arXiv Daily Digest - 2026-07-07

We propose NetinfoGC, a framework for graph classification that extends the Network Usable Information (NUI) paradigm to graph-level learning. Unlike conventional graph neural network approaches that rely on end-to-end training of black-box embeddings, NetinfoGC constructs a family of permutation-invariant graph representations derived from propagation-based mechanisms and classical structural descriptors, including graph centrality measures. To evaluate representation quality, we introduce a training-free NUI estimation procedure based on clustering consistency with ground-truth labels, providing a proxy for task-relevant information without supervised learning. We further exploit the same representations using sparse-group LASSO regularization, enabling automatic selection of informative structural descriptors while suppressing redundant ones. Experiments on benchmark datasets show that classical centrality measures are highly competitive with learned propagation-based representations, and in several cases yield superior performance. Moreover, we observe a strong correlation between estimated NUI and downstream classification accuracy, validating NUI as an effective measure of representation utility. Overall, NetinfoGC provides a unified and interpretable framework for evaluating and exploiting graph representations without requiring end-to-end neural training.

Engineering Digital Twins and Prognostics and Health Management (PHM) systems rely on robust perception modules to extract actionable information from heterogeneous and non-stationary time-series data. However, most existing approaches remain task-specific, data-hungry, and difficult to integrate into scalable monitoring and decision-making pipelines. Moreover, purely data-driven models often lack robustness and transferability across varying operating conditions. To address these challenges, this paper proposes a modular foundation model for time-series perception based on a collection of pretrained representation encoders. The framework leverages self-supervised learning on heterogeneous datasets to learn transferable and task-agnostic representations, which can be reused across multiple PHM tasks. A gating mechanism is introduced to dynamically select relevant encoders for a given target dataset, enabling conditional computation and adaptive model composition. The selected representations are projected into a shared latent space and aggregated using a Transformer-based self-attention module that explicitly models cross-encoder interactions. The resulting architecture supports multiple downstream tasks, including imputation, long-term forecasting, and few-shot learning, through lightweight task-specific heads, while keeping pretrained encoders frozen during adaptation. Extensive ablation studies demonstrate the complementary roles of self-supervised pretraining, encoder selection, representation alignment, and adaptive aggregation. Experimental results on the ETT benchmark show competitive performance across tasks, while a real-world industrial case study on virtual sensing for hydro-generator rotor temperature highlights the practical relevance of the approach.

The widespread adoption of facial masks, accelerated by COVID-19 and mandated in security-sensitive settings, has exposed limitations of conventional face recognition systems. Existing approaches relying on fixed cosine thresholds, non-adaptive CNNs, and purely data-driven features fail to generalize when facial regions are occluded, creating a gap between lab performance and real-world deployability. This paper proposes PLGSA-Transformer, a cross-modal face matching framework with three contributions. First, Periocular Landmark-Guided Spatial Attention (PLGSA) uses MediaPipe landmarks to compute Gaussian heatmaps over the eye, brow, and forehead regions, fusing them with EfficientNetB3 features via a learnable residual gate to direct attention toward discriminative visible regions. Second, a Hybrid CNN-Transformer Branch reshapes feature maps into tokens processed by a two-layer Multi-Head Self-Attention encoder, enabling cross-regional dependency modelling. Third, the Occlusion-Adaptive Cosine Threshold (OACT) is a jointly trained head that raises the matching threshold in proportion to predicted occlusion severity. The model is evaluated on 858 images from Zenodo MDMFR (60%), Kaggle CelebA-HQ masked collection (25%), and author-collected images (15%), spanning both genders, ages 21-75, with varied mask types, trained via a unified loss combining contrastive verification, identity classification, and occlusion cross-entropy. PLGSA-Transformer achieves 97.22% pair verification accuracy with ROC AUC 1.0000, surpassing VGG-16-based MUFM (Abdullah et al., 2025; 95.0%), HOG classifiers (Adnan et al., 2020; 85.0%), and Feature-based Structural Measure (Shnain et al., 2017; 86.61%). These results confirm that encoding periocular geometry into attention, with Transformer modelling and occlusion-adaptive thresholds, yields a robust, scalable solution for cross-modal masked face recognition.

Diffusion architectures now encompass convolutional UNets as well as transformer-based designs such as Diffusion Transformers (DiTs), inspired by Vision Transformers (ViTs), yet the effects of structured geometric perturbations within these architectures remain poorly understood. We study this question through a unified framework that applies reflection-based elements of the dihedral group to intermediate hidden states as controlled internal interventions, contrasting geometrically consistent and inconsistent variants. Using activation-level diagnostics, including Self-Consistency Shift (SCS), Activation Mass Scatter (AMS), and Drift, we analyze feature stability and geometric drift. We find that consistent transformations improve stability, while inconsistent ones induce predictable, architecture-specific failures. In the main Stable Diffusion 2.1 U-Net study, we evaluate seven intervention modes over three seeds and complement the internal diagnostics with image-level FID, KID, CLIP score, and LPIPS diversity. Taken together with supporting ViT and controlled DiT analyses, these results establish geometric consistency as a key principle for stable hidden-state interventions in spatially structured vision and diffusion models.

AI systems increasingly propose executable scientific models whose value depends on both their symbolic structure and their fitted continuous parameters. This makes parameter calibration the bottleneck of program-and-parameter co-search: an outer loop can generate thousands of candidate programs, but each needs an inner gradient-based optimization before it can be assessed. Staging each candidate into its own differentiable graph makes individual models fast but sacrifices the program-as-data property that keeps search fluid; interpreter-based approaches preserve programs as runtime data but pay interpreter overhead that dominates the numerical work. We present the Native Differentiable Virtual Machine (NDVM), a runtime representation that differentiates executable programs without compiling each candidate into a separate graph. NDVM separates symbolic structure from differentiable numeric state: tags, symbols, environments, and control remain native runtime data, while numeric payloads live in dense batched buffers with exact reverse-mode gradients recorded along the realized execution trace, so one evaluator walk is amortized across large populations of parameter vectors. A locked cost model of a real differentiable self-hosted Scheme interpreter motivates the design. We realize NDVM as a native runtime with forward and gradient equivalence to the reference backend, about 60x per-lane batch amortization, near-linear multicore scaling, and two independent front ends. In fixed-budget co-search over LLM-proposed programs, NDVM reaches high-quality solutions about 24x sooner in wall-clock time, suggesting runtime differentiation as a practical systems foundation for scientific discovery workflows.

Knowledge distillation is usually framed as a choice of what to match in the teacher - its logits, hidden features, or sample relations - which presupposes that the teacher's representation has absolute coordinates to match. It does not: a pretrained representation is identifiable only up to an orthogonal-and-isotropic-scaling equivalence class, so a student should learn the teacher's equivalence class, not its features. The organizing fact is that capability is the teacher's output function, a class invariant that factors through the quotient by the class action, so an objective recovers capability exactly when it is defined there. This makes absolute feature matching ill-posed, and admissible supervision a matter of targeting class invariants (Gram structure, CKA, principal subspaces) or aligning coordinates first, unifying feature matching, relational distillation, alignment, and grafting in one geometric account. We validate our framework on Qwen2.5 and Llama-3.1. A restoration study recovers a corrupted model's representation (CKA ~ 0.99) but not its capability, and an ablation isolates the cause: output-function (logit) matching drives capability, while matching hidden representations aligns geometry without restoring function. Recovery is confined to the corpus-covered region, and a graft study confirms that boundary overlap predicts transplant success but is necessary, not sufficient.

Objective: Accurate identification of acute ischemic infarcts on diffusion-weighted magnetic resonance imaging (DWI) is a critical prerequisite for reliable lesion quantification and effective clinical decision support in the management of cerebrovascular events. Methods: This study presents EPRA U-Net (Efficient Pyramid Residual Attention U-Net), a task-specific integrated architecture for efficient and accurate infarct segmentation of DWI images. In the proposed architecture, an EfficientNet-based encoder was used as a hierarchical feature extractor with a minimized parameterization. In addition, a Residual-Recurrent (R2) block (recurrent unrolling step t = 2, following the original formulation) and Atrous Spatial Pyramid Pooling (ASPP) were integrated to enhance the performance of spatial dependency modeling. Additionally, a dual attention mechanism was incorporated to highlight lesion-related activations while concurrently enabling the suppression of extraneous background responses. To prioritize lesion detection consistent with clinical imperative, a Tversky loss function was adopted, emphasizing the sensitivity of detection over its specificity during the optimization process. Results: Experimental evaluations were conducted utilizing an in-house dataset comprising 167 patients with 4,895 DWI slices; subsequently, the performance of the proposed EPRA U-Net was assessed in comparison with state-of-the-art models, specifically UNet++, DeepLabV3+, and TransUNet. The experimental results suggest that EPRA U-Net attained superior performance, evidenced by a pixel-aggregated Dice of 0.8984, a per-sample Dice of 0.9469, an IoU of 0.8155, a Recall of 0.8887, a Lesion F1 of 0.9378, and an HD95 of 11.62 px. Furthermore, a clear reduction in the rate of missed lesions, specifically by 16%, 25%, and 29%, was observed when compared with UNet++, DeepLabV3+, and TransUNet, respectively.

As AI models continue to develop powerful capabilities, it becomes critical that we are able to verify that their output is aligned with our intentions. A recent line of work focuses on verification via debate, a model of interactive proofs where two competing powerful provers, or AI models, debate each other to convince a weak verifier, or a human, of the correctness of their claim. However, debate assumes that the two AI models possess equal abilities and that one of them is truthful, which may not be realistic. In this work, we show \emph{how to avoid debate}: we initiate the study of \emph{single-prover} interactive proofs for AI safety. Prior results in single-prover interactive proofs do not immediately carry over to the AI safety setting: for example, they do not work when the computation has access to an oracle, such as to human judgment or an external database such as the web. We present doubly-efficient single-prover interactive proofs and arguments for oracle-aided computations (also known as relativizing proofs), in the settings where (1) the computation is robust, in the sense that the output does not change if at most a small fraction of the answers to oracle queries are incorrect, or (2) the oracle is a low-degree polynomial. These results suggest that interactive verification is possible even without debate, under structured or noise-tolerant oracle access.

Continuous video anomaly detection is dominated by reactive Multiple Instance Learning (MIL) that collapses spatiotemporal features into scalar scores. We introduce PULS (Predictive Unified Latent Space), a continuous semantic world-model pipeline comprising two modules: a 490M-parameter KSD Bridge (Kinematic-to-Semantic Distillation) and a 16.8M-parameter Anticipatory State Predictor (ASP). The KSD Bridge maps V-JEPA 2 physical tensors into the 2048-d Qwen3-VL-Embedding-2B text-aligned hypersphere, trained on a subset of UCF-Crime. This translation alone yields a chunk-level AUROC of 0.8994 for UCF-Crime and 0.8162 for out-of-distribution XD-Violence without MIL or hierarchical fusion. We introduce and validate the Latent Clarity Hypothesis: because JEPA's temporal predictor discards aleatoric pixel noise while preserving kinematics, anticipated future representations are more semantically separable than observed presents. The ASP sharpens these anticipated future latents, achieving 44.5% mean 14-way zero-shot VQA accuracy (exceeding observation baseline by +9.6 pp). Applying the ASP to Observation Tensors collapses accuracy to 7.3% (random chance), proving Anticipation and Observation occupy distinct sub-manifolds. A Triple-Track Lead-Time protocol with an L1-surprise gate yields a peak +8.9 pp anticipatory advantage at T-0.5s (p < 0.001, N = 1,000 permutation), separating physical anticipation from static scene priors. Zero-shot transfer to XD-Violence confirms that Newtonian-invariant kinematic representations generalize out-of-distribution.

Weight space learning aims to learn representations of neural network (NN) weights, enabling different downstream tasks. Existing approaches show promising performance, but lacking a way to shape these weight-space representations using information about the datasets the models were trained on, thus limiting downstream applications. We propose WeightCLIP, a method for learning a dataset-aligned latent space for neural networks, where datasets information is induced during training. The NNs are encoded as latent representations using an autoencoder, while dataset samples are encoded using a dataset encoder. The two representations are aligned using a contrastive objective, effectively reshaping the weight-space representations according to the datasets. We demonstrate that such representations can be used for different downstream tasks, including mapping dataset information to a weight-space representation that decode to strong models. In addition, we introduce a latent refinement process for generating models that outperforms standard fine-tuning. Overall, our results demonstrate that explicitly incorporating dataset information improves what can be achieved with weight-space representations across retrieval, generation, and refinement. Code will be available at https://github.com/HSG-AIML/WeightCLIP.

Human reasoning often operates through qualitative concepts expressed by linguistic labels such as high, low, expensive, or cheap, whose interpretation depends on context and is usually vague, despite being rooted in numerical data. This paper explores a novel fuzzy-logic-based qualitative extension of Answer Set Programming (ASP) to bridge numerical information and qualitative reasoning. The underlying language, formally introduced in a separate work, provides a principled framework that avoids rigid thresholds and supports robust reasoning under vagueness. Focusing on a representative use case, we illustrate how the framework integrates numerically grounded inputs (such as outputs of machine learning models) with symbolic reasoning over qualitative labels. Key features, including learning-based membership functions and semantically enriched predicates, enable the combination of expert knowledge, contextual factors, and subjective interpretations within a unified declarative setting.

Detecting mental health disorders in a timely manner is an important societal challenge. NLP and machine learning (ML) methods used to assist with detection rely on data collected primarily from social media. However, such datasets often have sampling biases and inherent ethical and privacy issues. One avenue to overcome these limitations is non-social media data. We present the first comprehensive review of non-social media, free-text datasets for mental health research. We use the PRISMA methodology to conduct our survey and we review datasets available in multiple languages. We find that non-social media free-text based datasets are predominantly focused on English and on detecting depression. These datasets also vary in demographics, platforms, data types, annotation techniques, and methodologies. This systematic review also reveals key gaps and highlights opportunities to develop more diverse, reliable and clinically-relevant resources.

We introduce MentalThink, a visual-symbolic reasoning paradigm that equips Multimodal LLMs (MLLMs) with an executable mechanism for "mental" visualization. The core of MentalThink is a think-with-SVG pipeline, where the model learns to generate, render, and interpret scalable vector graphics (SVG) code as an intermediate visual representation for multi-turn reasoning. By creating structured vector sketches, the model can externalize spatial hypotheses, inspect them through deterministic rendering, and reason within a constrained geometric space, effectively mimicking the human process of mental imagery. We instantiate this paradigm through a two-stage training framework, combining Supervised Fine-Tuning (SFT) for SVG syntactic alignment with multi-turn Reinforcement Learning (RL) to encourage iterative inspection, revision, and refinement of intermediate visual hypotheses. Extensive evaluations demonstrate that MentalThink achieves superior performance on spatial understanding and reasoning benchmarks (e.g., 55.1% on VSIBench, 76.0% on MindCube), showing that executable vector graphics provide a verifiable visual workspace for dynamic perspective taking, visual reflection, and compositional scene construction.

