. Overall, the ATRP technique is broadly applicable across nanotweezer substrates with appropriately chosen initiators.

Silicon has a unique combination of properties that makes it one of the best material choices for biosensor platforms: it is inexpensive, its native oxide is atomically smooth, its fabrication processes are CMOS-compatible and have been refined for more than three decades, and it can support many transduction mechanisms in biosensor design. Over the past thirty years, researchers and engineers have used silicon nanostructures to produce ion-sensitive transistors, ultrasensitive nanowire field-effect biosensors, refractive-index-based porous silicon films, microring photonic resonators, suspended cantilevers, luminescent quantum dots, and solid-state nanopores. These device families have demonstrated successful sensing capabilities at the single-molecule, single-virus, or sub-femtomolar level under laboratory conditions; however, they have rarely been widely deployed in clinical assays. This gap is mainly caused by several well-characterized bottlenecks: for nanowire BioFETs, device variability and Debye screening; for porous silicon, fouling, pore wetting, and surface stability; for silicon photonics, thermal drift, spectral readout, and packaging; and across all platforms, calibration, reproducibility, and validation in real biofluids. In this review, we trace the development of silicon biosensors from their early stages to their current state, search and organize the literature focusing on the three most mature platforms and a set of emerging directions, summarize and compare the performance and bottlenecks of different platforms, and argue that progress over the next decade will come primarily from integrated readout, interface engineering, and systematic benchmarking rather than from the discovery of new silicon nanostructures.

This white paper introduces the concept, prototype design, projected costs, and scientific goals of a mobile experiment for detecting geoneutrinos originating from uranium and thorium decay chains in the Earth's mantle. This will constrain the planet's radiogenic heat production and unearth its geochemical makeup. This design of a deep-ocean mobile neutrino experiment, which is not mirrored by any active or planned experiments, supports physics and geoscience's goal of multi-modal data on the Earth's internal composition and structure. Based on geoscientific studies, this design is expected to achieve a 50--100-fold reduction in crustal background compared to similarly sized continental detectors, thereby enabling direct measurements of mantle geoneutrinos. The multiple stereoscopic projections enabled by the detector's unique mobility can map spatial variations in heat-producing elements within the mantle. Beyond discussing the design, we report on our collaboration's most recent hardware developments in the active prototyping of this detector. We briefly highlight the potential multiuse and interdisciplinary nature of this detector.

Diamond color-center qubits integrated with photonic circuits can be initialized, manipulated, entangled, and read individually with high fidelity, making them attractive for large-scale modular quantum computers, quantum networks, and distributed quantum sensing. However, the limited size of heteroepitaxially grown single-crystal diamond (SCD) and photonic-grade diamond-on-insulator (DOI) substrates remains a challenge for integration with existing manufacturing processes. Here, we develop a plasma etch recipe to thin direct-bonded (100) SCD membranes (<50 $μ$m) into large-area, thin-film DOI substrates, and demonstrate free-standing photonic chiplets fabricated from the resulting DOI. The ICP-RIE recipe preserves diamond bonding, provides sufficient micromasking and surface-quality control, and enables thin-film DOI manufacture. We thin a 10 $μ$m diamond plate bonded to SiO$_2$/Si and obtain a photonic-grade DOI substrate with diamond thickness $\leq$300 nm. The DOI film is around 300 nm thick over 0.5 $\times$ 0.5 mm$^2$, with surface roughness < 0.5 nm, while the bonding interface remains intact. Diamond photonic chiplets are fabricated on this DOI substrate using a standard two-step lithography process, without complex thin-film transfer, under-etching, or pedestal formation. We also present a colorimetric study of diamond visibility on SiO$_2$ and quantify color differences across thicknesses in common colorimetric spaces. This analysis enables automatic diamond-thickness extrapolation from standard optical microscope images with 5 nm resolution, in good agreement with white-light interferometry (WLI) measurements. The DOI substrate and colorimetric thickness-evaluation method provide an effective fabrication platform and reliable validation route for scalable manufacturing of diamond nanophotonic devices, opening a path toward large-scale integrated quantum systems.

We use density functional theory and model Hamiltonians to reveal large spin splitting of bands localized at low-symmetry, ferromagnetic surfaces of bulk antiferromagnets (AFMs). There is great interest in finding new material platforms combining the robustness and ultrafast dynamics of AFMs with large, functional spin splitting which is often restricted to ferromagnets. Here, we show that a subset of AFM surfaces which have symmetry-allowed magnetization can host large spin splitting via bulk degeneracy lifting of sublattice-resolved exchange splittings. Using model Hamiltonians, we show that the spin splitting is maximized for two ferromagnetic surface motifs: terminations with single uncompensated magnetic sublattices, and two-sublattice surfaces whose sublattices are magnetically and electronically compensated in the bulk, but acquire distinct crystal field environments via surface truncation. The latter case can yield FM-like spin splitting magnitudes while also having vanishingly small uncompensated magnetization. In contrast, when surface magnetization arises from relativistic canting on symmetry-connected sublattices, the spin splitting is expected to be small. We confirm these predictions with first-principles calculations of $\mathrm{Cr_2O_3}$ and $\mathrm{FeF_2}$, finding splittings from $\sim10\mathrm{meV}$-$\sim1\mathrm{eV}$ depending on the surface in question. Our findings point to intrinsic surface symmetry breaking as a route to large, functional spin splitting in an expanded range of AFM materials.

Why would anyone help a stranger, knowing they may never meet again? Indirect reciprocity offers one of the most compelling evolutionary answers, yet its two canonical forms -- upstream reciprocity (experience-based), and downstream reciprocity (reputation-based) -- have been studied mostly in isolation. Their joint dynamics in finite and structured populations remain largely unexplored. Here, we fill this gap using agent-based simulations in which an agent is behaviourally either defector, upstream reciprocator, or downstream reciprocator, and the agents' population state is temporally updated using different evolutionary update mechanisms. We show that update mechanism plays a surprisingly decisive role in shaping the fate of downstream and especially upstream reciprocators. Whether agents' experiences and reputations are updated globally or locally can shift outcomes from rich behavioural coexistence to the dominance of downstream reciprocators alone. Intriguingly, we uncover a robust structural feature that persists across all the explored update rules and population sizes: an optimal network degree at which upstream reciprocity is maximized, reflecting a fundamental tug-of-war between cooperative clustering and exposure to defectors. Our results highlight that while downstream reciprocity can either foster or inhibit upstream reciprocity depending on the update mechanism, its net effect on cooperation remains largely positive.

Premise: Hybrid speciation and introgressive hybridization are increasingly recognized as important evolutionary phenomena across the tree of life. One widely used class of methods to detect introgression includes D statistics and related methods which employ the ABBA-BABA test using nucleotide site patterns. Recent studies have applied this theoretical framework to phylogenomic datasets using gene tree topologies instead, but no software packages using this method have been developed. Methods and Results: An R package was developed to facilitate the inference of introgression given a set of gene trees and a species tree. Using an ABBA-BABA framework, this package summarizes patterns of gene tree discordance to infer introgression across large phylogenies. Conclusions: Using gene tree topologies, quaint overcomes the limitations of site-based methods, enabling the detection of introgression across broad phylogenomic contexts. This R package provides an accessible and reproducible tool for researchers investigating reticulate evolution.

In the present paper, we apply the iterative mathematical method previously developed for the kinetics of template-directed multistate copolymerization to the kinetics of DNA replication by polymerases having multiple structural states. In particular, we study a two-state kinetic model for the T7 DNA polymerase. We obtain the mean velocity for the growth of the copy along the template, the error probability of DNA replication by the polymerase, and the local probabilities of base-pair formation along the template sequence. Furthermore, we show that the iterative method is more than a million times faster than usual numerical simulation methods. Results are also obtained in the approximation of homogenization of template heterogeneities.

We consider processes of template-directed multistate copolymerization by molecular machines such as polymerases or ribosomes, having multiple states of conformation or activation. We show that the kinetic equations of these processes can be exactly solved for the mean growth velocity, the sequence probabilities of the grown copy, and the local probabilities and fractions of monomeric units in the copy. Asymptotically, in the long-time limit, the kinetic equations are solved with a matrix factorization ansatz in terms of a backward iteration, forming an iterated matrix function system, and a complementary forward iteration, both running along the template sequence. The iterative method is very significantly faster than usual computational methods, as demonstrated with a numerical example.

We develop a theory of phase-locked activity in delayed spiking networks using the Haken Lighthouse model as an analytically tractable event-based description of neural dynamics. For networks with fixed delays, we derive self-consistency conditions for phase-locked states and an associated linear stability theory formulated directly in terms of spike-time perturbations. The framework is illustrated for a delayed autapse, a reciprocally coupled two-cell network, and spatially structured rings with distance-dependent coupling and conduction delays, where circulant symmetry allows stability to be decomposed into Fourier modes. We then introduce an activity-dependent white matter plasticity rule in which myelination modulates axonal conduction speed and hence communication delay. This leads naturally to a slow--fast system with state-dependent delays, in which frozen phase-locked branches organise the adaptive dynamics. The plasticity rule selects commensurate delay--period relationships, providing a mechanism for the emergence of synchrony, other frequency-locked states, slow switching between competing phase-locked patterns, and the organisation of heterogeneous delays into discrete delay--period classes. Direct simulations of the event-driven network support the analytical predictions and illustrate how adaptive conduction can reshape the attractor structure of a delayed spiking network and generate long-timescale transitions. These results provide a tractable mathematical framework for studying how activity-dependent myelination may regulate temporal coordination, synchrony, and communication through coherence in spiking neural systems.

We develop a deterministic mathematical model to quantify the distinct in vitro infection dynamics of Bunyamwera virus (BUNV) and Batai virus (BATV) in A549 cells, incorporating cell division and natural death, continued entry of virions into already-infected cells, and shortening of the eclipse phase driven by re-infection. The model parameters were estimated making use of viral decay data, growth curves at two different inoculum concentrations, and extra-cellular genome copy measurements (for BUNV) via Markov chain Monte Carlo. Genome copy measurements were essential for constraining estimates of the number of cells that can become infected per unit of infectious virus for BUNV. We found that BUNV exhibited substantially longer eclipse and infectious periods than BATV, while BATV showed a higher per-cell virus production rate. Re-infection was predicted to shorten the eclipse phase for both viruses, but the effect was markedly stronger for BUNV. Together, these results provide a quantitative comparison of the in vitro viral kinetics of BUNV and BATV and reveal substantial differences in their replication dynamics.

Affinity maturation is the Darwinian process by which antibodies improve antigen binding through somatic hypermutation and selection. The adaptive landscape, which defines the set of antibody-specific mutations that improve functional characteristics like antigen binding, has been explored in only a handful of antibodies. Identifying the sites of adaptive mutations in a given antibody sequence, and how these sites vary across the antibody repertoire, can inform the design of therapeutic antibodies. We develop a parameter-free population genetic framework that leverages the statistics of convergent affinity maturation in B cell lineages sharing similar naive sequences, called public clonotypes, to identify beneficial mutations. Applying this framework to more than 10,000 public clonotypes represented by multiple lineages across 20 healthy individuals, we identify widespread signatures of clonotype-dependent selection of individual mutations. We estimate the prevalence and typical fitness effects of mutations across the V gene at the single-site level, uncovering a general tradeoff between prevalence and fitness effect. These inferred landscapes broadly reproduce the statistics of convergent mutation in antibodies specific to SARS-CoV-2 and influenza. Finally, we use our framework to benchmark predictions from existing antibody language models, and show that while these models are dominated by non-selective signatures, a simple renormalization procedure can expose signatures of clonotype-dependent positive selection consistent with our predictions.

Remembering a visual scene requires organizing distinct details into a cohesive event. This study investigates whether relation-guided gaze transitions provide a behavioural marker of this cognitive organization during episodic encoding and retrieval. By applying scene graph annotations to eye-tracking data, we measured whether gaze moved between objects that were meaningfully related within complex scenes. This approach allowed us to quantify relational scanning within naturalistic environments, moving beyond prior methods that relied on simplified displays or isolated relation types. Participants showed above-chance relational gaze during both initial viewing and blank-screen retrieval, indicating that gaze actively tracks scene structure during first viewing and at recall. Additionally, relational scanning at encoding predicted subsequent free recall of both object and relational details, even after accounting for salience, fixation frequency, meaning, and image-level differences. In contrast, relational scanning at retrieval did not predict recall success, suggesting that relational gaze is most functional to memory during its formation. Together, these findings show that relational gaze can be measured in complex scenes and may serve as a marker of episodic encoding during natural visual exploration.

A cross-fused generative site-specific beamforming (GenSSBF) framework is proposed for low-overhead beam alignment in uniform planar array (UPA) systems. A decoupled channel sensing strategy is developed, where the azimuth and elevation domains of the UPA are probed independently, and the online sweeping overhead is reduced from multiplicative to linear complexity compared to exhaustive two-dimensional codebook sweeping. However, the resulting reference signal received power (RSRP) observations only contain marginal angular power information. The explicit azimuth-elevation coupling of the UPA channel is therefore lost. Beam generation from these separate observations becomes highly ambiguous. To address this issue, a bidirectional cross-attention encoder is designed to extract and fuse the latent dependency between the azimuth and elevation sensing branches. Conditioned on the fused feature, a conditional normalizing flow generator is proposed to generate a compact set of high-fidelity beam candidates. These candidates are further verified through lightweight pilot measurements for final beam selection. A task-oriented training objective is also introduced to encourage the generated candidate set to contain at least one high-gain beam, rather than fitting the full conditional beam distribution. Simulation results based on DeepMIMO scenarios show that the proposed framework consistently outperforms deterministic beam prediction and conventional discrete Fourier transform (DFT) codebook search. Compared with the full 1024-beam two-dimensional DFT search, normalized beamforming gain improvements of 83.6%, 74.6%, and 38.1% are achieved in the I2_28, O1B_28, and Boston5G_28 scenarios, respectively, while the sweeping overhead is reduced by 93.8%.

This paper proposes a novel method to reduce the peak-to-average-power-ratio (PAPR) of affine frequency division multiplexing (AFDM) signals without side information (SI). The method is based on circularly shifting the original transmit signal in the affine domain, and selecting the shifted candidate with the lowest PAPR. Next, a maximum-likelihood-based (MLB) receiver is derived, which exploits the position of the AFDM pilot and guard band to detect the shift applied at the transmitter without SI. Simulation results show that the proposed method can achieve 2.5 to 4 dB in PAPR reduction as compared to original AFDM and existing schemes, which can be translated into significantly lower error rate, depending on the quality of power amplifiers.

Spectral initialization is a critical yet challenging step in sparse phase retrieval. Existing spectral design theory is largely tailored to dense phase retrieval, where the objective is eigenvector estimation. In contrast, sparse initialization first requires a statistically distinct support screening step whose design remains much less understood. This paper develops a stage-specific design theory for diagonal support screening. We formulate each coordinate score as a scalar statistic for distinguishing support from non-support coordinates and adopt the deflection criterion as a tractable measure of screening quality. Within a Hilbert-space formulation, we characterize the optimal spectral preprocessors that maximize this criterion. In the Gaussian model, the unique optimum is the centered linear preprocessor. To obtain a bounded implementation, we introduce a spherical normalization and characterize its exact optimal preprocessor. Since the exact spherical optimum exhibits a boundary singularity, we construct a bounded surrogate preprocessor and establish its unique optimality under a surrogate deflection criterion. The surrogate optimum is shown to be the direction-only projection of the Gaussian rule, removing the unbounded radial factor while preserving the same first-order screening structure. We further establish a general finite-sample diagonal bridge that connects the exact and surrogate deflection quotients to the initialization sample complexity, and that replacing the unknown signal energy by its empirical estimate introduces only a lower-order perturbation. Numerical experiments are consistent with the ordering predicted by the design quotients and show that the Gaussian centered rule and its spherical counterpart behave nearly identically at both the screening and initialization levels.

Equalization-enhanced phase noise (EEPN) emerges as a key performance limitation in high symbol-rate coherent transmission systems. In this paper, we highlight recent advances in modeling EEPN and show that the temporal Gaussian noise model reproduces the characteristic burst-like SNR degradation, enabling efficient system simulation.

The Manakov equation governs the propagation of signals in dual-polarization systems. Its solution is usually approximated by regular perturbation on the nonlinear Kerr parameter. In this paper, we propose a novel regular perturbation on the group-velocity dispersion parameter for the Manakov equation.

In electromyogram (EMG)-based motion recognition, it is impractical to predefine all motions that may be required during deployment, necessitating class-incremental learning that sequentially adds new motion classes. The primary challenges in class-incremental learning are catastrophic forgetting, where previously acquired knowledge is overwritten when learning new classes, and the memory cost of retaining past data to counteract it. In particular, for EMG-based motion recognition intended for edge devices with limited computational resources, it is essential to suppress catastrophic forgetting and maintain low memory cost. In this paper, we conducted a comparative evaluation of eight class-incremental learning methods spanning generative and discriminative approaches, including both deep and non-deep learning methods, for EMG signal classification. Using four datasets, we evaluated each method in terms of classification accuracy, backward transfer, and memory cost. The results demonstrated that deep learning-based methods suffered significant accuracy degradation from catastrophic forgetting as the number of tasks increased, whereas generative models maintained stable accuracy with low memory cost. Among generative models, the scale mixture classification model (SMCM), which captures EMG signal variability, achieved the most favorable accuracy-memory trade-off while effectively suppressing catastrophic forgetting across all datasets.

Differentiable audio processors are habitually designed and optimised in machine-learning frameworks, but deploying them as real-time audio effects still often requires non-automatic implementation in a dedicated digital signal processing language. The translation is error-prone, demands an onerous verification process, and detaches research prototypes from usable production tools. That being so, we present ADAC, a compiler that lowers a trained model to a framework-agnostic intermediate representation and emits efficient FAUST code whose impulse response matches the source model to within floating-point arithmetic noise, direct paths included. The optimisation loop is made audible by replacing the model in a running plugin after each gradient step. The exported processor carries a small set of macro-controls that leave its stability intact. A stability certificate computed from the shipped parameters is checked before the plugin is built. At the demonstration, a feedback delay network is trained and exported to a working plugin.

Pixel antennas enable highly radiation pattern reconfigurability to enhance wireless systems, but its antenna coding design, that is optimizing the states of switches embedded in pixel antennas, remains an NP-hard challenge. Conventional approaches for antenna coding design typically rely on heuristic search algorithms, which suffer from high computational complexity. To overcome this issue, we propose a novel efficient data-free optimization algorithm called physics-informed neural optimizer (PINO) for antenna coding design. By integrating a deep convolutional neural network prior and a Gumbel-Sigmoid continuous relaxation into a differentiable physics engine, the proposed algorithm transforms the binary optimization problem into a continuous differentiable problem, which enables the antenna coding optimization problem to be efficiently solved via gradient descent. Simulation results demonstrate that the proposed algorithm outperforms the heuristic search based algorithms, reducing computational time while achieving higher average channel gain.

We demonstrate the first KAN-based DPD model for 5G analog RoF fronthaul link, achieving a 24.2% lower EVM than multi-layer perceptron and 29.6% lower than Volterra-based GMP at equivalent Bit Operations. To attain an EVM below 2%, KAN requires ~52% fewer BOPs than a perceptron.

The Unlimited Sampling Framework (USF) overcomes the traditional trade-off between dynamic range and digital resolution, achieving performance unattainable with standard ADCs. Its multi-channel extension (MC-USF) enables reconstruction from multiple folded measurements at critical sampling rates. Existing MC-USF methods typically rely on Chinese Remainder Theorem (CRT)-based unfolding, which requires strict channel-level sampling synchronization and is therefore vulnerable to timing mismatch, jitter, and drift. This paper introduces an asynchronous MC-USF architecture that eliminates the need for synchronization. By viewing spatial-temporal signal lifting as inducing smoothness over a graph of sensing channels, we develop a reconstruction strategy robust to temporal misalignment. Numerical experiments validate the approach, demonstrating accurate recovery and enabling more practical multi-channel USF implementations.

Spectrum sharing forces radars to vary carrier frequency and bandwidth on a pulse-to-pulse basis within a coherent processing interval (CPI). While the resulting range-Doppler distortion is well-studied, the corresponding velocity estimation limit is not. We show that in the resolved-bin slow-time model of coherent agile-CPI processing, the effective Fisher information for radial velocity is the SNR-weighted energy of the carrier-time lever arm that survives projection out of the range and phase nuisance subspace. The carrier sequence thus sets the projection geometry, while the bandwidth sequence enters only through SNR weighting. Two consequences follow. First, the carrier sequence inflates the bound by a closed-form factor governed by the correlation between carrier offset and slow time: randomized or orthogonalized hops are nearly harmless, while ramp-correlated hops can severely degrade velocity information. Second, under matched filtering at equal pulse energy, the velocity Cramer-Rao bound (CRB) is invariant to the bandwidth sequence; a corollary recasts the output-SNR loss of agile-CPI mismatched filtering as a processing cost entering only through a per-pulse mismatch loss. The bound is verified against a brute-force Fisher matrix and Monte-Carlo maximum-likelihood estimation. The result yields a design principle: carrier hopping should be chosen not only for spectral coexistence but also to preserve the velocity-information residual.

State-dependent proto-area data produced by approximate recovery need not be compatible with a single local bulk metric. Using recovery maps calibrated on the code channel and held fixed along a logical-state family, we derive exact finite-resolution criteria and, near the hyperbolic disk, necessary and sufficient conditions for a regular proto-area two-jet to arise from a metric two-jet on a time-reflection-symmetric asymptotically AdS$_3$ slice. Finite networks give a polyhedral realization problem with primal and dual certificates, stable reconstruction, and explicit witnesses of nongeometry. In the continuum, the geometric tangent space is the range of the rank-two geodesic X-ray transform. A metric-forced Jacobi equation determines the normal Hessian of the renormalized boundary-length image and yields a gauge-invariant quadratic obstruction. Under a split-regularity hypothesis, nearby geometric data form a local graph; the two-jet criterion itself is unconditional for regular data. Hamiltonian-skewed codes realize both first-order nongeometry and a response whose first obstruction appears only at quadratic order. The compatible metric perturbation is reconstructed modulo boundary-fixing diffeomorphisms.

Quantum platforms with beyond-planar connectivity provide new opportunities for fault-tolerant quantum computation (FTQC). While quantum low-density parity-check (qLDPC) codes offer high encoding efficiency, their direct implementation requires non-local couplings in every round of syndrome extraction, incurring additional physical error and implementation complexity. To reduce the frequency of such couplings, we propose the Hierarchical Logical Processor (HLP), which concatenates a high-rate quantum CSS code with the rotated surface code (RSC). HLPs can achieve beyond-RSC encoding efficiency while requiring long-range connectivity only once every $Θ(d_0)$ rounds of level-0 error correction, where $d_0$ denotes the base-code distance, substantially reducing the frequency of non-local couplings relative to direct implementations of qLDPC codes. HLPs introduce elongated RSC patches called shuttle buses. Using transversal hybrid-unit CNOT gates, a single shuttle bus can simultaneously couple to multiple standard RSC patches. This capability enables efficient level-1 syndrome extraction with suppressed level-1 error correlations and supports highly parallel logical Pauli measurements. We perform circuit-level simulations of several concrete HLP constructions and benchmark both logical memory and logical Pauli measurement performance. At a physical error rate of $10^{-3}$, an HLP based on the [[256,194,4]] code achieves 3-4 times higher qubit efficiency than the standard RSC. Compared with the yoked surface code on the same level-1 code, this HLP reduces the space overhead per logical qubit by 100-200 physical qubits and shortens the logical error-correction cycle time by a factor of 20-30.

We extend the Parameterized Spin Expansion Coefficients (ParSpec) framework for ringdown tests of black holes by simultaneously sampling the characteristic length scale $\ell$ and the scaling index $p$, where the two parameters are replaced by $\tilde{p}$ and $\tilde{\ell}$ in subsequent analysis, respectively. Unlike previous analyses that fix $\tilde{p}$ to theory-motivated integer values, this extension provides a theory-agnostic description inspired by effective-field theory to the Kerr ringdown spectrum. Using \texttt{pyRing}, we apply the framework to GW250114 with informative priors on $M_{\rm f}$ and $D_{\rm L}$, comparing the $220$-only and $220+221$ mode models under different conditions on $γ$. The posterior distributions of $\tilde{\ell}, \tilde{p}$ and $γ$ are presented with different ringdown start times. We show that the posterior of $\tilde{p}$ remains largely prior dominated, indicating that current ringdown data cannot distinguish different scaling behaviors of the correction. The inferred constraint on $\tilde{\ell}$ is instead mainly controlled by the geometry of the allowed parameter space induced by the $γ$ condition. In particular, $γ<1$ provides the most natural prescription in the $(\tilde{\ell},\tilde{p})$ plane and yields stable but weak upper bounds of $\tilde{\ell}_{90}\simeq 83\, {\rm km}$. An analysis based on the Kullback-Leibler divergence further shows that, after accounting for the effective prior associated with $γ<1$, the data-driven information gain on $\tilde{\ell}$ and $\logγ$ is very small. At the current SNR level, the $220$-only model provides a more informative constraint than the $220+221$ model. We also perform a joint analysis including GW231123 and find that the combined constraint is dominated by GW250114.

The quantum Hamming bound is the finite-length sphere-packing count for exact quantum error correction: the code dimension times the number of correctable local error patterns must fit inside the ambient Hilbert space. For nondegenerate codes this follows from disjoint error spheres. Degeneracy is the only obstruction, because distinct physical errors can coincide on the code subspace and turn sphere packing into an overcount. The central finite-length question has been whether this overcount can ever invalidate the Hamming inequality. Earlier linear-programming, asymptotic, and structural results left a pointwise finite-length problem for arbitrary exact subspace codes. Writing $Q=q^2$ for the Hamming-scheme alphabet, the nonbinary range begins at $Q=9$; here we prove the bound for every $q\ge3$, while the binary endpoint is governed by a distinct $Q=4$ charging geometry. For every nontrivial exact subspace code in this range, any possible violation reduces to a two-center Hamming-ball intersection inequality normalized by the Lloyd response. For $q\ge4$, the Lloyd-square linear program has a uniform half-gap after reduction to alphabet $Q\ge16$ and critical length $n=4t+1$. Qutrits form the boundary: the half-gap disappears, but the bridge is handled by a quadratic-filtered Lloyd square, an exact coefficient-certificate reduction, and a Stein-tangent positivity argument. Thus degeneracy may merge error sectors, but not enough to beat the Hamming count. This proves the nonbinary part of the finite-length quantum Hamming bound; together with the independent binary endpoint theorem, it gives the result in arbitrary local dimension.

A useful warp drive must change its velocity, yet known positive-energy constructions are static or constant-velocity, while accelerating ones require exotic matter; Bobrick and Martire noted that warp drives ``do not have any natural way of changing their velocities.'' First, a conservation law settles the principle: for any asymptotically flat drive with a confined dominant-energy source and standard peeling, the Bondi--Sachs four-momentum changes only through radiation to null infinity, so the drive cannot steer without radiating. Second, we construct a positive-energy spacetime that saturates this bound. Prescribing the passenger worldtube, we take the exact Kinnersley photon rocket as exterior and solve for the matching timelike shell; the exterior and steering law are exact, while the shell is certified perturbatively and numerically. The mechanism is photon-rocket recoil on a tidally protected, exactly flat cavity; the warp drive is the decoupled flat interior, not a warp field. The exterior energy conditions reduce to $n^2\ge0$, and steering obeys $-\dot m\ge 3m|a|$, paid for by Bondi mass loss. The shell satisfies the surface dominant energy condition for $2m/R<24/25$; admissible accelerating shells exist near the Schwarzschild--Minkowski anchor and, for slow burns, as time-evolved spacetimes. A Buchdahl-type frontier $a_{\max}R\le g(2m/R)$ caps the acceleration between a numerical lower bound $\sim\!0.2$ and a closed-form ceiling ${1\over2}(1-2m/R)$. On the gravitational-wave--silent class, the optimal maneuver is the Damour dipole. The wall is marginally stable, with strict stability available at no energy-condition cost; fully dynamical flux-coupled stability remains open. The drive is causal, subluminal, and energetically costly but positive-energy: steering is a problem of energy budget, not exotic matter.

Liouville pathways provide the conceptual foundation for interpreting multidimensional spectroscopies, yet are typically treated as fixed objects that evolve independently between optical interactions. In open quantum systems, this picture is incomplete. Environmental interactions continuously redistribute amplitude among pathways during every free-evolution interval, generating transport that leaves measurable signatures in the nonlinear spectroscopic response. We develop a geometric framework in which pathway transport is governed by a Liouvillian connection, its associated curvature, and the resulting observational holonomy. The framework applies to open quantum systems in which the environment selects a pointer basis distinct from the observational basis used to construct the spectroscopic response. This basis incompatibility induces transport among Liouville pathways, generating characteristic spectral distortions and a nontrivial Liouvillian curvature. Using a Duhamel expansion of the Liouvillian propagator, we derive a reconstruction procedure that identifies the transport operators responsible for the observed redistribution of pathway weight, accurate throughout the full range of basis misalignment. This perspective reframes spectral features as determined not only by which pathways exist but by how amplitude is transported among them. Spectral distortions, peak shifts, and otherwise-forbidden pathway contributions become geometric signatures of a curved Liouville-space manifold rather than phenomenological broadening corrections, identifying pathway geometry as a complementary layer of organization in nonlinear spectroscopy.

We develop a response-geometric framework for open quantum systems by decomposing the stationary-state response tensor into symmetric and antisymmetric sectors. The symmetric sector defines a metric-like response tensor governing local susceptibility, while the antisymmetric sector defines a curvature two-form associated with nonreciprocal response and geometric work. We show that the work curvature is precisely the antisymmetric component of the response tensor, establishing a fluctuation--response relation that extends the geometric structure of equilibrium thermodynamics to nonequilibrium steady states, and reveals a response geometry with both metric and symplectic sectors. In equilibrium, the antisymmetric sector vanishes by reciprocity, recovering the familiar metric geometry of thermodynamic response. Open quantum systems admit a broader structure in which reciprocal and nonreciprocal responses coexist on the same control manifold. We illustrate this with a driven dissipative qubit under pure dephasing, where finite curvature emerges from the misalignment between the Hamiltonian eigenbasis and the pointer basis selected by the environment, without requiring strong driving or engineered reservoirs. Comparison with the Bures metric shows that response geometry and information geometry characterize distinct properties of the stationary-state manifold: the former governs susceptibility and geometric work, the latter quantifies state distinguishability. Geometric work thus emerges as a measurable signature of nonreciprocal response in open quantum systems.