Large language models (LLMs) are increasingly deployed as critical decision-making components in high-stakes real-world AI systems, rendering LLM reliability a foremost practical concern. In this paper, we focus on enhancing LLM reliability through selective prediction (SP), a strategy that allows an LLM to only predict for inputs where it is likely to be correct (i.e., coverage) and hence reduce the error rate (i.e., risk) on that portion of inputs -- flagging the remaining inputs for future human discretion. In other words, SP improves LLM reliability by balancing the risk-coverage trade-off and enabling seamless human-AI collaboration. To integrate SP into LLMs, we focus on the LLM post-training alignment stage and propose to align LLMs with SP performance metrics, in contrast with existing LLM alignment methods that focus primarily on correctness or calibration metrics. Specifically, we propose a novel alignment framework, Reinforcement Learning for Selection Reward (RLSR), which targets the area under the risk-coverage curve (AURC) -- a popular SP performance metric -- as its alignment objective. RLSR achieves substantially better risk-coverage trade-off compared to multiple alignment baselines on both in-domain and out-of-domain tasks.

Game engines provide real-time simulation, rendering, physics, interaction, networking, and asset pipelines, making them valuable not only for games but also for 3D applications in healthcare, robotics, architecture, manufacturing, and related domains. Because game development is where these systems are most mature and publicly available, it offers a practical testbed for evaluating coding agents that must modify C++ code within stateful, interactive, real-time systems. We present GameEngineBench, a benchmark for evaluating coding agents on scoped C++ implementation tasks inside Unreal Engine 5 projects, built from nine real-world game repositories. The evaluation set consists of 110 tasks spanning gameplay mechanics, multiplayer behavior, AI and world orchestration, animation and movement, UI and session code, loading behavior, online-service integration, persistence, data serialization, XR behavior, and rendering-oriented plugins. These tasks require models to make native C++ changes that compile and satisfy behavioral tests within executable Unreal Engine projects. Across twelve evaluated configurations, the strongest model reaches 55.5\% pass@1, while 31 tasks remain unsolved by every configuration. Our results demonstrate that frontier coding agents continue to struggle with deeply integrated C++ development for real-time interactive software, highlighting game-engine benchmarks as a valuable complement to existing software engineering evaluations.

Code repair is an important capability for language models (LMs): given a buggy program and unit tests, an LM must produce a fixed program that passes the tests. Because code repair data is limited, we aim to scale supervision by using an LM to generate bug--fix tasks. We propose __generator--fixer self-play__, in which a single model is trained with reinforcement learning to generate bugs and fix them. As the fixer improves, the generator adapts to produce more difficult bugs, yielding an automatic curriculum. To test whether this curriculum generalizes, we introduce BugSourceBench, a repair benchmark spanning realistic bug sources: bugs in human-written code, LM-generated code, and human-edited LM-generated code. On BugSourceBench, we find that self-play drifts toward difficult but unrealistic bugs, improving on synthetic bugs but degrading on human-authored ones. We propose Anchored Self-Play (ASP), which anchors self-play with a small reference set by adding a code-embedding similarity reward for generation and mixing reference bugs into fixer training. Across bug sources, ASP achieves the best fix rates, improving average fix rate over standard self-play by $+24\%$ relative / $+7.0$ pp absolute, with gains on bugs from both LMs and humans.

Fine-tuning a single low-rank adapter on many domains at once is multi-task learning: the domains must be co-learned, and how they share the adapter decides whether they help or hurt one another. Most efficient fine-tuning pipelines ignore this and train on a fixed, uniform mixture, leaving two coupled questions unanswered: how much should each domain participate, and which domains should be co-trained given that some transfer positively and others interfere? We show that both answers can be read off cheaply and without labels. A forward pass of the current shared adapter over a small unlabeled probe yields, per domain, a competence signal whose level tracks remaining headroom and whose trajectory tracks learning speed; the drift of these probe representations yields a signed cross-domain affinity that predicts pairwise transfer. We fold both into CoDA, a co-adaptive controller that solves a small entropy-regularized quadratic program on the simplex to set each domain's participation -- jointly its loss weight and its share of the sampled data -- rewarding high-headroom, still-learning, mutually synergistic domains and damping interfering ones. The controller is forward-only, adds no trainable parameters, and wraps any multi-task LoRA pipeline. Across five heterogeneous domains and two backbones, CoDA improves the average over uniform mixing, learned mixtures, gradient-surgery multi-task optimizers, and online data selection while using half the data, and lowers cross-domain gradient conflict. We prove that the competence signal tracks domain risk, that the participation program has a unique fixed point reached by a contraction, and that its solution performs transfer-aware water-filling; analysis, ablations, and controls corroborate each claim.

We show that projected Adam for online optimization with arbitrary moment decay parameters $β_1,β_2\in[0,1)$ can have average regret bounded away from zero. A similar result of Reddi-Kale-Kumar from 2018 required $β_1<\sqrt{β_2}$. Similar to their result, we use a three-periodic sequence of linear functions on $[-1,1]$ with slopes $c,-1,-1$, though we use $c$ slightly larger than $2$. This nonzero average regret result extends to Adam variants such as AdamW, RMSProp, NAdam, Adan, AdaMax, Muon, and to an i.i.d. variant of the three-periodic sequence of slopes for Adam.

Brownian Bridge Diffusion Models (BBDM) offer an appealing framework for image restoration and inverse problems by constructing a stochastic bridge from the clean signal directly to the degraded observation, rather than to pure noise. Despite their promise, the choice of bridge schedule is typically inherited from heuristics, and a principled analytical framework for schedule design has been lacking. In this work, we develop such a framework by offering a novel analysis of BBDM reverse dynamics under a Mixture-of-Gaussians (MoG) prior. This setting yields a closed-form ideal posterior and a corresponding MMSE denoiser, while the BBDM-induced reconstruction law is captured analytically through a tractable surrogate. Building on these expressions, we formulate two complementary schedule-design objectives: a Wasserstein criterion targeting perceptual quality and an MSE criterion targeting reconstruction fidelity. Our work exposes an inherent tradeoff between the two and proves the existence of universal schedules for both that are independent of the degradation and prior. Extensive experiments on controlled MoG settings confirm full alignment between theory and practice, and experiments on the FFHQ dataset across inpainting, deblurring, and super-resolution tasks validate the practical value of our schedule-design criteria.

Enterprise artificial intelligence is moving from isolated experimentation toward operational dependency across copilots, retrieval-augmented generation systems, autonomous agents, and AI-enabled business workflows. As this transition accelerates, the primary enterprise challenge is no longer only model access or inference scale. It is governed intelligence operations: the ability to enforce authorization, preserve contextual lineage, control persistent memory, detect stale or conflicting knowledge, constrain agentic execution, and produce audit-ready evidence across distributed AI estates. This paper introduces AGL-1, the Enterprise AI Governance Layer, as a vendor-neutral reference model for the control plane that should operate across foundation models, retrieval systems, orchestration frameworks, enterprise memory, policy engines, observability systems, tools, APIs, and business applications. Building on governed knowledge-system principles introduced in GKS-5, AGL-1 generalizes the governance problem from retrieval-specific controls to full AI execution-path governance. It identifies recurring failure modes such as unauthorized retrieval, stale grounding, unmanaged memory, weak provenance, policy drift, fragmented observability, and uncontrolled autonomous execution. It then defines seven governance domains: identity-aware retrieval, policy enforcement, provenance management, memory governance, knowledge integrity monitoring, agentic execution control, and trust observability. The central claim is that durable enterprise value from AI will increasingly depend on the ability to govern intelligence at scale. In complex enterprises, trust is not a property of the model alone. It is a property of the system around the model: identity, knowledge, policy, memory, tools, human oversight, and evidence working together as a managed control plane.

In many machine learning applications, the most relevant items for a query should be efficiently retrieved. The relevance function is usually an expensive similarity model, making the exhaustive search infeasible. A typical solution is to train another model that separately embeds queries and items to a vector space, where similarity is defined via the dot product or cosine similarity. This allows one to search the relevant items through fast approximate nearest neighbor search at the cost of some reduction in quality. To compensate for this reduction, the found items (candidates) are re-ranked by the expensive ranking model. In this paper, we investigate an alternative approach to candidate selection that utilizes the scores of the expensive model to improve the representations of queries and items. The idea is to describe each query (item) by its relevance to a set of support items (queries) and use these new representations to obtain query (item) embeddings. We theoretically prove that such embeddings are powerful enough to approximate any complex similarity model (under mild conditions). We also investigate the choice of support items, which is a crucial ingredient of the proposed approach. The experiments on diverse academic and production datasets illustrate the power of our method.

We present AquaGen, the first all-atom, explicit solvent, periodic-boundary-condition-aware generative model that produces molecular configurations from the Boltzmann distribution at a fraction of the cost of molecular dynamics (MD). This is in contrast with existing generative models that remove degrees of freedom by operating on coarse-grained, vacuum, or implicit solvent systems. Operating at this resolution allows for post-processing through force field energy evaluations and MD simulations, and enables the prediction of relevant properties in a gray-box manner (as ensemble averages of potential energy evaluations over generated samples). We demonstrate the utility of this paradigm on absolute hydration free energy (AHFE), producing estimates 4-10x faster and with comparable accuracy to standard GPU-based MD. By generating uncorrelated samples from alchemical Boltzmann distributions, we create more accurate, interpretable, and refinable ensemble predictions with calibrated uncertainty estimates, unlike regression methods which are entirely black-box predictors. Our approach also yields predictable benefits from increasing train- and test-time compute, realized by scaling model size and generating more samples, respectively. We believe that this approach demonstrates the utility of high-resolution ensemble generation for free energy estimation, with future potential to replace MD in tasks such as the prediction of lipophilicity, membrane permeability, or absolute binding free energy (ABFE) -- whose grounding and interpretability may be critical for the development of new drugs and materials.

Enterprise artificial intelligence is moving from experimentation into operational workflows. Early programs focused on model access and retrieval-augmented generation, but enterprises are now beginning to deploy agents that plan, retrieve, remember, call tools, update systems, and coordinate work across applications. This changes the evaluation problem. Leaders are no longer asking only whether an answer is accurate or fluent. They need to know who authorized an action, which policy applied, whether evidence was current, whether memory was valid, whether a tool call was permitted, whether the decision can be replayed, and whether the agent can be stopped before it creates business impact. This paper introduces CAGE-1: Control, Assurance, and Governance Evaluation for Enterprise Agentic AI. CAGE-1 is an evaluation framework for deciding whether enterprise agents are ready for deployment. It evaluates authority, policy enforcement, retrieval quality, memory integrity, tool safety, auditability, human oversight, conflict handling, safe failure, Prebind Assurance, operational readiness, and business fitness. CAGE-1 introduces Prebind Assurance to describe the evaluated ability to prove that an agentic action is controlled before it becomes binding, effective, or operationally consequential. The framework tests whether a proposed action is admitted, held, narrowed, refused, escalated, quarantined, or made non-effective before protected consequence forms.

Graph-based semi-supervised learning (SSL) propagates a few labels over a similarity graph by minimizing a Dirichlet-type energy. The standard quadratic ($p=2$) energy reduces to a single graph-Laplacian solve, but it degenerates exactly where SSL is most useful when labels are scarce: gathering more unlabeled data drives the $p=2$ estimate to a near-constant function whenever $d\ge2$ (Nadler-Srebro-Zhou). Well-posedness requires the nonlinear $p$-Laplacian energy with $p>d$. Existing solvers reduce this to a sequence of weighted Laplacian solves, but their reference implementations use a direct sparse factorization or ichol-preconditioned CG instead. Plugging a near-linear Laplacian solver is not straightforward: at large $p$ the conductance weights degenerate near flat-gradient edges, making the system nearly singular and causing stagnation without a damped outer iteration. We close this gap. Recasting $p$-Laplacian SSL as a source-form nonlinear Laplacian flow $Bρ_p(B^\top x)=b$ and solving by damped chord-Newton continuation in $p$, every linearized system stays well-conditioned and can be delegated to a near-linear Laplacian engine. On size-scaled graph families the wall-clock is empirically $m^{0.96}$-$m^{1.02}$ per family (approximate Cholesky default), and a pooled fit across 228 SuiteSparse graphs gives $m^{1.19}$ vs.\ $m^{1.45}$ for direct factorization; the solver handles a $6.8\times10^7$-edge social network in minutes. Memory is the binding constraint: Cholesky fill reaches $10$-$280\times$ the graph nonzeros vs.\ our $O(m)$ hierarchy. Against the released FCL solver we are $1.5$-$14\times$ faster at matched accuracy. On MNIST $10$-NN, $p=3$ scores $64\%$ at one label per class vs.\ $36\%$ for $p=2$. Code: https://github.com/orenlivne/np.

Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT). This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning. But hidden from the output is not the same as hidden from us. On four task families (fact retrieval, parallel numeric composition, string manipulation, and in-context computation), two open-weights frontier models (DeepSeek V3, Kimi K2) compute over filler tokens in a structured, legible way: attention routes the question through the filler region to the answer, logit-lens readouts show retrieved facts emerging early and their composition crystallizing in late layers, and KV-cache transplants at filler positions causally swap outputs between examples. We introduce an unsupervised decoding pipeline that takes only hidden states as input and recovers intermediate values with 80-95% accuracy (best LLM judge) across both models and all four tasks, without ground-truth labels or training. Hidden computation that defeats behavioral CoT monitoring is, on these tasks, directly readable from the residual stream, suggesting monitorability is a property of the model's full computational trace, not just its surface tokens.

Copernicus, the European Union's Earth observation program, produces petabytes of Earth observation and climate data, offering immense potential for research, policy, and applications. However, access to these datasets requires advanced programming skills and familiarity with domain-specific formats such as NetCDF or GRIB. Moreover, general-purpose Text-to-SQL systems fail when applied naively to the meteorological domain due to a profound ``Symbolic-to-Numeric'' gap. To overcome these limitations, we present an end-to-end Text-to-SQL framework specifically engineered for real-world, scalable meteorological data exploration. Our system intercepts natural language to resolve spatial and semantic ambiguities \textit{before} SQL generation. We design STRATOS, a Spatio-Temporal Resolution Agent for Text-to-SQL to dynamically bridge the symbolic-to-numeric gap by mapping fuzzy concepts to a localized ontology and resolving spatial entities via external knowledge bases. Further, our complexity-aware query rewriter rewrites expensive spatial predicates, reducing execution times from hours to seconds. Last, we introduce the STRATOS Evaluation Workload, comprising 7,520 complex query pairs explicitly designed by domain experts to test scalability and symbolic-to-numeric translation across challenging spatio-temporal dimensions previously unexplored by Text-to-SQL systems.

We consider a swarm array of autonomous relays that seek to cooperatively forward a desired signal to a fusion center with the maximum possible fidelity while canceling out a number of interferers. We present a distributed algorithm for computing the optimal zero-forcing beamforming weights at the relays without requiring prior channel knowledge. Crucially, our algorithm is {\it scale-free} in the sense that the computational and bandwidth overheads are completely independent of the size of the array. We build on recent work that introduced the concept of a Collective Array that enables such {\it scale-free} computation by imposing a constraint that the array must always function as a {\it swarm} i.e. array elements can only ever communicate with external nodes collectively and never individually. While this is a very severe restriction, we show that it allows useful computations such as zero-forcing beamforming while being robust to noise and channel time-variations.

Mutual information (MI) estimation is a central problem in machine learning and statistics; however, existing benchmarks typically evaluate estimators on simplified, low-dimensional distributions, leaving their performance on complex, realistic data largely unexplored. We address this gap with a comprehensive benchmarking framework grounded in a unified copula-theoretic perspective that subsumes existing benchmarks as special cases. Within this framework, we propose two complementary families of tests: a copula-first family that systematically varies ground-truth MI, dimensionality, and marginal complexity using synthetic and flow-based transformations; and a marginals-first family that couples real-world image data with controlled dependency structures, extending the classic same-class-pairing paradigm. We use this suite to extensively evaluate three classes of estimators: non-parametric, discriminative, and generative. Contrary to prevailing assumptions, our results indicate that there is no universal winner: each category can systematically outperform all other estimators under specific setups. By analyzing these cases, we identify fundamental estimation barriers and propose new tests that more effectively stress these specific limitations. We share the open source code at https://github.com/VanessB/mutinfo.

In this paper, we present the Invariant-Variant Disentangled State-Space Model (IVD-SSM), our submission to SemEval-2026 Task 4 on Narrative Story Similarity and Narrative Representation Learning. Evaluating narrative similarity is a profound computational challenge that requires models to look past concrete, superficial elements such as specific names, actors, objects, or settings to isolate and compare abstract patterns of causality and plot progression. To model these extended causal chains without the quadratic bottlenecks of standard Transformers, we leverage a hybrid State-Space Model (Jamba-1.5-Mini). Building upon this backbone, we introduce the Structurally Gated Alignment (SGA) head, a novel, differentiable algorithmic architecture. The SGA head operates on two scales: a heavily strided Macro-path maps the coarse structural skeleton of a story, which then acts as a gating mechanism to filter a full-resolution Micro-path, actively suppressing semantic noise and superficial keyword overlaps. Evaluated on both pairwise comparative judgments (Track A) and dense representation learning (Track B), our approach demonstrates that explicitly disentangling structural invariants from lexical variants provides a robust, principled framework for deep narrative understanding.

Modernizing cancer registries with deep learning is opening new opportunities to automate labor-intensive tasks such as the coding of pathology reports. However, progress is constrained by the scarcity of report-level human-annotated training data. Cancer registries generate substantial volumes of expert-assigned labels as a routine product of their operations, but these exist at the patient level and are not linked to the individual pathology reports that informed them, limiting their direct use for training models. We develop an efficient framework for training deep learning classifiers by leveraging these operationally-generated labels without requiring per-report human annotation, demonstrated for tumor group classification at the BC Cancer Registry. We use Attention-Based Multiple Instance Learning (ABMIL) to recover the lost link between patient-level labels and the reports that informed them, leveraging the attention the model places on each report to distil a large, noisily-labeled corpus into a compact, high-quality per-report training dataset. A classifier fine-tuned on a distilled dataset achieved a macro F1 of 0.83, outperforming established baselines across most tumor groups. By turning routine operational labels into high-quality training data without additional annotation or large-scale computing infrastructure, ABMIL offers a practical and accessible route to automating cancer registry workflows.

Post-training of frontier language models is conducted on curated task suites, and inevitably leaves a distribution shift between training and deployment environments. This exposes developers to generalization failures, which are relatively poorly understood. To better understand such generalization failures, we believe the community should construct clean demonstrations under simplified conditions. To facilitate this, we propose a simple and flexible way to construct language models which fail to generalize in controllable ways when subsequently trained with Reinforcement Learning (RL) on a given distribution of training tasks. Our construction uses Supervised Fine-Tuning on a dataset of a mixture of transcripts corresponding to a collection of 'conditional policies', which can each independently be assigned certain behaviors on each different task distribution, to obtain a model that is then well approximated as a 'mixture of conditional policies.' We observe that RL training then selects for policies that obtain the highest reward on the training distribution. This can produce striking behaviors: in a controlled setting with two distributions containing identical questions prepended with two different 'trigger strings', RL training on either distribution actively degrades performance on the other to zero, even though the underlying task is identical. We also use our construction to illustrate two novel ways in which generalization may fail in future language models, corresponding to distribution shifts of task coverage and temporal context respectively. While our construction is deliberately simple and may not closely resemble 'natural' generalization failures, the resulting 'model organisms' are of interest for alignment stress-testing and generalization science, and can be used as existence proofs that training success and generalization can come apart in structured ways.

Inter-agent communication is critical for coordinating Multi-Agent Reinforcement Learning (MARL) agents under partial observability to perform effectively in cooperative games; however, real-world bandwidth constraints demand sparse interactions. Prior approaches primarily address this trade-off by optimizing information-theoretic surrogates. We argue that these statistical proxies are fundamentally misaligned with the true objective: a message can be highly informative yet irrelevant to the joint return of the task. In this work, we propose Message Unlearning for Targeted Efficiency (MUTE), a framework that views communication reduction as a value-guided machine unlearning problem. MUTE rigorously quantifies the Counterfactual Message Value using an attention-based estimator, and systematically unlearns the transmission of low-value messages from a policy trained without any communication constraints. This is achieved through a dual-objective mechanism that enforces communication sparsity while preserving the return of the original joint policy. We derive a theoretical upper bound on the performance gap induced by this sparsification, guaranteeing controlled return degradation. We also empirically evaluate MUTE on various complex multi-agent environments, achieving 80% to 90% bandwidth reduction while maintaining performance comparable to state-of-the-art baselines.

This study aims to predict Tumor, Node, and Metastasis (TNM) stage labels independently, with the Cancer Genome Atlas (TCGA) pathology report as the sixth shared task of SMM4H-HeaRD 2026. The problem is framed as three multi-label classification tasks. We explore both classical and deep learning approaches using Term Frequency-Inverse Document Frequency (TF-IDF) features and embeddings from ClinicalBERT, BioBERT, and PubMedBERT. These representations are used with Logistic Regression (LR), Light Gradient Boosting Machine (LightGBM), Feed-Forward Neural Networks (FFNN), and Wide Residual Networks (WRN). Our results show that individual embeddings perform similarly to the TNM label classification, while their combination improves its predictive ability. WRN achieves AUROC scores of 0.839 (T), 0.8502 (N), and 0.803 (M) with F1-scores of 0.622, 0.702, and 0.9337, respectively, for the training phase. LightGBM with TF-IDF performs best with AUROC scores of 0.9368 (T), 0.9524 (N), and 0.8311 (M) and F1-scores of 0.7559 (T), 0.7384 (N), and 0.7017 (M) during the training phase. Furthermore, the result of the Codabench for the test sets indicates a Macro-F1 score of 0.978, 0.957, and 0.879 for the T, N, and M categories respectively for test set 1; while test set 2 records a Macro-F1 score for T, N, and M is 0.807, 0.767, 1.0 respectively. However, performance declined during the evaluation phase of the test sets, a drop from 0.938 to 0.858 of test set 1 to 2, for the Macro-F1 score across all stages; suggesting limitations in model generalizability, sensitivity to class imbalance, and challenges in processing lengthy clinical documents. Although this study provides an efficient baseline model and a reproducible pipeline, further optimization and validation are required before it can be considered suitable for use in a real-world clinical setting.

World models aim to capture environment dynamics in ways that support perception, reasoning, and action, and have recently become a central direction in Vision-Language-Action-World (VLAW) modeling. Meanwhile, unified vision-language models have demonstrated strong multimodal generation capabilities, yet their potential as world models remains underexplored. In this work, we introduce \texttt{WorldBagel}, a unified VLAW framework built on BAGEL, a modern multimodal unified model, and use it to systematically investigate the role of unification in world modeling. Across multi-task robotic manipulation and cross-domain experiments, \texttt{WorldBagel} consistently outperforms task-specific alternatives and learns action representations that are more structured and semantically aligned with visual and linguistic context. Experiments on LIBERO, Language Table, and Franka show that unification is not only an architectural convenience, but also a key factor in learning effective VLAW models, leading to consistent empirical gains and deeper insights into multimodal world modeling. Code and model checkpoints will be released upon acceptance.

In this paper, we propose Adversarial Dynamics Priors (ADP) for perturbation-resilient humanoid locomotion control. Existing motion prior-based methods induce natural motion styles by imitating kinematic motion features, but they do not directly regularize dynamics features, such as CoM motion, centroidal momentum, contact forces, and contact states. To address this limitation, we replace kinematic motion-style feature with selected dynamics features extracted from locomotion trajectories as the target of adversarial regularization.To this end, we use trajectory optimization to construct a reference dataset and train a discriminator to evaluate whether policy-induced temporal windows are consistent with the resulting reference distribution.Without explicit motion tracking, ADP encourages policy rollouts to remain close to the reference support, even after perturbations. Experimental results show that, compared with AMP, the strongest baseline in our evaluation, ADP improves the $80\%$-success impulse threshold ($J_{80}$) by $16.7\%$, while reducing direction-averaged recovery time and velocity tracking error by $47.9\%$ and $35.4\%$, respectively.

Inference-time alignment methods, such as Best-of-$N$, offer a flexible alternative to training-based alignment by using reward models to select high-quality responses generated by a reference LLM. However, the efficacy of these methods is inherently limited by the response quality: if the reference LLM assigns negligible probability to high-reward responses, no selection strategy will succeed in finding aligned outputs. In this work, we propose Best-of-Better-$N$ (BoBN), an in context learning-based generation framework to address this challenge. Our method utilizes retrieval from high-reward examples relevant to the input query and task. Crucially, we introduce a restyling step where retrieved responses are rewritten by the reference LLM to align with the target task's format and style. These restyled examples are used in-context to shift the sampling distribution toward the high-reward region. We analytically characterize how in-context learning shifts the output distribution of pretrained transformers toward the high-reward region, resulting in provable benefits on the target task. We then evaluate BoBN on safety alignment and mathematical reasoning benchmarks across several reference LLMs. BoBN's higher-quality responses enable better performance to be achieved when the number of responses $N$ is fixed, and smaller $N$ required to achieve a target performance.

While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question unaddressed: What constitutes a minimal viable pipeline for skill optimization, where every component is justified by theory or empirical necessity? We formalize skill optimization via Zeroth-Order (ZO) optimization, mapping classical counterparts (central difference, trust regions) to recent literature. Noting that unlike blind numerical perturbations in classical ZO, skill trajectories serve as interpretable debugging feedback. Grounded in Claude Code philosophy and PAC learning, we establish three principles for convergence and generalization: file-system-based trajectory exploration, consensus attribute mining, and independent validation gating. Eliminating redundancies, we propose SkillOpt-Lite. It accelerates convergence and outperforms full SkillOpt: improving LiveMath by +8.8 points on GPT-5.5 and +25.4 points on GPT-5.4-nano, allowing the nano model to surpass standard GPT-5.4 optimized by SkillOpt. Finally, we integrate our framework into production coding agents like VSCode Copilot, enabling developers to evolve agent skills via one line of vibe. Because our framework treats all agent components simply as standard editable code, this minimal pipeline naturally generalizes to full harness optimization (HarnessOpt). On SpreadsheetBench, HarnessOpt enables GPT-5.4-nano to achieve 0.7758 accuracy, outperforming the larger GPT-5.5 running standard pipelines (0.7620). Code is available at https://github.com/EvolvingLMMs-Lab/SkillOpt-Lite.

Current Vision-Language-Action (VLA) models excel at robotic manipulation but often struggle with non-Markovian tasks requiring long-term memory and reasoning due to their reliance on immediate observations. Existing solutions face a ''frequency-competence paradox,'' where stronger reasoning models are too slow for real-time control, while faster models lack sufficient reasoning capabilities. To resolve this architectural misalignment, we propose HiMe, a Hierarchical Embodied Memory framework that decouples embodied intelligence into a high-frequency Executor for execution, a Sentry for working memory, and a Planner for long-term strategy. We also introduce a dynamic knowledge system based on cross-modal semantic schemas and active management mechanisms, allowing robots to maintain memory plasticity through ''Add, Update, and Delete'' operations. This hierarchical design effectively balances the conflict between real-time execution and slow thinking planning, significantly improving success rates in long-horizon tasks. Experiments demonstrate that this approach not only outperforms flat memory baselines but also exhibits the novel ability to self-correct its internal knowledge based on human preferences.

Knowledge graphs (KGs) that underpin Graph-based Retrieval-Augmented Generation (Graph-RAG) are increasingly built automatically by LLM-driven extraction rather than curated by experts. Proper evaluation would require instrumenting all pertinent stages: extraction, graph construction, and inference, coherently enough to localize failures, so that a failure at one stage is not discovered as a wrong answer at the end. We introduce TRIAGE, a stage-aware instrumentation framework for automated, document-grounded graph-RAG that asks not only whether the underlying graph can be trusted but at what cost it can be queried. TRIAGE attaches stage-specific, independently interpretable metrics to three stages: the KG Implementation (triple confidence, source coverage, and schema and canonicalization checks), the KG Validation by expert (graph-level structural quality, with correctness and completeness computed only as offline calibration when a reference is available), and the KG Usage (retrieval coverage, faithfulness, and retrieval cost); the deployed metrics need no gold annotations, the gold-requiring ones serving only as offline calibration. At usage time these metrics form a diagnostic chain of necessary conditions whose first broken link localizes the failure, and the diagnosis maps to the stage levers that can remedy it: extraction, graph and schema, or retrieval. TRIAGE is a theoretical framework with a proof of concept and a reproducible evaluation protocol.

LLM agents often degrade over long episodes: as trajectories grow, they revisit explored states, repeat failed actions, and lose strategies that previously worked. Test-time training (TTT) offers a way to adapt model weights to the evolving task state, but existing LLM TTT methods largely adapt once to a fixed input. We study continuous TTT in multi-turn agent episodes, where each update changes the policy that generates later training text. This creates a self-training loop that helps when new trajectory information appears, but can amplify drift when the agent gets stuck and repeatedly trains on similar text. We find that update-text repetition distinguishes these regimes and introduce Agentic Test-Time Training (aTTT), a token-level reweighting method that downweights the loss on tokens appearing in repeated $n$-grams from prior updates while leaving novel tokens fully weighted. To run such updates inside live episodes, we build a concurrent serving system using vLLM's runtime LoRA API, limiting overhead to 1.9$\times$ the no-TTT cost. aTTT improves success by up to 5.0 points on ALFWorld and 4.9 points on SWE-bench Lite. The gains concentrate where models already have task competence but drift over long trajectories, suggesting that aTTT mainly preserves existing competence rather than teaching new abilities.