We study the effect of subsystem decoherence to an extended cluster state which is a symmetry protected topological (SPT) phase. The model includes many subsystem $Z_2$ symmetries. We report that subsystem decoherence induces local charge fluctuations, leading to a mixed SPT state in the unaffected subsystems. If we start from the extended cluster state, hierarchical mixed-state SPT phases emerge in response to step-by-step subsystem decoherences. These mixed-state SPT phases keep strong symmetries the symmetry of which is protecting symmetries for the initial cluster SPT. Moreover, these SPTs can be characterized by Rényi-2 string orders. Then, as the subsystems are progressively decohered, the hierarchy of mixed-state SPT phases terminates in a $Z_2$ strong-to-weak spontaneous symmetry breaking (SWSSB) state on the final remaining subsystem, where a long-range entangled state appears, namely a glassy Greenberger-Horne-Zeilinger (GHZ) state. Our work demonstrates that decoherence is not merely a destructive process, but can induce and organize series of nontrivial mixed-states. This reveals a systematic route from mixed-state SPT order to SWSSB with the glassy GHZ-type long-range entanglement.

The spatial confinement of defect-induced bound states is conventionally governed by the effective mass in dispersive bands. More recently, Compact Localized States (CLSs) arising from exact destructive interference have been utilized to achieve confinement in flat bands. However, CLSs rely on pristine lattice symmetries and fine-tuned defect profiles. The introduction of a generic local impurity inevitably breaks these strict phase-matching conditions, resulting in extensive bound states whose fundamental length scale has remained an open question. Here, we establish a third regime of confinement: the quantum metric bound state. We provide a rigorous mathematical proof demonstrating that in the absence of kinetic energy and CLS protection, the exponential decay length of these states is lower-bounded by the quantum metric of the unperturbed flat band. We demonstrate the tightness of this geometric limit by constructing a family of highly tunable flat-band generators, and we verify its universality across diverse realistic architectures. Ultimately, this classification establishes the independently measurable quantum metric as a predictive design principle for engineering confined modes in synthetic wave platforms.

In this study, we numerically investigate Bondi-Hoyle-Lyttleton (BHL) accretion around Kerr-Sen black holes and examine how the charge-related deformation of the spacetime affects the shock-cone morphology, the variation of the mass accretion rate, and the quasi-periodic oscillation (QPO)-like temporal behavior. The relativistic BHL flow is solved numerically in the equatorial plane for two different black hole spin parameters, a = 0.9 M and a = 0.5 M. From the numerically computed mass accretion rate signal, we calculate the power spectral density (PSD) and perform multi-component Lorentzian fits to identify the dominant QPO-like modes excited around the black hole. The results show that the Kerr-Sen deformation shifts the characteristic frequencies, changes the coherence properties of the oscillation modes, and produces near-resonant harmonic structures close to 3:2 and 2:1. By using inverse mass scaling, the numerically computed frequencies are compared with observed QPOs from stellar-mass, intermediate-mass, and supermassive black hole systems. In particular, reasonable agreement between the numerical simulation results and observations is found for the sources GRS 1915+105, IGR J17091-3624, M82 X-1, NGC 5408 X-1, RE J1034+396, 1H 0707-495, and ESO 113-G010. This comparative analysis indicates that Kerr-Sen black hole shock-cone oscillations may provide a unified framework for interpreting timing features over a broad range of black hole masses and may additionally contribute to constraining the mass and spin parameters of sources whose properties are not yet fully established observationally. These findings further imply that combined hydrodynamical and timing diagnostics constitute a promising approach for assessing the extent to which deviations associated with the Kerr-Sen geometry can be empirically distinguished from those of the Kerr spacetime.

This dissertation aims to put forth a systematic construction of a fractional-calculus extension of holographic dark energy (HDE). We show that linking late-time cosmic acceleration to non-local or memory effects encoded in a fractional (Riesz) derivative within black hole thermodynamics produces deviations from standard HDE and can address some challenges of the Hubble cutoff. In particular, a Riesz fractional spatial derivative is introduced into the Hamiltonian constraint of a Schwarzschild black hole in quantum geometrodynamics, leading to a Fractional Wheeler--DeWitt equation whose solutions yield fractionally corrected thermodynamic quantities, notably fractional Bekenstein--Hawking entropy governed by the Lévy index \(α\), with \(1<α\leq2\). Using this entropy with Cohen's inequality, a new dark energy density is constructed, defining the Fractional Holographic Dark Energy (FHDE) framework. The cosmological implications of FHDE are then investigated. Within the Hubble cutoff, its late-time evolution is analysed through cosmological observables, and the model is reconstructed using effective field descriptions with spin-\(0\) and spin-\(1\) candidates, allowing kinetic and potential terms to be extracted as functions of redshift \(z\) and \(α\). The framework is then extended to BD, DGP, EPN, and Horndeski theories to derive the equation-of-state and deceleration parameters in terms of \(z\) and \(α\). In addition, the fate of the Universe is studied through late-time singularities, namely the big, little, and pseudo-rip, within the Granda--Oliveros FHDE setting. In short, this dissertation proposes FHDE as a theoretically motivated extension of HDE, bridging non-locality in quantum gravity with the late-time dynamics of the Universe, and offering a route toward understanding cosmic acceleration beyond \(Λ\)CDM.

Fault tolerance is an indispensable prerequisite for constructing large-scale universal quantum computers. Drawing philosophies from classical computer architecture, this paper presents a hardware-agnostic three-layer high-level architectural framework for generic fault-tolerant quantum computation. Guided by the real execution workflows of fault-tolerant quantum algorithms, the proposed model is decoupled from specific physical qubit hardware platforms and quantum error correction codes, serving as a universal abstract standard rather than a platform-specific implementation scheme. Special attention is devoted to the intermediate Fault-Tolerance Layer, which serves as the architectural bridge between application-level logical programs and hardware-level execution. We systematically characterize its five internal components, the interfaces and data exchanged among them, and the execution, correction, and adaptation paths that together enable logical synthesis, fault-tolerant resources management, decoding, and runtime fault-tolerant control. An end-to-end example is further provided to illustrate the full-stack operating pipeline of fault-tolerant quantum algorithms under this framework. Given the increasing emphasis on modular, heterogeneous, and cross-layer fault-tolerant quantum systems, our architecture provides a unified foundational model for organizing such designs.

We investigate the constraining power of scalaron-driven inflation on asymptotically safe scalar-tensor theories. Starting from a Horndeski-type theory and applying a renormalization group improvement procedure generates higher-derivative couplings which are fixed in terms of the microscopic parameters - a structure that is expected to occur also within first principle computations based on the asymptotic safety mechanism. The latter are taken to be the free parameters appearing at the Gaussian fixed point. We find that the free parameter initially associated with the non-minimal gravity-matter coupling is not confined to the gravity-matter sector of the theory and also enters the effective higher-derivative couplings in the gravitational sector. We review the setting of multi-field inflationary models which is appropriate to analyze the inflationary dynamics in this context and illustrate their applicability by working out the explicit bounds on the non-minimal gravity-matter coupling resulting from cosmological observations. Given the fixed point structure of asymptotically safe scalar-tensor theories, the results indicate that UV-completions by two of the three available non-Gaussian fixed points can be ruled out while pinpointing phenomenologically viable RG trajectories emanating from the third fixed point.

Experimental progress in quantum simulators highlights the role of long-range (LR) interactions in reshaping quantum criticality and stabilizing exotic phases beyond the short-range (SR) paradigm. We study ferromagnetic long-range quantum $O(n)$ models with interactions decaying as $1/r^{d+σ}$ and develop a perturbative renormalization-group expansion around the LR--SR boundary by setting $d=3-ε$ and $σ=2-δ$. In this parametrization, the full interacting LR window $2d/3<σ<2$ becomes $0<δ<2ε/3$, and is therefore perturbatively controlled. A two-loop calculation yields explicit expressions, in terms of $ε$, $δ$, and $n$, for the correlation-length exponent $ν$ and for the frequency and momentum anomalous dimensions $η_ω$ and $η_k$. The resulting exponents reduce to long-range Gaussian scaling at $σ=2d/3$ and to SR quantum Wilson-Fisher scaling in the $σ\to 2$ limit, thereby identifying $σ_*=2$ as the LR--SR boundary within the controlled $3-ε$ expansion. Combining the RG results with scaling boundaries and classical LR analogies, we propose a $(d,σ)$ universality diagram for ferromagnetic long-range quantum $O(n)$ criticality and use it as an organizing framework for the phase diagram of long-range quantum spin chains.

We study a quantum Otto cycle in a two-qubit non-Hermitian XY model with a staggered imaginary magnetic field. The energy spectrum of this system naturally decouples into a pair of working levels that depend on the external field and a pair of idle levels that are completely independent of it, thereby providing the first concrete microscopic realization of the idle-level quantum heat engine architecture proposed by de~Oliveira and Jonathan [Phys. Rev. E 104, 044133 (2021)] in a physical spin model. Tuning the non-Hermitian parameter eta_0 drives a continuous transition from a dissipative regime with negative net work and net heat absorption from the hot reservoir into a genuine heat engine mode, while simultaneously enhancing both output work and efficiency. As eta_0 increases within the stable PT-unbroken phase, the engine efficiency rises significantly, reaching a substantial fraction of the Carnot limit. This effect originates from the compression of the idle-level gap, which redistributes the level occupation weights in the hot and cold equilibrium states and thereby modulates the absorbed heat. The numerator of the net work expression is independent of eta_0, but the denominator depends on eta_0 indirectly through hyperbolic cosine functions -- this is the mathematical root of the idle-level control mechanism. We provide a detailed analysis of the robustness of these findings against parameter variations, a critical comparison of the non-Hermitian control with the Hermitian limit, and a concrete experimental proposal for trapped-ion quantum simulators. Our results demonstrate that non-Hermiticity serves as an indispensable tool for steering both the operation mode and the performance of a quantum engine.

We investigate the defocusing of null geodesics in dynamical, non-singular black-hole-to-white-hole transitions. Working at the level of spacetime kinematics, and without assuming any specific gravitational field equations, we show that the contraction and disappearance of a trapped region, as well as the subsequent formation and expansion of an anti-trapped region, necessarily require a violation of the null convergence condition. This conclusion follows directly from the behaviour of the null expansions across the trapping and anti-trapping horizons, and is therefore independent of the microscopic mechanism responsible for singularity resolution. We then illustrate this general argument by constructing a class of explicit bouncing geometries in generalised Painlevé-Gullstrand coordinates, obtained by promoting static regular black holes with de Sitter cores to time-dependent black-hole-to-white-hole transition models. For a Bardeen-type mass function, we show that the required violation of the null convergence condition is localised within the intermediate dynamical phase in which the trapped region evaporates and the anti-trapped region forms. Finally, we argue that the limiting case of an instantaneous black-hole-to-white-hole transition would require an unbounded violation of the null convergence condition, signalling a breakdown of the effective continuum metric description, and the need to appeal to a full quantum-gravitational description.

Superconducting quantum computing is one of the most mature solid-state platforms for quantum computation, with processors exceeding one hundred qubits. Yet further scaling toward fault-tolerant quantum computing is increasingly constrained by calibration complexity. Conventional scripts are brittle to anomalous signals, and expert judgment is bounded by cognitive bandwidth and serial operation time, failing to keep pace with system scale. Here we report Vibe Calibration, an autonomous calibration system orchestrated by large language model agents, which distills expert tacit knowledge into reusable Skills. Each Skill is organized as a decision tree that packages parameterized measurement commands, quantitative acceptance criteria, and audit records, enabling autonomous execution and self-healing. We capture this knowledge through a three-phase human-in-the-loop distillation process and fine-tune a large language model on validated trajectories. On a 112-qubit processor with frequency-tunable transmons, the system autonomously completes calibration of 108 out of 112 qubits in 4.7 hours, achieving a 4--5$\times$ speedup over manual calibration of the full 112 qubits. A cross-validated comparison with expert manual calibration on a 16-qubit subset shows agreement on 14 out of 16 qubits. More importantly, the model demonstrates transferable calibration workflows across devices. While low-level control scripts require minor interface adaptation for different hardware platforms, the core decision logic and task orchestration generalize to new processors, demonstrating a reusable laboratory interface rather than a memorized script.This work demonstrates, for the first time, fully autonomous calibration of a hundred-qubit superconducting processor through reusable and auditable Skills, removing a critical barrier to scalable quantum hardware operation.

Quantum machine learning has emerged as a promising research area for learning complex data patterns. However, most existing quantum feature maps employ fixed encoding strategies that do not explicitly consider the relationships among features within a dataset. In this study, we propose a Correlation Aware Quantum Feature Map (CAQFM) which integrates feature dependencies into the quantum encoding process. The proposed approach utilizes Pearson, Spearman, Kendall Tau, Mutual Information, and Distance Correlation measures to identify relationships among features. Dependencies exceeding a predefined threshold are incorporated into the quantum circuit through controlled quantum gates, enabling the construction of richer quantum representations that better reflect the underlying structure of the data. The proposed method is evaluated using a Variational Quantum Classifier (VQC) on three benchmark datasets, namely breast cancer diagnosis, credit default prediction, and student placement classification. Simulation results demonstrate that correlation based quantum encoding can improve classification performance compared to conventional encoding strategies. In particular, the Spearman and Kendall Tau based CAQFM variants achieved the highest predictive performance and consistently outperformed standard quantum feature maps. The findings indicate that incorporating dependency information from classical data into quantum feature maps facilitates the generation of more discriminative quantum representations and enhances the effectiveness of variational quantum classifiers.

The Fermi polaron, formed by an impurity interacting with a surrounding Fermi sea, exemplifies the canonical quasiparticle concept as a cornerstone in our description of quantum many-body systems across a wide range of energy scales. Experiments on atomic quantum gases have provided profound insights into the universal nature of the Fermi polaron. While most previous studies have focused on the case of zero impurity momentum, finite-momentum properties have remained largely uncharted. Here, we investigate the moving Fermi polaron by combining a novel Raman acceleration scheme with high-precision radio-frequency spectroscopy, exploring the quasiparticle dispersion relation over a wide range of momenta. We compare our measurements of energy shifts and spectral linewidths with a microscopic theory and reach quantitative agreement for all momenta. For low momenta, we find the energy of the moving polaron to be fully consistent with the Fermi liquid picture of a dressed particle with a constant effective mass. At high momenta, the polaron approaches the behavior of a weakly interacting bare particle, featuring small energy shifts and weak broadening. For intermediate momenta, broadening is generally larger and, most strikingly, the behavior differs for attractive and repulsive polarons. While the repulsive polaron exhibits a smooth connection between both regimes along with a monotonic change of the energy shift, the attractive case shows a peculiar non-monotonic behavior. With increasing momentum, the attractive polaron enters a regime where its energy deviates from the constant effective mass expression and broadening suddenly increases. By comparing this observation with theory, we show that this abrupt behavior coincides with the attractive polaron entering a molecule-hole continuum, where it is no longer the ground state. We interpret this as a motion-induced polaron-molecule transition.

A practical Hamiltonian approach to lattice gauge theories would provide access to several important areas of phenomenology that have been beyond the reach of conventional lattice methods. Quantum computers seem to be a natural platform for this approach. With near-term quantum computers in mind, our work considers a three-dimensional spatial lattice that can host fermions and non-Abelian gauge fields while needing fewer qubits than a simple cubic lattice. Specifically, the semi-simple cubic (ssc) lattice is obtained by removing half of the gauge links from a standard cubic lattice in such a way that every vertex becomes trivalent, which streamlines the handling of Gauss's law. The ssc lattice is topologically equivalent to the triamond lattice but, because the gauge links at each vertex span all three directions, the ssc lattice can accommodate a local fermion derivative. The case of staggered fermions with SU(2) gauge fields is presented here.

We investigate the head-on coalescence of charged binary black holes in Einstein-Maxwell-Scalar (EMS) theory using numerical relativity. The binaries are built from charged puncture initial data representing two Reissner-Nordström black holes immersed in a purely kinetic scalar perturbation: the scalar field initially vanishes, while its conjugate momentum provides a small seed for the instability. We evolve the coupled gravitational, electromagnetic, and scalar sectors and monitor the apparent horizons, the emitted radiation, and the scalar field on the horizons. Our simulations show that the nonminimal electromagnetic-scalar coupling can dynamically trigger the growth of scalar hair even when the individual black holes are initially scalar-free. The subsequent evolution depends on the coupling strength and on the charge retained by the remnant. For weak coupling, or when charge cancellation suppresses the electromagnetic source after merger, the scalar field is radiated away or absorbed by the final horizon and the system dynamically descalarizes. For sufficiently strong coupling and nonzero remnant charge, the scalar field remains finite and the final black hole approaches a scalarized configuration. The coalescence also excites scalar radiation whose time profile is qualitatively correlated with the dominant gravitational-wave mode during the nonlinear stage of the collision. These results provide a binary realization of scalarization/descalarization transitions in EMS theory and show that the fate of scalar hair is controlled by the interplay between the scalar coupling and the charge content of the remnant.

If one extends the AdS/CFT extrapolate dictionary to large but finite $N$, we are expected to obtain non-perturbative violations of bulk micro-causality. Previously this was achieved by implementing a late boundary time cut-off, while smearing the boundary operator via the HKLL kernel. By performing explicit bulk reconstructions in the backgrounds of near-horizon modified AdS$_2$ and BTZ black holes, we recover the same non-locality estimates as above. For these black hole mimickers, the near-horizon modification is controlled by a throat parameter which sets the scale of this non-locality. In three bulk dimensions, probe dynamics also exhibits a dip-ramp-plateau structure in their spectral form factor when averaged over the throat parameter. Such structure has also been found recently in the background with a stretched horizon or a brick wall.

Asymptotically safe quantum gravity is an approach to quantum gravity. It is based on the premise that quantum field theory can describe the quantum nature of gravity in our universe. At its core lies quantum scale symmetry. This review provides an introduction to the key ideas of the approach and surveys the current status of the field. Over the last years, the field has taken large strides towards an increasingly realistic setting: First, compelling evidence for quantum scale symmetry exists in four-dimensional, Euclidean, pure gravity, establishing the Reuter fixed point robustly. Second, matter fields, including the Standard Model as well as beyond-Standard-Model-candidates, have been studied in depth, with increasingly conclusive evidence for quantum scale symmetry. Most recently, the final gap to a realistic description of quantum gravity is being closed, because Lorentzian spacetime signature can now be accounted for. As a consequence of quantum scale symmetry in the ultraviolet, the approach is highly predictive at all scales. This review discusses the physics of asymptotic safety across all scales. Predictive power for particle physics, black holes and cosmology provides a clear pathway to confronting quantum gravity with current and near-future observations. The review closes by discussing the connection to other approaches to quantum gravity. It advocates the perspective that such connections between approaches may lead us to an understanding of universal physical features of quantum gravity.

Wave-optics interference in strongly lensed gravitational waves is a new interferometric probe of dark matter substructure: a subhalo population threading a galaxy-scale lens imprints frequency-dependent distortions on the amplification factor of each macro image. In a companion paper (arXiv:2603.04267), we computed these imprints for the magnified minimum image. Here, we extend the calculation to the saddle-point image and we assess the detectability of the combined signal with the Laser Interferometer Space Antenna (LISA). Evaluating the amplification factor at a saddle is numerically delicate, because the equal-arrival-time contours are open and the subhalo signal is a small difference of large terms; we present a time-domain method that resolves it. Across a Monte Carlo ensemble of cold dark matter subhalo realizations, subhalos induce percent-level amplitude and phase modulations in both image parities, while the mean (de)magnification splits by parity: the minimum is net magnified and the saddle net demagnified. Demodulating the macro-image interference recovers the per-image modulations, and a matched-filter analysis that projects out the lens parameters yields a combined detection above $5σ$ in $62\%$ of realizations for fiducial massive-black-hole-binary sources of total mass $\sim10^{6}\,M_\odot$ at redshift $1.5$, provided the source lies close to the lens caustic at small impact parameter $y_{\rm src}\lesssim0.1$. Folding these naive per-event significances through optimistic strong-lensing rate forecasts yields $10$-$20$ substructure detections over the LISA mission. Strongly lensed gravitational waves are thus a sensitive, complementary probe of substructure at $10^{4}$-$10^{7}\,M_\odot$ scales inaccessible to electromagnetic observations.

Distributing multipartite entanglement is a prerequisite for scalable quantum networks. Networks restricted to local operations and shared randomness (LOSR) avoid the quantum-memory and latency costs associated with the real-time classical communication required by LOCC-based networks. However, when only bipartite sources are available, LOSR networks cannot prepare useful GHZ states. In earlier work, we conjectured that multipartite sources overcome this limitation and supported this claim with a single numerical example. In this work, we prove the conjecture for regular and uniform networks of arbitrary size. By identifying the hyperedges of the network with the coordinates of a linear code, we show that whenever the edges incident to each node form an information set, a fixed collection of local unitaries, namely the local decoders of the code, transforms the source state into an (N)-party GHZ state with fidelity (d^{m-M}), without requiring any classical communication. Here, (m) denotes the number of edges incident to each node and (M) the total number of edges in the network. For the complete ((N-1))-uniform hypergraph (K_N^{(N-1)}), this fidelity reduces to (1/d). We further prove that this value is optimal among all local-unitary strategies and exceeds the best fidelity achievable using only bipartite sources. For example, in the four-node case, the optimal fidelity is (1/2), compared with the bipartite bound of (1/8). These results demonstrate that multipartite sources, together with shared randomness, can replace real-time classical communication for entanglement distribution.

For slowly-varying noise, time-frequency methods offer a natural middle ground between the efficiency of the frequency domain and the generality of the more expensive time domain. Despite growing adoption, such methods remain less well documented and less familiar in the gravitational-wave literature, compared to the ubiquitous frequency domain. Aimed at gravitational-wave analysts, in this paper we present a self-contained account of time-frequency methods, detail derivations for a specific basis, namely the Wilson-Daubechies-Meyer (WDM) basis, and share intuition and lessons learned. We document key concepts: the orthogonality and good time-frequency localization of the basis, the edge effects at the DC and Nyquist frequencies, the forward and inverse transforms and their practical implementation, and the noise covariance matrix and likelihood in the time-frequency domain.

We consider the classification of toric self-dual Einstein gravitational instantons with negative cosmological constant. As is well known, any Killing vector field on a self-dual Einstein manifold defines a local conformal Kähler structure. We prove that if the conformal Kähler structure associated to one of the torus Killing fields is global and extends to an ALE manifold with no additional fixed points, then the corresponding self-dual Einstein instanton is precisely given by the infinite class of multipole solutions constructed by Calderbank, Pedersen and Singer.

We study extreme-mass-ratio inspirals (EMRIs) evolving in a scalar cloud environment that may form through superradiant instabilities, using a fully relativistic perturbative framework that allows for eccentric and inclined orbits. EMRIs, consisting of a stellar-mass compact object inspiraling into a supermassive black hole, are key sources for space-based gravitational-wave detectors such as LISA. Previous relativistic studies of EMRIs in scalar clouds have been restricted to circular, equatorial motion. Here, instead, we focus on a Schwarzschild black hole background to incorporate eccentricity and orbital inclination. By computing the scalar energy and angular momentum scattered off to spatial infinity and absorbed at the event horizon, we show that orbital eccentricity can induce a dense spectrum of resonances near the last stable orbit, associated with strong relativistic apsidal precession. We further find that orbital inclination can significantly modify the orbital energy and angular momentum losses. In particular, we identify a critical inclination angle below which, at sufficiently small orbital radii, there is a net transfer of energy from the scalar cloud to the orbit. Moreover, for sufficiently large eccentricities, resonances associated with relativistic apsidal precession persist across the full range of inclinations, although their structure changes significantly between prograde and retrograde orbits. These results provide a foundation for future studies of EMRIs in scalar cloud environments on fully generic orbits around spinning black holes.

The dimensionless tidal deformability, $Λ$, of neutron stars (NSs), inferred from gravitational-wave (GW) observations, has thus far been used primarily to constrain the pressure of dense matter near twice nuclear saturation density, leaving the core equation of state (EOS) largely inaccessible to inspiral-phase GW observations. We show that the core EOS can be probed directly through $Λ$ using a perturbative analysis of the dimensionless stellar-structure and tidal-response equations formulated in terms of scaled intrinsic variables, without invoking any specific EOS model. We uncover a remarkable EOS-insensitive scaling relation between $Λ$ and the central EOS parameter $\mathrm{X}\equiv P_{\rm c}/\varepsilon_{\rm c}$, where $P_{\rm c}$ and $\varepsilon_{\rm c}$ denote the central pressure and energy density, respectively. The relation is validated against a broad ensemble of physically viable EOSs. Applying it to tidal deformabilities inferred from events such as GW170817 enables a direct determination of $\mathrm{X}$. We further derive a tight lower bound, $Λ_{\rm{TOV}}\gtrsim 9.2^{+1.2}_{-1.2}$, for maximum-mass NSs along stable mass-radius sequences, quantitatively demonstrating that even the most compact stable NSs remain distinctly separated from black holes, for which $Λ_{\rm{BH}}=0$. These findings reveal a previously unrecognized connection between inspiral-phase tidal deformability and the core EOS, establishing a direct link between GW observables and the microphysics of ultradense matter in the strong-gravity regime. The resulting scaling establishes inspiral-phase tidal deformability as a direct and largely model-insensitive probe of the EOS of NS cores.

Quantum entanglement in cavity magnomechanical system has a wide range of applications in modern quantum technologies. In this work, we propose a theoretical scheme to generate and enhance quantum entanglement through cross-Kerr nonlinearity and mechanical parametric amplification (MPA) in a magnomechanical system. Our system is made of a magnonic mode that is simultaneously driving the acoustic phononic and the center-of-mass motion (CMM) phonon in a yttrium iron garnet sphere. The acoustic mode and the center-of-mass mechanical (CMM) mode are weakly coupled via the phonon hopping rate $J_m$. Moreover, the magnonic and phononic modes interact through cross-Kerr interaction, while the phononic mode is additionally driven via a Mechanical Parametric Amplification (MPA). Without the mechanical coupling ($J_m = 0$) and the MPA, the generation of entanglement among the subsystems requires a relatively strong effective cross-Kerr coupling. However, when phonon hopping and MPA are accounted, quantum entanglement can be generated even for weak values of the cross-Kerr coupling strength, revealing the key role of these interactions in the engineering of quantum correlations in our proposal. Furthermore, the related purity of the generated entangled states remains high for the same parameter's regime, revealing that the generated quantum entanglement is established without significantly increasing the mixing of the involved states in the system. Our work suggests how robust and stable quantum correlations can be engineered in magnomechanical structures based on nonlinear interactions. These results are useful for modern quantum applications including quantum information processing, quantum communication, and quantum computational tasks.

Low-entropy motional states of ion Coulomb crystals are an essential prerequisite for a plethora of applications and are typically prepared by laser cooling. As larger crystals are operated in the quantum regime, it remains unclear, and has recently become debated, whether increasing the ion number can be beneficial for cooling. Here, we investigate theoretically and experimentally many-ion sideband cooling and the role of collective effects in different spin-motion coupling regimes. For weak coupling, the many-body effects are insignificant. In the strong-coupling regime, however, the spin and motional subsystems undergo a coherent state swap, enabling cooling by a suitably timed laser pulse. Using planar Coulomb crystals with up to 91 ions, we demonstrate that the residual mean phonon occupation after one such pulse scales as $1/N^2$ with the number of ions. By iterating the pulses, we measure mean phonon occupations $<2\cdot10^{-4}$. For large crystals in the coherent regime, we further show that the spin-motion dynamics becomes largely independent of the initial phonon statistics. Through spin measurements, the state-swap mechanism can be utilized to probe the phonon distribution in the mode.

We investigate device-independent certification of global randomness based on Bell inequality violations in bipartite scenarios with three outcomes per party. Our goal is to determine whether multi-outcome measurements allow one to surpass the amount of randomness achievable with binary outputs in realistic scenarios. We begin by analyzing several known Bell expressions and evaluating their robustness against noise for randomness certification. We then introduce a systematic method for generating new Bell expressions within structured families and perform a large-scale numerical study. We find that a substantial number of inequalities certify significant amounts of min-entropy. In particular, we identify simple inequalities that achieve near-maximal global randomness while involving a reduced number of measurement settings, thus improving the balance between certified randomness and number of inputs. Moreover, the vast majority of nontrivial certificates exhibit robustness against realistic noise, maintaining positive certified randomness away from the ideal regime. These results demonstrate that strong device-independent randomness expansion in multi-outcome scenarios is not restricted to carefully engineered inequalities, but arises generically within suitably constructed families of Bell expressions.

Device-independent random number generation (DIQRNG) is the most secure form of generating private randomness using quantum physical processes. Its strength lies in producing numbers that are impossible to predict by any eavesdropper restricted by the laws of quantum theory. Moreover, security is proven solely from observed measurement statistics, without the need to characterise or trust the devices used in random number generation. Implementing DIQRNG is, however, costly, as it requires high-quality entangled systems. It is therefore important to make the best use of available resources. In this work, we show that using projective measurements -- which are most readily implementable experimentally -- one can certify $2\log(d)$ bits of device-independent randomness from a bipartite system of local dimension $d$ for every $d \ge 2$, thus reaching the theoretically maximum possible rate of DIQRNG. We provide explicit protocols reaching $2\log(d)$ bits based on mutually unbiased bases. Furthermore, we compute numerical bounds on the rate for the case of imperfect implementations, showing that our protocols are robust to experimental noise.

We present two novel families of bipartite Bell inequalities designed to achieve optimal global randomness certification for an arbitrary number of outputs $d$. We first use symmetry arguments to argue that their maximal quantum violations certify $2\log d$ random bits. For the first family, we construct a quantum realization using $d\times d$ maximally entangled states which provides a quantum violation that we conjecture to be optimal for any $d$. It is then numerically shown that the obtained quantum violation certifies optimal global randomness, up to numerical precision, for $d=3,4$. For the second family, we provide the optimal quantum violation and its quantum realization for any $d$, again using $d\times d$ maximally entangled states and projective measurements over at least two unbiased bases on one of the parties. We self-test this realization for $d=3$, which implies the optimal certification of two fully random trits.

We propose and characterize a protocol for rapid near-ground state cooling of the center-of-mass (c.m.) mode of a large trapped ion crystal. When the initial mean thermal occupation of the mode $\bar{n}_i$ is small compared to the number of ions $N$, a red sideband drive implements a beam-splitter type SWAP operation between the mode and the collective spin of the $N$ ions, with the latter effectively serving as a quantum harmonic oscillator. Subsequently, a reset of the spins removes the entropy, leading to near-ground state cooling of the c.m. mode. We term this protocol as quantum beam-splitter cooling (QBSC). We analyze the impact of several practical imperfections on the final temperature achievable under QBSC, including finite ion number, off-resonant carrier and blue-sideband contributions, and the impact of the sideband drives arising from spectator modes. In addition, we outline practical strategies to eliminate the carrier drive. Furthermore, we show that measuring the population statistics of the ions at the end of the SWAP operation can enable near-optimal quantum beam-splitter thermometry (QBST), with the classical Fisher information approaching the quantum Fisher information of a thermal state. We discuss the connection of QBSC with continuous sideband cooling and compare QBST with a recently proposed rapid adiabatic passage-based thermometry scheme. Our work constitutes an example of harnessing many-body effects to open new routes to laser cooling and thermometry in large trapped ion crystals.