Routing among large language models (LLMs) promises better quality at lower cost, motivated by the reported gap between learned routers and a per-instance oracle. But that oracle is computed from a single correctness label per (query, model), so under stochastic decoding it is one Bernoulli draw, not a reproducible property. We recast the question structurally: the expected per-instance oracle decomposes as $O^{\exp}=O^{\mathrm{repro}}+Δ$, into reproducible single-commit headroom $O^{\mathrm{repro}}$ and a non-negative single-commit selection floor $Δ$. Our main result is a recoverability asymmetry: this floor is closed by no single-commit router, yet is recovered by test-time sampling -- best-of-$K$ on the committed model, at the oracle's own budget, dominates the independent-pool single-draw oracle. The cap needs no cross-model independence; we prove it with the exact decomposition and noise-share bounds that shrink as the budget grows. The procedure adds no new router, only resampling. The floor's magnitude is a prospective, conservative localization, not an audit: our primary target LLMRouterBench (33 models, 391,645 instances) defines its oracle as a per-query union over single $T=0.2$ generations -- by construction a union of stochastic single draws. Since $O^{\mathrm{repro}}$ is non-identifiable from the released $k=1$ matrix, we estimate the noise share by fresh $k\ge20$ resampling under one-sided, dependence- and guessing-floor-corrected bounds, recasting 'model-recall failure' as thin-support union inflation. On a controlled open-model re-generation, single-draw noise is a substantial minority of the gap -- larger on an unsaturated benchmark, approaching half on the hardest queries where no model is reliable -- while the majority remains recoverable specialist advantage. We release a multi-sample oracle evaluation protocol for routing benchmarks.

Large language models (LLMs) exhibit strong reasoning and world-knowledge capabilities, yet often struggle to gather information effectively across the multi-turn interactions required in sequential decision-making settings. We introduce Amortised Sequential Information Gathering (ASIG), a fine-tuning approach that amortises Bayesian Experimental Design (BED) into LLM policies via a multi-turn extension of Group Relative Policy Optimisation with an Expected Information Gain reward. Evaluated on the 20 Questions task, ASIG more than doubles the success rate of the 7B base model and reduces inference cost by over $25\times$ relative to BED-LLM, a competitive inference-time baseline. Applied to MediQ, a medical diagnosis benchmark unseen during training, ASIG improves information-seeking performance at the 7B scale, suggesting that the learned strategies can transfer out of distribution. Our findings show that amortising BED into LLM policies provides an effective and computationally efficient approach to sequential information gathering.

Algorithmic recourse addresses the challenge of providing tailored recommendations to users affected by unfavorable machine learning decisions, in potentially high-stakes scenarios. Traditional approaches to recourse often rely on the closest counterfactual explanations or assume a priori knowledge of a user's causal structure, resulting in interventions that overlook individual contexts and specific feature interactions. To overcome these limitations, we study a human-in-the-loop framework that iteratively approximates the user's structural causal model through interactive queries via Bayesian inference before producing recourse recommendations. This framework exploits humans' feedback to improve the identification of causal effects, allowing personalized recourse that is plausible, cost-effective, and aligned with the actual causal dependencies of each user. As a proof of concept, we evaluate this framework through simulated human responses. Our simulations across linear and non-linear causal models show promising results, though challenges remain in capturing complex, non-linear structures, emphasizing the importance of accurate approximations and robust noise distribution modeling.

Modern AI agent implementations such as frontier coding agents chain multiple tools at runtime that create a security surface that per-tool guardrails are unable to address, as individually permitted tools can violate organizational policies when composed. We propose the Dynamic Security Control Compositor (DSCC), a two-phase approach to compositional security for multi-tool agent chains. In Phase 1, at session checkout, a Most Restrictive Set (MRS) algorithm composes per-tool security policies into a single effective policy with a formal monotonicity invariant that extending a chain can only tighten the result, blocking incompatible combinations before any tool executes. Outputs of any tool call propagate their classification constraints into a session-level taint state, so subsequent invocations must satisfy the most restrictive constraints seen so far. In Phase 2, at runtime, the system tracks the sensitivity of data the agent touches through a monotonic taint state and revokes the session if the accumulated exposure would make a subsequent tool call a policy violation. Together, these phases provide defense in depth, where static composition prevents unsafe chains from starting, and runtime taint tracking catches violations that emerge from the specific data used. We provide a reference implementation on 32 tools governed by 16 NIST SP 800-53 aligned policies and evaluate it under two composition modes. In the default clearance mode, permitted combinations are partitioned into classification-level clusters, blocking 79.2% of policy pairs and 95.5% of triples. The alternative taint mode admits mixed-classification chains within the exfiltration boundary, blocking 42.5% and 60.5% respectively. We discuss the governance implications for organizations deploying multi-tool agents, including the utility-security tradeoff and the changes needed to operationalize chain-aware policies.

This research paper examines how Knowledge Components (KCs) - fine-grained concepts or skills required to solve programming tasks - can be used as interpretable signals for understanding assignment difficulty and student struggle in introductory programming courses. While prior work has focused on predictive models based on programming behavior, such models are often difficult to interpret and therefore hard to use for instructional decisions. We analyze KC-based metrics, including the number of KCs per assignment and changes in KC coverage between consecutive assignments. We examine correlations between the number of KCs and student performance on the assignment, and analyze changes in KCs across assignments to identify cases where performance declines without new concepts being introduced. Selected assignments are then qualitatively inspected to understand potential design issues. Our results on data from three introductory programming course datasets show that assignments involving more KCs are generally associated with lower performance, and that sudden shifts in required KCs can coincide with disruptions in learning progression. We also identify assignments where performance declines even though no new KCs are introduced, suggesting potential issues in task design or instruction, which could be examined with qualitative analysis. We propose an interpretable framework for analyzing programming assignments using KC-based metrics, with practical implications for instructors and course designers who want to better understand where and why students struggle, and how course materials might be improved. Our method can use KCs defined by an expert or extracted by an LLM, offering instructors an additional way to assess assignment quality beyond average correctness. It can be applied to any course with ordered assignments and measurable performance.

A trustworthy and GDPR-compliant deepfake audio detector must base its decisions on acoustic artifacts, not on what is being said or who is speaking. We present a large-scale study of semantic independence for Resemble AI's detector, DETECT-3B-Omni. Using 10,240 audio samples from diverse US English speakers across 30 states, generated through 8 different AI voice-cloning systems, we test whether detection accuracy depends on spoken content (benign versus malicious), speaker gender, speaker age, or speaker region. Using equivalence testing, our results show that the accuracy difference between any two of these groups is at most 2 percentage points, at 99% confidence. The detector therefore identifies AI-generated audio with equivalent accuracy regardless of what the audio says or who the speaker is.

This paper presents an approach to the SemEval-2026 Task 3: Dimensional Aspect-Based Sentiment Analysis. We investigate methods for moving beyond traditional categorical sentiment (e.g., positive or negative) to predict fine-grained, real-valued scores for sentiment "valence" (positivity) and "arousal" (intensity). We participate in two subtasks: predicting these scores for given aspects (Subtask 1) and extracting full sets of sentiment details, including aspects, categories, and opinions alongside their scores (Subtask 3). Our approach for the regression task involves a weighted ensemble of transformer-based encoder models. For the Russian language, we further enhance the input by using a large language model (LLM) to generate synthetic sentiment descriptions. For the extraction task, we fine-tune a decoder LLM to perform structured prediction, allowing the system to identify sentiment elements and estimate their numerical scores simultaneously.

Jamming and spoofing threaten wireless and satellite navigation by disrupting or manipulating radio frequency (RF) signals, undermining availability, integrity, and trust. Robust interference monitoring (i.e., detection, classification, characterization, and direction finding) is therefore essential to identify and localize anomalous signals. While machine learning (ML) promises improved performance in complex environments, its development and validation depend on large-scale datasets that capture realistic signal and channel variability. Collecting such data in the real world is difficult because intentional jamming is illegal and ground-truth attribution is confounded by propagation, hardware, and environmental effects. To address this gap, we create and publish S-ICDF, a large-scale indoor interference dataset generated with Sionna, a GPU-accelerated simulation library for physical-layer wireless communications. S-ICDF covers 102 interference configurations, including diverse antenna array patterns, bandwidths, and simulation settings such as noise level and reflection depth. We further provide baseline results by benchmarking S-ICDF with classical estimation and direction finding (DF) methods (MUSIC, ESPRIT, and CAPON) and with modern ML approaches. The dataset is publicly available at: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss/sicdf_dataset

We describe the 2026 Methods for Open Agent Systems Evaluation Initiative (MOASEI) Competition, a benchmark event for evaluating multi-agent decision-making under open-system conditions. Building on the inaugural 2025 competition, the 2026 edition retained wildfire fighting, cybersecurity, and ride-sharing domains while adding a bonus wildfire track with frame openness, in which agent equipment states such as suppressant capacities and firefighting range vary over time. The competition also expanded its reporting metrics to emphasize total task completions, mean task-completion time, and mean value of completed tasks. Participation in 2026 was limited: eight teams registered, but only one team submitted a final entry, and that entry targeted the ride-sharing track. The submitted DLC approach used planning and replanning to solve routing problems across agents as passengers appeared. This report summarizes the 2026 competition design, highlights differences from the previous year, and reports ride-sharing evaluation results against baseline policies. DLC is recognized as the 2026 ride-sharing track winner among submitted teams.

Text-to-Image diffusion models often propagate harmful bias inherited from the training data. Existing bias mitigation techniques typically intervene only at the text encoder or provide inference-time guidance, often leading to generations that collapse into semantically incoherent outputs. To address these limitations, we introduce CO-ALIGN (Concept Ontology Alignment), a novel bias mitigation approach based on concept-graph alignment that operates on the model's internal concept ontology. By aligning concepts within the text encoder and denoiser, CO-ALIGN achieves substantial bias reduction while preserving generative integrity. We demonstrate the effectiveness of concept-graph alignment across three paradigms: text-encoders, denoisers and joint text-denoiser ontology alignment. CO-ALIGN outperforms the state of the art, improving fairness by $30\%$, $ΔFID=11.4$ in image quality, $2.8\%$ in image fidelity, all while reducing semantically incoherent outputs by $88\%$. Beyond bias mitigation, we show that CO-ALIGN benefits other downstream tasks as well. In particular, our experiments demonstrate that better-aligned internal ontologies enhance concept unlearning robustness across multiple unlearning techniques.

Fuzz testing is a popular approach to the security testing of proprietary software. Efficient testing strategies rely on execution feedback to guide the input generation process, particularly when the basic blocks in the binary can be directly observed and instrumented. Unfortunately, collecting such feedback is impossible in scenarios such as in-situ fuzzing of black-box devices and the fuzzing of obfuscated compiled binaries. In this work, we discuss approaches to guide the fuzzer using feedback derived from a control-flow-graph-like (CFG-like) structure constructed from runtime execution. We start by outlining a simple divergence-detection approach that identifies unique execution traces, and then present an improved approach based on an Execution Divergence Graph (EDG). We implement both approaches and demonstrate that they outperform a baseline blind fuzzer. In addition, we discuss particular challenges, such as repeated code execution in loops, and show that the EDG-based approach handles them effectively. We then demonstrate that our approach enables effective fuzzing of a number of obfuscated targets, and compare its performance in scenarios where static instrumentation is impossible. While we focus on a scenario in which full instruction traces are directly observable by the attacker, our scheme can also be applied in scenarios with other feedback channels, such as power consumption.

Real time location data derived from mobile applications is a powerful tool for addressing various urban challenges, including tourism planning, parking management, bus route optimization, and resource allocation. Besides, it offers invaluable insights for shaping strategic decisions in commercial domains such as location based services, market share analysis, and behavioral profiling. In this expansive study, we aim to address all of the aforementioned challenges by investigating the behaviors and patterns of smartphone users within urban environments, particularly in the domains of tourism, transportation, and retail. Our approach encompasses the development of a sophisticated data platform from inception to implementation, which includes the formulation of use cases, architectural design, and implementation of modules. We employ state of the art techniques and technologies, including data anonymization, ETL pipelines, and utilizing Google BigQuery and Vertex AI for data processing and machine learning model development. A modular architecture based on reusable analytical building blocks was developed to generate data products that support multiple stakeholder driven use cases. Additionally, we apply interactive data visualization techniques via Power BI to facilitate the effective interpretation of analytical findings by stakeholders. The developed models address a wide range of mobility analytics tasks, including mobility profiling, frequent trajectory mining, area of influence analysis, traffic anomaly detection, and origin destination pattern analysis. The results demonstrate the framework's ability to capture user mobility dynamics at fine spatial and temporal resolutions, providing actionable insights for urban planning and strategic business decision making.

The ever-increasing collection of personal data has created mounting pressure to develop technologies that protect sensitive aspects of individual identity. Differential privacy (DP) provides a principled framework with strong formal guarantees and has already achieved practical success. However, releasing high-dimensional data, such as images, has remained elusive: releasing uncompressed privatized data requires significant storage. At the same time, no effective data compression scheme exists that can compress high-resolution data with privacy guarantees. We address this challenge with DP-DiPP, a compression pipeline that combines stochastic codes with diffusion models. DP-DiPP is highly flexible: the practitioner has direct control over the compression rate-privacy-utility tradeoff. As the theoretical backbone, we extend the Poisson private representation (PPR) to encode the outputs of privacy mechanisms. We then combine it with DiffC, a diffusion-based lossy data compression method, to obtain a differentially private image compressor. Our experiments on privatized image classification on CIFAR-10 demonstrate that DP-DiPP significantly outperforms the baseline, achieving a 10-30 times better compression while retaining comparable privacy guarantees and utility.

Agentic AI systems are increasingly used to edit, refine, and repair decision policies, but evaluating these edits is difficult when per-state expert action labels are unavailable. We study this problem in a hotel-pricing simulator where an agentic policy editor receives only region-level diagnostic feedback: summaries of how its price distribution differs from a benchmark policy across time, inventory, and market regions. The editor cannot observe benchmark actions, benchmark source code, reward numbers, or held-out outcomes, and may only propose constrained edits to a target-action table. On 5,000 held-out episodes, a multi-restart LLM editor reaches RevPAR 108.47 (95% CI 107.61 - 109.34), close to the benchmark policy's 108.75 (107.81 - 109.68), with paired gap (LLM minus benchmark) -0.276 and 95% CI [-0.692, 0.146]. A cheap diagnostic projection already recovers much of the revenue (107.90), so the LLM editor's distinctive gain is not raw revenue lift alone: it also reduces episode composition distance from 1.153 to 0.609. This is the strongest non-benchmark repair result. This profile is not explained by restart search alone: non-semantic proposers with up to 2,500 evaluations fall 8.77 - 14.57 RevPAR points short. Nor is it explained by plausible prompt format: a shuffled-diagnostic control breaks region-error correspondence and falls to RevPAR 94.30. The match is genuine but partial. A tree editor achieves stronger pooled alignment, 0.214 versus 0.266, and stronger reference-state D1, 0.328 versus 1.197, yet revenue falls to 98.91. These results show that agentic policy repair should be evaluated by whether diagnostic feedback becomes reliable closed-loop outcome, not by a single behavioral distance.

Policy learning has received substantial attention with the goal of learning policies from observational data for decision-making. A majority of work in this space has focused on developing algorithms for computing policies that minimize regret compared to the optimal policy. However, in many practical settings, there is insufficient data to obtain low regret. As a result, recent work has shifted attention to alternative objectives, most notably, studying whether it is possible to learn an improving policy that statistically significantly outperforms baseline policies. We argue that there is substantial merit in studying a broader range of policy learning problems. When there is insufficient data to learn an improving policy, there may still be useful questions that can be answered. To this end, we provide a mathematical framework for studying the relationships between policy learning problems. We formalize three problems within our framework: beyond the optimal policy problem and the improving policy problem, we also propose the policy existence problem, which aims to determine if an improving policy exists. Within our framework, we show that the policy existence problem reduces to the improving policy problem, which in turn reduces to the optimal policy problem; these reductions prove that each problem is at least as easy as the next one (in sample complexity). A key question remains: is this hardness strict? We provide partial answers. First, the gap between the optimal policy and improving policy problems is strict. For the improving policy and policy existence problems, we prove that a sublinear polynomial gap exists under natural conditions on improving policy learning algorithms. Thus, we may be able to answer questions about the existence of an improving policy even when we cannot find one. These results highlight the value in studying a broader range of policy learning problems.

Scientific papers follow rhetorical structures that organize content into sections such as Introduction, Methods, Results, and Discussion. Automatically identifying these sections at scale enables granular analysis of scientific writing patterns. We present a dataset of section-level annotations for millions of scientific papers from the Semantic Scholar Open Research Corpus (S2ORC). Using a rule-based classification algorithm, we identified and labeled major sections across 15.6 million papers after quality filtering. The dataset covers primarily STEM disciplines, with strong representation in medicine and biology. We provide comprehensive human and LLM-based validation showing that classifier agreement with human annotators is on par with human inter-annotator agreement. This dataset enables large-scale computational studies of scientific discourse and writing patterns.

The rapidly growing repository of publicly available large language models (LLMs) presents significant challenges for systematic management and quantification at scale, such as model lineage tracing, licensing, and evaluation. However, task-specific benchmarks are insufficient for this setting, as LLMs differ widely in architectures, scales, and training procedures. To address this challenge, we adopt spectral shape-based metrics for managing and quantifying LLMs based on Heavy-Tailed Self-Regularization theory. Our approach uses the shape information of the weight empirical spectral density as a compact spectral signature of each model. This signature captures intrinsic properties of pretrained models and remains robust during post-training, making it suitable for model-level analysis. In addition, this metric is data-free, computationally-efficient, and scale-invariant, enabling large-scale analysis in practice. Moreover, we curate a large and diverse model corpus consisting of major open-source LLM families, and use it to systematically benchmark spectral and non-spectral metrics across models and downstream tasks. We show that our spectral signature supports the tracking of the model lineage, the unsupervised clustering of similar models, and the quantification of the model performance. Overall, the proposed spectral signature provides a meaningful proxy for broad performance trends across LLMs, enabling efficient organization, comparison, and analysis of large model collections.

Vision-language models (VLMs) increasingly read news and web content as images, where the publisher's identity is visually present. We show that VLMs carry a strong source-credibility prior keyed on outlet identity, and study it along three axes. (i) Cross-model benchmark. We introduce CueTrust, a cross-model diagnostic that measures which surface source cue overrides an article's content evidence via a Source-Override Index (SOI). Across seven VLMs and five cues, the vulnerability profile is model- and scale-dependent, and the override is outlet-identity-specific and encoding-invariant, firing from the masthead name, the logo image, or the bare domain, but not from a named author, in-text authority, or page layout (clean negative controls). (ii) Mechanistic account. For the brand cue, we give a full mechanistic account: swapping only the masthead moves credibility across an approximately 11 log-odds range that tracks professional ratings (rho = 0.88 with Media Bias/Fact Check). The prior is dual-coded (name and logo), strengthens with scale, is causally formed at layers 19-21, carried by interpretable seed-stable sparse-autoencoder features, and recurs at the same relative locus in a second model family. It overrides content (about 1.8x) as a signal-magnitude effect within a shared pathway, not a privileged route. Steering the localized direction selectively reduces the override (41% reduction) and generalizes to held-out outlets, confirming the prior is causally used, not merely decodable. Deployed VLMs may thus defer to source identity over the evidence in front of them, a reliability failure we can measure across models, localize, and causally probe. We release the stimulus suite and CueTrust.

We propose \textbf{CaSPECT}, a causal spectral clustering framework for discovering causally homogeneous subgroups from observational data. Rather than clustering in covariate space, CaSPECT defines similarity through the topology of a learned directed acyclic graph (DAG); a bootstrap-stabilised PC algorithm recovers the causal skeleton; a novel \emph{Orientation Validation Score} (OVS) combines PC bootstrap evidence with DirectLiNGAM to orient edges robustly; directed edges are weighted by backdoor-identified average treatment effects estimated via OLS or double machine learning. Chung's directed Laplacian provides a spectral embedding in which individuals close together share the same causal propagation pathways. We establish almost-sure consistency of the full pipeline and validate the method through a controlled simulation study and on LaLonde CPS1, IHDP, and 401(k) datasets, where CaSPECT recovers a positive and statistically significant treatment effect within the causally comparable subpopulation and corrects for severe confounding without requiring a pre-specified propensity score model.

We study how visual information is routed in vision-language models (VLMs). Using causal patching on controlled synthetic and natural datasets, we find that models rely on two distinct pathways to solve visual tasks: A direct pathway, where visual information is retained in image token representations and read out by the final token at later layers, and a text-mediated pathway, where visual information is first transferred to the query tokens and then read out by the final token. Across three visual tasks, we show that pathway selection is task-dependent, and that data distribution and prompt design can also modulate which pathway is used to solve the image-based query. Moreover, using attention knockouts and corrupted-input patching, we find that these pathways are flexible, under certain interventions, models can rely on the text-mediated pathway as a fallback when the usual pathway is ablated. This behavior unifies findings in prior work and shows that ablation-based interventions can reveal what models could do rather than what they normally do. Together, our results provide a mechanistic characterization of visual information flow in VLMs and highlight the flexibility of their internal mechanisms under intervention.

Malicious Python packages have become a major threat to software supply chain ecosystems due to the widespread adoption of open-source repositories such as PyPI. Existing learning-based detection methods struggle to capture the hierarchical organization and heterogeneous interactions among different program entities. Although Large Language Models (LLMs) have demonstrated strong capabilities in code understanding and semantic reasoning, they are rarely integrated with structural program representations for fine-grained malicious behavior analysis. In this paper, we propose an LLM-enhanced hierarchical heterogeneous graph representation learning framework for malicious Python package detection. The framework constructs a hierarchical heterogeneous code graph that explicitly models heterogeneous code entities and different types of structural dependencies. LLMs are further leveraged to infer function-level semantic roles, introducing an additional layer of semantic heterogeneity. Based on this graph, we develop a hierarchical heterogeneous graph neural network that performs type-aware message passing over different node and edge categories, effectively modeling malicious behavior propagation for accurate package-level classification. The framework also incorporates a function-level attribution mechanism which, combined with LLM reasoning, automatically identifies suspicious functions and localizes fine-grained malicious behaviors without human expert intervention. Extensive experiments on real-world datasets show that our framework consistently outperforms traditional machine learning methods, graph-based detectors, and state-of-the-art LLMs across packages with varying sizes and dependency complexities, while providing accurate, robust, and interpretable malicious behavior localization.

The accuracy of consumer-grade inertial navigation is bottlenecked by the stochastic noise of Micro-Electro-Mechanical Systems (MEMS). Traditional deterministic neural architectures often succumb to ``estimation jittering,'' sacrificing high-frequency kinematic fidelity for numerical stability. We propose PedestrianDiffusion, a multimodal spectral-domain generative framework reformulating dense 6D state estimation as a continuous conditional denoising process. By operating in the frequency domain, our formulation bounds the spectral covariance, acting as a mathematical preconditioner to stabilize the reverse diffusion trajectory. Furthermore, we introduce a zero-shot semantic conditioning mechanism leveraging vision-language embeddings as categorical priors to generalize across heterogeneous sensor noise profiles. To address the computational intractability of generative tracking, we deploy a single-step deterministic probability flow ODE solver ($T=1$). This yields high-capacity asynchronous batch trajectory refinement, establishing the viability of generative architectures for asynchronous batch trajectory refinement on edge hardware. Extensive evaluations on the OxIOD, RIDI, RoNIN, and TLIO benchmarks demonstrate that PedestrianDiffusion achieves state-of-the-art performance, exhibiting unprecedented robustness to impulse perturbations and coupled 6D kinematic drift. This work provides a rigorous algorithmic blueprint for next-generation Neural Inertial Measurement Units (N-IMUs).

We consider the Multiscale Single-Index Model (MSIM), first introduced in \cite{oymak2021learning}, as a stylized model for hierarchical learning with \emph{scale separation}. Each layer extracts a shared single-index feature at one physical scale and passes it to the next, thus defining a tractable setting in which to study how deep architectures learn multiscale representations. Under non-degeneracy and delocalization assumptions on the link function and planted features respectively, for fixed depth $K$ and local scale $d$, the first Wiener chaos of the target behaves as a perturbed spiked tensor, where the perturbation of order $d^{-1/2}$ comes from the non-linearity -- revealing the MSIM as a natural non-linear analogue of the Tensor PCA model \cite{montanari2014statistical}. While this perturbative picture is sufficient to enable efficient spectral recovery based on Tensor unfolding (as already observed in \cite{oymak2021learning}), it is not precise enough for the analysis of backpropagation gradient-based methods. In this work, we address this limitation by performing a fine-grained analysis of the Wiener chaos using Edgeworth expansions. In the first chaos, this gives a finite-rank hierarchy at scales $d^{-q/2}$. In higher chaoses, balanced flattenings exhibit staircase singular-value plateaus of size $d^{-ρ/2}$ and multiplicity $d^ρ$ under a natural higher-chaos non-cancellation condition. Using this higher-chaos structure, and under an additional slow Hermite-energy tail condition, we first establish shallow-network approximation lower bounds, quantifying the benefit of depth in this model. Next, and most importantly, we prove that online SGD on the correlation objective, where all layers evolve in the same timescale, achieves $1 - o_d(1)$ recovery with $n = \widetilde{O}( d^{K-1})$ samples, recovering the same sample complexity as in the linear counterpart.

Accurate and efficient dataset valuation is essential for enabling fair and transparent data marketplaces, especially when multiple contributors provide data for training multi-task models. Most existing valuation methods, however, are limited to single-task settings, overlooking scenarios where a buyer aims to optimize performance across multiple downstream tasks. Moreover, traditional valuation approaches, such as Shapley-based or retraining-based methods, are computationally expensive and poorly suited for decentralized environments without a trusted central coordinator and with strict privacy constraints. We propose DMVM (Decentralized Multi-task Valuation via Model Merging), a novel framework that bypasses retraining and data sharing by leveraging task arithmetic to infer dataset contributions directly from model combinations. Instead of retraining or sharing raw data, DMVM quantifies how models trained on different datasets combine in parameter space to infer each dataset's marginal utility across multiple tasks. This formulation yields a valuation process that is scalable, computationally efficient, and explicitly aligned with multi-task generalization behavior. To support decentralized deployment, we introduce a secure aggregation protocol that enables collaborative valuation without revealing individual model parameters or private data. We also provide theoretical error bounds characterizing the approximation quality of DMVM and validate our framework through comprehensive experiments on computer vision and natural language processing tasks.

Learning dissipative dynamics from discrete observations is essential for reliable long-horizon prediction and physically meaningful parameter identification. For linearly damped Hamiltonian systems, the exact flow is generally not symplectic but conformally symplectic, contracting the canonical symplectic form by a scalar factor that reflects the net dissipation. We propose Conformal Symplectic Networks with damping identification (CSympNet-ID), a discrete-time map-learning framework that learns the one-step flow map directly from snapshot pairs while enforcing exact discrete conformal symplecticity by construction, without penalty terms or projection. The architecture composes an exact symplectic neural core with explicit diagonal scaling layers whose factors are parameterized exponentially by a scalar damping-rate parameter, thereby guaranteeing positivity and interpretability of the learned dissipation factor. We establish a scaling-conjugacy factorization for conformal symplectic maps and derive a pointwise-in-step density result for CSympNet-ID. We evaluate an irregular-step damped oscillator, a damped spring-mass chain, a damped nonlinear cubic oscillator, and additional high-dimensional extensions. CSympNet-ID gives the most favorable overall results among the compared models in the reported experiments, particularly in data-scarce regimes, target contraction-law recovery, and high-dimensional tests where unstructured baselines degrade rapidly.

The federated learning (FL) paradigm fosters distributed pervasive computing combined with artificial intelligence techniques, allowing for optimized data usage and improved mitigation of privacy concerns. Indeed, model training occurs on the client's local devices, and model parameters are subsequently shared with a centralized server. However, there is a need to find a tradeoff between models' personalization and generalization capabilities. In this paper, we design and implement several testing scenarios devoted to evaluating and comparing the centralized, local, and federated paradigm performances. We also design and implement a scenario that emulates a change in clients' data. We then present experimental results of the FedAvg algorithm applied to the Human Activity Recognition (HAR) domain to understand the trade-off between personalized and generalized accuracy. Results show that, although FedAvg confirms a higher degree of personalization capabilities while keeping a high degree of generalization with respect to the traditional centralized learning, this result is not so obvious under stressful conditions, such as when varying class distribution over clients.

LLM agents are becoming a common interface for research, coding, and question answering, yet their Thought-Action-Observation loop is often serial: the model reasons, emits a tool call, then idles the GPU until the result returns. This wait consumes 16-37% of wall time in our workloads and 35-61% in prior reports. Speculative tool execution can hide this wait, but existing systems need auxiliary predictors, historical traces, or static workflow graphs, leaving a gap for training-free, day-one deployment. We observe that the model can be its own predictor: a probe forked at the start of generation predicts Qwen3-32B's upcoming tool name with 74.6-99.6% accuracy across five benchmarks. We present SPORK (Self-sPeculative fORKing), a training-free controller that dispatches the speculated tool call early, overlapping its execution with the remaining chain-of-thought decode. A cost model captures when speculation breaks even, and each component improves one of its terms: a prefix-cache fork cuts probe cost, a confidence gate filters mispredictions, and partial-token accept turns rejected probes into speculative-decoding drafts. On acceptance, the tool result is ready when reasoning ends; on rejection, SPORK falls back to serial execution with no correctness penalty. On real-tool benchmarks, SPORK cuts Qwen3-32B's GAIA P95 by 18% (131.9 to 108.1 s); the mechanism holds across model sizes from 4B to 32B and across dense and mixture-of-experts models, with task accuracy within 1 pp of baseline or better wherever measured. SPORK deploys as a thin controller over standard completion APIs (no retraining, no auxiliary models, no offline traces) and is orthogonal to token-level speculative decoding. SPORK is open source at https://github.com/baihuajun24/spork.