We develop a framework for quantum measurement verification (QMV) using only local state preparations. For locally transitive and irreducible projective measurements, we prove that symmetry reduces locality constrained QMV to quantum state verification of a single basis state, thereby reducing protocol design to the optimization of homogeneous verification operators. We apply the framework to generalized Bell measurements, single-parameter measurements on two qubits, elegant joint measurements, and stabilizer state induced measurements, and derive explicit local protocols together with closed form verification operators, success probabilities, and sample complexities. We further show that homogeneous QMV protocols can estimate measurement fidelity directly from observed passing frequencies.

We study the stability of a generalization of the GHS-GM black hole in the presence of a dilaton and a monopole field. We find that the thermal behaviour of system depends on the scalar charge of the dilaton field and as this parameter is decreasing the system becomes more thermally stable. We also find that, as the charge of the black hole is increasing, both the real and the imaginary parts of the quasi-normal frequencies decrease in absolute value. The overtone modes die out faster than the fundamental modes and no positive imaginary parts appear, indicating the stabilty of the system

A distinctive feature of Quantum Reservoir Computing (QRC) is the ability to directly embed quantum input states into the reservoir dynamics. However, the resulting output is fundamentally linear for a single input state, preventing QRC from naturally computing nonlinear functionals such as purity or entropy. We overcome this limitation with a quantum-classical hybrid architecture combining a qubit quantum reservoir with a classical echo state network (ESN), allowing both nonlinear functional approximation and effective temporal processing. We systematically study performance under two information regimes: full-tomography and partial information (single-axis measurements), with the latter demonstrating that the hybrid system outperforms its standalone components in both linear and nonlinear tasks due to the enhanced information retrieval provided by the quantum reservoir. Building on these results, we apply an online monitoring protocol that explicitly accounts for measurement back-action and finite measurement ensembles, enabling a realistic assessment of performance under experimental conditions. These results establish hybrid quantum-classical reservoir computing (HRC) architectures as a practical and scalable route for enhanced quantum machine learning on near-term qubit hardware.

We demonstrate a dual-chip InP-SiN photonic QKD system with on-chip pulse generation and reconfigurable decoding, implementing time-bin BB84 with finite-key security against coherent attacks. The system sustains a QBER below 4% and delivers secret keys at kbps rates over 150-250 km of fiber.

We show that adaptivity, also called feed-forward, is a powerful resource in reducing the error of measurement circuits that combine multiple uses of a noisy measuring device. In previous work, it has been shown that error in measurements can be mitigated by using the measuring device multiple times. So far, the most effective protocols have been parallel measurement schemes, where all measurements are simultaneous. We extend this idea to adaptive measurement schemes, where the results of previous measurements can influence our choice of processing further down the circuit. We show that adaptive measurement circuits can in general reduce overall measurement errors further than any non-adaptive measurement circuit could with the same number of measurements. In particular, we show that for a large class of noisy two-outcome qubit measurements, such an adaptive advantage can exist when as few as three measurements are used. We also show that the adaptive advantage is unbounded across the class of noisy two-outcome qubit measurements, as the number of uses of the device increases. As part of this work, we devise and use methods for finding optimal measurement circuits in both the non-adaptive and adaptive cases. In addition, we prove general results about the limits of such circuits, both in measuring a qubit, and more generally, a qudit.

We investigate the ground-state phase diagram of a frustrated spin-1/2 Heisenberg ladder in the transverse magnetic field with an anisotropic inter-rung exchange coupling $α$. This bond-anisotropic parameter continuously interpolates among the Ising-like limit $α=0$, the isotropic point $α=1$, and the XY-dominated regime $α\gg 1$. Using density-matrix renormalization group calculations within a matrix-product-state framework, we analyze bipartite entanglement, magnetization, and spin correlations to characterize the emergent quantum phases. We identify six distinct ground-state phases, including rung-singlet, Haldane-like, Tomonaga-Luttinger liquid, canted Ising-ordered, XY-polarized, and ordered ferromagnetic states. While magnetization and local correlations provide a first insight into phase classifications, the finite-size scaling of the entanglement measures offers a more sensitive and unified diagnostics of both gapped and gapless regimes. In the gapless regime, we extract the central charge $c \simeq 1$, confirming Tomonaga-Luttinger liquid behavior, while at the transition between the canted Ising and ferromagnetic phases, we observe $c \simeq 1/2$, consistent with the Ising universality. Finally, we find that both the Haldane-like behavior and the extended critical Tomonaga-Luttinger liquid regime are strongly confined to the vicinity of the isotropic point $α= 1$.

In variational quantum algorithms, QAOA, and quantum annealing, the cost Hamiltonian defines the optimization landscape explored by the quantum hardware; however, in many application-driven formulations, this Hamiltonian is a simplified proxy for the true task-level objective. Using lattice-based quantum protein structure prediction as a case study, we investigate whether the contact-energy cost Hamiltonian commonly used in this setting is sufficiently aligned with structural accuracy, as measured by RMSD against experimentally determined structures. Through this specific problem, we show the importance of studying the reliability of the cost Hamiltonian independently from the quantum approach used. This work shows that, for small peptides and on average, the energy landscape of the considered cost Hamiltonian is not correlated well enough to the actual error to provide meaningful predictions. Moreover, this correlation was estimated through Monte-Carlo sampling for larger instances. It shows an increase of said correlation for larger problem instances and when more interaction shells are considered. This investigation illustrates the meaningfulness of investigating cost Hamiltonian relevance independently from the quantum algorithm used.

Fault-tolerant quantum computing based on lattice surgery requires place-and-route compilation with low spacetime overhead. Routing, in particular, faces a basic tension between suppressing path conflicts through greater spatial allocation and exploiting the time direction to realize ancilla-efficient spacetime routing. Existing approaches do not fully resolve this trade-off while retaining compatibility with inner factory layouts and termination guarantees. Here we introduce double-slice routing, a constant-depth spacetime-routing method that uses two consecutive time slices with a guarantee that its kink-parity correction terminates under both planar and stacked architectures. We numerically benchmark the resulting compiler on Hamiltonian-simulation workloads to show that double-slice routing reduces compilation cost by up to a factor of 2.4 over a single-slice baseline. Compared to projective routing, an existing method that allows an unbounded number of time slices per path, double-slice routing achieves smaller circuit volume with only a marginal execution-time penalty. Combined with a cultivation-compatible mapping optimization, the overall improvement reaches up to 7.5-fold over a naive single-slice compilation baseline. These results identify double-slice routing as a practically useful operating point in lattice-surgery compilation and show the substantial benefit in joint optimization of mapping and routing.

In this paper, using a novel geometrical approach, we relate the concept of the thermodynamic temperature of the free gravitational field, to the non-affinity of real null geodesics in a Newman Penrose tetrad. This naturally links various temperature functions like Clifton, Ellis and Tavakol temperature, Hawking temperature, Unruh temperature etc., in their respective proper limits. Although our analysis is done within the realm of local rotational symmetry, we show that the result can be extended to other Petrov type D geometries, like the Kerr spacetime. We also obtain the geometrical and causal transport equations for this temperature function, in the form of a hyperbolic wave equation with a forcing term, sourced by Weyl curvature and matter. Finally, as a possible physical interpretation of the non-affinity, we relate the geometrical temperature with the gravitational red/blue shift of light rays travelling along null geodesics.

We reexamine the quantum circuit model of black hole evaporation proposed in Class. Quantum Grav. 35, 235013 (2018). This model incorporates the tripartite systems: black hole, just radiation and early radiation. We apply the scrambling unitary matrix with a single parameter $θ$ to the ground state of the qubits in infalling matter towards a black hole in order to generate initial qubit states of the black hole that are more general than those in Class. Quantum Grav. 35, 235013 (2018). Specifically, the scrambling unitary matrix reduces to no scrambling and maximum scrambling when $θ=0$ and $θ=π/2$, respectively. Our aim is to explore the role of quantum monogamy in the firewall formation between the black hole and radiation. In this model, entanglement and firewall formation depend on the black hole mass $M$ and the frequency of Hawking radiation $ω$. For the initial state with $θ=π/2$, a firewall emerges at an earlier stage of the evolution than with $θ=0$. We also find that a firewall structure emerges between BH and JR, and that the information is carried away by radiation for all values of $Mω$, provided that $θ$ lies within a certain analytically determined range. Following the unitary gate dynamics, the initial black hole qubit state can be retrieved from its imprint on the final radiation state, which was originally hidden behind the black hole's horizon. These results may provide insight into the properties of multipartite entanglement due to the different initial states in the evolution of a quantum circuit model for black hole evaporation.

The maximum field-free orientation attainable from a \emph{thermal} molecular ensemble within a finite rotational subspace has not been characterised analytically. Here we derive a Schur--Horn-type upper bound on the field-free orientation $\avg{\cosθ}$ achievable by \emph{any} $M$-conserving unitary control acting on a Boltzmann ensemble truncated to rotational levels $J\le\Jmax$. The bound is the sum, over magnetic-quantum-number sectors, of the sorted Boltzmann weights paired with the sorted spectrum of $\cosθ$; it is purely kinematic, set by temperature, the rotational constant, and $\Jmax$ alone, and interpolates exactly between the zero-temperature subspace eigenvalue and a finite-temperature ceiling fixed by the rotational partition function. Benchmarked on LiH against the analytical $N$-subpulse resonant protocol of Hong \textit{et al.} [Phys.\ Rev.\ Research \textbf{7}, L012049 (2025)], it reveals three regimes: the protocol saturates the bound to within about $1.6\%$ for $T\le\SI{5}{\kelvin}$, loses roughly $10\%$ of its zero-temperature orientation at $T=B/\kB\approx\SI{10.8}{\kelvin}$, and leaves a $10$--$40\%$ gap above $\SI{10}{\kelvin}$. Optimising the subpulse areas and carrier phases within the fixed layout closes only $\approx7\%$ of that gap, the optimal phase offsets vanishing, which localises the dominant loss in the rigidity of the analytic layout rather than in the choice of areas or phases. The bound is a control-independent target for coherent-control design, applies unchanged to any linear polar molecule in a $^1Σ^+$ state, and is mapped across the $(\Jmax,T)$ plane.

We revisit thawing quintessence models with nearly flat scalar-field potentials using a cosmographic framework. Earlier work indicates that the cosmographic reconstruction of the slope $λ=-(dV/dφ)/V$ of the quintessence potential in the general case requires the knowledge of the cosmographic paremeters up to the jerk parameter $j$. In this work we show that the slow-roll conditions $[(dV/dφ)/V]^2 \ll 1$ and $|(d^2V/dφ^2)/V| \ll 1$ allow the reconstruction of the slope of a nearly flat potential with knowledge of only the deceleration parameter $q$ (and the density parameter $Ω_φ$). Confronting the assumption of near-flatness with the cosmographic data after DESI DR2, however, reveals possible tension between the two. We further show that these models exhibit attractor behaviour in the $w$--$Ω_φ$ and $w$--$w'$ phase planes, corresponding to a universal thawing evolution with $w \approx -1$ at early times. We also derive the corresponding relation in the cosmographic $q$--$j$ plane and show that different cosmological expansion histories can produce the same thawing evolution. Nevertheless, all viable trajectories remain close to the $Λ$CDM limit $j=1$.

Newtonian noise is expected to be the dominating noise source for low frequencies at the Einstein Telescope. It originates from seismic waves that cause density fluctuations in the rock around the interferometer. The mitigation strategy for Newtonian noise relies on an array of seismometers, placed at depth in boreholes, which provides measurements of the seismic wave field. We optimize the positions of the individual seismometers for the mitigation capabilities of the array for a full corner of the Einstein Telescope. We find that the mitigation capabilities of arrays with multiple seismometers in each borehole match the capabilities of only somewhat smaller arrays but with only one seismometer per borehole. Mitigation is further improved by extending the array with seismometers in the interferometer tunnels. Such configurations may hence provide a cost-effective way towards realizing an efficient seismic array. In each case, we quantify the broadband mitigation performance in the range from 1 to 10 Hz for arrays that are optimized for a frequency of 10 Hz, as well as the robustness of the arrays with respect to variations from their optimized positions. We find that larger arrays with several seismometers per borehole and additional seismometers in the tunnels provide promising broadband performance above 3 to 4 Hz and that such arrays are particularly stable against variations in the seismometer positions with mitigation factors $>6$ for an array of 20 boreholes with 3 seismometers each and $>15$ for a large array of 50 boreholes with 10 seismometers.

The holographic time-like entanglement entropy (TEE) extends entanglement to time-like boundary subregions. While its definitive holographic dictionary remains debated, one concrete proposal utilizes piece-wise extremal surfaces. In this work, we adopt this geometric prescription as an exploratory framework to holographically investigate the late-time ($τ_0\to \infty$) growth of TEE in asymptotically AdS black holes with a space-like singularity and no inner horizon. By assuming a Kasner geometry near the space-like singularity and using null energy condition, we analytically show that a critical extremal surface $\mathcal{A}_c$ inside the event horizon completely governs the late-time linear growth of the TEE. This result suggests that the late-time behavior of TEE is tightly constrained by the geometry of black hole interiors. Using numerical results from Einstein-scalar theory, we find a robust behavior: the vacuum Schwarzschild-AdS geometry sets an upper bound on the growth rate of the real part and a lower bound on the imaginary part. We prove these bounds in static planar symmetric case under dominant energy condition and conjecture that it should be true in more general cases.

With increasingly large numbers of gravitational-wave events, population inference is now beginning to move beyond predictions of marginal distributions and to probe correlations between compact-binary parameters such as masses, spins, and redshifts. These correlations have strong constraining power for both astrophysics and tests of general relativity. In this paper, we present an idealized analytical model to study the interplay between single-event correlations, systematic biases, and population-level correlations. With this, we investigate the potential emergence of false-positive measurements of population-level correlations. We quantify how the presence of correlations at the single-event level between a pair of parameters increases the uncertainty of population-level correlations for those parameters, potentially obscuring the true underlying population correlation if present. We also find that if waveform systematics lead to biases that are correlated across the catalog (which is likely, because certain regions of the parameter space are more difficult to model), this can be effectively absorbed by a population analysis that targets correlations and can be misinterpreted as such. This simple Gaussian-based model may serve as a broad compass for future, more detailed explorations.

The phonon laser realized in a trapped ion, i.e., a self-sustained mechanical oscillator, has demonstrated the unique characteristics in practically detecting externally applied electric signals without the prerequisite of sideband cooling. Entering the quantum regime via sideband cooling is expected to further improve its sensing performance. Here we report the first experimental realization of a quantum-regime single-ion phonon laser ($\bar{n}<10$) using a trapped $^{40}\mathrm{Ca}^+$ ion and demonstrate electrometry based on its phase-space symmetry-breaking response to weak resonant electric fields. By tuning the phonon-laser parameters, we reveal that the sensing performance is fundamentally governed by the finite-time relaxation dynamics of the underlying open quantum system. We find that a slow Liouvillian relaxation, correlated with the finite experimental interaction window, effectively enhances the dynamic susceptibility while maintaining the structural robustness of the limit cycle. This regime, when applied to the detection of electric fields, produces a shot-noise-limited peak sensitivity of $14.15 \pm 0.77~μ\mathrm{V/m}/\sqrt{\mathrm{Hz}}$ and a minimum detectable field variation of $δE_{\mathrm{min}} \approx 1.83~μ\mathrm{V/m}$. Our results establish quantum phonon lasers as a practical platform for advanced sensing and highlight the central role of Liouvillian dynamics in non-equilibrium electrometry.

Timelike orbits constitute a crucial probe for exploring the intrinsic properties of curved spacetimes, and the carried gravitational radiation signals provide a direct window into strong field gravity. In this paper, we develop a versatile computational framework based on Mathematica and the OpenMP parallel architecture to simulate the rational orbits of timelike particles and their gravitational radiation in static spherically symmetric spacetimes. Specifically, requiring only the user defined covariant metric, this numerical tool can efficiently calculate rational orbits across various configurations, as well as the corresponding gravitational wave polarization states and characteristic strains. The package presented here offers a highly efficient and comprehensive one-stop solution for investigating the properties of curved spacetimes and their potential observational signatures. To demonstrate the reliability and capability of our code, we apply it to the Schwarzschild spacetime as a test case, illustrating the functionality of the code across several key aspects, including the effective potential, stable orbital regions, rational and irrational orbits, and gravitational wave signals. Furthermore, we show that the gravitational waves emitted by an extreme mass ratio inspiral system composed of an intermediate mass black hole and the Galactic Center supermassive black hole have the potential to be identified by future space detectors.

The Unruh effect predicts that accelerated observers perceive the inertial vacuum as populated by particles, providing a flat-spacetime analogue of Hawking radiation. Its direct observation, however, remains experimentally challenging, since an Unruh temperature of $1\,\mathrm{K}$ requires accelerations of order $10^{20}\,\mathrm{m/s^2}$. Here, we show that the Lamb shift of a centripetally accelerated atom inside a high-$Q$ cavity provides a sensitive spectroscopic probe of the Unruh effect at dramatically lower accelerations. The cavity reshapes the electromagnetic density of states and converts otherwise tiny noninertial corrections into tunable level shifts. Depending on the atomic angular velocity and cavity detuning, the Lamb shift can be enhanced, strongly quenched, or completely screened. Remarkably, for experimentally realistic parameters, a rotation-induced shift of order $10\;\mathrm{Hz}$ can arise already at accelerations as low as $0.5\,\mathrm{m/s^2}$, more than twenty orders of magnitude below the acceleration scale conventionally associated with direct Unruh detection. These results identify cavity-controlled Lamb-shift spectroscopy as a viable route toward laboratory tests of the circular Unruh effect in the ultralow-acceleration regime.

In this work, we show how simulations performed on classical computers such as those of Numerical Relativity can be tackled by quantum algorithms for solving systems of partial differential equations. We develop a proof-of-principle quantum algorithm for solving Einstein Field Equations in the Wahlquist-Estabrook-Buchman-Bardeen(WEBB) tetrad Numerical Relativity formalism [1], and test it by evolving the Schwarzschild Black Hole spacetime in the WEBB Numerical Relativity formalism [2], perturbing it to obtain gravitational Quasinormal Modes [3]. We program the algorithm components for a gate-based, digital quantum computer using the Qiskit software [4] and run it on classical simulators and physical IBM quantum computers through the UKRI National Quantum Computing Centre (NQCC) Quantum Access program and quantify the computational resources and runtime.

Current era quantum computers continue to grow in both capability and capacity. Despite these advancements, errors induced by environmental noise severely limit practical applicability. Current research into error mitigation and correction to bridge the gap between current-era quantum computers and the execution of noise-sensitive workloads. These methods have significant performance and resource overheads, thereby greatly limiting the real-world benefits of their use. Circuit replication, as a naive form of error mitigation, is not new and has largely been ignored given the resource constraints of current quantum hardware. However, its simplicity is attractive as a means to supplement modern methods, reducing the overall performance overhead while still preserving error-mitigation capabilities. In this paper, we profile the effects of simple circuit replication under real-world noise profiles to better establish replication's limits as a supplemental mitigation strategy. Quantum Approximate Optimization Algorithm (QAOA) for the Maxcut problem is explored for the analysis. For small graphs, we found that the average inference strength decreases by approximately 21.8% while the average standard deviation decreases by 108.8% compared to 6 replicates. For larger graphs, inference strength decreases by 35.4% while the average standard deviation decreased only 20.5%. Fewer replications did not affect smaller graphs, but degraded inference strength, with comparable benefits to standard deviation in larger graphs. These results show that replication has potential uses as a supplemental mitigation strategy for large-depth, highly variable workloads.

The rapid appearance of supermassive black holes at high redshift motivates a reassessment of non-baryonic growth channels. We develop a loss-cone framework for collisionless dark-matter (DM) capture by early black-hole seeds, with particular attention to phase-space depletion and refilling. The calculation combines Eddington-inverted NFW-like halos, a relativistic direct-capture boundary, and an orbit-averaged Fokker-Planck treatment of angular-momentum transport. Primordial black holes (PBHs) are treated as massive perturbers whose refilling strength depends on both their abundance and individual mass. Collisionless refilling by triaxial or chaotic centrophilic orbits is included as a phenomenological upper-envelope channel. We show that ordinary stellar relaxation gives negligible DM-driven growth for the fiducial high-redshift seeds. PBH-driven granularity can yield order-of-magnitude growth in sufficiently compact halos, while triaxial or full-loss-cone supply can produce a rapid early burst. In the self-consistent calculations, however, the evolution generally becomes supply-limited. Once the accessible low-angular-momentum phase-space reservoir is depleted, the capture rate collapses and the black-hole mass saturates. Fixed-background NFW calculations therefore overestimate sustained growth, especially in the full-loss-cone limit. A TNG50-calibrated NFW benchmark gives negligible growth even under optimistic refilling assumptions. Collisionless DM capture is therefore unlikely to solve early SMBH growth in generic NFW-like halos, but it can provide a radiatively dark upper envelope in rare compact environments with efficient angular-momentum refilling.

We present a systematic, model-agnostic analysis of out-of-equilibrium effects, including viscosity and thermal conductivity, in non-radial oscillations of relativistic stars. Extending the Lindblom-Detweiler formalism, we construct, to our knowledge, the first general framework for linear, non-radial relativistic stellar perturbations that incorporates generic nonequilibrium corrections to the perfect-fluid sector in both the even- and odd-parity channels. Our framework is formulated in terms of the tensorial structure and thermodynamic decomposition of generic corrections without relying on any specific constitutive relations, thereby allowing us to elucidate, at a structural level, how these effects enter the perturbation equations and contribute to geometric deformations and fluid fluctuations. As an application, we consider the Bemfica-Disconzi-Noronha-Kovtun fluid and perturbatively investigate shifts in the frequencies and damping times of modes connected to their perfect-fluid counterparts in the limit of vanishing transport coefficients. We also identify structural features of the closed eigenvalue problem that can give rise to additional mode families. Our formalism provides a unified framework for analyzing how different relativistic fluid theories modify the structure of non-radial stellar perturbations.

Successive interference cancellation (SIC) is an important technique for 5G/B5G wireless receivers to resolve interfering signals from multiple users. While SIC has been proven to approach the channel capacity limit in many settings of multiple access networks, it remains unknown if this limit can be surpassed by an advanced yet practically implementable quantum receiver technique. In this work, we answer this quest by proposing a full-stack quantum receiver that integrates front-end quantum sensing with back-end quantum signal processing. The proposed technique leverages simple qubit ensembles without using any complex entanglement resources to parallelly extract interfering multiuser signals, with a channel capacity beyond the limit of SIC in some operational corners. In performance evaluation, we report notable quantum advantage over the ultimate multiple-access channel capacity limit and practical SIC algorithms in low-SNR regimes in terms of spectral efficiency and detection efficiency due to the receiver's ability to exploit quantum correlated measurements and superposition-enabled parallel processing.

Why do unrelated approaches to quantum gravity nearly all find spacetime two-dimensional at the shortest scales? Each theory answers only within its own dynamics; we highlight a single kinematic route to the same value, one assuming no field equation and living in the geometry of the space of states alone. That route is concentration of measure on the Fubini-Study geometry of pure states, which forces the pairwise distances of a random sample to equalize as the dimension grows, so any finite sample collapses to an equidistant dust whose thresholded metric graph is the complete graph. Handed this dust, a diffusion probe reads it as two-dimensional in the large-sample limit, the value the running spectral dimension takes at the dust's single relaxation scale, a property of the measurement rather than the structure; this convergence on two is not, by itself, evidence that spacetime is two-dimensional. Whether a given two is such an artifact is governed by the Laplacian spectrum near zero, and whether that reading carries across an emergence map is the condition we call spectral faithfulness; a single relaxation scale encodes no spectral dimension that tells one structure from another. The collapse, the probe value, and the eigenvalue-density criterion are machine-checked in Lean 4 against Mathlib, resting on the standard Beta law of overlaps; a power-law tail of small eigenvalues reads a genuine dimension, a single scale above a gap reads two at its own clock, and a gapped two-scale band reads off the universal line. These classes are run on graph-Laplacian proxies, and whether a link-graph reading carries to the physical nonlocal operator is left open. The spectral test reads the eigenvalue density near zero and separates, on a given structure, a measurement artifact from a dimension the structure genuinely expresses.

The dark bubble model explains the existence of a positive cosmological constant by making explicit use of the instabilities underlying the de Sitter swampland conjectures to make the accelerated expansion of the universe inevitable. A distinctive consequence of the construction is a unique hierarchy connecting cosmological, gravitational, string, and higher-dimensional scales. In particular, the model naturally predicts the existence of a dark dimension of micron size, an idea that has been argued for on independent grounds in the literature. The same framework also predicts a weakening of gravity at distances of order the dark-dimension scale, leading to a fading of the gravitational force at micron distances. We argue that the dark bubble therefore provides a concrete realization of both the dark dimension proposal and Sundrum's fat graviton scenario, in which gravity effectively ceases to probe shorter distances. Additional predictions include a string scale of order tens of TeV and a measurable positive spatial curvature of the universe. We review these key aspects of the model, discuss their implications for gravity and cosmology, and highlight its key predictions.

We study the dynamical and scattering properties for the Euler-Heisenberg BH surrounded by perfect fluid dark matter. The geometry contains a compact non-linear electrodynamic correction governed by the EH coupling and a logarithmic dark-matter contribution governed by the surrounding PFDM halo. We study the scalar, electromagnetic and a particular effective axial spin-2 channel constructed on the fixed EH plus PFDM background, acting as a proxy for the gravitational-like perturbation problem and not for the fully coupled gravitational perturbation problem. We compute the quasinormal-mode spectrum employing a thirteenth-order WKB method supplemented with Padé resummation and compare it with the eikonal prediction calculated in terms of the angular frequency of the photon sphere and the Lyapunov exponent. Moreover, we study the near-extremal configurations and derive a purely imaginary branch of quasinormal frequencies in the near-horizon region, whose damping rate increases with the PFDM parameter and is nearly spin-independent. We then compute exact greybody factors by direct numerical integration of the radial wave equation and compare them to analytical lower bounds. We also analyze the absorption cross sections and the Hawking emission spectra. We also calculate the relativistic enhancement of the neutrino-antineutrino annihilation channel outside the EHPFDM black hole. We find that the PFDM parameter contracts the optical structure, increases the oscillation frequency, enhances the damping rate and suppresses transmission. On the other hand, the Euler-Heisenberg correction leads to a weaker near-horizon deformation whose effect becomes relevant for sufficiently large charge. These results provide a common scattering framework for comparing the impact of dark-matter environments and nonlinear electrodynamics

The interface between the quantum and the classical is an intriguing and, at times, hotly contested subject of ongoing research. The quantum regime is characterized by interference, made possible by the superposition principle, while such phenomena are absent in macroscopic, everyday experience. Here, we investigate the link of this absence (or, as we will argue, unobservability) to computational complexity. We show how the assumption that quantum systems cannot solve NP-complete problems efficiently implies that certain formally valid quantum measurements on finite-dimensional systems are unperformable. We study several consequences of this restriction. First, Pauli matrices in an inconveniently transformed basis are a simple example of unobservables. Furthermore, some quantum states are not connected by any physically realizable time evolution. Finally there are quantum states whose coherence cannot be observed, i.e. superpositions of pure quantum states which are indistinguishable from mixtures. We discuss the connection of this phenomenon to the presence of superselection sectors. Our results suggest that the apparent classicality of macroscopic systems may be partly due to limitations on measurements and time evolutions imposed by computational complexity.

Quantum low-density parity-check (QLDPC) codes are strong candidates for fault-tolerant quantum computation, but efficient decoding remains a major challenge due to short cycles, degeneracy, and the poor convergence of standard belief-propagation (BP) decoders. We propose a reinforcement learning-based list sequential (RL-LS) BP decoder for QLDPC codes by extending the reinforcement-learning-based sequential variable-node scheduling (RL-S) framework with list-based search. At each step, the learned policy selects the next variable node to update; the decoder then retains the ordinary RL-S trajectory while also exploring a competing branch obtained by softly biasing the post-update LLR pair toward the second-most likely Pauli symbol, recomputing the incident local BP messages, and setting the visited variable node to that second-best symbol. Candidate trajectories are ranked and pruned using our proposed cumulative path metric. The resulting decoder extends the learned decoder by combining the improved convergence of learned sequential scheduling with list exploration. Numerical results on representative QLDPC benchmark codes over the depolarizing channel show that our proposed method improves the decoding performance of the underlying decoder and compares favorably with existing BP-based decoding methods.

The renormalization of the primordial inflationary power spectrum has long raised the possibility that ultraviolet effects could significantly alter predictions for cosmological observables. We demonstrate that inflation dynamically suppresses the entire renormalization sector: while super-Hubble perturbations freeze after horizon crossing, renormalization contributions decay rapidly during inflation. As a consequence, the observable primordial spectrum is remarkably insensitive to renormalization ambiguities, providing strong evidence for the robustness under renormalization of standard inflationary predictions at observable scales.

We investigate a K-essence like cosmological model whose scalar-field potential is constructed from a negative power-law Sáez--Ballester potential. By means of a suitable field redefinition from $φ$ to $\varphi$, we show that the resulting field equations acquire a mathematical structure analogous to that of a previously solved Friedmann-Lemaître-Robertson-Walker (FLRW) cosmological model. This correspondence allows us to obtain exact classical solutions for both the scale factor and the scalar field within the Hamiltonian formalism. The resulting cosmological dynamics exhibits a late-time accelerated expansion, with the deceleration parameter approaching the asymptotic value $q\rightarrow -1$, characteristic of a de Sitter phase. At the quantum level, the corresponding Wheeler-DeWitt (WDW) equation is derived and exact quantum solutions are obtained. These results provide a consistent classical and quantum description of the cosmological evolution generated by this class of K-essence models. In this formalism, the scalar field remains as a cosmic background where the universe unfolds, which is glimpsed from the quantum solution perspective.

Since the Fermi satellite's identification of a candidate $γ$-ray burst (GRB) temporally coincident with GW150914, several tentative, and often debated, associations between electromagnetic (EM) transients and gravitational-wave (GW) signals from binary black hole (BBH) mergers have been reported. One such event, S241125n (later confirmed as GW241125_010116), was identified during the fourth observing run (O4) of Advanced LIGO and found to be spatially (within large GW localization uncertainty) and temporally coincident with a subthreshold GRB detected by the Swift Burst Alert Telescope Gamma-ray Urgent Archiver for Novel Opportunities (BAT-GUANO). Here, we present results from a radio follow-up campaign targeting the BAT-GUANO localization region, carried out with the Karl G. Jansky Very Large Array (VLA). We also re-analyze Swift/XRT observations of the field, and combine these results with optical upper limits. Our analysis constrains the isotropic kinetic energy of a putative relativistic jet launched in the BBH merger to $\lesssim 3 \times 10^{50}$ erg for $n_{ISM} = 1.0 cm^{-3}$. We also discuss both the challenges and the diagnostic power of radio follow up in assessing candidate BBH-GRB associations, and present projections for analogous radio studies in the LIGO-Virgo-KAGRA observing run 5 (O5), and in the era of next-generation ground-based instrumentation. The enhanced sensitivity and localization capabilities of detector networks such as Cosmic Explorer and the Einstein Telescope, paired with the enhanced sensitivity of next-generation radio interferometers such as the next-generation VLA and the Square Kilometre Array, will significantly strengthen coordinated multi-messenger follow-up of BBHs. These next-generation facilities are likely to provide an answer to whether BBHs host relativistic ejecta powered by mini-disk accretion.

We investigate the dissipative dynamics of multilevel quantum systems strongly coupled either to a lossy cavity mode or directly to a bosonic environment. By deriving spectral localization bounds, we establish conditions under which strong system-bath coupling gives rise to a hierarchy of population relaxation timescales. Our approach builds on the reaction-coordinate polaron-transform framework. By mapping the original strong-coupling problem onto an effective weakly dissipative model, we analyze the spectrum of the resulting Liouvillian superoperator through localization bounds. For the generalized V model, we find that strong system-bath coupling gives rise to a bright-dark structure in the effective system-bath coupling operator: a single collective mode remains strongly coupled to the environment, while the remaining modes become progressively dark. Consequently, the dynamics separate into fast and slow sectors and, at finite coupling strengths, develop a hierarchy of population relaxation timescales. Numerical simulations based on both secular and non-secular quantum master equations corroborate the emergence of timescale separation and the pronounced slowing down of dissipative dynamics at strong coupling. Our results reveal a general mechanism underlying anomalously slow relaxation in strongly coupled open quantum systems and provide a route for engineering long-lived states through system-environment interactions.

We advocate an alternative paradigm to operate quantum computers that utilizes multipartite entanglement generated in dedicated auxiliary systems during idle times. This stored entanglement enhances the future capabilities of the quantum processor, as it can be flexibly used to assist and speed-up computations when needed. We identify classes of multipartite entangled resource states whose computational power are related to their entanglement features, and in turn to the complexity to generate them. During idle times, one can thus continuously work towards generating more and more powerful auxiliary multipartite entangled states. Idle times include both times prior to the start of a computation, but also any step during the execution of an algorithm where parts of the processor are not actively involved. To illustrate our approach, we consider architectures with limited connectivity, e.g. corresponding to a 1D geometry. We show that $d$-dimensional cluster states allow one to flexibly perform multiple long-distance two-qubit gates in parallel, where both the complexity to generate them, as well as the number of achievable gates increases with $d$.

This work presents a comprehensive account of a gravitational model based on the gauging of the $\mathrm{ISIM}(2)$ Lorentz subgroup parameters through the so-called soft algebra formalism. After presenting a closed form for $\mathfrak{sim}(2)$ and its inhomogeneous extension, we proceed to develop a complete torsionless gravitational model, culminating in an invariant action.

Efficient simulation of strongly-interacting fermionic systems on quantum processing units (QPUs) is a challenging task due to nonlocal mode entanglement generation. However, it is not yet well understood how the structure of entanglement governs the hardness of large-scale quantum chemistry simulations or the scaling of distributing such workloads. Here, we introduce an algorithm for estimating the distribution complexity of hybrid quantum-classical simulation for electronic structure Hamiltonians over heterogeneous high-performance architectures. Our algorithm relies on efficient analytical evaluation of the low entanglement boundaries for the orbital rotations and dephasing-induced localization within tensor fragments, in a double-factorized representation. Our entanglement estimation scales as $O(N^3)$ for each fragment, where $N$ is the number of orbitals. When QPUs are communicating via a quantum network, the cost of distribution per fragment is reduced quadratically from $O(N^2)$ to $O(N)$. Similarly, for hybrid quantum-classical approaches, with access to only conventional HPC interconnects, the worst-case cost is reduced from $O(\exp(N^2))$ to $O(\exp(N))$. We show that emergent entanglement patterns are induced by the interplay between coherent Gaussian orbital rotations and disordered Coulomb interactions. We discuss the underlying physical mechanisms that govern distribution complexity and introduce model systems that are tunable based on the localizability of fragments and the overlap of interfragment rotations. We characterize three different regimes of hardness for distribution complexity and classical simulability. The framework introduced here enables novel and more efficient quantum-classical application workflows towards utility-scale quantum computing.

We present a variational approach to quantum field theory based on wavefunctions parameterized by neural networks. While variational methods have a celebrated history across many fields, their application to quantum field theory has been limited by well-known challenges. We show that neural-network wavefunctions, combined with modern machine-learning techniques, enable competitive variational calculations in nontrivial field theories. As a demonstration, we reproduce the essential features of the two-dimensional nonlinear $σ$-model: asymptotic freedom, dynamical mass generation and the model's step-scaling function.

We perform 3D hydrodynamics simulations of an equal-mass quasi-circular live $10^6~{\rm M}_\odot$ massive black hole binary (MBHB) embedded in a prograde, locally isothermal circumbinary disk (CBD) with $0.1$ aspect ratio. The binary evolution is driven by the gaseous torques and its dynamics is described with $2.5$ post-Newtonian corrections. This approach allows us to track the influence of the CBD on a gravitational-wave (GW) driven MBHB inspiral from $55$ to $46$ Schwarzschild radii, i.e., at its early evolution in the LISA band at redshift $z\sim1$. For the first time for the $0.1$ aspect ratio disk, we report the measurement of gravitational and accretion torques with and without concurrent GW emission. We also report how the morphology of the accretion time series onto the MBHB modestly alters when GW emission is the dominant binary evolutionary mechanism. Lastly, we find that the gas-induced orbital phase-shift is $0.12$ rad over $600$ orbital cycles, which LISA should detect at $z=1$. Our results have implications for multi-messenger astronomy, since observation of accretion rate modulation by LSST/Roman surveys and phase-shift by LISA will provide crucial information on the complex environment surrounding MBHBs.

We present a pedagogical introduction to the cosmological constant problem that requires only basic knowledge of quantum field theory and general relativity. A massive real scalar field is used to illustrate how the quantum vacuum energy density and pressure can be calculated both in flat spacetime and in an expanding universe. Detailed computations are provided for dimensional, cutoff, and adiabatic regularizations. No attempt is made to address quantum gravitational effects, and the expanding-universe background is treated classically. We point out that although the commonly cited discrepancy of 120 orders of magnitude between theory and observation is based on an estimate that does not account for regularization and renormalization, fundamental principles of quantum field theory nevertheless lead to a huge mismatch. In addition to this large discrepancy, we emphasize that there are also conceptual challenges related to cosmic expansion, such as the choice between comoving and physical scales in certain contexts and the non-uniqueness of vacuum.

In the decay of quantum particles under field theoretic consideration, the decay rate is typically a convolution of the density of modes the primary field is allowed to decay into and the allowed probability density for the field to decay into such modes. In free space, though many such processes show high amplitude of such transitions towards the infrared sector, the depletion of allowed mode density in that regime arrests the efficacy of such decays at low energies. Therefore in free space, in order to enhance the decay rate, one needs the transition probability density to be rich enough towards the high energy sector where mode density support is also high enough to make the rate sufficiently large. In this work we argue that in the controlled boundary condition environment e.g. in a cavity, the mode functions of product field receive significant support towards their infrared sector, boosting the probability (and hence rates) of low energy processes. The cavity geometry offers sweet spots in terms of resonant geometry around which the interaction of a primary field with product fields receives dramatic enhancement, significantly enlarging its decay rates. Therefore, a judicious selection of cavity geometry serves as a potential substitute to studying interesting processes at high energy. The results have direct relevance for the study of QED processes and implications for the study of exotic new physics are also discussed.

Over the past century, both theoretical advancements and experimental observations have established General Relativity (GR) as the most successful framework for describing gravitational phenomena. However, observations from multiple cosmological probes over the last two decades have provided compelling evidence for the accelerated expansion of the Universe. The rapid progress in observational astronomy and precision cosmology has highlighted several challenges that motivate the search for extensions or alternatives to General Relativity. In this thesis, we investigate alternative formulations of gravity based on non-Riemannian geometry, with particular emphasis on (f(Q)) gravity in the presence of bulk viscosity. We first explore the existence of exact cosmological solutions for viscous fluid models within the framework of (f(Q)) gravity and constrain the free parameters of these solutions using observational datasets. The resulting constrained models are then employed to study the evolutionary history of cosmic expansion. Our analysis demonstrates that the proposed viscous (f(Q)) gravity models can successfully account for the observed late-time acceleration of the Universe. Furthermore, we examine several classes of nonlinear viscous (f(Q)) gravity models through both dynamical systems and observational analyses. Particular attention is given to the cosmological viability of these models and their ability to reproduce the different evolutionary epochs of the Universe, ranging from matter-dominated eras to accelerated expansion phases.

Quantum error-correcting codes (QECs) are essential components of quantum computation and have deep connections to quantum phases of matter. A key obstruction to passive self-correcting QECs is the presence of string logical operators, which can generate logical errors through constant-energy-barrier processes. Haah's Codes (fracton codes) showed that three-dimensional stabilizer codes can forbid such string logical operators, but their translation-invariant structure supports self-similar fractal logical operators with a logarithmic energy barrier. We introduce the qutrit random cubic codes, a family of local qutrit Calderbank-Shor-Steane stabilizer Hamiltonians with similar cube-check structure as Haah's Code 1 but built from spatially varying stabilizers. We prove that these models retain the no-string property and numerically observe that they have properties distinct from translation-invariant fracton codes: the smallest ground-state degeneracy exponent is $k=2$ for odd $L$ and $k=4$ for even $L$; noncontractible plane-logical operators span the entire logical space; and charge-push diagnostics show that the self-similar fractal operators are absent. These results demonstrate that constrained randomness can fundamentally change the nature of stabilizer codes and improve their self-correction properties. They further point to broader families of quantum error-correcting codes and quantum phases beyond canonical topological and fracton orders.

We present a quantum algorithm for simulating Generalized Parton Distributions (GPDs) in the Schwinger model. Unlike the staggered fermions widely utilized in current quantum simulations, we employ Wilson fermions for lattice discretization. This choice is critical for the quantum computation of GPDs due to their strict preservation of charge conjugation symmetry. We construct a comprehensive algorithmic framework that includes the preparation of hadronic states with non-zero momentum and the measurement of light-cone correlation functions incorporating Wilson lines. We provide a complexity analysis, demonstrating that the resources required for our algorithm scale polynomially with both the number of qubits and the desired precision $\varepsilon$. Finally, we benchmark our approach using exact diagonalization, extracting mass spectra and GPDs (also parton distribution functions) that are consistent with theoretical expectations and fundamental physical constraints.

We study O+O collisions at $\sqrt{s_\mathrm{NN}} = 5.36$ TeV within a fully three-dimensional $\text{McDipper}$+$\text{MUSIC}$ model, which allows us to describe the experimentally measured dependence of charged hadron multiplicity on centrality and pseudorapidity. We show that the initial elliptical eccentricity is mainly driven by the fluctuations of the energy deposition and thereby varies considerably event-by-event within a fixed centrality class. This also holds for elliptic flow $v_2$, whose origin in O+O thus differs from that in collisions of heavy nuclei. Using a decomposition of initial states in an average event and uncorrelated modes, we find that despite the large size of fluctuations we can reproduce the joint probability distribution of eccentricity and elliptic flow with a reasonable accuracy with only a small set of fluctuation modes.

We study collisions of domain walls in $SU(N)$ gauge theories using the Polyakov-loop effective potential models. We find that string junctions play a crucial role in the dynamics of $Z_N$ domain walls. In $SU(3)$ gauge theory, the merger of two non-planar $Z_3$ domain walls into a single wall proceeds via the creation of a vortex--antivortex pair in $2+1$ dimensions. In $SU(4)$ gauge theory, low-energy collisions of $Z_4$ walls result in the formation of a single domain wall without the creation of vortices. At higher collision energies, two $Z_4$ domain walls can either bounce back or scatter into another pair of domain walls through the formation of vortices. The creation of vortex--antivortex pairs generalises to the creation of string loops in $3+1$ dimensions for both gauge theories. These results demonstrate a direct dynamical role for topological strings in the evolution of center-domain-wall networks and reveal a new aspect of defect dynamics in non-Abelian gauge theories.

Motivated by recent experimental progress and theoretical developments, we investigate the Cabibbo-favored $b\to c$ governed ${\overline B}_{d,s}$ ${\to}$ $D_{d,s}^{(*)+}S^{-}$~($S$$=$$K_0^*(1430)$, $a_0(1450)$) weak decays by considering the next-to-leading (NLO) contributions within QCDF framework. With the updated values of $B_{(s)}\to D_{(s)}^{(*)}$ transition form factors obtained from a covariant light-front quark model, branching ratios are estimated in two scenarios for scalar mesons. It is found that the branching ratios for ${\overline B}^0{\to}D^{+}{a_0^-}$ and ${\overline B}_{s}^0{\to}D_{s}^{+}{a_0^-}$ decays can reach up to the order of ${\cal O}(10^{-4})$ in scenario-2 by assuming that the scalar mesons are lowest-lying p-wave states, which deserve high-priority experimental searches and may be observed in the ongoing LHCb and SuperKEKB experiments.

Despite many years of development in integration-by-parts reduction, reconstructing all reduction coefficients, which are rational polynomials of kinematic variables and the space-time dimension, remains a non-trivial problem. The main difficulty comes from the large number of unknowns in a general ansatz, which can lead to a linear system that is too large to solve. In this paper, we present an algorithm for reconstructing reduction coefficients through a sequence of intermediate bases. The resulting analytic reduction coefficients are products of a few analytic matrices, whose non-zero entries are simple rational polynomials. We demonstrate the efficiency of this algorithm with two cutting edge examples: a three-point massive box-triangle and a four-point massive pentagon-triangle. Reconstructing all reduction coefficients for the box-triangle (pentagon-triangle) requires 3289 (13013) numerical samplings, significantly fewer than the number of unknowns in the general ansatz, 1407406 (21638331).

In \cite{Lindstrom:2026quz} it is pointed out how the tensionless string with a gauged scale symmetry discussed in the recent articles, \cite{Sheikh-Jabbari:2026cnj,Sheikh-Jabbari:2026vqh,Sheikh-Jabbari:2026tpf}, is a reduction of the conformal string \cite{Gustafsson:1994kr} to Minkowski space. Here we corroborate this by choices of slices in Dirac $d+2$ dimensional conformal space. We perform a Dirac reduction of the model and its algebra of constraints and see how they map to the constraints in $d$ dimensions, including how the semidirect product of the Virasoro algebra with a su(1,1) Kac-Moody algebra becomes the Corrollian-Weyl symmetry in $d$ dimensions.

The measurement of the Higgs boson self-coupling is one of the key physics goals of the LHC program, providing direct access to the shape of the Higgs potential and offering sensitivity to physics beyond the Standard Model (BSM). Higgs pair production probes the trilinear self-coupling, while triple Higgs production is sensitive to the quartic coupling. In addition, resonant searches for heavy scalars decaying to pairs of Higgs bosons or to a Higgs boson and a new scalar are motivated by a broad class of BSM models. Recent results on multi-Higgs production searches from the ATLAS and CMS experiments at the Large Hadron Collider (LHC) are presented, covering both resonant and nonresonant production modes across a variety of final states.

We investigate a holographic s+p superconductor model coupled to nonlinear electrodynamics in the probe limit. The equations of motion are solved numerically, and the condensates as well as the grand potential curves for phase transitions at different values of the nonlinear parameter $b$ are illustrated. It is found that as $b$ increases, both the pure s-wave and p-wave condensates are suppressed. From the $b-T$ phase diagram, we observe that the region of the pure s-wave phase gradually shrinks with increasing $b$, which is attributed to the stronger suppression on the s-wave condensate compared to the one on the p-wave. Moreover, a smaller charge ratio $q_p/q_s$ is needed for the s+p coexistent phase to appear as $b$ grows. A particularly interesting feature is that, due to the nonlinear self-interaction of the electromagnetic field, charge accumulates spontaneously outside the event horizon from the bulk perspective even in the absence of scalar and vector condensates, thereby invalidating the conventional formula for the superconducting charge density. We improve the definition of the superconducting charge density as the background-subtracted value of the accumulated charge outside the horizon with respect to that in the normal phase. Furthermore, the optical conductivity in the normal phase is also modified by this accumulated charge for finite $b$, and its imaginary part develops a minimum at a finite frequency. This minimum persists in the superconducting phase near the critical point, confusing the extraction of the gap frequency.

We derive compact analytical formulae for the spectrum of nonlinear Compton scattering in a finite plane-wave pulse with a smooth temporal envelope. The strong-field QED probability is reduced to finite-pulse phase integrals, which are evaluated asymptotically for multicycle pulses with a broad class of smooth envelopes. We use the uniform approximation to remove the caustic divergences that appear at the nonlinear edges of broadened harmonics. Away from the caustics, it reduces to the standard saddle-point result. The behavior near the linear edge is further improved by an envelope-corrected saddle-point approximation. The approach retains the harmonic substructure in the spectral-angular region carrying the dominant part of the emitted radiation. The locally monochromatic approximation is recovered by averaging the finite-pulse interference. Within their asymptotic domain of applicability, the resulting formulae agree with direct numerical calculations and can be used to evaluate spectra from an electron beam.

In the framework of effective field theory, we study the anomalous $t{\bar q}Z$ interaction through the process $μ^+μ^- \to t{\bar q}Z$ at future muon colliders with $\sqrt s= 3, 10, 14\,\text{TeV}$. Based on the top quark decay modes involving $W$ and $Z$ bosons, we first divide the signal into six cases. Then, in order to obtain the limits on the corresponding branching ratios, we perform a detector simulation for both signals and Standard Model backgrounds. To enhance the signal significance, we exploit the polarization of the muon beams and employ the fat jet method to reconstruct signals in hadronic final states. For $\sqrt s= 14\,\text{TeV}$ with $20\,\text{ab}^{-1}$, we find that the upper limit on the branching ratio for $t\to qZ$ can reach the order of $\mathcal{O}(10^{-8})$, which exceeds the limits provided by the CMS and ATLAS collaborations by 2 to 3 orders of magnitude. Our study thus demonstrates that TeV-scale muon colliders can provide an efficient and complementary platform for probing rare top quark interactions.

We propose a novel model for pseudo-scalar dark matter (PSDM) by extending the Standard Model (SM) with a dark gauged $U(1)_X$ symmetry, but without dark charged fermions. We impose a $Z_2$ symmetry to ensure the stability of pseudo-scalar dark matter and regard the $U(1)_X$ symmetry as being broken dominantly by a large VEV of the singlet scalar field. The would-be Goldstone associated with the $U(1)_X$ gauge boson is almost orthogonal to the direction of PSDM. As a result, we show that PSDM appears as a stable pseudo-Nambu-Goldstone boson receiving the mass from the $U(1)_X$ invariant mixing potential and the corresponding cross section for direct detection gets suppressed even for the weak-scale mass of PSDM. We also show that the correct relic density can be explained by the PSDM annihilations into the SM particles or into a pair of light Higgs-like scalars, being compatible with the bounds from Higgs invisible decay, Higgs data and indirect detection.

The simplest nontrivial amplitude in planar $\mathcal{N}=4$ super-Yang-Mills theory is six-gluon scattering in the maximally-helicity-violating configuration. It has been computed to 8 loops with the help of antipodal duality, which relates it to the three-point form factor of a protected operator, the chiral stress tensor super-multiplet, represented also as ${\rm tr} φ^2$. In this talk, we describe the computation to 8 loops of another three-point form factor, for the operator ${\rm tr}φ^3$. This form factor lives in the same restricted space of polylogarithms as the ${\rm tr}φ^2$ form factor. We also report on all-order patterns for sequences of coefficients in the symbols of these polylogarithmic results, for the leading discontinuity of the ${\rm tr}φ^3$ form factor.

Based on $(10.087 \pm 0.044) \times 10^{9}~J/ψ$ events collected at the center-of-mass energy $\sqrt{s} = 3.097$~GeV with the BESIII detector, we search for the charged lepton flavor violating decay $η\to e^{\pm}μ^{\mp}$ through the process $J/ψ\to γη'$ with $η' \to π^{+} π^{-} η$. No signal is observed, and an upper limit on the branching fraction is determined to be $\mathcal{B}(η\to e^{\pm}μ^{\mp}) < 6.8 \times 10^{-7}$ at the 90\% confidence level. This result improves the previous best limit by one order of magnitude.

Line bundle cohomology on complete intersection Calabi--Yau threefolds is an important input in string phenomenology. Previous work, based on direct extrapolation and on machine-learning analyses of finite data sets, has shown that these cohomology dimensions often admit chamber-wise polynomial descriptions on the Picard lattice. In this paper we explain this structure through the Hilbert functions associated with the non-trivial maps in the Koszul spectral sequence. Once Bott--Borel--Weil theory determines the non-vanishing ambient cohomology groups, the rank defects of the relevant Koszul maps are governed by Hilbert functions of cokernel or kernel modules. This turns many empirical chamber formulae into explicit rank-defect statements, with proofs that are either analytic or certified on finite boxes. Using this approach, we recover analytically almost all of the piecewise formulae appearing in the existing literature. For the remaining cases, we prove finite-box certificate theorems for infinite families of line bundles in the specified regions. We also identify new wall structures on certain CICYs which refine the chamber decompositions suggested by finite-range data. Finally, we use the same framework to construct line bundle cohomology formula libraries for two CICY threefolds which had not previously been analysed in this way. These results suggest that Hilbert-function methods can provide promising faster inputs for future string model-building scans based on line bundle constructions.

We propose a noninvasive SQUID-based polarimeter for the polarized proton beam in the Electron-Ion Collider (EIC) Hadron Storage Ring (HSR), exploiting the collective magnetic dipole moment of the bunches rather than scattering. The six-snake HSR lattice has synchronous-particle spin tune $ν_s = 1/2$, placing the in-plane spin-precession signal at half the revolution frequency ($\sim$39 kHz), in the DC SQUID band. Three pickup channels (cosine-$θ$ and sine-$θ$ saddle loops for the transverse components, a coaxial axial gradiometer for the longitudinal one) reconstruct the full polarization vector $(P_x, P_y, P_z)$ in two complementary modes. Static mode, the default for continuous noninvasive monitoring, reads all three components: $P_y$ at the revolution frequency and the residual in-plane components at $ν_s f_\mathrm{rev}$, bunch by bunch over an hours-long fill, including $P_z$, inaccessible to single-spin scattering polarimetry by parity conservation. Dynamic mode gives a precise polarization-magnitude measurement: a longitudinal kicker tips a small fraction of the polarization into the horizontal (ring) plane to produce a free-induction-decay (FID) signal, and many phase-locked tip-$π$-echo-restore cycles are summed coherently via a matched filter across all bunches, with $\mathcal{O}(α^2/π^2) \sim 10^{-4}$ loss per cycle, negligible over a full $δP/P = 1\%$ measurement. For tipping angle $α= 30$ mrad, polarization $P = 0.7$, and effective rms spin-tune spread $σ_{ν_s}^\mathrm{eff} = 10^{-3}$ (coherence time $\sim$2 ms), the integration time to reach $δP/P = 1\%$ is about 18 s at injection and 5 min at flattop. The architecture extends to deuteron and $^3$He beams via species-specific spin-magnetic factors, with applications to storage-ring EDM searches.

This paper investigates the rare decays of the Z boson into S-wave quarkonia within the Bethe-Salpeter formalism. Both the production of double S-wave quarkonia and radiative decays into single S-wave quarkonium are analyzed. For double quarkonia production, i.e., $Z\to VV$ and $Z\to VP$ (where $V$ and $P$ represent vector and pseudoscalar quarkonia, respectively), we consider the leading order contribution from both QCD as well as electromagnetic transition via virtual photon (QED) amplitudes. The heavy quark limit is adopted to simplify calculations. Additionally, we have introduced another possible leading order channel for Z-boson decays to double S-wave quarkonia, where the Z boson decays into bottomonium plus charmonium via QED amplitude. Such processes may include $Z\to J/ψ+Υ(1S)$, $Z\toη_{b}+J/ψ$ and $Z\toη_{c}+Υ(1S)$. Moreover, we have also studied radiative Z boson decays to S-wave quarkonium, namely $Z\to X(Q\bar{Q})γ$, where $X(Q\bar{Q})=J/ψ$, $Υ$, $η_{c}$, and $η_{b}$. Interestingly, for double charmonium and radiative charmonium production, our results are larger than the NRQCD finding, while for the bottomonium case, our findings are comparatively smaller. This shows that charmonium is a relativistic particle, while bottomonium is a non-relativistic particle.

The proton's unpolarized $u(x)-d(x)$ parton distribution function (PDF) has been studied by a number of lattice QCD groups through large momentum expansion. However, due to lattice artifacts (excited state contaminations, unphysical pion masses, and discretization effects) and less-advanced theoretical analysis (renormalizations, large-distance extrapolations, and large-log resummations), the resulting PDFs cannot be compared strictly with experimental data. By using the state-of-the-art theoretical tools and mitigating the lattice artifacts empirically, we reanalyze the available datasets in the literature and find that the new PDF in the physical limits is consistent with global fittings within $\sim1σ$. This provides compelling evidence that large momentum expansion is capable of accurately predicting the $x$-dependence of the PDFs when ideal lattice data become available.

Experimental results in high-energy physics are inherently hypothesis-dependent, as collider analyses are designed to test specific theoretical frameworks. The proliferation of theoretical models creates a mismatch between the pace of theory development and experimental validation, motivating the reinterpretation of existing results in terms of alternative hypotheses. Existing strategies face limitations: full reanalysis requires enormous computational resources and experimental expertise, while simplified methods sacrifice accuracy by neglecting kinematic variations and systematic uncertainties. This thesis develops, validates, and applies a novel reinterpretation method that preserves statistical rigour while dramatically reducing computational requirements, enabling rapid inference on alternative theoretical parameters through distributable likelihoods. The method reweights the joint distribution of reconstruction and kinematic variables according to the ratio of theoretical predictions of alternative to null hypotheses. Starting from a histogram-based likelihood, this produces updated templates for any theoretical model while preserving correlations and systematic uncertainties. The method is validated through examples and applied to the Belle II $B^+ \to K^+ ν\barν$ measurement ($2.7σ$ tension with the Standard Model). Two analyses are presented: a Weak Effective Theory reinterpretation constraining Wilson coefficients, and a light new physics search via ${B^+ \to K^+ X}$ two-body decays. The public release of the resulting model-agnostic likelihoods establishes an important precedent for open science in experimental particle physics. By facilitating rapid theoretical validation, this approach democratizes experimental data access, enhances scientific return on investments, and advances FAIR Data principles in high-energy physics.

The conventional and topological phase transitions of a four-dimensional asymptotically AdS black hole with a non-minimal coupling term $F^{αβ}F^{γλ}R_{αγ}R_{βλ}$ are investigated. Using a perturbative approach to first order in the coupling $ε$, the thermodynamic quantities are derived and the first law and Smarr relation are verified. Intriguingly, while the $ε= 0$ limit yields the standard Reissner--Nordström--AdS black hole belonging to the $W^{1+}$ topological class ($W = +1$), switching on the non-minimal coupling fundamentally transforms the topology to the $W^{0-}$ class ($W = 0$). This transition occurs while the van der Waals-type first-order phase transition survives in the intermediate region, embedded within the overall Hawking--Page pattern, rendering the system a hybrid black hole thermodynamic system. The coupling $ε$ thus acts as a topological deformation parameter that alters the universal classification of the system, despite the perturbative nature of the solution.

Warm inflation is usually expected to completely deplete the inflaton condensate by dissipating its energy into radiation. We show that this expectation fails in a simple and observationally viable regime. In a strongly dissipative warm inflationary scenario, the dissipative ratio, $Q=Υ/(3H)$, can fall rapidly after the end of inflation as the system approaches radiation domination, thereby suppressing further energy transfer to the thermal bath. This leads to a residual inflaton condensate, which subsequently evolves as an effectively non-dissipative scalar field. For potentials with a stable quadratic minimum, this remnant inflaton manifests as a cold dark matter component. We establish this mechanism for the minimal renormalizable potential, with a dissipative coefficient $Υ\propto T^3$. In this case, current cosmological data allow strong dissipation while leaving the inflaton mass weakly constrained by inflationary observables. The observed dark matter abundance then fixes its mass to be $m \approx 0.02\,{\rm MeV}$, while larger masses overclose the Universe. The transition to matter-like scaling occurs well before BBN, avoiding a long-lived inflaton dark radiation component. Relic inflaton dark matter therefore turns the post-inflationary dynamics of warm inflation into a new late time constraint on its parameter space.

We use numerical minimization of the lattice action of trace-deformed Yang-Mills theory on $T^4$ with twisted boundary conditions to find the classical minimum action configurations of fractional topological charge. We vary the twists and ratios of torus periods to interpolate between different $R^{4-k} \times T^k$ geometries. This allows us to see how the corresponding minimum action saddle point configurations -- monopole-instantons ($k=1$), center vortices ($k=2$), and fractional instantons ($k=3,4$) -- morph into each other. We also study how the transition between them depends on the presence of a deformation potential. In particular, we argue that the recent analytic picture of chains of monopole-instantons collimating their flux into center-vortex sheets, while technically relying on the deformation potential, also holds in pure Yang-Mills theory, for tori whose shape causes the abelianization due to the deformation to align with the one due to the twists. Our results also indicate that with nonzero deformation potential, some transitions between different minimal-action fractional charge configurations may be discontinuous and involve level crossing.

We give a summary of our work [3] and [4] with Geoffrey Compère and Sk.Jahanur Hoque. After quickly reviewing the $Λ$-BMS solution space and why it is not trivial to define an analogue of Bondi mass loss, we discuss how this can be resolved by studying the linearized solutions. Then we describe how to solve the linearized Einstein s equations in the generalized harmonic gauge and the set of assumptions such as adiabaticity and being at a large distance from the source that allows us to write the metric perturbations in terms of the source multipole moments. We point out how to consistently truncate the multipole expansion. We then discuss the net change of the metric perturbation at future infinity due to a source that is time-varying in a finite time interval that will be connected to the memory effect. We explain how to define charge flows using the holographic stress tensor and what form they take in terms of multipole moments at second order in perturbations. We conclude with a some remarks on the results.