Coding agents often finish a session by returning one squashed patch that mixes feature implementation, bug fixes, refactorings, tests, and configuration edits. While the final code may be correct, collapsing unrelated edits into one patch removes the history structure needed for review, selective revert, and later maintenance. We study retrospective commit-history reconstruction: given a completed squashed change, an agent groups its hunks into commits and materializes a replayable commit sequence. We formalize the task as hunk-to-commit partitioning with a replay requirement and build AtomicCommitBench, containing 800 real consecutive-commit episodes from 10 Python projects. Because multiple decompositions may be reasonable, we evaluate outputs using complementary metrics: PPAR for replay validity, ARI for reference-based grouping quality, and TCR for failure containment on scoreable modified-test episodes. Natural retrospective reconstruction proves substantially harder than replay checking or synthetic tangling. Although nearly all non-random methods achieve replay validity (PPAR >= 0.988), grouping quality ranges from 0.03 to 0.46 ARI. Matched synthetic composites are much easier than real same-author squashed diffs (+0.333 ARI). In our evaluation, the GPT-5.4 setup driven by Codex CLI (0.46 ARI) and the GLM-5 setup driven by Claude Code (0.43 ARI) outperform MiniMax (0.31) and Kimi (0.29). Qualitative analysis identifies same-file lumping and support-hunk drift as recurring failure modes. Dependency-Aware Commit Evidence (DACE) improves the lower-scoring setups by 0.05 to 0.08 ARI, indicating that dependency cues and hunk-role information help agents avoid locality-driven grouping errors. AtomicCommitBench enables evaluation of the commit histories produced by coding agents alongside the final code.

Conventional uniform convergence bounds and empirical risk minimization break down in massive over-parameterized models, such as large language transformers and biological sequence networks. With near-infinite unconstrained internal degrees of freedom, their optimization landscapes develop flat vertical gauge valleys, rendering classical generalization metrics vacuous and inducing severe pathologies, specifically generative hallucination and catastrophic forgetting. We introduce the Statistically Meaningful Geometry (SMG) framework, an information-geometric paradigm lifting deterministic parametric models into infinite-dimensional non-parametric Orlicz statistical manifolds. Modeling the total state space as a differential fiber bundle ($\mathcal{M}, \mathcal{B}, π, \mathcal{V}, \mathcal{H}, ω$), we establish a Two-Fold Inference Paradigm. We formalize an Ehresmann connection 1-form $ω$ as a dynamic geometric filter that strips away vertical gauge noise (Structural Internal Directions, or SID) and isolates learning trajectories along the strictly non-degenerate horizontal distribution (Statistical Variational Directions, or SVD$χ$). We prove that under connection-filtered pre-training, out-of-distribution predictive variance is strictly upper-bounded by the finite diameter of the identifiable quotient base manifold $\mathcal{B}$, establishing a hard geometric containment of generative hallucinations. By projecting downstream updates onto the orthogonal complement of the historical horizontal carriage, we formalize the SMG Sequential Adaptation Flow, proving the total non-asymptotic elimination of catastrophic forgetting. SMG replaces empirical fine-tuning heuristics with coordinate-free topological constraints, bridging advanced differential geometry with structural reliability in AI.

We present an automated pipeline that decomposes Italian tax-court judgments into individual legal issues and extracts, for each issue, a structured XML representation grounded in the IRAC framework and the legal syllogism. The pipeline targets a corpus of approximately $330{,}000$ first- and second-instance decisions of the Italian tax courts and is built around a capable yet cost-efficient general-purpose model (DeepSeek V3), a choice driven by the need to process several hundred thousand documents at a sustainable cost. To address the well-documented unreliability of large language models on legal citations, we couple the extraction step with an automatic hallucination-detection filter that compares the references produced by the model with those identified in the judgment text by a dedicated parser (Linkoln), normalised to standard identifiers (URN-NIR, ECLI, CELEX). We validate the pipeline on $50$ judgments annotated by two PhDs in tax law, computing inter-annotator agreement and LLM-vs-expert agreement on both issue extraction and legal citations, together with a stand-alone evaluation of the hallucination filter. To the best of our knowledge, this is the first issue-level, expert-validated structured extraction pipeline with hallucination control for Italian tax-court decisions, and it provides a concrete starting point for downstream applications such as issue-level retrieval, citation-network analysis, and the construction of large-scale datasets of legal reasoning.

Eco-friendly energy management for artificial intelligence data centers (AIDCs) is crucial because of the significant increase in energy consumption-induced carbon emissions from AIDCs resulting from the rapid expansion of AI applications. This paper proposes a hierarchical carbon-aware multi-agent reinforcement learning (CA-MARL) framework for robust and efficient operations of AIDCs under uncertainties while ensuring low-carbon operation of power distribution systems. The framework comprises a workload manager (WM) agent and multiple local AIDC agents trained using a multi-agent transformer method, corresponding to a global AIDC aggregator and a local AIDC operator, respectively. Leveraging AIDC operation data along with nodal carbon intensity (NCI) calculated from the carbon emission flow-integrated distribution system operator problem, the WM agent spatially allocates AI training and inference jobs among all AIDCs. Based on the jobs allocated from the WM agent and NCI information, each AIDC agent schedules economical and eco-friendly operations of the AIDC by performing the following tasks: i) temporal shifting of training jobs, ii) spatial allocation of training graphics processing unit (GPU) blocks and inference GPUs within the AIDC, and iii) control of the supply air temperature of the cooling system. The effectiveness of the proposed framework was assessed using an IEEE 33-node power distribution system.

Masculinity in nineteenth-century fiction is not a single ideal but a field of competing scripts. Drawing on 150 British and American canonical novels from the txtLAB Novel450 corpus, published between 1771 and 1930, this paper examines the changing relative prominence of competing models of masculine authority. To focus the analysis on masculine characterisation, the study extracts male-character-centred text windows by using coreference resolution to group names, nominal mentions, and pronouns into character-specific reference chains. It then fits an unsupervised structural topic model with publication year and author gender as topic-prevalence covariates. The model identifies six distinct masculine formations: aristocratic-chivalric, Christian manhood, gentlemanly respectability, country squire, professional-commercial, and imperial/adventure. Across the corpus, formations tied to inherited rank and sacred authority decline, while those organised around paid work and adventure rise. The largest increase occurs not in professional-commercial breadwinning but in imperial/adventure masculinity, particularly the frontier-wilderness register. The trajectory points to a reallocation from inherited and sacred status towards achieved, commercial, and expansionary forms of masculine authority. Adventurous and commercial formations are also more prevalent in novels by authors recorded as male. Because these formations emerge without a seeded vocabulary yet align with categories established in independent scholarship, the article offers a reproducible method for measuring the reorganisation of gendered authority across the long nineteenth century.

Agentic code review, where autonomous agents provide code review comments on pull requests, is increasingly integrated into development workflows, yet there is limited empirical evidence on how developers respond to such comments in practice. In this paper, we present an empirical study of agentic code reviews using CodeRabbit as a case study. Through an empirical study of 31,073 pairs of code reviews and developer feedback from 10,191 pull requests across 239 GitHub repositories, our results show that agentic reviews receive mixed reception: 36.4% were accepted and 7.3% triggered discussion, while 56.3% were rejected. Rejections were primarily associated with invalid suggestions that were false positives, redundant, or out of scope, as well as misalignment with developer intent and coding practices. We further found that agentic reviews tend to focus more on functional concerns than evolvability-related comments, yet they were more likely to be invalid. To improve effectiveness in review practices, we explored various LLM-based approaches for predicting review rejection. We found that lightweight learning-based methods achieve up to 76% F1 score, suggesting learnable patterns exist between code reviews and their corresponding feedback. Our results highlight the current state of CodeRabbit's agentic code reviews, showing opportunity gaps for improvement, as well as shortcomings hindering its effectiveness.

A very large number of datasets are made available via Hugging Face (HF). These datasets are used to train a vast number of AI models, also available via HF. There are regulatory issues that might arise in the AI supply chain when those datasets do not follow, for instance, data privacy practices, or do not disclose their data sources, cleaning and preparation processes. In this preliminary work, regulatory compliance-readiness is examined as a quality attribute in the framework of HF datasets, with a main focus on data privacy (e.g. GDPR, CCPA). Towards this direction, an analysis on regulatory compliance (e.g. with privacy laws) of the dataset using the dataset card and the dataset structure overview as starting points has been performed. We collected 11,682 HF datasets that have a dataset card, are not gated and have at least 500 downloads and analyzed the datasets using automated techniques and manual analysis on a sample of the dataset. The results show that a very small number of datasets are explicit on the datasets creation processes and mention data sources, while some make mentions to Personally Identifiable Information (PII) or other sensitive data in their data schemas. These results show the need for more detailed examinations of regulatory compliance in AI datasets and models, and research on tools that provide guidance to practitioners and automate the compliance process.

Reliable analysis of bird vocalisations in passive acoustic monitoring requires models handling multiple, imbalanced annotation targets. We extend BirdCallNet for joint species and call-type classification on the long-tailed WiWa dataset and investigate how task-loss balancing interacts with pretrained representations and adaptation depth. We evaluate four bird-domain encoders, ConvNeXtBS, EAT, BirdMAE, and ProtoCLR, with separate species and call-type heads under linear probing, attentive probing, and full fine-tuning. A manually tuned fixed objective is compared with homoscedastic uncertainty weighting and Dynamic Weight Averaging across all three adaptation regimes, while GradNorm is evaluated only under full fine-tuning. Results indicate that the factorised multi-task formulation yields the most consistent improvements over the combined single-task baseline for call-type recognition, while its effect on species recognition depends on the adaptation regime. Full fine-tuning is not consistently optimal: ConvNeXtBS achieves the highest mean species performance under linear probing, whereas BirdMAE provides the strongest call-type performance under attentive probing. Adaptive weighting benefits species recognition more consistently than call-type recognition. Uncertainty weighting is particularly effective for species recognition under attentive probing, whereas Dynamic Weight Averaging is generally stronger for the same task under full fine-tuning. GradNorm achieves competitive call-type performance for selected backbones but consistently underperforms other weighting strategies for species recognition and incurs higher computational and memory costs. Overall, the preferred loss-balancing strategy depends on the backbone, adaptation regime, and target task, while frozen-backbone adaptation can provide a more favourable performance-efficiency trade-off than end-to-end fine-tuning.

While Large Language Models (LLMs) can provide personalized support in learning, several studies have raised concerns regarding their use in education. Importantly, learning depends on how students engage with LLMs. This study examined how two types of LLM-based tutors shape students' prompting practices, learning, and subsequent LLM-use: a Socratic-Guidance (SG) tutor, which structures interaction through dialogic questioning, and a Prompt-Refinement (PR) tutor that guides the formulation of effective prompts. We conducted a two-phase study in a graduate-level mobile robotics course: 66 students used either the SG or PR tutor during a 6-week intervention, followed by 52 students using an unconstrained LLM during a 3-week course project. Results show that while the SG- and PR tutors led to similar task performance and prompting patterns during guided use, they differ in learning outcomes and later LLM-use. SG-students, relative to PR-student, achieved higher learning gains in later sessions, and were more likely to adopt understanding-driven prompting strategies, which are predictive of higher understanding, when using an unconstrained LLM. Although learners perceived the SG tutor as less efficient, the findings suggest that Socratic guidance supports the development of students' capacity to learn with LLMs over time, highlighting its importance for LLM tutor design.

Real-time cardiac cine MRI enables visualization of the beating heart during free breathing, but severe undersampling and motion make reconstruction highly challenging. A central challenge for reconstruction is incorporating powerful priors of cardiac anatomy while remaining computationally efficient. We propose Piecewise Dynamic Diffusion Regularization (PDDR), a reconstruction method that integrates a spatiotemporal diffusion model as a generative prior within a variational reconstruction framework for cine MRI. The model employs dedicated spatial layers to encode anatomical structure and temporal layers to capture cardiac motion learned from gated cine data. PDDR leverages the dynamic prior in a piecewise manner, enabling the efficient use of spatiotemporal diffusion models for processing of long real-time sequences. Experiments on retrospectively accelerated and prospective real-time cine MRI demonstrate that PDDR outperforms classical, unsupervised, and diffusion-based methods, delivering high-quality reconstructions with substantially reduced computation time compared to state-of-the-art baselines. These results highlight PDDR as a practical and scalable solution for free-breathing, real-time cardiac MRI. Code is available at https://github.com/MLI-lab/pddr.

Foundation models for Earth systems have so far been trained primarily on physical climate and weather data, with limited representation of the human systems that both drive and respond to environmental change. The lack of a unified global training resource that combines climate, land, ocean, cryosphere, infrastructure, hazards, and socioeconomic data on a common grid hinders progress toward truly multimodal Earth system foundation models. We present WorldTensor, a harmonised global dataset that aligns hundreds of environmental and socioeconomic variables to a standardised 0.25$^\circ$ spatial grid and annual temporal framework. WorldTensor integrates reanalysis products, remote sensing, emissions inventories, land use reconstructions, hydrological observations, infrastructure and hazard datasets, and socioeconomic indicators within a single representation designed for machine learning workflows. To build the dataset, we regridded inputs across heterogeneous native resolutions and projections, rasterised point and vector datasets into spatially meaningful gridded fields, and reconciled temporal coverages ranging from daily observations to sparse multiyear socioeconomic snapshots. All outputs are distributed as NetCDF files with standardised coordinates, variable metadata, and a common CF metadata convention. WorldTensor provides a reproducible resource for training and evaluating foundation models that learn coupled dynamics across environmental and human systems at planetary scale.

Crossmodal correspondences between sound and taste are well established in psychology and neuroscience, but largely absent from content-based multimedia retrieval. We formalise taste-from-audio prediction as a content-based music information retrieval benchmark over a perceptually validated multi-source corpus, comparing ten frozen audio encoders from the four HEAR families under a shared multi-task regression head, with gated late-fusion as a configurable variant. In order to assess the effectiveness of the models, we compute absolute error and rank correlation. The strongest systems predict the five tastes within a macro RMSE of 0.134; on held-out real music their error is less than half a single rater's deviation from the consensus (RMSE 0.13 vs. 0.28), so the model tracks the group consensus more closely than an average human rater, and well below the previous state of the art baseline (0.219). On absolute error the encoders are statistically flat, with a single VGGish matching the best fusion, but gated late-fusion's advantage is confined to rank correlation (macro Pearson r 0.724 vs. 0.666). Operationalised as a content-based retrieval index, the predicted taste space ranks a 309-item pool far more faithfully than a CLAP-text baseline, which sits at chance; ridge probes and an audio-bandstop knockout read the strongest representations against documented sound-taste correspondences.

Embodied intelligence systems require not only end-to-end policy models, but also reusable functional modules that transform multimodal observations, robot states, human demonstrations, and task contexts into structured representations, decisions, trajectories, control references, and system services. This work defines these modules as embodied operators and studies them as independent yet composable units in embodied intelligence pipelines. We clarify their definition boundary, emphasizing task semantics, standardized input-output contracts, deployability, reusability, and multi-layer optimizability. We further construct a taxonomy covering five categories: detection and segmentation, spatial localization and 3D understanding, hand motion recovery, embodied foundation models and task-decision operators, and planning, control, and system support operators. For each category, we summarize representative functions, technical paradigms, application roles, and practical limitations. Beyond taxonomy, we propose a multi-dimensional benchmark framework that evaluates embodied operators in terms of correctness, end-to-end efficiency, resource usage, temporal stability, portability, interface compatibility, deployment reliability, and downstream task utility. We also discuss workflow-level operator acceleration and open challenges in operator composition, data standardization, world models, VLA safety, edge deployment, and real-world application value. Overall, this work argues that embodied operators should be optimized and evaluated as holistic deployable components, providing a foundation for reusable, scalable, and verifiable embodied intelligence systems.

Accurate regional weather prediction requires resolving fine-scale structure while remaining consistent with global dynamics. Traditional limited area models rely on computationally expensive simulations, while many learning-based approaches frame the problem as super-resolution, overlooking statistical and physical mismatches across scales. We propose a foundation-model-driven downscaling framework that learns regional refinements of global forecasts by augmenting a pretrained weather model backbone with lightweight, multi-scale prediction heads operating directly in its latent space. Despite being trained on substantially coarser inputs, the pretrained backbone supports regional adaptation at resolutions corresponding to a two-order-of-magnitude increase in grid-cell resolution, without the need for retraining. The proposed approach uses regional numerical simulations as training targets and is evaluated not only against gridded datasets but also against ground-based weather station observations, enabling analysis of systematic biases between global reanalysis, regional simulations, and in-situ weather station observations. Our experiments show improved accuracy in comparison to NWP on most of the metrics at the fraction of computational cost. Moreover, we observe that building on a latent space of globally pre-trained weather foundation model offers better downscaling capabilities than the standard image-based super-resolution approaches.

Topological data analysis (TDA) is a machine learning technique that uses topology to extract patterns from data and has shown the potential to exhibit quantum advantage. A key concept in TDA is persistent homology, which measures the robustness of topological information at different lengthscales. In this paper, we introduce and study the problem of normalized persistence, a practically motivated and easily interpretable version of persistent homology that counts the fraction of holes that persist at different lengthscales. We prove that a variant of normalized persistence is $\mathsf{DQC}_1$-hard and contained in $\mathsf{BQP}$, giving evidence of an exponential quantum speedup for TDA under the standard assumption that $\mathsf{DQC}_1 \not\subseteq \mathsf{BPP}$. These are the first $\mathsf{DQC}_1$-hardness results that are directly applicable to TDA instances. We also find a close connection between normalized persistence and the complexity of estimating spectral quantities in the low-energy subspace of local Hamiltonians. We study a family of such problems, including a low-energy normalized subtrace and spectral density. We show that these are $\mathsf{DQC}_1$-hard for $O(1)$-local Hamiltonians, strengthening previous results that required log-local interactions. We also introduce a variant of $\mathsf{DQC}_1$ with perfect completeness ($\mathsf{SDQC}_1$) to characterize the hardness of problems normalized by an exact kernel. This includes normalized persistence for $O(1)$-local Hamiltonians, which we show is $\mathsf{SDQC}_1$-hard.

Although machine learning (ML)-enabled software systems seem to be a success story considering their rise in economic power, there are consistent reports from companies and practitioners struggling to bring ML models into production. Many papers have focused on specific, and purely technical aspects, such as testing and pipelines, but only few on socio-technical aspects. Driven by numerous anecdotes and reports from practitioners, our goal is to collect and analyze socio-technical challenges of productionizing ML models centered around and within teams. To this end, we conducted the largest qualitative empirical study in this area, involving the manual analysis of 66 hours of talks that have been recorded by the MLOps community. By analyzing talks from practitioners for practitioners of a community with over 11,000 members in their Slack workspace, we found 17 anti-patterns, often rooted in organizational or management problems. We further list recommendations to overcome these problems, ranging from technical solutions over guidelines to organizational restructuring. Finally, we contextu-alize our findings with previous research, confirming existing results, validating our own, and highlighting new insights.

As hematoxylin & eosin (H&E) staining constitutes the primary entry point in routine diagnostic workflows, computer-aided diagnosis from whole-slide H&E images is of particular clinical relevance. However, substantial variability in specimen preparation, staining protocols, and scanning conditions, together with inherent uncertainty in expert pixel-level annotations, makes automated analysis of H&E-stained images challenging. In this study, we propose a semantic segmentation-based framework for image-level diagnosis, grounded in the clinically motivated assumption that each histopathological image corresponds to a single cancer type. Image-level predictions are obtained by assigning the class of the dominant pixel-level label in the segmentation output. To ensure clinical relevance, we adopt the nnU-Net architecture and train it on a publicly available dataset collected in our study with pixel-level annotations for three liver cancer types: hepatocellular cacrcinoma (HCC; 55 images from 30 patients), cholangiocellular carcinoma (CCA; 55 images from 29 patients), and colorectal metastatic adenocarcinoma (CMA; 60 images from 30 patients). Annotations were independently provided by four pathologist. We hypothesize that the combination of stain normalization and semantic segmentation mitigates domain shift and reduces sensitivity to annotation noise. Five-fold cross-validation yielded balanced accuracy of 0.975 (HCC), 0.950 (CCA), and 1.000 (CMA), comparable to results obtained with immunohosthochemical staining and superior to several deep learning models trained on patch-level annotations. The proposed framework has the potential to support pathologists in prioritizing immunohistochemical marker selection, thereby reducing diagnostic costs and turnaround time. Integration with immunohistochemical findings improve overall diagnostic reliability.

The alignment of large language models with human preferences is commonly achieved through Reinforcement Learning from Human Feedback or Direct Preference Optimization. However, these methods are vulnerable to the significant noise prevalent in real-world preference datasets. To address this critical issue, we present a theoretical framework for unbiased alignment, introducing the Unbiased Reward Model (URM) loss and the Unbiased Direct Preference Optimization (UDPO) loss. By mathematically correcting the distortion induced by preference noise, our novel objectives enable unbiased model training directly from noisy datasets, without requiring clean ground-truth supervision. We provide rigorous theoretical analyses demonstrating that our methods are noise-tolerant, parameter downward compatible, and classification-calibrated. Comprehensive experiments across diverse datasets demonstrate that our approaches outperform state-of-the-art baselines. Code available at: https://github.com/cswjl/unbiased-alignment.

High-throughput plant phenotyping generates valuable data that often remains trapped in unstructured text and isolated RGB images. To bridge this semantic gap, we propose a framework for constructing a multimodal granular Knowledge Graph (KG) to monitor genotype-phenotype interactions across time and experiments. In this work, we focus on wheat Triticum aestivum as a representative target crop to validate our methodology across complex canopy environments. Our pipeline first distills noisy field notes to extract entities and relations, dynamically constructing the KG by converting unique instances into hierarchical class entities via RDF-typing. These graph nodes are then aligned with standardized ontologies (PO, RO, WTO) using PlantDeBERTa. To visually ground the constructed graph, a Vision-Language Model paired with a wheat-segmentation ViT generates attention-based softmaps, linking specific KG entities directly to image pixels. We introduce a central observation node Plant_Obs_Id to connect these multimodal subgraphs temporally. Evaluated on 500 curated WisWheat samples using Pointing Game accuracy, Visual Word Sense Disambiguation (VWSD), and rank-based metrics, our neuro-symbolic approach successfully maps complex field observations to a structured graph. This enables automated field note auditing, temporal stress monitoring, and precise spatial trait localization for wheat breeders.

As Large Language Model (LLM) agents become integral to modern software systems, ``skills'' have emerged as a novel unit of software reuse, enabling developers to package workflows, decision procedures, and prompt-based policies. While skills are intended for reuse, downstream developers frequently modify published skills to fit local contexts, yet little is known about the nature of such adaptations. This paper presents the first empirical study of downstream skill adaptation in public forks, to understand how published skills are adapted, and to provide implications for researchers and engineers on improving skill design, evolution, and orchestration. Specifically, we analyze 1,126 skill-adaptation instances from six widely adopted skill repositories and develop a taxonomy comprising 46 adaptation patterns organized into 13 families. Our key findings reveal a reuse paradox: although skills are intended to be easily imported and reused, developers spend a lot of effort rewriting what the skills do, fixing skill discoverability, and translating them for different tools and languages, indicating a need for better abstractions, standardized interfaces, and automated support for skill adaptation. Furthermore, adaptations are highly interdependent, with changes in one component often requiring coordinated updates elsewhere, motivating automated support for detecting inconsistent modifications. We also find that nearly one-fifth of adaptations introduce security-sensitive content within the same instruction text that governs behavior.

Diffusion language models (DLLMs) generate text by iteratively denoising masked positions, exposing a trajectory of predictive distributions rather than a single instantaneous belief. Most existing decoders ignore this trajectory and commit tokens from the current snapshot alone, conflating confidence with commitment readiness: a transient top-1 peak under incomplete context can be locked in, while candidates with consistent cross-step support are delayed. We propose Trajectory-Aware Commit Gating (TACG), a training-free gate-level decoder that anchors token identities to the base posterior and uses trajectory-aware signals only to decide whether the current proposal is ready to commit. TACG combines Temporal Implicit Logits Guidance (TILG), which keeps an exponential moving average of past logits as a self-reference and contrasts the current logits against this reference in natural-parameter space, with a History Gate (HG) that enforces short-term proposal persistence before commitment. Together with a capped extra-promotion budget, these components yield a stability-constrained commit rule without auxiliary networks or extra forward passes. We evaluate TACG on LLaDA, Dream, and LLaDA2-Mini across code (HumanEval, MBPP) and math (GSM8K, MATH500) benchmarks; it typically improves or preserves accuracy while reducing denoising steps and increasing tokens per forward (TPF). The code is publicly available at https://github.com/Clarence-CV/TACG-DLLM.

The rapid growth of publicly available digital information has rendered manual open-source intelligence (OSINT) analysis insufficient for modern intelligence, cybersecurity, and cyber investigation. Large language models (LLMs) and agentic AI systems, capable of tool use, multi-step reasoning, and iterative intelligence generation, have emerged as promising solutions, yet evaluation frameworks have not kept pace with reported capabilities. This survey systematically reviews 74 studies and makes four contributions. First, it establishes agentic AI as a distinct analytical category rather than an extension of LLM prompting, organising the literature through an 11-category taxonomy covering LLM foundations, agentic architectures, retrieval-augmented generation (RAG), knowledge graphs, prompt engineering, domain adaptation, evaluation benchmarks, and risk. Second, it identifies the hallucination-validation gap as a corpus-level finding: although hallucination is recognised as a major reliability concern in over twenty studies, end-to-end hallucination is empirically measured in only one OSINT-specific RAG-based system, non-reproducible conditions, while related reasoning and factual-correction studies evaluate general-domain question answering rather than OSINT. Third, it maps existing research to the OSINT lifecycle, showing strong support for collection and analysis but limited coverage of verification, reporting, dissemination, and decision support. Fourth, it derives a ten-point research agenda addressing evaluation, benchmarking, hallucination measurement, adversarial robustness, dark-web coverage, multimodal intelligence, and governance. It concludes that a human-AI co-pilot model, where LLMs assist collection and triage while analysts retain responsibility for verification and decision-making, represents the most defensible near-term deployment architecture.

Modern AI applications have expanded beyond text-only interaction into a wide range of multimodal scenarios, making multimodal large language models (MLLMs) crucial for both research and industry. However, compared with traditional decoder-only LLM training, large-scale MLLM training often shows much lower MFU. We analyze the key pain points in MLLM training and introduce Mpipe, which uses a schedule algebra to derive concrete runtime behavior from a compact schedule specification. From this algebra, Mpipe derives transpose, a multimodal-aware heterogeneous parallel schedule that remaps modality-encoder computation into otherwise idle pipeline regions. On Ascend 910C NPU clusters, Mpipe achieves 2.70x speedup in a small-scale setting and 1.21x speedup in a 512-card large-scale setting.

LLM-based agents offer new opportunities for automating business process execution beyond the limits of rule-based systems. However, general-purpose LLMs lack the organization-specific knowledge required for reliable execution, which is typically fragmented across human-oriented artifacts such as policies, process models, and standard operating procedures. While such knowledge can technically be encoded in individual prompts or agent-specific retrieval setups, this approach does not scale in enterprises, as it gives rise to knowledge silos and rule duplicates, and makes consistent updates and learning across agents difficult. We argue that this calls for an organizational memory for agentic business process execution: a shared, governed, and agent-consumable reference layer of evolving organization-specific procedural knowledge about how work should be executed. We derive requirements for such a memory, propose an architecture for its curation and consumption, and demonstrate its effectiveness in a proof-of-concept based on a procurement scenario.

This paper presents Round-Trip Mutation Testing (RTM), a novel approach that generates mutants from LLM mistranslations between a program code and its intent. Leveraging the generative capability of LLMs from and to programming and natural language, and given an input program, our approach predicts its intent, that is used to generate programs, which when different from the original one, constitute the output mutants. The approach produces additionally mutants, stemming from artificially provoked mistranslations, by mutating the intent prior to the final programs (mutants) generation. Originating from the propagation of small changes in the intent to the code, our intuition is that these programs would present subtle semantic differences from the original one, simulating likely-to-occur faults that could result from specification misunderstandings, and enabling mutation testing. To evaluate RTM, we run it on 40 real buggy methods and evaluate its effectiveness and cost-efficiency in guiding testing towards detecting the bugs. Our results demonstrate the potential of round-trip mutation testing to produce syntactically more diverse mutants, potentially exposing faults that traditional mutation operators fail to reveal. More interestingly, RTM outperforms traditional pattern-based mutation in producing smaller and stronger test-suites, detecting on average over 4 and 1.7 times more faults when selecting only 4 and 30 tests respectively.