We construct magnetic and electric Carrollian Maxwell theories by performing Kaluza-Klein-like null reduction of a complex Maxwell field in a Bargmann deformed light-cone background with manifest gauge symmetry. The procedure preserves a first-class U(1) Gauss constraint throughout the Carrollian limit. Gauge invariance is therefore maintained in our Hamiltonian formulation. By choosing different scalings, we obtain standard magnetic Carrollian theory and electric Carrollian theory. However, a scalar field could appear in the Carrollian theory in a coupled or decoupled way, which has not been found by previous methods. This result fully reveals the diversity of Carrollian theories accessible through the deformed light-cone Kaluza-Klein-like null reduction method. Furthermore, our work provides an explicit example of the correct application of this approach, thereby broadening the scope of its applicability to gauge theories.

We investigate a discrete warped Chern-Simons description of three-dimensional warped gravity based on boundary holonomies and monodromy sectors. Starting from the lower-spin SL(2, R) + U(1) gauge structure associated with warped AdS(3) holography and warped conformal field theories (WCFTs), we construct a discrete boundary framework in which ordered products of link holonomies replace continuous gauge connections along noncontractible cycles. In this setting, boundary monodromies become the primary gauge-invariant observables characterizing the physical sectors of the theory. We show that the hyperbolic, eliliptic, and parabolic sectors naturally arise from the conjugacy classes of the discrete SL(2, R) monodromy, while the additional U(1) holonomy supplies the warped contribution to the boundary charges. Using these monodromy invariants, we derive a discrete entropy relation entirely from boundary holonomy data without relying on a smooth geometric thermal background. The resulting entropy reproduces the characteristic warped black-hole and WCFT structure in the continuum limit. We further demonstrate that the continuum warped holonomy conditions are recovered from the large-lattice limit of the ordered boundary products, establishing a direct correspondence between discrete monodromies and continuous Wilson loops. Our analysis suggests that warped gravitational thermodynamics may be understood from a fundamentally holonomy-based perspective in which boundary monodromy sectors provide an alternative organizational description of the physical states within the discrete warped framework. Keywords: warped Chern-Simons theory, boundary monodromies, holonomy sectors, Wilson loops, warped thermodynamics.

We investigate a quintessence dark-energy model with an exponential potential and an exponential coupling to cold dark matter (CDM), hereafter referred to as the CQ-EXP model, using Planck CMB, DESI BAO, and DES-Dovekie supernova observations. We also examine how variations in the neutrino mass sector affect the constraints. When the neutrino mass sum is fixed at $\sum m_ν=0.06$ eV, the data favor a coupling between quintessence and CDM, with the coupling parameter $β$ deviating from zero at more than $3σ$. In particular, the observations favor the $β<0$ branch, where the energy transfer between the two dark sectors changes sign and the effective equation of state (EoS) of dark energy crosses the phantom divide, $w=-1$. When the effective neutrino mass parameter $\sum m_{ν,\mathrm{eff}}$ is treated as a free parameter, the data show a preference for negative values of $\sum m_{ν,\mathrm{eff}}$. This additional freedom weakens the preference for the coupling between quintessence and CDM and leads to nearly identical values of $χ^2_{\rm min}$ for the CQ-EXP models with $β>0$ and $β<0$, corresponding respectively to models without and with phantom-divide crossing in the effective EoS. Both values are slightly larger than that obtained in the $w_0w_a$CDM model, indicating that the CQ-EXP model cannot be statistically distinguished from the $w_0w_a$CDM model with the data considered here. Therefore, when $\sum m_ν$ is fixed, current observations favor the CQ-EXP model with phantom-divide crossing. In contrast, when negative values of $\sum m_{ν,\mathrm{eff}}$ are allowed, a CQ-EXP dark energy without crossing $w=-1$ can also provide an effective explanation of the latest observations.

Monte Carlo event generators are essential tools in collider physics, providing the link between theoretical predictions and experimental measurements through fully exclusive event simulation. Future collider programmes, particularly high-precision lepton colliders will place significantly increased demands on their accuracy and scope. This contribution reviews key challenges in MC generator development, including electroweak corrections, initial-state radiation, beam dynamics, perturbative QCD, and non-perturbative modelling. The discussion is not exhaustive and reflects a selective choice of topics.

We present a comprehensive investigation of single $Λ$ hyperon production via the lepton-nucleon deep inelastic scattering (LNDIS) process, ${\ell}^{-}$ $+$ $p$ ${\to}$ $ν_{\ell}$ $+$ $Λ$, in the experimental environment of electron-positron colliders. Our approach utilizes incident leptons originating from the decays of resonances (${J/ψ}$, $ψ(2S)$, $Υ(1S)$, $Υ(2S)$, and $Z^{0}$) produced in $e^{+}e^{-}$ collisions, which then scatter off stationary protons in the surrounding detector materials. The differential and total cross sections are calculated using baryonic transition form factors parameterized with the $z$-expansion scheme within both the quantum chromodynamics (QCD) sum rule and lattice QCD frameworks. Our results indicate that the cross section increases with center-of-mass energy and is highly sensitive to the choice of form factors, resulting in significant theoretical uncertainties. This study highlights the experimental challenges in observing the LNDIS process at $e^{+}e^{-}$ colliders and underscores the need for improved determination of baryonic form factors. It serves as a valuable reference for future experimental searches and suggests that an anomalous observation of single $Λ$ hyperon production at $e^{+}e^{-}$ colliders could indicate new physics.

We study the resurgent structure of short-time heat kernel asymptotics from the viewpoint of Picard-Lefschetz theory. For a real analytic Riemannian manifold, we show the heat kernel admits a 1-Gevrey small-time expansion whose Borel transform detects complex-geometric data beyond the real geodesic sector. We formulate an infinite-dimensional Picard-Lefschetz problem of Morse-Floer type for the holomorphic energy functional on the complexified path space, and propose a heat-kernel analogue of the Picard-Lefschetz/Alien correspondence. In this framework, pointed alien operators acting on the asymptotic expansion associated with the real geodesic are predicted to produce the formal heat-kernel sectors associated with other holomorphic geodesics, with coefficients given by signed counts of connecting trajectories of the Morse flow. We perform a confirming test of this proposal on the hyperbolic plane $H^2$.

Scalar-induced gravitational waves are usually modeled as a classical stochastic background sourced by primordial curvature perturbations. We investigate whether residual quantum-information properties of the scalar sector can survive decoherence and leave imprints in the induced tensor background. Using the covariance-matrix formalism, we describe primordial curvature perturbations as decohered two-mode squeezed Gaussian states and identify the anomalous scalar coherence that may remain after scalar entanglement has vanished. We then derive the leading scalar-to-tensor transfer relations for opposite-momentum induced tensor modes. The ordinary tensor power is sourced by scalar power contractions, whereas the opposite-mode tensor coherence is sourced by anomalous scalar-coherence contractions. This tensor coherence controls the induced Gaussian discord and generates connected and phase-sensitive observables, including a connected power covariance $κ(k)\propto |γ_k|^2/α_k^2$. Thus the robust signature is not a universal shift of the gravitational-wave spectrum, but a correlated tensor background with nontrivial covariance and phase structure. We discuss phenomenological templates and provide an illustrative Fisher estimate for future gravitational-wave observations. Our results suggest that scalar-induced gravitational waves may offer a new probe of primordial quantum correlations beyond entanglement.

Scalar-induced gravitational waves are usually treated as a classical stochastic background sourced by phase-random curvature perturbations. We show that this description can miss residual quantum information. Starting from a decohered two-mode Gaussian scalar state, we derive explicit transfer relations between the scalar anomalous coherence and the covariance matrix of induced tensor modes. For a localized scalar power spectrum, the ordinary tensor power is sourced by scalar power contractions, whereas the opposite-mode tensor coherence is sourced by scalar anomalous-coherence contractions. This coherence can generate nonzero tensor discord and a connected tensor-power covariance even after scalar entanglement has vanished. We identify the connected covariance and phase-sensitive strain correlations as probes of primordial quantum coherence in secondary gravitational-wave backgrounds, and discuss their possible relevance for future space-based interferometers and pulsar timing arrays.

Recent baryon acoustic oscillation measurements from DESI provide important new clues for reassessing whether the standard $Λ$CDM model offers a sufficient description of the late-time expansion history of the Universe. When combined with cosmic microwave background and type Ia supernova data, these measurements show an apparent departure from the $Λ$CDM model, commonly described as dynamical dark energy (DDE) with equation of state crossing the phantom divide (i.e., quintom behavior). This review examines the current status of the DESI-motivated indications for DDE and their possible implications for physics beyond $Λ$CDM. We discuss how the strength of the preference for DDE depends on the adopted parametrization and dataset combination, and how residual systematics or internal tensions among datasets may affect its interpretation. At the background level, several mechanisms beyond $Λ$CDM can produce similar expansion histories. We therefore further discuss how the same effective departure from $w=-1$ may arise from physically distinct scenarios, including interacting dark energy, non-minimally coupled gravity, and non-standard dark matter. Meanwhile, these different new-physics interpretations may have different implications for current cosmological tensions, especially those involving $H_0$, $S_8$, and $\sum m_ν$. In conclusion, the question posed by DESI is not merely whether dark energy evolves with time, but rather how, within the framework of precision cosmology, to disentangle new physics scenarios from systematic errors.

We study what Lorentz-equivariant jet taggers learn internally, using equivariance tests, linear probes and grade ablations across five models including L-GATr, L-GATr-slim and LLoCa-T. Linear probes show that equivariant models suppress frame-dependent pseudorapidity to zero while encoding jet mass and N-subjettiness strongly. Grade ablations on L-GATr reveal that bivector channels are negligible for top-quark tagging while vector-like channels are dominant but seed variable, consistent with the network exploiting multiple representational pathways. These results characterize which physical features and algebraic grade structures carry discriminative information in equivariant taggers and may inform future development of such models.

We numerically calculate the shear viscosity of quark-gluon matter via the Green-Kubo relation with an improved ZPC model. We include all $2\leftrightarrow 2$ parton cross sections at finite temperature, which are based on perturbative QCD and screened with thermal masses, and consider massless quark-gluon systems with Boltzmann statistics in chemical equilibrium. We then compare the Green-Kubo results with the analytical results from the leading-order Chapman-Enskog method for the same parton cross sections over the temperature range $150-600$ MeV. We also examine the simpler case of isotropic and constant parton cross sections. Overall, we find that the two methods agree rather well. Specifically, the Green-Kubo results are greater than the Chapman-Enskog results by an average of $\sim 9\%$ for isotropic and constant cross sections and by an average of $\sim 3\%$ for finite-temperature pQCD cross sections, where the difference between the two methods is presumably due to higher-order corrections to the leading-order Chapman-Enskog results.

We investigate the gravitational form factors (GFFs) and the generalized parton distributions (GPDs) of the deuteron within the framework of the hard-wall AdS/QCD model. The momentum dependence of the GFFs obtained here is in good agreement with the results of the soft-wall AdS/QCD model. The value of the gravitational mean square radius in this model agrees with the experimental data. GFFs provide a holographic description of the. Deuteron GPDs are obtained from GFFs via sum rules and have shapes of plots similar to those for GPDs extracted from the electromagnetic form factors (EFFs) calculated in the framework of the soft-wall AdS/QCD model.

We present analytic results for the next-to-next-to-leading-order QCD corrections to heavy-quark production in charged-current deep-inelastic scattering, retaining the exact dependence on the charm quark mass. We compute the complete partonic coefficient functions for the structure functions $F_2$, $F_L$, and $F_3$ in the quark and gluon channels, including contributions with up to three heavy quarks in the final state. Working within the reverse-unitarity framework, we use integration-by-parts and canonical differential-equations techniques to express all contributions with at most two final-state heavy quarks in terms of manifestly real Goncharov polylogarithms which allow for a robust and efficient numerical evaluation. The three-heavy-quark contribution involves elliptic structures for which we give a general representation in terms of Chen iterated integrals, as well as expressions in terms of rapidly convergent expansions that are valid in the perturbative $Q\gtrsim 5~{\rm GeV}$ region and also allow for a flexible and fast numerical evaluation. We validate our results against known exact results at lower orders, massless NNLO coefficient functions, and existing leading-power expansions in the asymptotic limit where the virtuality $Q$ is much larger than the charm mass.

We investigate the domain structures of the G$_2$ vacuum within the framework of the domain model, analyzing the interactions among vacuum domains. Refinements to the static potentials in the intermediate regime reveal the crucial role of SU($3$) and SU($2$) subgroups in the confinement mechanism of the G$_2$ gauge group. Inside vacuum domains of G$_2$ with nonzero aggregate flux, repulsive interactions among subgroup vortices are found. Models containing either SU($3$) or SU($2$) domains, as well as vacuum domains with zero aggregate flux, exhibit favorable Casimir scaling and convexity properties at intermediate distances; notably, the SU($3$) version shows no concave regions in the potential at all. For a Wilson loop of increasing size, all subgroup center vortices belonging to a vacuum domain with nonzero aggregate flux, when fully enclosed by the loop, do not contribute a nontrivial phase, supporting the viability of this picture as a model for the vacuum structure of G$_2$ theory. The long-range behavior of the static potential yields the expected ordering of the (constant) potential values at large distances with respect to representation.

We investigate Krylov complexity for single-trace operators dual to open strings attached to giant gravitons in planar $\mathcal{N}=4$ super Yang-Mills theory. We show that in protected and few-body sectors, Krylov dynamics is governed by orthogonal polynomial theory associated to the seed spectral measure, leading to bounded Lanczos coefficients determined solely by spectral support. In particular, for fixed magnon number $M$ and open-string length $L\rightarrow\infty$, we derive $a_n=2Mg$ and $b_n\rightarrow Mg$, demonstrating integrable, band-limited dynamics. This establishes that such sectors are insufficient to test recently proposed gravity-side universality of operator complexity growth. We therefore formulate a finite-density program in which magnons scale with system size, and propose a concrete universality test: whether the leading Krylov growth depends only on coarse thermodynamic data $(ρ,\varepsilon)$ and not on microscopic probe structure. This provides a precise boundary-field-theory framework for testing gravitational universality conjectures.

Dark matter constraint repositories are critical community infrastructure, giving experimentalists and theorists a shared landscape of existing bounds. Yet the most widely-used repositories are maintained by individual volunteers, creating a sustainability risk as the pace of new results accelerates. We present a large language model (LLM) pipeline that monitors arXiv, extracts limit curves from papers, integrates them as code, and opens pull requests (PRs) for human review. On a 346-paper benchmark whose ground truth is the upstream-curated repository itself, the pipeline classifies the coupling type correctly for 90.5% of papers and reaches a median coupling residual of 0.33 dex (a factor of two for 48% of curves), with 76% mean mass-range coverage. This is driven by treating each extraction as a noisy sample reconciled through consensus voting, a physics convention canonicalization layer built with the agentic physics assistant Get Physics Done (GPD), and a scoring methodology that separates genuine extraction error from non-comparability. The remaining difficulty is concentrated in rare coupling types with idiosyncratic conventions (macro-averaged residual 1.1 dex). The pipeline is deployed and has generated limit proposals; none have merged. Governance of AI-generated scientific data is itself an unsolved problem.

Vacuum decay at zero temperature is generically described by a real $O(4)$-symmetric Coleman bounce. When the scalar field driving the decay carries a conserved global charge, this picture changes qualitatively: the path integral must be projected onto a definite charge sector, the Euclidean field obeys twisted boundary conditions, and the saddle is complex. For the simplest case of a $U(1)$ global symmetry, we first reformulate this problem in a two-field real Euclidean description with a real saddle. We then solve the resulting two-dimensional partial differential equation problem describing the decay of a homogeneous charged medium to a deeper vacuum via bubble nucleation. At finite charge the bounce departs from $O(4)$ symmetry, the barrier between vacua is lowered, and the decay rate increases. Continuing the solution to real time, we find that charge rearrangement around the expanding wall costs phase-gradient energy and drives the bubble to a subluminal terminal velocity even in vacuum. We also clarify how the fixed-charge construction interfaces with finite-temperature and finite-chemical-potential descriptions.

Extra-dimensional axions have coupling strength related to fundamental, ultraviolet physics. It has been proposed that the properties of such axions imply a bound on the quantum gravity cutoff: $Λ_\mathrm{QG} \lesssim 2π\sqrt{S} f$, where $f$ is the axion decay constant and $S$ is the instanton action. In the context of weakly-coupled string theory, we identify $Λ_\mathrm{QG}$ with the string scale $M_s$. In this paper, we carry out a quantitative study of this bound on the string scale in the context of Calabi-Yau compactifications of Type IIB string theory, considering both $C_2$ and $C_4$ axions. We show, both analytically and numerically, that the bound holds even near boundaries of the Kähler moduli space, including those where the co-scaling relationship for axion strings fails. This evidence bolsters previous arguments, based on naturalness and on unitarity, that the bound is a general feature of extra-dimensional axions in quantum gravity.

Experimental analyses of the semileptonic decays $ B \to D^{(*)} \ell \barν$ typically rely on the assumption that the missing energy originates from a massless neutrino, as predicted by the Standard Model. However, this assumption may not hold in scenarios where the invisible final-state particle is instead massive, such as a sterile neutrino or a dark-sector fermion. In this work, we explore how the presence of a massive dark sector fermion modifies the kinematic and angular distributions of these decays. Our analysis is carried out within the framework of a general weak effective theory, and we also discuss effective and simplified models in which these interactions may arise. In addition, we study the implications of these effects for the extraction of the CKM matrix element $ |V_{cb}|$. Overall, our results show that relaxing the standard assumption of a massless neutrino can lead to observable effects and provide a framework for systematically investigating their impact on semileptonic $B$- decay distributions.

Modern cosmology is often taken to provide an increasingly accurate description of the universe's underlying ontology through progressively refined mathematical models. I challenge this interpretation by arguing that the empirical success of cosmology underdetermines not only ontology but also the mathematical and conceptual frameworks used to represent observational data. I propose instead that the objective content of cosmology is best identified with cross-representational informational invariants-features of observational structure that persist across empirically adequate descriptions. These invariants can be characterized in information-theoretic terms, including correlation structure, statistical distinguishability, and limits on accessible information, and formalized using tools such as the Fisher-Rao metric on model space. On this view, cosmological models are best understood as efficient encodings of observational structure rather than uniquely privileged descriptions of fundamental reality. Scientific progress, accordingly, consists not in convergence toward a fixed ontology, but in the progressive refinement of the informational structure accessible to observation.

We present a cosmological reassessment of light sterile-neutrino scenarios, examining whether current observations exclude sterile neutrinos as a class or primarily constrain the fully thermalized case. We consider three distinct realizations: (i) a fully thermalized sterile species (FTS), (ii) a thermal relic with a suppressed temperature relative to the active neutrino background (DTS), and (iii) a Dodelson-Widrow-like (DW) sterile neutrino with reduced phase-space normalization. Constraints are derived within both LambdaCDM and the CPL dynamical dark-energy framework using combinations of Planck CMB data, DESI DR2 BAO measurements, and the PantheonPlus and Union3 Type Ia supernova samples. For baseline data combinations without a local H0 prior, the FTS scenario is strongly disfavored in both cosmological models. Adding the local H0^DN prior allows LambdaCDM+FTS to accommodate the high local H0 value and become statistically competitive with standard LambdaCDM once SNIa data are included, although the sterile-neutrino mass remains consistent with zero. By contrast, partially populated sterile-neutrino scenarios remain viable: the DW realization is broadly compatible with current observations, while the DTS scenario yields the weakest cosmological pressure among the cases considered. Overall, cosmological data mainly require a strongly suppressed effective sterile abundance, leading to tight constraints on m_eff^sterile while allowing substantially weaker bounds on the physical sterile mass. We conclude that current observations do not generically exclude sterile neutrinos, but rather place strong pressure on fully thermalized or highly populated scenarios, highlighting the importance of production history and phase-space distribution when interpreting cosmological constraints.

We present theoretical predictions for heavy-quark pair-production in deep-inelastic scattering (DIS) at next-to-leading order (NLO) in quantum chromodynamics (QCD), matched to a parton shower in the POWHEG framework. We revisit the NLO heavy-quark pair-production cross section and implement a consistent matching to parton-shower evolution, with careful treatment of heavy-quark mass effects and the avoidance of double counting between fixed-order and parton-shower radiation. In addition, we compare the virtual NLO corrections available in the literature to one-loop amplitudes obtained through massification in the small-mass limit. This provides an independent validation of the virtual contributions. The study is presently restricted to the gluon-initiated channel, which dominates the kinematic region of interest at the HERA collider and remains important for the future Electron-Ion Collider.

We investigate non-relativistic limits of four-dimensional maximally supersymmetric Yang-Mills theory (4d MSYM) and their relation to the nonperturbative $\operatorname{SL}(2;\mathbb Z)$ S-duality of the relativistic theory. We construct a general family of non-relativistic limits using a Type IIB brane set-up with a D3-brane and $(p,q)$-strings and show that the resulting theories are topological deformations of supersymmetric Galilean Yang-Mills theory or quantum mechanics on the moduli space of BPS monopoles. The deformations of the Galilean Yang-Mills theory are the familiar $θ$-term and a coupling to the monopole charge, while in the moduli space theory the only deformation is a $θ$-term. This family of theories fit together into a three-dimensional moduli space with nontrivial topology, on which $\operatorname{PSL}(2;\mathbb Z)$-valued dualities act in a richer and more complex way than in the relativistic parent theory. In the Abelian case, we establish the duality directly using the path integral, while in the non-Abelian case we support our claim by matching the one-particle spectrum as well as the Galilean spacetime symmetries and electric/magnetic invertible one-form symmetries.

The shape function and the light-cone distribution amplitude of heavy meson jointly characterize the nonperturbative structure of the heavy meson on the light-cone, with the former being essential for theoretical predictions of inclusive decays and the latter for exclusive decays. While first-principles lattice QCD results for the heavy meson LCDA have become available in recent years, lattice results for the shape function remain absent. In this work, we establish a two-step factorization scheme -- known as the HQLaMET framework -- for computing the $B$-meson shape function on the lattice, which fully disentangles the effects of the disparate scales $P_B^z$, $m_b$, and $Λ_{\textrm{QCD}}$. For illustration, starting from a phenomenological model for the shape function in HQET, we provide a graphical presentation of the entire procedure of this framework. The results of the current work lay the foundation for nonperturbative lattice QCD determinations of the shape function in the near future.

We establish the calculus structure on Hochschild invariants of open-closed homotopy algebras. We further define the Getzler-Gauss-Manin connection and show that it is flat up to chain homotopy on the open-closed periodic cyclic chain complex.

This dissertation investigates a supercooled phase transition (PT) in the early Universe. Using high-temperature dimensional reduction (DR), we compute the NLO thermal bubble nucleation rate. By explicitly evaluating fluctuation determinants, we provide a state-of-the-art description of thermal bubble nucleation. As a case study, we consider the SU(2)cSM, an extension of the conformal Standard Model with an additional SU(2)$_X$ gauge sector and scalar field that acquires a vev through radiative symmetry breaking. This symmetry breaking proceeds via a supercooled first-order phase transition. The first part of the thesis introduces the theoretical framework, including effective actions, RG improvements, finite-temperature quantum field theory, effective field theory techniques, and thermal bubble nucleation. The second part applies these methods to the SU(2)cSM. We establish a consistent power-counting scheme, construct the leading-order effective potential, analyse symmetry breaking and parameter space, derive an RG-improved potential, and incorporate thermal corrections. We then apply high-temperature DR to a classically scale-invariant model for the first time, derive the corresponding three-dimensional EFT, and compute the NLO nucleation rate. A detailed numerical evaluation of fluctuation determinants enables a comparison of different approximation schemes and their limitations. Finally, we present the phenomenological implications. We determine phase transition parameters, perform parameter scans, and predict the resulting gravitational wave signals. We find that the supercooled phase transition in the SU(2)cSM produces a strong signal detectable by LISA throughout the parameter space considered, making the model experimentally testable. We also demonstrate that higher-order corrections can significantly affect both phase transition dynamics and gravitational wave predictions.

Recent advances in artificial intelligence (AI), particularly large language models (LLMs), have created new opportunities for natural-language interaction with scientific software, but reliable theoretical model building still requires deterministic symbolic calculations. We present \texttt{bsm_agent}, an open-source symbolic framework for beyond the Standard Model (BSM) model building that combines a deterministic physics backend with an LLM chat interface. Starting from the SM field content and a user-specified set of additional scalars and/or fermions, the package constructs renormalizable Lagrangian, performs gauge-anomaly checks, expands operators into component fields, and derives electroweak symmetry breaking stationary conditions and tree-level mass matrices. The key novelty of the framework is that all of these tasks are performed automatically once the user specifies the quantum numbers of the new fields through a natural-language interface, eliminating the need for manual model construction. The symbolic calculations are performed entirely by the Python backend to ensure the correctness and reproducibility of the physics results; the LLM is used only as an orchestration layer that interprets natural-language requests, manages confirmation steps for ambiguous inputs, triggers backend tools, and formats report-ready summaries. The package supports three provider classes: local Ollama inference, remote self-hosted model servers accessed through the implemented remote provider interface, and commercial hosted APIs via OpenAI and Anthropic. This separation between conversational control and deterministic computation preserves reproducibility while making interactive BSM model construction substantially more convenient.

Recently, Benini, Mamroud, Reis and Serone have presented exact results for the Gross-Neveu model at finite density, both at finite N and in the large N limit. Generalizing previous studies, they introduce a chemical potential acting only on a subset of fermion flavors - the two-component Gross-Neveu model. Here, we take up this idea and extend the semiclassical study to finite temperature. The full (large N) phase diagram including inhomogeneous phases is constructed without solving the thermal Hartree-Fock problem. All phase boundaries can be found on the basis of various kinds of stability analyses. As a result, we find a qualitative change of the phase diagram at some critical filling fraction.

We study spin correlations and quantum entanglement in the process $γγ\to t\bar t$ with the photons coming from Compton backscattered laser beam. We present predictions for the cross sections and spin observables including next-to-leading order (NLO) QCD corrections under various beam-polarization configurations. The NLO QCD corrections significantly enhance the total cross section while having only a minor impact on spin observables. Using the spin density matrix of the $t\bar t$ system, we further investigate quantum entanglement and Bell nonlocality. We find that the entanglement is the strongest near the $t\bar t$ invariant-mass threshold, while both entanglement and Bell nonlocality are highly sensitive to the initial beam polarization. Our results provide a theoretical basis for future studies of quantum correlations in top-quark pair production at photon colliders.

Events characterized by a high multiplicity of charged particles have been a central focus in the study of collective behavior across both large and small collision systems. A previous measurement of two-particle angular correlations in $e^+e^-$ collisions at center-of-mass energies up to $\sqrt{s} = 209$ GeV, using LEP2 data, revealed discrepancies with Monte Carlo (MC) predictions at high multiplicity, suggesting the possible emergence of long-range near-side correlations even in the simplest collision system. Unlike at lower energies, where quark-antiquark production dominates, $W^+W^-$ processes become increasingly important at multiplicities above 30. On the one hand, the observed excess in long-range correlations may reflect the more complex color-string configurations arising from $W^+W^-$ production. On the other hand, it can simply arise from the higher final-state multiplicity made possible by the increased collision energy, independent of the underlying production mechanism. To discriminate between these competing interpretations, we present a measurement of two-particle angular correlations in $e^+e^-$ collisions at $\sqrt{s} = 183-209$ GeV, with a focus on enhancing the contribution from $W^+W^-$ processes. The analysis uses data collected by the ALEPH detector during the LEP2 program. Correlation functions are evaluated across a broad range of pseudorapidities and full azimuth, in bins of charged-particle multiplicity. A ridge-like modulation is seen for multiplicity above 50, deviating from the MC reference. In addition, the correlation functions are further decomposed into a Fourier series, and the resulting harmonic coefficients $v_n$ are compared with predictions from the archived Monte Carlo sample. For multiplicity starting from 30, the signed $v_2$-like proxy goes from negative to positive, also deviating from the MC baseline.

In this work we consider a particular codimension-1 region of a holographic spacetime which we call the entanglement wedge polygon (EWP). For a pure state and a partition of the boundary into a number of regions $A_i$ the EWP is defined as the region external to all the individual homology regions $r_{A_i}$ which consists of the intersection of the entanglement wedge EW($A_i$) with the time slice. In vacuum AdS$_3$ the quantity is topological as a direct consequence of the Gauss-Bonnet theorem. In higher dimensions we make progress by considering a number of concrete examples including vacuum, black brane, and soliton solutions of AdS$_{d+1}$ as well as spacetime geometries with end of the world branes dual to boundary conformal field theories. We provide a suitable generalization to mixed states and comment on possible connections between the EWP and measures of multi-partite entanglement.

The era of precision ab initio nuclear theory has arrived, enabling uncertainty-quantified predictions for nuclear structure and for interactions with external probes directly from the underlying nuclear force and electroweak currents. This review highlights recent breakthroughs that extend ab initio calculations to the heavy nucleus $^{208}$Pb, to medium-mass systems with complex deformation, and to weakly-bound nuclei near the driplines. We also summarize ab initio calculations of nuclear responses for dark matter direct detection. Together, these advances demonstrate how ab initio methods can substantially reduce nuclear-physics uncertainties in searches for physics beyond the Standard Model, providing a more robust interpretation of current and forthcoming precision experiments.

Understanding the structural properties of the pion is essential for elucidating the mechanisms of mass generation within the Standard Model and their role in the emergence and properties of the hadronic matter. Light-front wave functions encode extensive information about the internal structure of these systems and provide the link to measurable quantities such as generalized parton distributions and transverse-momentum-dependent distributions. Guided by recent progress in continuum Schwinger methods, we derive well-founded and practical representations of these quantities, enabling the exploration of several facets of the pion structure, including distribution amplitudes and distribution functions, elastic and gravitational form factors, and the associated momentum and spatial distributions. The results presented here are consistent with expectations and can be tested at modern experimental facilities, including the new generation of electron-ion colliders.

We investigate composite neutrino models where heavy neutrinos emerge as bound states from a near-conformal strongly coupled sector. Standard Model neutrinos mix with these composite singlets via an inverse seesaw mechanism, where the anomalous scaling dimensions of the composite-sector operators naturally suppress light neutrino masses to sub-eV scales. Matching the conformal dynamics onto low-energy theory yields enhanced electromagnetic transition dipole operators with couplings $d_{μN} \sim 10^{-6}$-$10^{-8}\,\mathrm{GeV}^{-1}$, parametrically larger than the loop-level predictions of minimal Dirac or Majorana models. We carry out a dedicated event-level simulation of the production-and-decay chain $νX \to \mathcal{U} X \to νγX$ and compute the resulting event rates at MiniBooNE and MINERvA within the model, accounting for the composite production cross section and decay kinematics in detail. We further present predictions for the photon energy, angular, and multiplicity distributions. For the benchmark scenarios accessible at these experiments the radiative signal is predominantly single-photon; the composite structure additionally permits fragmentation of the up-scattered state into multiple heavy neutrinos, each decaying as $N\toνγ$, with multi-photon final states emerging for lighter compositeness scales or higher beam energies as a qualitatively new probe of the composite dynamics.

Generalized transverse-momentum-dependent parton distributions (GTMDs) are the most complete two-parton correlation functions in QCD, encoding the joint spatial and momentum structure of hadrons. Through appropriate projections and limits they yield generalized parton distributions (GPDs), transverse-momentum-dependent distributions (TMDs), parton distribution functions (PDFs), and phase-space (Wigner) distributions. We construct conformal moments of unpolarized gluon GTMDs at strong coupling using gauge/string duality. For fixed even conformal spin $j$, we distinguish the local boundary limit at $b_T=0$ from the finite-separation regime $b_T>0$, where the planar semiclassical amplitude is governed by a minimal worldsheet. There the GTMD moment factorizes into a universal staple-worldsheet soft factor and a stripped spin-$j$ Witten amplitude carrying target dependence. The cusp of the renormalized minimal area generates the rapidity-logarithmic Collins-Soper structure. We derive universal ultraviolet and infrared endpoint reductions. As $b_T\to0^+$, the finite-separation sector matches onto the local conformal moment through a universal overlap kernel. At large $b_T$, after cusp/perimeter subtraction, it factorizes into target projections and infrared transfer kernels. The ultraviolet endpoint is universal within the leading saddle, whereas the infrared tail depends on the holographic completion: soft-wall, gap-matched hard-wall, and repulsive-wall backgrounds generate algebraic, exponential, and Gaussian falloffs, respectively. Analytic continuation in $j$ yields the low-$x$ Regge regime governed by the holographic Pomeron spectral curve. The framework describes hadron tomography, transverse structure, rapidity evolution, and Reggeization for GTMD moments and provides a unified starting point for holographic studies of observables relevant to the Electron-Ion Collider.

We introduce and demonstrate an event-level analysis framework for quark-gluon imaging. For a first application we use it for the inference of parton distribution functions from synthetic deep inelastic scattering data. This framework removes the need for unfolding of detector effects and the binning of events, and therefore eliminates two key sources of information loss. We contrast this event-level framework with the traditional histogram approach by performing a closure test for parton distribution functions from event data obtained from a known ground truth. In this study we assume a perfect detector, which makes unfolding straightforward. The elimination of binning in the event-level framework is demonstrated to have important benefits over the traditional histogram approach, and performs better in the closure test, particularly for a smaller number of events. For example, defining a mean-squared error distance metric, we find that the event-level framework performs around $35\%$ better than the traditional approach for a moderate number of events. The benefits of an event-level framework should increase for inference associated with 3D quark-gluon imaging, because these differential cross sections are of higher dimension and the comparative number of measured events is significantly reduced.

The extraction of the leptonic charge-parity (CP)-violating phase $δ_{\rm CP}$ from long-baseline neutrino oscillation experiments rests on the assumption of charge-parity-time (CPT) conservation. We show that CPT violation, parametrized as an asymmetry $δΔm^2_{31} \equiv Δ\bar{m}^2_{31} - Δm^2_{31}$ between neutrino and antineutrino mass splittings, induces an effective, energy-dependent phase shift $φ_{\rm eff}(E)$ that is functionally degenerate with $δ_{\rm CP}$ in the appearance asymmetry $\langleΔP\rangle$. This has a profound implication for long-baseline experiments, where the tension between T2K and NO$ν$A CPT-conserving best-fit $δ_{\rm CP}$ values can be significantly alleviated by a CPT-violating truth; and a CPT-conserving fit can miss the true CP phase entirely for $|δΔm^2_{31}|\gtrsim 0.3\times10^{-3}~\text{eV}^2$ for DUNE. We then demonstrate that atmospheric neutrino telescopes provide the natural tool to resolve this degeneracy: using existing data from IceCube-DeepCore (7.74 yr) and KM3NeT/ORCA-6 (433 kt-yr), we derive a world-leading constraint on CPT-violation at $| δΔm^2_{31}|\leq 0.57\times10^{-3}~\text{eV}^2$ at 90% CL. With the IceCube Upgrade and full ORCA detector, we can reach a $1σ$ constraint at $10^{-4}~\text{eV}^2$ within a decade, providing the independent CPT constraint needed to ensure that DUNE's $δ_{\rm CP}$ measurement is unambiguous.

The existence of multiple axion species is a generic prediction of a number of extensions of the Standard Model. If more than one axion couples to photons, their combined signal in helioscope experiments may mimic that of a single axion with different parameters. This raises a fundamental question: if a next-generation helioscope such as IAXO detected a signal, would we be able to disentangle whether it originated from one or multiple axions? To answer this question, we first recast current CAST bounds and derive IAXO/IAXO+ projections in the two-axion parameter space, identifying the regions where a signal could be observed. Then, we analyze the spectral signatures of axion flavor oscillations in both the quasi-degenerate and hierarchical mass regimes, and point out where IAXO can discriminate a two-axion signal from the single-axion hypothesis given the expected energy resolutions of the detector. Finally, we show that these results extend to a broad class of $N$-axion systems.

We introduce the \emph{Centauric 1-jettiness}, $τ_1^C$, a generalized event shape for Deep Inelastic Scattering (DIS) with adjustable beam and jet reference vectors and thus beam and jet regions. We demonstrate that a specific choice of weights allows this observable to exactly reproduce the geometric boundaries of the Centauro jet algorithm in the Breit frame. Within the framework of Soft-Collinear Effective Theory (SCET), we derive a factorized cross section in the small-$τ_1^C$ region in terms of known perturbative ingredients. This allows the resummation of large logarithms to N$^3$LL accuracy, which we then match to fixed-order NLO QCD ($\mathcal{O}(α_s^2)$) predictions from \texttt{NLOJet++}. We establish that the soft measurement reduces to a rescaled hemisphere measurement, placing Centauric 1-jettiness in the same universality class, for the leading non-perturbative corrections, as DIS thrust and jet mass. As a consequence, the leading non-perturbative shift depends on the same universal first-moment-shift parameter $Ω_1$ and scales exactly as $1/R$ with the jet radius, thanks to the boost invariance of the Centauro algorithm along the photon axis in the Breit frame, a scaling that we test using \textsc{Pythia} simulations. These results open new strategies for determining the strong coupling from DIS event shapes, with $R$ providing a handle to break the degeneracy between $α_s$ and the universal non-perturbative shift parameter $Ω_1$.

We study warm fermionic dark matter (DM) in the framework of freeze-in at stronger coupling, in the minimal Higgs portal scenario. The reheating temperature is taken to be low, so that DM production from the Standard Model thermal bath is Boltzmann-suppressed and the DM stays out of equilibrium even for a sizeable coupling. This opens the possibility of observable signatures, in particular invisible Higgs decays. We compute the DM relic abundance including both the pre- and post-reheating contributions. We find that the fermionic DM production reaction is strongly velocity suppressed, requiring larger reheating temperatures than those obtained for scalar DM in order to reproduce the correct relic abundance. The resulting DM momentum distribution is strongly non-thermal and its shape is not captured by the common $αβγ$-parametrization. We find that the Lyman-$α$ constraint excludes DM masses below about $100 -180\,\mathrm{keV}$, depending on the reheating history.

We compute resummed and matched predictions for jet angularities in hadronic Z+jet events, where the jet is initiated by a b-quark. The analysis is performed both with and without grooming the candidate jets using the SoftDrop algorithm. Mass effects are consistently included at both fixed-order and resummed levels. Our theoretical predictions also incorporate non-perturbative corrections from the underlying event and hadronization, implemented through parton-to-hadron transfer matrices extracted from dedicated Monte Carlo simulations with Sherpa. Finally, we compare results for b-jets with the ones from light-flavour jets, in order to quantify the impact of finite-mass effects.

Measurements of charge-parity (CP) violation in quark transitions provide stringent tests of the current theory and sensitive probes of particles or interactions beyond it. We report the first measurement of mixing-induced CP violation in $B^0\toπ^0π^0$ decays. The measurement uses 190 million $Υ(4S)\to B^{0}\overline{B}{}^0$ events collected with the Belle II experiment at the SuperKEKB asymmetric-energy electron-positron collider. A new approach that exploits the quantum entanglement of the $B^{0}\overline{B}{}^0$ pair enables a decay-time-dependent analysis without reconstructing the $B^0 \to π^0π^0$ decay position. We measure the mixing-induced CP-violating coefficient to be $S_{π^{0}π^{0}} = 0.61^{+0.75}_{-0.79}\,(\mathrm{stat}) \pm 0.11\,(\mathrm{syst})$, attaining a precision that would require a data set twenty times as large with a conventional analysis. We also measure the direct CP-violating coefficient to be $C_{π^{0}π^{0}} = 0.05 \pm 0.28\,(\mathrm{stat}) \pm 0.07\,(\mathrm{syst})$. These results substantially improve the constraints from $B \to ππ$ decays on the CP-violating phase $φ_2$ of the quark-mixing matrix using far less data than had previously been assumed necessary.

The Stanford-Witten-Norbury generalization of Mirzakhani's volume recursion computes $V^{(2m)}_{g,n}(\{b_i\})$, the Weil-Petersson volumes of the moduli space of $N=1$ supersymmetric Riemann surfaces of genus $g$ with $n$ Neveu-Schwarz boundaries of geodesic lengths $b_i$ ($i{=}1,\ldots,n$), and $2m$ Ramond punctures. Recently, a spectral curve has been derived that allows their Laplace transforms $W^{(2m)}_{g,n}(\{{\hat z}_i\})$ to be computed using topological recursion. We prove that the Stanford-Witten-Norbury volume recursion is directly derivable from the spectral curve. An alternative volume recursion can also be derived from it. The difference comes from whether the Ramond information is in the initial data, or in the volume recursion's kernels. The latter invites a geometrical understanding.

We present a minimal dark matter scenario: a single complex scalar carrying baryon and lepton number, with no new exact stabilizing symmetry. Its leading interaction is a dimension-7 semileptonic portal that, below confinement, generates a low-energy Yukawa coupling with the proton and electron. Requiring absolute stability of both the proton and dark matter forces the dark matter mass into a narrow window around the proton mass, which may be anthropically selected. Despite its minimal field content, the model can be probed by many observables: proton burning in stars, hydrogen decay, brown dwarfs and neutron star heating, and nucleon decay-like signatures in direct detection. UV-dominated freeze-in produces the observed relic abundance. This framework provides a unique testable example of dark matter arising from a minimal extension of the Standard Model.

We present the first study of Leggett-Garg inequality violation in polarized muon spin precession. We formulate a procedure to reconstruct temporal correlators of the longitudinal muon polarization from measured time-dependent muon decay spectra and apply it to publicly available Fermilab Muon $g-2$ data corresponding to approximately $10$ billion muon decays. Using a simplified model of the detector acceptance and efficiency, the Leggett-Garg inequality is found to be violated with a single-bin significance of $5.5σ$, while combining neighboring bins further increases the significance. While our analysis is limited by systematic uncertainties associated with the detector modeling, a dedicated experimental analysis could reduce these uncertainties toward the statistical level, $\mathcal{O}(10^{-3})$, potentially enabling one of the most precise measurements of temporal quantum correlations.

The ATLAS experiment presents an observation of a centrality-dependent dijet transverse momentum imbalance in O+O and Ne+Ne collisions at a nucleon-nucleon center-of-mass energy of 5.36 TeV at the Large Hadron Collider. The measurement uses 8.0 nb$^{-1}$ of O+O and 1.0 nb$^{-1}$ of Ne+Ne data collected in 2025, together with 386 pb$^{-1}$ of \textit{pp} data at the same energy used as a reference. The dijet momentum balance is quantified using the ratio of the sub-leading jet transverse momentum to that of the leading jet, $x_J$. For dijets produced azimuthally back-to-back, the self-normalized $x_J$ distributions exhibit increasingly large deviations from the \textit{pp} reference as collisions become more central, corresponding to an increasing overlap of the colliding nuclei. The observed centrality dependence is consistent with medium-induced partonic energy loss in O+O and Ne+Ne collisions, demonstrating that such effects persist in collision systems considerably smaller than Pb+Pb and Xe+Xe. These results establish a new regime for investigating the path-length dependence of jet quenching and constrain the onset of quark-gluon plasma effects in small nuclear collision systems.

We present a unified, multi-scale study of non-Gaussianity of Galactic foreground emissions using Minkowski Functionals and generalized skewness-kurtosis parameters, focusing on the characterization of small-scale non-Gaussianity and its underlying physical origin. We find that all foreground components studied exhibit a remarkably universal non-Gaussian nature dominated by excess kurtosis, whose shape remains stable across angular scales despite large differences in emission physics. Focusing on thermal dust, we perform a detailed comparison between observed maps (GNILC and Planck 545 GHz) and dust model realizations (PySM and filament-based models) to assess the performance of state-of-the-art models in reproducing the observed non-Gaussian properties. At the global level, GNILC and PySM display closely matched kurtosis behavior over the angular scales where the GNILC reconstruction is reliable, while the filament-based model produces substantially weaker skewness and kurtosis signals. For PySM, however, a patch-based analysis reveals statistically significant regional variations, indicating that while the model reproduces the overall non-Gaussian amplitude and scale dependence, it does not fully capture the spatial variability of the observed kurtosis signal. Using simple PDF-based toy models, we demonstrate that the universal kurtosis signature arises from the combination of heavy-tailed one-point statistics and steep large-scale spatial correlations, while its detailed amplitude and scale dependence depend on the underlying foreground physics. These results identify excess kurtosis as a robust statistical fingerprint of Galactic foregrounds and provide a practical framework for validating small-scale foreground models for future CMB analyses.

We present a study of two tailed radio galaxies in the core of a massive, dynamically young galaxy group - an early group-group merger. Using VLA (3 GHz and 1.4 GHz), MeerKAT (1.35 GHz), and GMRT (610 and 325 MHz) observations, we investigate their radio spectral properties, spectral and dynamical ages. Radio morphologies show clear evidence of interaction with the intragroup medium (IGM). One galaxy is a wide-angle tail (WAT) source, while the other is most likely a head-tail (HT) galaxy. Both galaxies exhibit high radio luminosities, and we find spectral indices of $α=0.8\pm 0.1$ (WAT) and $α=0.6\pm 0.2$ (HT). Spectral index analysis reveals spectral steepening with distance from the core in both galaxies, with localized flattening in the WAT lobes and hotspots along the northern jet, and indications of such flattening in the middle of the HT tail. Spectral ages derived using Jaffe-Perola model are $33.80\substack{+7.63 \\ -7.23}$ Myr (WAT) and $20.86\substack{+10.07 \\ -17.17}$ Myr (HT), significantly lower than dynamical ages of $420\pm60$ to $700\pm100$ Myr (WAT) and $140\pm20$ Myr (and possibly up to $280\pm40$ Myr, for HT), yielding dynamical-to-spectral age ratios of $\sim12-20$ and $\sim7$ (and up to $\sim14$), respectively. The discrepancy may be reduced by using more complex dynamical age models, incorporating interactions with the IGM, which requires deeper X-ray observations of the group. Spectral age estimates may be affected by mixing of electron populations, and could be better constrained with future deep, high-resolution broad-band radio observations at both MHz and frequencies above 3 GHz. The combination of extended radio structures, spectral signatures of radiative aging with localized re-acceleration, and activity timescales up to hundreds of Myr indicates that galaxies are actively interacting with, and likely depositing energy into their environment.

We present an investigation of the Tully-Fisher (TF) relation solely for galaxies hosting an active galactic nucleus (AGN). Using 22 galaxies with primary, z-independent distances, we find that active galaxies exhibit significantly larger scatter about all TF relations compared to each respective calibration for (largely) inactive galaxies. The larger scatter persists despite removal of the AGN contamination from the photometry of the Type 1 AGNs via 1) careful surface brightness decompositions or 2) employing SEDs to constrain the light contribution of the AGN. These results suggest that the influence of an AGN on its host galaxy's surface brightness may extend beyond the nucleus. We also calculate the percentage difference between TF and primary distances, and find that TF-based distances are biased towards overestimation of the primary distances to active galaxies by anywhere from 5-10 percent for the optical/near-infrared and approximately 15 percent for distances predicted from inverting the Baryonic TF (BTF) relation. As TF-based distances (especially the I-band) are relied on heavily for analysis and modeling of the local peculiar velocity (Vpec) field, we suggest that active galaxies be removed from future Vpec modeling samples.

We investigate whether embedded stellar-mass black holes (sBHs) in active galactic nucleus (AGN) disks can leave observable optical/UV variability signatures through migration-trap-driven magnetic heating. This mechanism operates when sBHs migrating toward torque-balance radii pile up near migration traps, triggering localized, stochastic magnetic reconnection that heats the disk atmosphere. It is potentially important because it provides a physical source of non-coronal disk heating and directly links optical/UV continuum variability to otherwise hidden compact-object populations. By coupling a one-dimensional sBH population synthesis model with a corona-heated accretion-disk reprocessing variability framework, we show that migration traps concentrate sBHs at preferred radii and generate localized, stochastic reconnection heating. The resulting heating is self-regulated: sBH pile-ups enhance the reconnection rate, while gap opening reduces the local gas density and partially suppresses the reconnection power. This heating produces excess short-timescale optical/UV variability, flattened short-term structure functions, and deviations from the standard $τ\proptoλ^{4/3}$ lag-wavelength relation, which describes the time delay between variability at different wavelengths for a standard thin accretion disk. These signatures are strongest at low-to-moderate Eddington ratios, and related observations could provide indirect evidence for embedded compact-object populations in AGN disks.

Comprehensive characterization of numerical dissipation is essential for high-fidelity simulations of magnetohydrodynamic (MHD) turbulence. In this work, we present an a posteriori framework for directly estimating numerical dissipation in MHD turbulence from simulation data without invoking a priori assumptions. Implemented in the open-source Python package PyMHD, the framework is applied to simulations of Alfvénic turbulence, turbulent small-scale dynamos, and MRI-driven turbulence, yielding a systematic characterization of the anisotropy and spectral properties of numerical dissipation across these regimes. The results indicate that numerical dissipation primarily dissipates energy transferred by the turbulent cascade at small scales, consistent with the conventional interpretation. However, its spectral properties are distinct from those of physical viscosity and resistivity, such that it cannot simply be represented by effective dissipation coefficients. In addition, numerical dissipation inherits the anisotropy of the underlying turbulence, and can even exhibit anomalous anti-dissipative behavior under certain circumstances. Moreover, this framework enables identification of the conditions under which physical dissipation dominates numerical dissipation across all scales, thereby providing practical guidance for achieving high-fidelity simulations of astrophysical MHD turbulence.

To advance our understanding of dust properties in class 0/I young stellar objects, it is crucial to resolve their structures at multiple wavelengths and investigate how grain growth and environmental effects shape their spectral properties. We present 0.5 arcsec resolution ALMA observations of the GSS 30 complex at 1.2-3.0 mm from the FAUST large programme, achieving a linear resolution of 69 au. We analyse the dust continuum emission and perform modelling to constrain dust properties and disk structures. For IRS3, the spectral index increases radially from 2.0 at the centre to 2.5 at the disk edge, while decreasing to 1.6-1.8 along the outflow direction. The asymmetric low-alpha region towards the northeastern blueshifted lobe may result from cold outer envelope dust obscuring warmer inner regions. SED fitting suggests maximum grain sizes of tens of microns and a dust mass of 650-1510 M_earth. IRS1 is associated with an extended north-eastern structure, which may represent an outflow-disk complex, a trailing structure linked to a circumbinary disk, or a gas streamer accreting onto IRS1. The central IRS1 shows alpha < 0.8, consistent with marginally optically thick free-free emission. IRS2 displays brightness variations over 420 s, and multi-epoch comparison suggests a flare lasting tens of minutes, likely caused by magnetic activity in the protostar. Our results highlight the importance of environmental effects, including dust obscuration and streamer structures, in shaping the observed properties of young disks, and reveal millimetre variability associated with possible protostellar magnetic flares.

We present the first measurement of the angular two-point correlation function \(w(θ)\) of radio sources from the MeerKAT Large Area Synoptic Survey (MeerKLASS) UHF on-the-fly (OTF) continuum Data Release~1. DR1 provides interferometric Stokes-\(I\) imaging at a reference frequency of 816\,MHz over \(\sim 800~\mathrm{deg}^2\) within the DESI footprint. We detect a positive clustering signal over \(0.02^\circ \lesssim θ\lesssim 10^\circ\). The measurement is stable to reasonable variations of the depth mask and flux threshold on intermediate and large scales. Modelling the intermediate-scale signal (\(0.112^\circ\leθ\le 1.36^\circ\)) with a fixed-slope power law (\(γ=1.8\)) yields \(A(1^\circ)=(1.434\pm0.475)\times10^{-3}\), corresponding to \(\log_{10}A=-2.843^{+0.124}_{-0.175}\). We infer an effective large-scale bias by fitting \(Λ\)CDM projected-matter templates \(w_{\rm DM}(θ)\) computed with \textsc{CAMB} and Limber projection, including an integral-constraint correction evaluated from random--random weights. Using two bracketing T-RECS redshift-distribution priors, we obtain \(b_{\rm eff}=1.998\pm0.350\) (AGN prior) and \(b_{\rm eff}=1.530\pm0.265\) (TOTAL prior), demonstrating that the dominant modelling uncertainty arises from \(N(z)\). As a derived summary we Limber-invert the power-law amplitude to obtain \(r_0=6.18\pm1.13\) and \(5.59\pm1.02~h^{-1}\mathrm{Mpc}\) for the AGN and TOTAL priors, respectively. These results establish MeerKLASS UHF DR1 as a new wide-area, intermediate-frequency dataset for radio-continuum clustering. As MeerKLASS expands and overlapping optical/IR spectroscopy provides improved redshift calibration, future releases will enable population-split clustering and bias evolution measurements with substantially reduced modelling uncertainty.

We present a global unbinned-likelihood analysis of the rest-frame ultraviolet (UV) quasar luminosity function (QLF) over $0.1\le z\le7.5$, using a final analysis sample of 70,960 Type~1 quasars selected from the homogenized compilation of Kulkarni et al. (2019). We test a shape-preserving luminosity and density evolution (LADE) framework, in which the local luminosity function (LF) shape is fixed and the redshift evolution is described by separate density- and luminosity-evolution factors. For a double-power-law (DPL) local LF, we search 81 LADE models built from nine density-evolution and nine luminosity-evolution functions, and compare them with two flexible double-power-law (FDPL) reference models. The fiducial shape-preserving LADE model gives the lowest Akaike information criterion (AIC) and Bayesian information criterion (BIC) values among the main models considered, and 12 DPL-based LADE models have lower AIC and BIC values than both FDPL reference models. Repeating the model-grid analysis with a modified-Schechter local LF gives the same preferred evolutionary structure, indicating that the result is not driven only by the assumed local LF shape. The fitted evolution functions further show that the luminosity-evolution component is the more stable part of the LADE decomposition: it rises rapidly to $z\simeq2$--3 and then flattens or slowly declines. The density-evolution component is more model dependent, but the preferred LADE models consistently require a declining effective normalization toward high redshift. Taken together, we conclude that a shape-preserving framework offers a statistically efficient and compact empirical description of the global UV QLF.

Fast radio bursts (FRBs) are luminous, millisecond-duration extragalactic radio transients that have emerged as a powerful, complementary cosmological probe for investigating the late-time cosmic evolution, offering unique advantages over conventional probes such as Type Ia supernovae, baryon acoustic oscillations, and cosmic microwave background radiation. This review systematically summarizes the cosmological applications of FRBs, focusing on their critical roles in measuring the Hubble constant ($H_0$) and constraining dark energy properties. Benefiting from the precise dispersion measure (DM) - redshift relation of localized FRBs, the integrated electron density of the intergalactic medium (IGM) along the line of sight can be tightly modeled, enabling independent and low-redshift measurements of the cosmic expansion rate. Current FRB samples consisting of localized and non-localized events provide competitive $H_0$ constraints, offering an independent method to measure $H_0$. FRBs also serve as effective tracers to constrain dark energy equation-of-state parameters. We comprehensively discuss key limiting factors for FRB cosmological precision, including uncertainties in Galactic and host galaxy electron density models, and IGM inhomogeneities. With the rapid growth of high-precision FRB surveys and localized FRB samples, FRBs are promising to provide stringent constraints on late-time cosmic acceleration, dark energy evolution and cosmic baryons.

The Fermi bubbles have been identified as a collimated bipolar outflow emanating from the Milky-Way center, tracing strong forward shocks which extend $\sim10$ kpc from the Galactic disk, originating from a $\sim10^{56\pm1}$ erg outburst. These shocks are sufficiently strong, extended, and energetic to produce a detectable flux of $\gtrsim 10$ TeV neutrinos, especially in their denser north and east, but early IceCube data were insufficient for identifying the signal. We find that IceCube high-energy starting events (HESE 12-year) correlate with the $\textit{Fermi}$-LAT sky map outside the Galactic plane ($>3σ$). Testing for neutrinos coincident with the bubble shells, localized using $\textit{eROSITA}$ data, we detect ($>4σ$) both bubbles at high ($|b|>30^{\circ}$) latitudes, with a local excess ($>5σ$) mainly in their X-ray bright eastern shells. The signal matches the anticipated secondaries of relativistic ions carrying $\sim 10^{54.5}$ erg (with factor $\sim3$ uncertainty) in each bubble, shock-accelerated to $>$PeV energies. The results verify the strength of the shocks, suggesting an ion acceleration efficiency of order $\sim10\%$. We also present preliminary evidence for neutrinos from the even larger shells of the eROSITA bubbles, which encapsulate their younger Fermi-bubble counterparts, carrying a similar energy and confined by shocks nearly as strong.

We report the discovery of a bow-shock pulsar wind nebula (PWN), dubbed Thunder, powered by the radio pulsar PSR J1631-4722 and projected within the Galactic supernova remnant (SNR) G336.7+0.5 (Nimbus). The system was first identified in observations from the Australian Square Kilometre Array Pathfinder (ASKAP) Evolutionary Map of the Universe (EMU) survey and further characterised using MeerKAT Galactic Plane Survey data together with follow-up observations at 5.5 and 9 GHz obtained with the Australia Telescope Compact Array (ATCA). Assuming a distance of 7 kpc, the radio images resolve an elongated ~80 arcsec (2.7 pc) cometary nebula, indicative of a high velocity pulsar. An X-ray counterpart extending ~50 arcsec (1.7 pc) is detected in archival XMM-Newton data. The flat radio spectrum ($α$ = -0.27 $\pm$ 0.05) and hard X-ray photon index ($Γ$ = 1.6 $\pm$ 0.4) indicate synchrotron emission from relativistic particles injected in the pulsar wind. Polarisation analysis reveals a highly ordered magnetic field aligned with the nebular flow, with fractional polarisation reaching up to 30% in the tail. An equipartition estimate gives a PWN magnetic-field strength of Beq $\approx$ 54-140 $μ$G. Pulsar timing over a ~2.2 yr baseline reveals strong timing noise and a small spin glitch with amplitude $Δν/ν$ = 1.10$\times$10$^{-8}$. The SNR shows no clear diffuse X-ray counterpart. The morphology and multiwavelength properties of the Nimbus-Thunder system, along with evolutionary models, constrain the system's age to approximately 30-45 kyr, placing the remnant in the late Sedov phase, approaching the transition to the radiative stage.

We investigate cosmic anisotropy within the updated Pantheon+ sample using both the dipole fitting (DF) and hemisphere comparison (HC) methods. With the DF method, the dipole signal within the full sample is statistically weak. However, the low-$z$ subsample yields a dipole signal of $A_{\mathrm{D}} = 0.952^{+0.454}_{-0.403} \times 10^{-3}$ at $\sim 2σ$ significance, pointing towards $(l,b) = (149.77^\circ, -12.20^\circ)$. This signal is predominantly driven by a combined subset of surveys 5, 56, 63, and 150, which is characterized by an amplitude of $A_{\mathrm{D}} = 1.730_{-0.715}^{+0.554} \times 10^{-3}$ towards $(l,b) = (153.05^\circ, -1.25^\circ)$. For the HC method, the full sample yields a maximum anisotropy level of $\mathrm{AL}_{\mathrm{max}} = 0.289 \pm 0.052$ oriented towards $(l,b) = (127.97^\circ, 17.90^\circ)$ with a $1.56σ$ significance. This preferred direction is primarily determined by the highly inhomogeneous SNLS subsample, whereas the low-$z$ and high-$z$ subsamples act to suppress the anisotropy level along this axis. These subsample-dependent results suggest that the apparent anisotropy arises from local structures or the inhomogeneous distribution of the datasets rather than an intrinsic cosmic anisotropy.

We present a novel method to conduct targeted gravitational-wave searches for compact binary mergers using the GstLAL inspiral pipeline. By incorporating sky localization and timing information from external electromagnetic triggers, we enhance the sensitivity of the search for sub-threshold gravitational-wave signals associated with events such as short gamma-ray bursts. Our approach modifies the standard likelihood ratio ranking statistic to include a sky localization prior, allowing for a more focused analysis on specific regions of the sky. We demonstrate the effectiveness of this method through injection studies, comparing the performance of the targeted search against the standard all-sky search configuration. The results show a significant improvement in detection efficiency for signals consistent with the provided sky location and timing, while maintaining control over false alarm rates. This targeted search framework enables rapid follow-up of electromagnetic transients, facilitating multi-messenger astronomy efforts in the era of advanced gravitational-wave detectors.

IGR J17480-2446 is a low-mass X-ray binary, harboring an exceptional accreting pulsar (a neutron star) with an unusual spin frequency of 11 Hz and a very slow post-outburst crust cooling. The former may imply that it is observed at an early stage of recycling, while the latter was shown to indicate the presence in the outer crust of a low thermal conductivity layer, possibly made of glass. Here we argue that the glass layer formation is a natural result of accretion induced failure of pristine cold crystal crust. This allows us to determine the mass of the accreted material as $ΔM \approx 2.4\times 10^{-6}~M_\odot$, confirming very early accretion stage for this neutron star. An analysis of spin and thermal state reveals a peculiar set of neutron star birth properties which is commonly associated with `recycled' neutron stars, i.e.\ those that have been experiencing prolonged periods of accretion from a companion. We speculate that such birth properties may represent the outcome of neutron star formation in an electron-capture supernova.

We present the multiwavelength variability of the BL Lac object PKS 0735+178, associated with the high-energy neutrino event IC211208A. The light curves cover the radio (1-230 GHz), optical, and Fermi-LAT $γ$-ray bands over a $\sim$30 years time-scale. The light curves are correlated, with delays from 0 to 1200 days increasing towards lower frequencies, consistent with emission from an opacity-stratified jet. The bright flare after the IC211208A event indicates emission from a compact, optically thick region with enhanced activity and more efficient particle acceleration. Short optical and $γ$-ray bursts have been detected very close to the neutrino event, within a few days. The $\sim$12-day lag between the $γ$-ray and optical emissions is detected for the first time, suggesting that the emission regions are not co-spatial. A characteristic variability time-scale of $\sim$10-11 yr is robustly detected in the radio-mm bands ($\geq3σ$), while the optical and $γ$-ray data show weaker, shorter-period signals. The independent constraints from the VLBI core-shift measurements and radio time delays yield consistent estimates of the jet geometry and disturbance propagation, supporting variability governed by jet propagation effects. The long-term modulation is consistent with a slow variation in energy release at the jet base, while individual flares arise from shocks propagating downstream. Jet precession may contribute to the long-term modulation; however, the required viewing angles are inconsistent with the VLBI constraints, indicating that precession alone cannot explain the observed variability.

(Abridged.) The local environment of a Type Ia supernova (SN Ia) encodes information on progenitor age, star formation, dust, and chemical enrichment that may influence both its light-curve properties and its standardized luminosity. We assemble a homogeneous explosion-site dataset to test how near-ultraviolet (UV) imaging combined with optical integral-field spectroscopy improves the characterization of SN Ia environments and their connection to SN observables. We analyze 340 low-redshift SN Ia sites observed with HST/WFC3 F275W and matched VLT/MUSE spectroscopy. For each site we measure local photometry, extract a 1 kpc-aperture spectrum, model the stellar continuum with a joint UV+optical fit, derive local stellar and gas-phase properties, and fit the SN light curves with SALT3-NIR. UV emission is detected at 235 of the 337 sites with valid HST imaging. The UV- and Halpha-based SFR surface densities are strongly correlated, with the UV estimates systematically higher by about 0.5 dex. Including the UV constraint shifts local luminosity-weighted ages toward older values by a median of +0.20 dex for UV detections and +0.45 dex for UV non-detections, showing that UV data help break age-dust-metallicity degeneracies. Consistently, F275W-r is very strongly correlated with the joint-fit stellar age. Within the 169-SN Ia calibration sample, the joint stellar age, the Halpha specific SFR, and the UV specific SFR all correlate strongly with the SALT3 stretch parameter x1, with significances above 6.8 sigma. After standardization, Hubble-residual trends are weaker, with amplitudes of about 0.04-0.06 mag. The results support a primary age-stretch mechanism largely absorbed by standard light-curve corrections, while any residual luminosity dependence on local environment is weaker and subtler.

Magnetic fields thread the interstellar medium from the largest to the smallest scales and play an important role in molecular cloud evolution and star formation. Quantifying this requires measurements of the field strengths, and the most direct way to measure them is via the Zeeman effect in spectral lines. The effect is subtle for the typical field strengths expected from theory, from a few $μ$G in diffuse molecular clouds to a few 10s of mG in dense star-forming regions, and detections are scarce. Existing measurements of magnetic field strength suggest dense clouds and cores are marginally supercritical (cannot prevent collapse, but can inhibit it), but may be biased due to small sample sizes. Zeeman effect measurements tracing different scales and densities within molecular clouds can reveal the variation of field strengths, providing critical measurements to address the question of whether star formation is primarily regulated by magnetic fields or turbulence on different scales. Observations with SKA precursors such MeerKAT and FAST are beginning to increase the number of Zeeman effect detections in nearby star-forming regions. The SKA will extend their reach to many regions within our Galaxy that are best representative of where most stars form, while zooming in on the densest star-forming regions, providing a statistical basis for the role of magnetic fields in molecular cloud evolution and star formation. We present predictions and plans for Zeeman effect observations with the SKA telescopes, demonstrating the significant advances they will provide for studies of magnetic fields in molecular clouds.

We analyse physical and chemical structure of the hot molecular core RCW120 S2, based on high-sensitivity (2-4 mK) APEX observations at 1.3 mm. The analysis reveals a rich molecular inventory, including complex organic molecules (COMs) such as CH$_3$OH, CH$_3$CHO, CH$_3$OCHO and CH$_3$OCH. We derive gas temperatures (20-300 K), H$_2$ densities (10$^4$-10$^7$cm$^{-3}$), and molecular column densities. The detected emission probes a radially stratified envelope, with cooler (<= 60 K) and less dense (10$^4$-10$^5$cm$^{-3}$) outer layers traced by SO, SO2, and c-C$_3$H$_2$, while warmer (60-100 K) and denser (10$^5$-10$^7$cm$^{-3}$) inner regions are traced by H$_2$CO, OCS, and low-excitation CH$_3$CN. The hot gas (<= 100 K) exhibits broad (8-10 km/s) lines from high-excitation CH$_3$OH, CH$_3$CN, HDO, and CH$_3$OCH$_3$. Relative molecular abundances of COMs generally agree with astrochemical hot-core models, while methanol appears underabundant and CH$_3$CN overabundant compared to predictions. We attribute these discrepancies to the need for interferometric observations at intermediate spatial scales to resolve the core's true filling factor and radial gradients.

PKS 1725+123 is a flat-spectrum radio quasar (FSRQ) with a redshift of $z=0.586$. The detection of this object in the TeV band was reported by the MAGIC telescopes and H.E.S.S. in August 2025. Subsequently, we promptly initiated Target-of-Opportunity observations using the Space-based multi-band astronomical Variable Objects Monitor (SVOM) satellite. By analyzing the observational optical data from SVOM-VT and comprehensively examining the Fermi-LAT and Swift-XRT observational data, it was found that the source is in a high-flux state across the optical, X-ray, and GeV $γ$-ray bands around the time of the TeV detections. Its optical flux reaches a historically unprecedented high level and shows significant variability on timescale as short as minutes. The variability is accompanied by changes in the color index, exhibiting a \textit{bluer when brighter} behavior during the high-flux state. Based on the simultaneous multi-wavelength data, we construct the broadband spectral energy distribution (SED) of the source in the high-flux state. PKS 1725+123 demonstrates a remarkably high synchrotron peak frequency, which is distinctly different from that of other FSRQs. We propose a two-zone spine-sheath jet model to reproduce this SED. The optical--X-ray emission is generated by the synchrotron process of the relativistic electrons within a compact zone. The inverse Compton (IC) scattering processes of the same electron population contribute to the low-energy end of the Fermi-LAT spectrum, while the high-energy end of the Fermi-LAT spectrum is ascribed to the IC scattering of the synchrotron photons within the compact zone by the higher-energy electrons in an extended region.

We present an analysis of the intraday flux and spectral variability of BL Lacertae from 2020 September 30 to 2024 December 7, covering its highest recorded brightness and low, intermediate, and high flux states. Our study involved 62 nights of multi-band (BVRI) optical monitoring using four ground-based telescopes located in Egypt, Türkiye, and Bulgaria. We assessed intraday flux variability using the power-enhanced F-test and the nested ANOVA test. Significant variability was detected in 88 out of 117 light curves, consistent with previous studies of this object during active epochs. The maximum variability amplitude is 44.5 per cent in the B band. Spectral analysis reveals a bluer-when-brighter trend during intraday flares, supporting a synchrotron origin for the variable emission. For a subset of well-sampled flares, we model their profiles with a double exponential function, deriving rise and decay time-scales, thereby constraining the characteristic times of particle acceleration and cooling processes within the relativistic jet. Assuming a turbulent jet model, we determined limits on the radii and magnetic field strengths of the emitting regions. We detected soft time lags for a multi-band flare and from their analysis, derived the Doppler factor and magnetic field strength of the corresponding emitting region. Our long-term, high-cadence study confirms that BL Lacertae was in an exceptionally active phase during the 2020-2024 period, with intraday variability being a common phenomenon. The results underscore the dynamic nature of the jet emission region and provide valuable observational constraints for models of blazar variability and jet physics.

We present a method for determining the spin parameter $a/M$ and inclination angle $i$ of a non-extremal Kerr black hole from segments of the critical curve identified in black hole images. Although the critical curve itself is not directly observable, higher-order photon rings accumulate near it, and in realistic observations localized portions of the resulting brightness enhancement may be available for identifying segments of the critical curve. We introduce standardized segments of the critical curve and define three observables that characterize their geometry. We show that these observables uniquely determine $(a/M,i)$, together with an auxiliary parameter $r_{nl}\in[0,1]$ specifying the location of the identified segment along the critical curve, within the domain considered. Thus, even a segment of the critical curve contains sufficient geometric information to constrain the black hole spin and inclination without reconstructing the full critical curve. The framework is naturally suited to realistic observations and may be extended to more general rotating black hole spacetimes.

Tessellation methods are extensively employed in the analyses of cosmic large-scale structure (LSS). However, these techniques are highly sensitive to perturbations in both densities and positions of points, often leading to substantial rearrangements of tessellation configurations. As a result, considerable additional statistical errors are introduced in various tessellation-based statistics, thereby weakening their cosmological constraints. In this work, we identify this issue and propose an efficacious measurement scheme through subsampling and averaging to enhance the stabilities of tessellation-based statistics. As a case study, we apply the new scheme to measure multiple primary void statistics [i.e., void size function (VSF), void two-point correlation function (VTCF), and void power spectrum (VPS)] in two distinct classes of voids, based on Delaunay and Voronoi tessellations, respectively. We notice that the statistical uncertainties in void statistics can be predominantly attributed to tessellation instabilities. Through rigorous testing, we demonstrate that the proposed method can substantially eliminate these scatters to deeply mine the statistical power of void statistics. Specifically, we find that our method can dramatically boost the signal-to-noise ratios (SNRs) of void Baryon Acoustic Oscillations (BAOs) and significantly improve the constraining power of void statistics on cosmological parameters. These findings showcase enormous application potentials of our new method in maximizing extraction of cosmological information from galaxy surveys. Importantly, our method is simple yet highly potent with broad applicability, hopefully evolving into a standard framework for measuring tessellation-based statistics in the future.

Impulsive plasma dynamics in the laboratory are often driven by rising electric currents, yet their quantitative plasma response has not been well established. By means of fully kinetic particle-in-cell simulations and laser-driven capacitor-coil experiments, we show that a rising current expels plasma, forming an expanding magnetic bubble and accelerating particles. The expansion front velocity scales with the Alfvén speed determined by the magnetic field at its inner edge and the plasma density at its outer edge. This mechanism establishes impulsive current drive as a fundamental way that generates plasma flows and accelerates particles in laboratory plasmas, with potential relevance to astrophysics.

Broadband observations with the {\em Fermi} mission reveal that a large fraction of gamma-ray burst (GRB) spectra are dominated by non-thermal emission, while a small fraction are dominated by thermal/quasi-thermal emission, likely indicating the difference in jet composition among GRB. By selecting two typical long GRBs (080916C and 090902B) and two short GRBs (130603B and 090510), we present a comparative analysis to investigate whether the composition of prompt GRB jets influences the afterglow emission for bursts originating from both massive star collapse and compact binary mergers. Incorporating emission from both primary and cascade electron populations, we fit the multi-wavelength afterglow lightcurves of these GRBs with the standard forward shock model and analyze the particle acceleration and radiation physics of the jets. Our results show that the afterglow lightcurve evolution with the characteristic parameters is not related to the composition of the GRB fireball but rather depends on the ambient medium density. The early UV-optical afterglows of the two long GRBs are dominated by synchrotron emission from cascaded $e^{\pm}$ pairs produced via the $γγ$ annihilation process, whereas this is not the case for the two short GRBs. These results suggest that the internal energy of the fireball is converted into jet kinetic energy during the prompt phase, and that the fireball composition leaves no detectable footprint on the afterglow jet. Instead, the density of the ambient medium plays an essential role in shaping the afterglow emission.

The understanding of the mechanism(s) by which massive stars collapse and possibly explode is rapidly maturing. Uncertainties in the structure of the stellar core at the onset of collapse are often dominant in numerical simulations, and a limited number of progenitor grids are available. This is especially true for explosions where rotation and magnetic fields play a significant or primary role. We present a grid of 113 single-star models with initial masses $M_{\rm ZAMS}=30-90\,M_{\odot}$ and initial rigid rotation $ω_{\rm ZAMS}=0.5-0.99\,ω_{\rm crit}$ computed at $Z=0.001$ with the open-source stellar evolution code \textsc{MESA}. We adopt a 128-isotope nuclear reaction network capable of following the weak reactions deleptonizing the core during and after silicon core burning. By construction, these models experience rotationally-induced chemically-homogeneous evolution, and reach the onset of collapse ($v_{\rm infall}\lesssim -300\,\mathrm{km\ s^{-1}}$) with large and structured amounts of angular momentum, possibly sufficient to form accretion disks on a proto-compact object. Therefore, these progenitor structures provide a homogeneous set of models with updated input physics and improved algorithmic accuracy to understand stellar explosions of (some types of) stripped-envelope supernovae, possibly jetted and/or broad-lined, collapsars or magnetar-powered, and/or long $γ$-ray bursts.

Wide-field surveys like Euclid mark a new era of extragalactic stellar stream studies. With a large number of streams, it is now possible to constrain the dark matter halos of galaxies in a cosmological volume and draw comparisons to theoretical expectations for the geometry of dark matter halos. This study combines Euclid imaging with visual detection and segmentation annotations to analyse streams. We use projected stream morphologies to constrain the shape and centre-of-mass position (CoM) of each host galaxy's potential, jointly probing baryonic and dark matter distributions. These inferences complement weak lensing methods, with sensitivity to halo profile and geometry on sub-virial scales. The method enables both stacked, population-level constraints on halo flattening and CoM position, and constraints on these quantities for individual halos. We also present a novel method for transforming segmentation maps of stellar streams into smooth, curvature-preserving tracks optimised for fast and robust dynamical inference. This approach enables rapid modelling of stream morphology, supports a statistically rigorous combination of constraints across multiple streams within a single galaxy, and enables joint inference across galactic hosts. From our study of 13 galaxies with prominent tidal streams, we find agreement with spherical halos, albeit a mild preference for flattening with $q = 0.95^{+0.05}_{-0.10}$ at 68\% confidence. This is promising early agreement with $Λ$CDM predictions. With thousands more discovered streams expected across \Euclid's mission, our programme will enable precise measurements of halo shapes and CoM positions across large samples and redshifts, offering constraints on the geometry of dark matter halos.

We report an ab initio and master-equation investigation of the gas-phase reactions of thioformaldehyde (H2CS) with CN and CCH radicals, motivated by the recent detection of the S-containing species HCSCN and HCSCCH in cold interstellar environments. Structures and frequencies were obtained at the DSD-PBEP86/aug-cc-pVTZ level, with energetics refined by CCSD(T)-F12a calculations and kinetics treated using an energy-grained master equation. The CN + H2CS reaction proceeds through orientation-dependent entrance channels. Two barrierless addition pathways lead to a connected multi-well network that preferentially forms cyano thioformaldehyde, HCSCN + H, whereas abstraction-type channels leading to HNC + HCS or HCN + HCS contribute only marginally. The calculated kinetics indicate capture-controlled low-temperature reactivity and a strong preference for HCSCN formation, suggesting that this reaction should be considered in astrochemical models of cold clouds. For CCH + H2CS, barrierless capture gives access to two connected entrance adducts. Although the cyclic product is the most exothermic channel, its formation is kinetically hindered by a high-lying late transition state. The flux is shared between propynethial formation, HCSCCH + H, and the HCCH + HCS channel, with the latter being favored. These results show that subtle differences in radical structure, bonding preferences, and entrance-channel topology strongly affect product branching in S-containing radical-molecule reactions. The computed mechanisms and rate coefficients provide useful input for astrochemical models of sulfur chemistry in cold molecular clouds and for interpreting recent molecular detections in sources such as TMC-1.

The James Webb Space Telescope (JWST) has recently revealed evidence of supermassive black holes (SMBHs) forming in the cores of high redshift galaxies that grew in mass extremely rapidly during the first billion years of the universe. In this study we present a new solution where the growth mass factor of BHs, $γ(t) \propto \dot M_{BH}/M_{BH}$, varies with time, following the changes of radiative efficiency, $ε$, and Eddington ratio, $λ_{Edd}$, caused by the variation in accretion flow. By optimizing the accretion rate, $\dot M_{BH}$, we obtained a solution that is universal, that is, which applies to any BH at any redshift. This suggests that the mass of the BH is a thermodynamics state function, in good agreement with the no hair theorem and the four laws of mechanics of BHs.

It is commonly assumed that the spin-orbit angles of binary black holes (BBHs) originating from dense stellar environments rapidly converge to an isotropic distribution following a number of strong gravitational encounters. We challenge this assumption by modeling the evolution of the BBH orbital angular momentum orientation through successive binary--single encounters as a random walk on the unit sphere, yielding an exact solution for the orientation distribution after $n$ encounters and a closed-form expression for the number of encounters required to reach isotropy. To characterize the step distribution, we conduct a large suite of Newtonian point-particle scattering experiments with an equal-mass binary, varying the mass of the single, and obtain semi-analytic expressions for both the mean step size and the full distribution of steps. Applying these results to BBHs with initially aligned spins, as may arise from the evolution of primordial massive binaries, we find that spin-orbit alignment can survive several strong encounters before being erased. This has direct implications for the slight trend toward spin-orbit alignment reported in GWTC-5.0 as well as for the retention of hierarchical merger products, such as the components of GW231123.

Resolving the impact of magnetic field instabilities in triggering small scale turbulent flow and the associated rearrangement of the field is of critical importance in understanding multimessenger observables in binary neutron star mergers, and angular momentum transport in neutron stars and accretion disks. Direct simulation of these instabilities are unfeasible, however large-eddy simulations can incorporate the impact of this turbulence with a subgrid model. We present the first machine-learning-based subgrid model for special relativistic magnetohydrodynamics, trained using a neural network. We demonstrate its performance in online simulations of the 3D Kelvin-Helmholtz instability through both a priori and a posteriori tests. Evaluated in a low resolution simulation, our model captures magnetic field amplification of a simulation at 4 times the resolution with a speed-up of a factor 44. This demonstrates the applicability of such methods in general relativistic simulations of neutron star mergers and other scenarios.

We revisit the growth of central black holes via tidal disruption events (TDEs) and plunges of stellar-mass black holes (sBHs). Our model incorporates the current understanding of mass segregation, where sBHs sink to the center, enhancing the rates of both TDEs and plunges. We demonstrate that in dense cluster cores, with densities exceeding $10^6\,\mathrm{M_\odot\,{pc}^{-3}}$, seeds of initial mass $M_0\gtrsim100\,M_\odot$ undergo runaway growth. This runaway terminates once the black hole radius of influence surpasses the core radius, or equivalently, most of the core mass has been consumed. This typically results in an intermediate-mass black hole (IMBH), within a few $\mathrm{Gyr}$. Subsequent growth proceeds as a power law. In contrast to observed supermassive black holes (SMBHs), which tightly follow the famous $M \propto σ^β$ with $β\cong5$, our derived sBH accretion rates, integrated over a galactic lifetime, predict final masses of $M\approx10^5\,M_\odot \times(σ/50\mathrm{km\;s^{-1}})^{2.5}$. While our prediction for the contribution of plunges to the growth of the IMBH is robust, the TDE contribution can be negligible if only a small fraction of their mass is actually accreted or up to 3 times higher than the plunges contribution if half a stellar mass gets accreted in each event. Below $M\sim10^5\,M_{\odot}$ this accreted star and sBH mass exceeds the extrapolation of the observed $M$-$σ$ relation. This predicts that the $M$-$σ$ scaling must flatten below $M \sim 10^5\,M_{\odot}$ to a shallower, $2.26<β<2.5$ profile. If confirmed by observations, this would indicate capture-driven growth.

We present the evolution of the resolved mass-weighted age, stellar metallicity, and sSFR of 872 Milky Way Analog (MWA) progenitors up to redshift $z=5$ from the Canadian Unbiased Cluster Survey (CANUCS). The metallicity and mass-weighted ages were obtained via spatially resolved SED-fitting with the non-parametric code Dense Basis. We split the sample into mergers versus non-mergers using the merger parameter from the Gini-$M_{20}$ plane obtained through Gini-$M_{20}$ analysis of the morphology of the stellar mass maps with Statmorph. Across our redshift range, non-mergers have negative or flat average age gradients from $-0.022$ to 0.005 dex/kpc, and positive or flat sSFR gradients from $-0.089$ to 0.092 dex/kpc, consistent with inside-out assembly. The average $\log(Z/\Zsun)$ gradients for non-mergers range from $-0.029$ to 0.044 dex/kpc, however, positive gradients only appear between $2 < z < 3$. At every redshift epoch, mergers typically have flatter age gradients, more negative sSFR gradients, and similar metallicity gradients compared to non-mergers. We divide the property maps of ongoing mergers into separate regions based on their component galaxies, and find little to no difference between the components' average ages or metallicities, but the less massive of the merging system is on average $0.1-0.4$ dex higher in sSFR. Our results point to major mergers contributing some momentary disruption to the general trend of inside-out mass assembly, but does not upend the overall picture of MWA disks growing inside-out over cosmic time.

A systematic Bayesian inference study of the equation of state (EOS) of dense matter with strangeness is presented, extending five relativistic mean-field (RMF) models with both constant and density-dependent couplings to include the full baryon octet. The hyperon-nucleon couplings in the scalar channel are varied within ranges informed by hypernuclear data, while vector isoscalar couplings are fixed by the SU(6) symmetry quark model. Observational constraints from NICER (PSR J0030, J0437, J0740) and GW170817, theoretical constraints from chiral effective field theory ($χ$EFT) and perturbative QCD (pQCD), and experimental constraints from nuclear saturation properties are imposed simultaneously. We find that the inclusion of hyperons systematically reduces the maximum neutron star mass by $0.05$-$0.10,M_\odot$ across all models while increasing the radius at $1.4,M_\odot$ by $0.5$-$0.8$ km. The speed of sound exhibits a characteristic softening at densities $2$-$3,ρ_{\rm sat}$ coinciding with hyperon onset. All hyperonic models remain consistent with the $2,M_\odot$ constraint. Models with a more flexible isovector channel span a larger proton fraction when only nucleons are included. However, the extra flexibility is visibly suppressed by hyperons, meaning that the average proton distribution is independent of model flexibility when hyperons are included. Less flexible models show comparable or slightly increased proton fractions due to EOS stiffening when hyperons are included. Only a residual number of hyperonic equations of state give rise to a mass-radius curve with a negative slope at low masses. A $1.8,M_\odot$ neutron star with a radius larger than or similar to the radius of a $1.2,M_\odot$ star would provide strong evidence that the star contains baryonic degrees of freedom beyond nucleons.

X-ray observations are a primary probe of the intracluster medium, widely used to infer galaxy cluster masses and scaling relations. We present and validate a pipeline to generate X-ray emission maps of galaxy groups and clusters from large cosmological simulations, and use it to study the origin of deviations from self-similarity across the group-to-cluster transition. We apply this pipeline to the Illustris-TNG300 and MillenniumTNG simulations, constructing \mbox{X-ray} emission maps and spectra based on APEC cooling functions. For simulations that do not explicitly track individual chemical abundances, we introduce a metallicity-based prescription that accurately reproduces the full spectral emission. We derive the $L_{\mathrm{X}}$--$M_{500}$, $M_{\mathrm{gas}}$--$M_{500}$, and $T$--$M_{500}$ scaling relations over $10^{12.5} \leq M_{500} \leq 10^{15.5}\,\mathrm{M}_\odot$, compare them with observational data, and quantify the hydrostatic equilibrium and spectroscopic temperature biases through synthetic X-ray analyses. The simulated scaling relations are in good overall agreement with observations and are best described by broken power laws with a pivot at $M_{500}=10^{13.67}\,\mathrm{M}_\odot$. At high masses, the slopes are close to self-similar expectations; at lower masses the relations steepen significantly, reflecting the growing importance of AGN feedback. X-ray hydrostatic masses are systematically underestimated by $15\%$, independently of cluster mass. When spectroscopic effects are included, the bias becomes mass-dependent, ranging from $15\%$ at low masses to $21\%$ at high masses. The recovered X-ray luminosity, measured in the $0.15$--$1\,R_{500}$ aperture, is also mass-dependent: high-mass clusters are underestimated by $18\%$, while low-mass systems show discrepancies of up to $33\%$, driven by single-temperature spectral modelling of the gas outside the core.

Hadronic stars and strange quark stars could coexist within the so-called two-families scenario. In this respect, hadronic matter and strange quark matter correspond to two distinct equilibrium phases described by two different equations of state. We perform here the first detailed Bayesian analysis that makes use of astrophysical and laboratory data in order to constrain the equations of state adopted within the two-families scenario for hadronic and strange quark matter. In particular, in hadronic matter we consider the possible formation of hyperons and delta resonances (beside nucleons) within a class of non linear relativistic mean field models and in quark matter we consider the possible formation of a color-superconducting phase within a bag-like model. Results of the analysis indicate that while at the moment both the one-family and the two-families scenarios are compatible with the data, by comparing the Bayes factors of both models, the two-families scenario is favored with respect to the one-family scenario. Specifically, the two-families framework naturally relieves the tension between the intermediate-density softness of the equation of state required by small-radius objects and the high-density stiffness needed to support massive pulsars. Ultimately, future detections of even more massive compact objects, very compact ordinary-mass objects, or precise measurements of two distinct masses with the same radius, will provide strong indications in favor of the two-families scenario.

The radio--X-ray correlation and Fundamental Plane (FP) of black hole activity can serve as a diagnostic tool for the origin of X-ray emissions. There was a scaling relation between radio and X-ray emissions for High-synchrotron-peaked BL Lacertae objects (HBLs), i.e., $L_{\rm{R}}\propto L_{\rm{X}}^{0.64}$, which can be explained by ADAF-dominated mode or synchrotron cooling (Syn-c). However, many results of studying blazar physics show that the X-ray emissions of HBLs are mainly produced by the synchrotron process of jets. Therefore, Syn-c appears to provide a plausible explanation for this relation. To further clarify the origin of X-ray emissions of HBLs, we constructed a sample containing 69 HBLs in this paper to re-investigate their radio--X-ray correlation and FP. Considering the Doppler beaming effect, we find that the intrinsic radio--X-ray correlation and FP of HBLs are $L_{\rm R,int}\propto L_{\rm X,int}^{0.68}$ and $\log L_{\rm R,int}=(0.57\pm0.06)\log L_{\rm X,int}+(0.33\pm0.11)\log M_{\rm BH}+(12.65\pm2.00)$, respectively. Our results agree with the scaling relation mention above, which suggests these scaling relations are not artificial. By employing the theoretical model of Syn-c, we find that these shallow radio--X-ray correlations and FP are caused by Syn-c, which implies that the X-ray emissions of HBLs may be produced by rapidly cooling, high-energy electrons accelerated at a shock. This is consistent with results of the recent X-ray polarization observations of HBLs. Our results provide the observational evidence of $L_{\rm R}\propto L_{\rm X_{\text{Syn-c}}}^{0.6\sim0.7}$.

Context: With Gaia's third Data Release (DR3), low-resolution BP/RP (XP) spectra became available for more than 219 million sources. In previous work, we developed a fast method for detecting absorption and emission lines while measuring their equivalent widths directly from the XP spectral coefficients. Aims: Our aim is to conduct a systematic search for H$α$ emitters in the Gaia DR3 XP spectra. Methods: We measured the equivalent width of the Balmer H$α$ line for all sources with Gaia XP spectra, and selected an initial sample of 556100 sources depicting an H$α$ emission-like feature at 2$σ$ level or above. Using a semi-supervised classifier complemented by manual spectral labelling, we removed sources that merely masqueraded as emission-line stars. Results: We find that 95% of the initial candidates are most likely cool M-type stars, where a molecular absorption band gives rise to a local maximum that can mimic H$α$ emission. Subsequently, 28394 sources are flagged as bona-fide H$α$ emitters using the Gaia DR3 XP spectra. The majority are Be-like candidates (26287), active M-type stars (717), Herbig Ae/Be candidates (525), quasi-stellar objects (204), Wolf-Rayet stars (177), cataclysmic variables (51), and carbon stars (18), among other sources. Our equivalent widths and stellar classifications are in good agreement with results from higher-resolution studies and recent Gaia XP spectra-based catalogues. Ground-based spectroscopic follow-up for some of the newly found emitters reinforces our confidence in the catalogue.

We report the discovery of [CII] 158 micron emission in and around the most distant known massive quiescent galaxy RUBIES-UDS-QG-z7 at z = 7.27. Observed with ALMA in band 6, the [CII] line independently confirms the spectroscopic redshift from JWST/NIRSpec spectra at low and medium resolution. The emission extends over an effective radius R_eff,[CII] = 8 +/- 3 kpc, well beyond the compact stellar body traced by JWST/NIRCam (R_eff = 209 (+33/-24) pc), with a significant fraction of approximately 70% of the flux arising from a circumgalactic halo. No dust continuum is detected at rest-frame ~160 micron, setting an upper limit on the infrared luminosity of L_IR < 1.4 x 10^11 Lsun, overall consistent with expectations from rest-frame UV to near-infrared SED modeling under energy balance. Converting the galaxy-scale [CII] emission into cold gas mass, we find log(M_mol/Msun) = 9.53 (+0.32/-0.31) and log(M_HI/Msun) = 9.46-10.34, depending on the assumed calibration and metallicity. Despite being approximately 10x more gas-poor than typical star-forming galaxies at fixed redshift, stellar mass, and [CII] to gas mass conversion, RUBIES-UDS-QG-z7 retains a substantial cold gas reservoir with fractions f_gas >~ 20% and long depletion timescales across most assumptions. The extended [CII] halo carries approximately twice as much gas as the galaxy alone and shows a blueshifted velocity offset consistent with the tentative gas outflow detected in MgII absorption in previous work, suggesting a past episode of AGN-driven gas expulsion possibly linked to the suppression of star formation. The presence of a large gas reservoir in and around a massive quiescent galaxy just 700 Myr after the Big Bang implies that whatever mechanism is suppressing star formation must be remarkably effective at maintaining a low star formation efficiency on ~100 Myr timescales, even in the presence of abundant fuel.

We present a new all-sky model of low-frequency diffuse Galactic emission in the regime where synchrotron, free-free, and spinning dust dominate. The model extends the Planck 2015 diffuse component-separation analysis by incorporating recent radio and microwave surveys. We fit 35 full- and partial-sky maps at 1 degree resolution, including S-PASS at 2.30 GHz, C-BASS at 4.76 GHz, and QUIJOTE at 10-20 GHz, together with reprocessed WMAP and Planck LFI data from the Cosmoglobe collaboration and Planck HFI channels. Using a Bayesian parametric approach with Commander, we derive spatially varying amplitude and spectral parameter maps for the dominant low-frequency foreground components in total intensity. The main products are a full-sky synchrotron amplitude and spectral-index solution, an all-sky characterization of spinning dust emission with a single-component log-normal spectral model, and a reconstructed all-sky total-intensity map at 4.76 GHz tracing diffuse synchrotron emission with reduced systematics relative to Haslam 408 MHz. The revised low-frequency anchoring increases the recovered synchrotron amplitude: at 4.76 GHz, it is approximately a factor of two higher than the Planck 2015 prediction. The model achieves RMS temperature residuals below 10 $\mathrmμ$K over 95% of the sky up to 353 GHz, with fractional residuals below 1.5% in the Galactic plane and below 5% across QUIJOTE bands. Residual angular power spectra lie more than two orders of magnitude below the CMB spectrum. These products describe the transition between radio and microwave emission and provide a new reference for foreground modeling and sky-simulation applications.

Gamma-ray emission has now been detected from a variety of source classes in the Galaxy, including clusters of young massive stars. RCW 38, a very young embedded massive star cluster, is a case of particular interest: its gamma-ray emission detected up to hundreds of GeV, provided the first observational evidence of high-energy particle acceleration powered exclusively by stellar winds. In this work, we aim to characterize the emission mechanisms responsible for the gamma-ray flux in RCW 38 and to provide estimates of the acceleration efficiency, as well as the fraction of accelerated electrons compared to protons, $K_{ep}$. We present the most comprehensive multi-wavelength study of a single star cluster to date. Our analysis ranges from MHz radio observations obtained with the GLEAM-X survey from the Murchison Widefield Array (MWA) to GeV gamma-ray data from Fermi-LAT, and includes GHz and THz measurements from Parkes, Planck, and IRAS. We model the thermal and non-thermal emission of RCW 38 using an eight-parameter model constrained by the Markov chain Monte Carlo method. Our results support an interpretation in which the gamma-ray emission from RCW 38 is produced by hadronic interactions with the host molecular cloud. We derive robust constraints on the electron-to-proton ratio, with $K_{ep} \lesssim 10^{-3}$, and on the acceleration efficiency, estimated to be $\gtrsim$1%, consistent with the values required to explain the cosmic-ray composition, and in particular its $^{22}$Ne anomaly. These results strengthen the idea that stellar clusters play a significant role as contributors to cosmic-ray protons in our Galaxy at least up to energies of a few TeV. Future investigations with the next generation of ground-based detectors will determine whether they also play a relevant role at higher energies, particularly in the context of the cosmic-ray knee.

Testing the validity of the cosmological principle's assumption of large-scale isotropy remains crucial for modern cosmology. We investigate the angular distributions of gamma-ray bursts using the GRB catalog from Neil Gehrels Swift Observatory (Swift) for an independent probe of isotropy. Using the HEALPix spherical harmonic decomposition, we estimate the dipole and quadrupole amplitudes and compare them against the null hypothesis obtained from 500 isotropic Monte Carlo realizations. Our results show 2.9$σ$ dipole and 7.2$σ$ quadrupole amplitude when applied to the raw data. To account for observational biases, we then create an exposure map using the pointing history, roll angle, and the partial coding fraction of the Swift Telescope. Reevaluating the null hypothesis using this map reduces the significance of these anisotropies to less than $1σ$. Therefore, our findings confirm statistical isotropy of the GRB sky using the Swift data, consistent with previous studies. We have also made the Swift exposure map publicly available.

We report the first XRISM observation of the low-mass AGN in the nearby dwarf galaxy NGC 4395 ($M_{\rm BH}\sim10^{4-5}\,M_\odot$), complemented by a simultaneous NuSTAR observation. We constrained the continuum by jointly fitting the XRISM/Resolve (2-12 keV) and NuSTAR (3-30 keV) spectra while excluding the Fe K band (5.5-7.5 keV). Relative to this baseline continuum, the time-averaged Resolve spectrum revealed an unresolved neutral Fe K$α$ core with a velocity width of $\lesssim$110 km s$^{-1}$ and an adjacent redward wing. The red wing was well reproduced by an additional relativistically broadened Fe K component. Furthermore, time-resolved spectroscopy with $\approx$87 ks bins showed that the diskline profile varied significantly over the $\sim$400 ks observation. This evolution can be interpreted in terms of changes in the inner radius of the line-emitting region, together with a possible inclination modulation with a period of $\approx$210 ks. If interpreted as Lense-Thirring precession of a tilted inner flow, the observed period would favor the low end of the black hole mass estimates ($M_{\rm BH}\approx9\times10^3\,M_\odot$) and imply a moderate spin ($a\gtrsim0.6$). These results highlight the capability of XRISM to track relativistic disk dynamics in AGNs.

We present a multiwavelength investigation of the molecular cloud complex G12.79+0.43, which extends over $\sim18'$ on the sky. Several infrared- and radio-bright regions are arranged along an irregular rim, surrounding a central region characterised by diffuse 24~$μ$m emission. CO molecular line observations reveal three prominent velocity components along the line of sight. Low-frequency radio continuum observations at 666 and 1300~MHz show diffuse emission spanning $\sim10.5'$ ($\sim$7.3~pc), predominantly filling the central region enclosed by the infrared-bright structures. We identify 70 compact radio sources and six \hii~regions across the cloud complex, which are likely powered by early B-type ZAMS stars. Using infrared data, we identify a total of 82 YSO candidates, including 28 Class~I sources, distributed across the cloud complex. On larger scales, the kinematics of the molecular gas over a $2^\circ\times2^\circ$ field indicate that G12.79+0.43 is located along the rim of a larger molecular superbubble (diameter $\sim50$~pc) that also encompasses the well-known W33 region. The inferred expansion age of this superbubble is $\sim0.3$~Myr. While the spatial association between G12.79+0.43 and the superbubble is evident, the current data do not allow us to establish a clear causal connection between the superbubble evolution and the ongoing star formation within G12.79+0.43.

The observed abundance of galaxy clusters as a function of mass and redshift provides a powerful route to precision cosmology; a key challenge for cluster cosmology is to establish the relation between cluster observables and cluster masses, for which cluster weak gravitational lensing has become the standard tool. A key challenge for cluster lensing is that the shear signal near cluster centers can reach the non-linear regime, where many shear estimators rely on perturbative assumptions that must be explicitly validated. In this work, we use image simulations to test the performance of the shear estimator AnaCal for cluster weak lensing under conditions representative of the 10-year LSST data. We find that AnaCal recovers the input shear with minimal bias even at mildly high shear, $|g|\sim 0.15$. We discover a radially decreasing mean shear response as seen previously in data, driven by the radial dependence of the convergence field; if unmodeled, this effect can bias shear inference. We also find a positive shear-estimation bias at third order in the reduced shear near the cluster center. However, because only a small fraction of galaxies lie in the high-shear regime and those measurements are further downweighted by the covariance matrix, the resulting mean cluster-mass bias for cluster lens masses in $[10^{14} M_\odot, 10^{15} M_\odot]$ -- adopting a scale cut of $\sim 0.2$ Mpc at $z=0.25$ -- is $0.24 \pm 0.26\%$ under ideal settings. These results demonstrate that AnaCal is a robust tool for accurate cluster mass calibration in the LSST era.

We present a study of dwarf galaxy candidates in the deepest optical imaging yet obtained of the field surrounding the nearby galaxies NGC 5068 and NGC 5084/NGC 5087, the latter two located in the peripheries of the Virgo Supercluster. This field, covering $\sim 2.6$ deg$^2$, was observed as part of the multi-band, wide-field and very deep data from VST-SMASH, a distance-limited program ($D < 11$ Mpc) that reaches $g$- and $r$-band surface brightness depths of $μ\sim 30$ mag arcsec$^{-2}$. Using a two-step visual inspection procedure, we identify 47 dwarf galaxy candidates and perform the surface photometry of the sample and the fitting procedure with 1D Sérsic model on their profiles. Only 4 galaxies were previously reported in the literature, augmenting by one order of magnitude the number of dwarfs discovered in these regions. The colors (median $g-r = 0.57$ and $r-i = 0.24$ mag) and structural properties of the dwarf candidates are consistent with the literature, as are their scaling relations with effective radius, Sérsic index ($n < 2$), and absolute magnitude. We also investigate their central colour gradients, which exhibit significant scatter, and discuss them within the broader context of galaxy formation. We finally analyze their spatial distribution relative to potential host galaxies. We identify reasonable associations with NGC 5084, NGC 5087, and NGC 5068 as likely hosts for a significant fraction of the sample. Several candidates are at physically credible distances from NGC~5068, despite what their offset size-luminosity relation alone might indicate. Future spectroscopic and deeper imaging follow-up is required to determine distances and velocities, enabling robust association with hosts, studies of satellite distributions and counts, and comparisons with cosmological expectations for planes of satellites and dark matter models. (abridged)

High-redshift little red dots (LRDs) are compact sources characterized by V-shaped spectral energy distributions (SEDs), broad emission lines, and often prominent Balmer breaks. Their high number density and apparently large black hole masses suggest that they are essential to the early evolution of galaxies and supermassive black holes (SMBHs); however, the nature of their central engines remains uncertain. Here, we propose that LRDs are the supermassive, high-redshift analogs of the hyper-Eddington accreting Galactic microquasar SS~433, viewed at high inclinations. By scaling the hyper-Eddington accretion physics from stellar-mass black holes to supermassive scales, we show that the observed LRD features, including X-ray weakness, soft optical SEDs, apparent sub-Eddington accretion ratio, and Balmer breaks, emerge naturally from the self-shielding geometry of a puffed-up accretion disk. In this framework, the broad-line regions are ionized by anisotropic radiation escaping from the inner disk, analogous to the unseen UV/X-ray emission revealed by the W50 nebula in SS 433. Their low-inclination or lower-accretion-rate counterparts would appear as little blue dots (LBDs) or normal active galactic nuclei. Our model predicts that the Balmer break strength positively correlates with the broad-line width, that the emission lines are more variable than the optical continuum, that LRDs are intrinsically more luminous than observed, and that LBDs are more variable than LRDs. This unified-scale model redefines LRDs as the essential laboratories for observing the rapid accretion-driven growth that shaped the early assembly of galaxies and their central SMBHs.

We present the morphological, photometric, and spectroscopic properties of a z ~ 1 galaxy, "lil gal", serendipitously detected in JWST Mid Infrared Instrument (MIRI) images of nearby galaxy VV 340. In the MIRI F560W and F770W images, we identify what appears to be a spiral galaxy with a central bulge. However, in the F1500W image, a second peak appears ~0.7" northwest (NW) from the central bulge, calling into question the nature of this source as a clumpy disk, a high-redshift interloper, or a galaxy merger. Multi-band analyses of the three MIRI and four Hubble Space Telescope (HST) images suggest a photometric redshift of ~0.92. Spectroscopic analyses of data from the Keck Near-Infrared Echellette Spectrometer (NIRES) reveal two sets of [N II] and H-alpha emission lines corresponding to the two observed sources. A redshift of z = 0.9248 is identified for the NW companion. Fainter emission lines are identified from the underlying galaxy at z = 0.9225, suggesting a merging galaxy pair at a projected separation of ~5 kpc. From the emission line ratios from Keck NIRES and Keck Cosmic Web Imager (KCWI) spectra, we classify the system as hosting regions of active star formation, likely attributed to merger-induced starburst activity. The results demonstrate the necessity of resolved, spectroscopic follow-up analyses of galaxies found in deep JWST images to disentangle the role of galaxy mergers from clumpy disk galaxies at z ~ 1 to cosmic noon and beyond.

Globular clusters (GCs) host multiple stellar populations with distinct chemical compositions, but their properties among very low-mass stars remain poorly constrained. The James Webb Space Telescope (JWST) enables precise infrared studies that are highly sensitive to abundance variations in cool stars. We initiate a homogeneous survey of Galactic GCs, based primarily on deep JWST GO-8960 observations and complemented by archival JWST and Hubble Space Telescope data, to characterize multiple populations across a wide range of cluster properties. In this first paper, we present the survey and initial NIRCam results. We analyze eleven GCs, deriving high-precision photometry and astrometry to measure proper motions. Multiple populations are detected among low-mass stars in all clusters, with diverse behaviors. We find discrete main sequences in NGC 288, NGC 6723, and NGC 2808, and more continuous distributions in NGC 104 and the Type II clusters NGC 1851 and NGC 6656. The bulge clusters NGC 6528, NGC 6553, and NGC 6440 show patterns consistent with varying helium and oxygen abundances that do not scale simply with cluster mass. In Terzan 5 and Liller 1, we identify populations spanning different ages and helium variations within the old population of Terzan 5. We also detect an M-dwarf gap in NGC 104 around 0.35 solar masses, consistent with the Jao Gap of field stars and open clusters. This work establishes the foundation for a homogeneous JWST survey of Galactic GCs and provides a valuable dataset for studies of cluster evolution, Galactic stellar populations, and background extragalactic sources.

Aims. Classical Cepheids (CCs) are fundamental standard candles for the extragalactic distance scale through their period-luminosity (P-L) relations. Although well calibrated in broadband systems, their behavior in narrowband filters remains poorly explored. Combining broad- and narrowband photometry can improve constraints on extinction and dust distributions. Methods. We derived multiband P-L relations for more than 8000 CCs in the Magellanic Clouds (MCs) using Southern Photometric Local Universe Survey (S-PLUS) observations in five broad and seven narrow bands. Single-epoch magnitudes were corrected using a framework that simultaneously accounts for relative reddening and line-of-sight depth. P-L coefficients were determined through iterative fitting and maximum-likelihood optimization, with uncertainties estimated from Bayesian Markov chain Monte Carlo analyses for both fundamental- and first-overtone-mode Cepheids. Results. We present the first Cepheid P-L relations in the S-PLUS photometric system. The relations provide consistent distance and reddening estimates across all 12 bands. Their dispersion decreases toward longer wavelengths, while the slopes vary from $\sim-2.0$ to $-2.8$ in the bluest bands and from $\sim-2.9$ to $-3.2$ in the reddest bands. The resulting reddening maps reveal strong spatial extinction variations and recover prominent structures, including 30~Doradus and the H\,I supergiant shell SGS~12 (LMC~3). The inferred orientations of the MCs is consistent with previous studies. Conclusions. Accurate distances and reddening can be recovered from single-epoch multiband photometry when combined with light-curve corrections, enabling high-precision three-dimensional mapping of the MCs.

Feedback from massive stars in the form of radiation and winds impacts the associated host molecular cloud. Feedback can disperse cloud material and lead to the destruction of the cloud. Recent observations of the ionized carbon CII 158 micron line in high-mass star-forming regions have demonstrated that this line is an excellent tracer of the gas dynamics in such environments. Expanding CII shells have been detected, along with high-velocity gas escaping the natal cloud through low-density channels. Motivated by these results, we conducted a systematic analysis of spectrally resolved CII maps obtained with SOFIA towards ten high-mass star-forming regions hosting at least one O-type star. Across all regions, we identify high-velocity CII line wings with velocities that exceed the cloud escape velocity, indicating that this gas is not gravitationally confined. We show that the high-velocity gas exhibits a complex velocity structure and cannot be attributed solely to a single, coherent expanding CII bubble. The amount of material in these erosion flows depends on the evolutionary stage of the molecular cloud and its associated HII region. Once the initial bubble around the cluster ruptures, typically after 0.1 Myr, gas is expelled from the cloud. The resulting cloud erosion timescales based on these directly observed mass ejection rates typically vary between 2 and 10 Myr after the formation of the first O stars, similar to other indirect measures of molecular cloud life times. These results suggest that stellar feedback is able to remove enough molecular gas to terminate the star formation in the host cloud.

We present first results from the JWST LEGGOS Survey (LEnsing and Galaxy Growth: Observing Substructures), aimed at studying the physics of clumpy star formation and quenching in eight lensed galaxies at $z\sim2$--4. LEGGOS combines multiple Cycle 2 JWST GO programs (GO 4125, GO 3843) and Cycle 1 archival data, and utilizes strong gravitational lensing with NIRCam imaging and NIRSpec integral-field spectroscopy. LEGGOS targets UV-bright, highly magnified systems to resolve $\sim$10--200 pc regions in both rest-frame optical continuum and nebular emission. This overview paper describes the survey design, data reduction and calibration strategy, and science-quality data products, and highlights early examples demonstrating how spectroscopy breaks key degeneracies inherent to photometry-only clump studies, including identifying recent quenching in previously-thought UV star forming galaxies. We introduce a uniform analysis framework that jointly models lensing reconstruction, multi-band photometry, and integral field spectroscopy to disentangle multiple stellar populations within individual clumps and their surrounding diffuse regions. Using maps of Balmer recombination lines and key emission line diagnostic ratios, we connect star formation histories, dust attenuation, and nebular conditions on sub-kpc scales -- LEGGOS galaxies range from uniform metallicities across the whole galaxy, to having higher clump metallicities and harder ionization conditions relative to diffuse regions. The full survey dataset, with simultaneous flux and morphology constraints on clumpy source-plane regions, and a flexible spectrophotometric SPS modeling approach, provides a direct bridge between parsec-scale star formation physics and galaxy assembly at and beyond cosmic noon, offering a robust and efficient means of resolving star formation in the first galaxies.

We present the results from an extensive broad-band (radio to X-rays) observing campaign of the engine-driven Type Ib SN 2012au in the first 13 years of evolution. The early-time ($δt\leq{190}$ d) radio and X-ray evolution is well-described by conventional models of a forward shock interacting with a wind-like circumstellar medium ($ρ_{\rm{CSM}}\propto{r}^{-2}$). However, starting at $δ{t}\approx{6.7}$ yr, we detect a significant radio re-brightening. This late-time emission is dominated by a luminous component characterized by a broad and rapidly evolving spectral peak and a shallow optically thin spectral slope, $F_ν\proptoν^{-0.31\pm0.02}$. These properties imply a compact emitting region ($R\lesssim{10}^{16}$ cm) expanding at a remarkably slow velocity ($\lesssim{500}$ km/s) into a high-density environment ($\geq{10}^4 \rm{cm}^{-3}$), accompanied by a hard electron power-law index $p\approx{1.6}$. No soft or hard X-ray emission is detected at any epoch, indicating that high-energy radiation is either strongly absorbed or intrinsically absent. In the context of aspherical shock-CSM interaction models, these observations imply extreme properties of the CSM (geometry, density, total mass) that lack clear astrophysical motivation. Instead, we show that the emergence of radiation from a newborn Pulsar Wind Nebula (PWN) naturally explains the radio spectral evolution and high-energy limits, where the emission is governed by the adiabatic expansion of a relic pair plasma. We conclude that SN 2012au represents the most compelling candidate for a young, newborn PWN discovered to date, a scenario that can be directly tested with pending Very Long Baseline Interferometry (VLBI) observations.

We present a LensMC lensing analysis of Euclid Quick Release 1 images that were made available in March 2025. We measured shapes, positions, weights, and other morphological parameters of galaxies with a surface number density of $26\;\text{arcmin}^{-2}$ for $I_{\scriptscriptstyle\rm E}<24.5$, achieving 75 arcmin$^{-2}$ for $I_{\scriptscriptstyle\rm E}<27$, in 63 deg$^2$ of Euclid VIS images. This is the first shear measurement catalogue produced with \lensmc in anticipation of the Euclid Data Release 1. To within the scope of this work and availability of survey volume, we found no spatial dependency in the additive biases. However, we developed an empirical bias correction based jointly on galaxy sizes and magnitudes, which was applied to the measurement of the lensing signal over the whole area. In order to validate the quality of our measurements, we calculated two-point statistics and cluster profile measurements, and cross-checked results with external cluster catalogues from WISE and the Dark Energy Survey. Additionally, by stacking shear profiles of random clusters we found that the low-redshift, large radii bins may be still contaminated by residual systematic effects. Thanks to Euclid image resolution and depth and overall good control of systematic errors, we are able to constrain the lensing profiles of clusters with masses of $10^{14}M_\odot$ out to $z\approx2$ over nearly 10 Gyr of evolution history.

Dual nuclei systems are galaxy merger remnants or closely merging galaxies that have two distinct stellar cores separated by ~ 10pc to 10kpc. They are important laboratories for probing the co-evolution of stellar populations, galaxy dynamics, and central black holes during the hierarchical assembly of galaxies. In this study, we present a spectroscopic analysis of a sample of dual nuclei from the GOTHIC survey, using the penalized pixel-fitting (pPXF) code. The sample consists of star forming nuclei pairs, dual active galactic nuclei (DAGN) and mixed pairs. Using the SDSS spectra, we extracted stellar kinematics, emission line fluxes, the star formation history, metallicity of the nuclei, and derived important properties such as the supermassive black hole (SMBH) masses, accretion rates and SMBH ratios. We compared different properties of the nuclei in the dual systems, such as stellar velocity dispersion, stellar masses, black hole masses, age and metallicity. Our results show that the SMBH masses are higher for BHs in galaxy mergers compared to single nuclei for a given stellar mass, thus revealing that SMBHs grow during the galaxy merging process and not only due to the merger of SMBHs. Our study provides new observational constraints on the dynamical and evolutionary states of dual-nuclei systems, offering a deeper understanding of the role these systems play in galaxy evolution and central black hole growth.

We present the first cosmological model for the lifecycle and luminous properties of PAHs in galaxies as they evolve from z=6-->0. We model 40 zoom-in galaxies, coupled with an on-the-fly model for the evolution of dust grains in the ISM. We assume that PAHs are ultrasmall (a < 13 Angstrom) carbonaceous dust grains, and couple this model with single-photon excitation calculations to compute the emergent mid-infrared spectra. (1) If we assume that dust is large upon formation, then PAHs are naturally able to form in situ in the ISM via grain-grain shattering. Interstellar collision velocities increase in low density, diffuse gas in our model; as galaxies evolve, the increase in fractional mass of diffuse gas drives an increase in grain-grain collision velocities and a corresponding rise in the PAH mass fraction (qPAH) from ~5 x 10^{-4} at z~4 to ~10^{-2} at z~0. (2) Increased PAH production in the diffuse ISM results in an inverse relationship between qPAH and the molecular gas fraction. (3) The PAH light-to-mass ratio scales linearly with the radiation field intensity (LPAH/MPAH ~ G_0) but anti-correlates with qPAH, because high-Sigma_SFR galaxies have a denser ISM that suppresses shattering. This means the physical qPAH and observed LPAH/LFIR do not evolve in lockstep. (4) The PAH-metallicity relationship (PZR) arises naturally in this framework: galaxies enrich and grow their diffuse ISM fraction simultaneously, linking rising metallicity to rising qPAH. Our models represent the first to reproduce the PZR observed across z=0-2. (5) The LPAH-SFR and LPAH-M_mol relations emerge from two effects: more massive galaxies have larger PAH reservoirs, and higher-SFR galaxies excite their PAHs more efficiently per unit mass. Taken together, these results suggest that grain-grain shattering in the diffuse ISM is the main driver behind the evolution of cosmic PAH abundances.

Strong gravitational lensing by galaxy clusters combined with the resolution of JWST enables studies of star formation on ~10-100 pc scales in galaxies at z~2-4. As part of the LEnsing and Galaxy Growth: Observing Substructures survey (LEGGOS), we present an updated strong lensing model of the galaxy cluster SDSSJ1110+6459 (z=0.659), which lenses the clumpy star-forming galaxy SGASJ111020.0+645950.8 at z=2.481 into a highly magnified giant arc. Using JWST NIRCam imaging, NIRSpec spectroscopy, and archival HST data, we confirm and refine the identification of four multiply imaged background sources, including one newly identified system, and map over 20 luminous regions between each image of the primary arc. Spectroscopy confirms that several previously ambiguous edge "clumps" belong to the main arc at z=2.481. Despite the limited number of strongly lensed sources in the field, the resulting lens model has high precision, owing to the high density of JWST-resolved clump constraints that tightly probe the lensing potential near the giant arc. The model yields a projected lens mass of $M(<250~\mathrm{kpc}) = 1.21^{+0.09}_{-0.04} \times 10^{14}~M_\odot$, an Einstein radius of $θ_\mathrm{E} = 10.8^{+0.3}_{-0.4}~\mathrm{arcsec}$, and a total effective magnification of $μ_\mathrm{tot}=24.2^{+3.4}_{-1.2}$ for the giant arc. Across the arc, individual clump magnifications span $μ_\mathrm{clump}\sim4-19$, with fractional magnification uncertainties of $σ_μ/|μ_{\rm best}|\sim0.03-0.09$. We report a $\sim2-8\times$ improvement in magnification precision over previous models. Ongoing and future analyses of this arc will enable robust measurements of star-forming structure, building on the lensing foundation established here for LEGGOS studies of galaxy growth and feedback during cosmic noon.

Strong gravitational lensing combined with JWST's spatio-spectral resolution enables resolved studies of star-forming regions in $z\sim$ 2-4 galaxies, but identifying and characterizing such regions in lensed integral-field and multi-band data remains a manual, observer-dependent process. We present $\texttt{SUMAC}$ (Software for the Uniform Manifold Approximation of Clumps), an unsupervised learning pipeline that segments JWST imaging and spectroscopy at the "spaxel" level by combining $\texttt{UMAP}$-based manifold embedding with $\texttt{HDBSCAN}$ density clustering applied to spectral energy distributions/spectra. We demonstrate the pipeline on JWST/NIRSpec PRISM IFS observations of the lensed galaxy SGAS111020.0+645950.8 at $z = 2.481$, recovering six physically distinct stellar/nebular populations. The cluster median SEDs separate cleanly on the presence and strength of H$β$+[OIII], H$α$+[NII], $β_{NUV}$ slope, Balmer break strength, and the Balmer decrement, with bluer clusters tracing unobscured star-forming regions and progressively redder clusters tracing dusty star-forming regions.

We investigate the role of obscured active galactic nuclei (AGN) in recently quenched post-starburst galaxies (PSBs), using a sample of 65 photometrically selected PSBs in the PRIMER-UDS field at $1 < z < 2$. Combining JWST/MIRI 7.7 $μ$m and 18 $μ$m (F770W and F1800W) imaging with eight NIRCam and three HST/ACS bands, we probe hot dust emission to test for hidden AGN or dust-enshrouded star formation. We find strong differences between the low- and high-mass PSBs. Most high-mass PSBs ($ > 10^{10}\textrm{M}_\odot$) show no excess infrared emission (consistent with the quiescent population), indicating little or no dust-obscured activity, while low-mass PSBs display enhanced emission at 18 $μ$m, which we attribute to residual star formation. AGN template modelling indicates that the absence of mid-IR excess in massive PSBs limits any dust-enshrouded AGN to Eddington ratios of $ < 1\%$. In addition, we show that the F770W--F1800W colour alone is a highly effective diagnostic for separating passive and star-forming galaxies, particularly at high stellar masses. Overall, our results provide further evidence for distinct quenching pathways within the PSB population, and confirm that massive PSBs show no evidence for excess AGN activity relative to older passive galaxies.

Within the highly successful $Λ$CDM paradigm established with cosmic microwave background (CMB) anisotropy measurements, the optical depth through reionization $τ$ is the most uncertain due both to the difficulty in measuring large-angle polarization and the assumptions made in their interpretation. Currently, for the Planck primary data in the flat $Λ$CDM cosmology with slow-roll inflation and standard reionization, the one-sided 95% upper limit for $τ$ is $τ_{\rm max}=0.0696$. Yet when all current CMB measurements excluding large-angle polarization are combined with baryon acoustic oscillation (BAO) measurements, the one-sided 95% lower limit is an incompatible $τ_{\rm min}=0.074$. If the long-standing low-power feature of the temperature measurements is interpreted as physically originating from inflation then $τ$ inferred from large-angle polarization becomes larger. Marginalizing over templates of the low-power feature based on the generalized slow-roll formalism of inflation raises the Planck maximum to a more compatible $τ_{\rm max}=0.075$ which further increases to $τ_{\rm max} = 0.082$ with the inclusion of all CMB+BAO data. This marginalization does not assess the statistical significance of the low-power feature itself; rather, it shows that allowing a higher $τ$ is a consequence of interpreting the anomaly as a physical feature instead of a statistical fluctuation.

Gravity models in which the dark energy equation of state crosses $w=-1$, also known as the phantom divide, have received extensive interest due to recent analyses favouring this behaviour. We introduce a new subclass of Horndeski scalar-tensor models capable of generating phantom crossing, whilst remaining minimally coupled to matter: the Asymptotic Cubic Galileon (ACG) models. We show that ACG models can jointly fit the expansion history inferred from observations of the Planck cosmic microwave background, baryon acoustic oscillation measurements from the Dark Energy Spectroscopic Instrument, and distance-ladder supernovae measurements from the Dark Energy Survey. We then demonstrate that perturbative observables, including the galaxy-ISW cross-correlation and void force profile, provide powerful constraints that confine viable and testable ACG models to a well-defined region of the broader Horndeski landscape. Model comparison metrics, including $χ^{2}$ and Bayesian evidence, favour both ACG and $w_{0}w_{a}$CDM models over $Λ$CDM, with ACG providing a fit of comparable quality to $w_{0}w_{a}$CDM. Crucially, ACG models ground the observationally preferred $w_{0}w_{a}$CDM behaviour in a robust Lagrangian formulation. This enables interpretation beyond mere phenomenological fits, and motivates further tests of these models on nonlinear scales.

Galaxy formation models predict that galaxies grow inside-out, becoming larger over time. While observations broadly support this paradigm, the nature and timescales of this growth remain poorly constrained. We report the discovery of an extremely faint and young (~600 Myr) stellar component in the outer regions of the nearby galaxy Messier 74 (M74). Using deep optical imaging from the TST telescope at the Teide Observatory, reaching surface brightness limits of ~30-31.5 mag arcsec^-2 in the g, r and i bands, we detect stellar emission extending well beyond the previously known disc radius of ~14 kpc. This newly identified component reaches galactocentric distances of ~30 kpc, effectively doubling the known size of the stellar disc and matching the extent of the HI disc. The revised size of M74 places it in the upper envelope of the mass-size relations. The young age of the outer stellar population suggests a recent episode of disc growth, potentially occurring on timescales shorter than ~1 Gyr. We discuss a possible scenario in which a past flyby interaction with UGC 1176 may have triggered this extended star formation. Further studies of galaxies with similar deep imaging will be key to determining whether such rapid outer disc growth is common or exceptional.