Recent tools such as OpenClaw have extended the capabilities of LLM-based agents from simple dialog-based systems to fully autonomous agents. These systems allow personalization of the agent through modifiable internal files and the installation of skills. While this enables deployment in a wide range of settings and the automation of diverse tasks, greater capability and autonomy increases the risk of malicious actions being executed unintentionally. In this work, we explore the interplay between agent configuration and the risk of executing dangerous actions without explicit instruction. To this end, we propose CONfiguration Tree-search for Red-teaming Agents (CONTRA), an LLM-assisted tree-search algorithm that discovers agent configurations resulting in the execution of malicious actions. CONTRA works by reasoning about benign yet dangerous configurations and evaluating them in a simulated environment. We construct a dataset of the 473 most popular skills from a public repository, along with 2-5 corresponding malicious target actions per skill. In a large-scale analysis, we find that 75.1% of skills have at least one configuration resulting in the execution of a malicious action, most of which have not been detected as containing malicious content by existing scans. Overall, CONTRA successfully identifies a configuration leading to the execution of the target action in 39.2% of all tested cases. Our findings demonstrate that current agents provide insufficient safety with respect to personalization.

Uncertainty quantification of hydrological predictions is necessary to inform operational decisions. Recent generative machine-learning methods have advanced probabilistic streamflow prediction, but have remained confined to lumped models that predict a basin outlet directly. At the same time, deterministic LSTM runoff models are increasingly applied at grid or catchment scale and routed through river networks to produce spatially continuous, physically consistent discharge fields. This technical note argues that moving probabilistic prediction from lumped to distributed models introduces a specific new requirement: the joint distribution of upstream runoff generation must be sampled jointly. In lumped inference, the model predicts the outlet distribution directly and can modulate spread from basin attributes. In distributed inference, downstream discharge is obtained by routing many upstream runoff predictions, so independent local sampling averages uncertainty away. Using Japan as a case study, we train two probabilistic basin-scale runoff LSTMs and route their runoff through a Hayami routing scheme. Randomly matching upstream ensemble members produces severely under-dispersed downstream ensembles, whereas a simple quantile matching strategy restores much of the spread of the direct basin-scale reference. The shift from lumped to distributed probabilistic hydrology therefore requires explicit attention to the spatial joint structure of runoff uncertainty.

Artificial intelligence has spread across the whole of the security lifecycle. The same family of models now writes application code, hardens it, and probes it for weaknesses, so that a single generative substrate increasingly performs all three roles at once. Enthusiasm for this convergence tends to treat full autonomy as the natural end point of partial assistance. This article argues that it is not. When the system that builds an artifact is drawn from the same distribution as the systems that defend and test it, the three roles inherit a common set of blind spots, and the independence that makes verification meaningful is quietly lost. Removing the human does more than raise the automation level: it collapses the external oracle against which machine output is judged, outruns the point at which a person could intervene, hands adversaries a predictable and poisonable target, and dissolves the locus of accountability when something fails. Drawing on evidence from autonomous code generation, adversarial machine learning, software fault tolerance, and the first all-machine hacking tournaments, we argue that the human belongs in the loop not as a temporary scaffold but as a permanent structural requirement, and set out what a defensible division of labour between people and machines should preserve.

We present OpenGlass, an open-source, privacy-oriented, local-first system for low-latency multimodal visual assistance, with a primary focus on blind and low-vision users. Cloud MLLM assistants offer strong visual understanding, but often require uploading first-person visual data and can suffer multi-second network delays; wearable glasses are ideal for sensing, but cannot host large models under tight compute and power budgets. OpenGlass addresses this gap with a sensing-computing split: an ESP32-based glasses-side unit captures visual context, while a nearby consumer-grade device performs local MLLM inference and local speech output, reducing cloud reliance and keeping raw egocentric visual data on user-controlled devices by default. We evaluate response quality, query-ready-to-audio latency, safety-aware abstention, and auditable logs. Under real ESP32 Wi-Fi capture, OpenGlass reaches 993 ms median user-to-audio latency with resized payloads and 1625 ms with raw 1280 x 720 payloads; 97.5% and 93.3% of trials fall below 2 s, respectively. OpenGlass is a user-initiated visual-assistance reference platform for obstacle/hazard awareness, sign/object queries, and image-quality self-checking, rather than a certified navigation aid. We release source code, hardware instructions, prompts, evaluation data, and logs.

Standard machine learning training presents data as discrete endpoint pairs, omitting the structure of the space between them. This paper introduces Transition Information Density (TID) -- the information content recoverable from structured intermediate states between categorically distinct training endpoints -- and Positional Identity, the defined location of an intermediate state on the A-to-B continuum. Both constructs are grounded in three empirical contexts: grapheme-color synesthesia, the Synesthesia Grid (a boundary-contour morphing algorithm instantiating TID in visual morphological space), and a four-condition training experiment across four representational mediums. Probes trained on structured interpolation at defined Positional Identities (C3) exhibit substantially lower intrinsic dimensionality than volume-matched controls (C2) in Phonetic/Linguistic (C3: 3.33 vs. C2: 10.81) and Semantic Description (C3: 4.59 vs. C2: 8.67) mediums. Visual and cross-modal mediums do not show this effect, establishing a modality boundary condition. A fixed-N=50 comparison confirms that Positional Identity structure, not sample count, drives the effect. Resolution N scales monotonically with representational richness. Pooled TwoNN analysis reveals globally collapsed representations in visual space (0.075) and globally consistent representations in phonetic space (0.977). The paper contributes a formal definition of TID and Positional Identity, a nine-metric shape characterization framework, and a four-condition experimental design isolating trajectory structure, data volume, and Positional Identity as distinct factors.

Automatic speech recognition (ASR) has advanced remarkably for standard speech, yet speech affected by neurological conditions remains a challenge. We present S-DiverSe (Spanish Diverse Speech), a corpus of 3.2 hours of in-the-wild Spanish speech from 22 speakers with amyotrophic lateral sclerosis, Parkinson's disease, and stroke. The dataset contains 444 manually transcribed audio segments with metadata on speaker sex, disease type, and intelligibility. S-DiverSe is designed to support ASR evaluation and development for neurologically affected Spanish speech. We describe the dataset, analyze its composition, and report baseline ASR results alongside initial adaptation experiments. Our findings reveal that heuristic text post-processing is more robust than fine-tuning for out-of-domain neurological Spanish speech. This underscores the need for dedicated in-the-wild Spanish benchmarks.

Long-form recordings (LFRs) of child-centered audio are ecologically valid sources for studying early language development, but three problems limit their use. First, LFR corpora are collected across sites with heterogeneous formats and consent structures, making cross-corpus use non-trivial. Second, without standardized benchmarks, assessing whether tools generalize across languages and conditions is hard. Third, ML workflows rarely respect privacy constraints governing sensitive child speech. This paper presents a framework addressing all three: a standardized collection of 27 child-centered datasets built with open-source tools (S1); a replicable pipeline for four speech-processing benchmarks (S2); and ELSI, a role-based ecosystem embedding ethical governance into the ML workflow (S3). We demonstrate the framework via a voice type classification case study and show the three solutions are mutually dependent.

Origin-destination (OD) flow prediction is central to urban analytics, yet deep models trained on raw counts remain vulnerable to distribution shift. The core problem is that raw count supervision cannot distinguish transferable choice mechanisms from environment-specific shortcuts. Raw OD count mixes two objects: how much demand an origin produces and how that demand is allocated across destinations. We argue that the transferable object is the exposure-to-choice law that maps spatial conditions to relative destination preferences. We propose OpFlow, a mechanism-constrained framework that learns row-centered choice potentials and reconstructs flows by combining the induced allocation with a separately calibrated origin scale. Under distribution shift, spatial exposures and the induced allocations are allowed to vary; what transfers is the conditional map from exposure states to relative choice potentials. Theoretically, we characterize the identifiable row-centered potential and show that classical spatial interaction laws are restricted log-potential cases. Controlled synthetic shifts and a real-world experiment show OpFlow improves robustness under environment shifts.

World models -- compressed latent representations of an environment that support action-conditioned prediction and planning -- are typically presented as a product of modern self-supervised learning. This paper argues that the functional anatomy of a world model was independently developed, deployed, and formally analyzed decades earlier in the model-order-reduction (MOR) and control literature, under different names and for a different purpose: the real-time operation of physical systems. We trace the anatomy across three communities. Low-dimensional models of turbulence built on proper orthogonal decomposition (POD) supplied latent dynamics learned from data of a chaotic environment; eigenface methods in early computer vision supplied the encoder-decoder half, including a primitive runtime validity check; and measurement-based POD frameworks for facility thermal control assembled the complete loop -- POD coefficients as latent state, parametric dependence on actuator setpoints as action conditioning, modal reconstruction as decoding, and, critically, a priori analytical error bounds as a verification layer that certified when the model's predictions could be trusted in closed loop. We then examine what each tradition possesses that the other lacks: MOR contributes verification, physical grounding, and extreme data efficiency; learned world models contribute nonlinear representation, transferability, and horizon. We argue that the outstanding obstacle to deploying world models in systems that cannot fail -- power, thermal, process control -- is not predictive fidelity but verifiability, and we outline a research agenda for physics-grounded, verifiable world models that unifies the two lineages.

WiFi-based human pose estimation (HPE) enables the detection and interpretation of human body positions and movements without the need for wearable devices while preserving individual privacy concerns. Implementing this solution requires enhancing model performance and maintaining efficiency, especially on resource-constrained devices. This paper introduces a novel framework, WiLHPE, for lightweight and efficient human pose estimation using WiFi CSI signals. Empowered by a camera-based model during training, WiLHPE processes raw WiFi signals directly to estimate human poses in the testing phase. It employs a novel neural network architecture to dynamically learn convolutional kernels and apply attention mechanisms across channel and frequency spaces. This innovative method diversifies the kernels to improve the recognition capabilities of WiFi signals without adding complexity, ensuring efficiency. Additionally, the Tree-Structured Parzen Estimator algorithm is employed to optimize the critical hyperparameters of the neural network efficiently, minimizing the time required for optimal hyperparameter search compared to heuristic methods. Results from experiments on both the MM-Fi and WiPose datasets highlight the superiority of WiLHPE over state-of-the-art approaches, achieving 85.96% and 94.27% at PCK50, respectively, with minimal computational overhead. Notably, WiLHPE performs impressively even under challenging conditions, maintaining around 80% at PCK50 under AWGN noise with an error variance of 0.5.

Complex Java methods remain challenging for automated unit test generation because achieving high coverage and fault detection often requires satisfying branch-specific testing requirements that are not directly visible from a focal method. Recent LLM-based approaches, such as KTester, PANTA, and MUTGEN, leverage project context, static analysis, coverage feedback, or mutation guidance. However, they do not explicitly represent and track individual testing requirements across iterations. As a result, generation may repeatedly target satisfied requirements while overlooking unresolved branches and weak assertions. Existing approaches also optimize structural coverage and mutation effectiveness separately. We present TATG, a tracking-aware LLM-based unit test generation approach. TATG introduces a unified objective representation that captures testing requirements derived from static analysis and dynamic feedback. The representation enables fine-grained tracking of satisfied and unresolved requirements throughout generation. TATG further employs a two-stage workflow: structural rounds improve coverage, followed by mutation-guided hardening rounds that strengthen assertions and improve fault detection. We evaluate TATG on 141 complex Java methods, including the 110 KTester subjects and 31 additional challenging methods. Compared with KTester and PANTA, TATG improves line coverage, branch coverage, and mutation score by 22.15, 20.14, and 37.66 percentage points on average. On a selected subset of focal methods, TATG also achieves performance comparable to a proprietary industrial test generation tool while achieving higher line coverage and mutation score.

Calibrating a superconducting transmon chip is a sequential decision problem under noise, drift, and a finite budget: an expert must choose experiments, read ambiguous plots, judge fit quality, and revise stale beliefs as the chip drifts. We study whether a vision-language agent can close this loop and specialize itself to one physical device without weight updates, via three co-designed artifacts. The first is a physics-grounded simulation environment for transmon chips: calibration observables derive from circuit-quantized parameters via scqubits, with realistic flux-line distortion, wall-time-scaled and mid-scan drift, and gate leakage, concerns a toy simulator would omit; each tool call advances a modeled clock so drift accrues by wall time, not call count. The second is a vision-language agent that runs the loop end to end, calling tools, reading plots, maintaining a structured notebook, and submitting parameters without hidden truth, scored against hidden parameters and gate fidelities measured on the device. The third is gradient-free online adaptation: a reflector reads truth-free anomaly signatures from past attempts and grows a small, human-readable device note appended to the prompt, admitted by a paired-snapshot accept gate that isolates strategy improvement from drift. On a hard-tier chip under budget pressure, six iterations raised the worst-case CZ fidelity from 0.678 to 0.787 and cut its variance, reproducing at four-qubit scale; a single accepted note raised CZ fidelity from 0.678 to 0.913 on its paired snapshot. A planted-fault study confirms the note is causal, diagnosing a hardware fault truth-free, its principal value raising the failure floor and cutting variance. The agent, scoring, and reward transfer to real hardware via a measurement-backend swap; only the accept gate is a simulation affordance, reducing to a held-out-slice or repeat-and-average form.

Spiking neural networks (SNNs) run today on a fragmented landscape of hardware: dedicated neuromorphic processors, application-specific FPGA accelerators, and large-scale neuroscience simulators, each typically built around a fixed neuron model, execution strategy, or workload class. We present AIGOR, a modular, event-driven neuromorphic architecture for spiking neural network inference. AIGOR organizes neurons into timestep-synchronized processing cores that exchange spikes as packets over a packet-switched communication layer, and it is assembled from a library of parameterized compute, memory, and communication IP blocks rather than as a one-off design for a single network. The neuron model, numeric precision, the folding of neurons onto hardware, and the partitioning across cores are configured per instance rather than committed at design time; a single declarative specification then generates the cores, neuron kernels, and synaptic-memory images that realize a given network. We validate a first prototype on the AMD Versal VPK180 across two deliberately different workloads mapped onto the same cores: a feedforward image classifier trained in snnTorch and a recurrent bal anced random network modeled in NEST. The classifier reproduces its snnTorch reference accuracy, and the recurrent network matches its NEST reference at spike-level precision across multiple cores spanning two FPGAs. We report post-implementation resource utilization and validate the multi-node synchronization scheme in simulation up to one thousand cores on a three-dimensional torus. The prototype's measured limits localize the throughput bottleneck in the synaptic-delivery datapath and the global timestep barrier, and motivate a set of datapath refinements, now in development, that the configurable structure of the architecture admits as changes to the same cores.

Scenario-based transportation analysis specifies future assumptions through aggregate population targets, whereas generative population synthesis models produce detailed individual-level realizations. When scenario targets are imposed on generative models, current practice relies on deterministic marginal calibration, implicitly assuming that the targets are compatible with the model's learned structural support. However, whether scenario-level constraints lie within the generative support--and how strongly they distort structural uncertainty--remains largely unexamined. We propose an ensemble-based Bayesian updating framework to quantify scenario compatibility in conditional population synthesis. A population-aware conditional variational autoencoder is developed to learn a distribution over plausible population structures while preserving aggregate fidelity. An ensemble of realizations sampled from the learned prior provides an empirical approximation of structural uncertainty. Scenario targets are treated as probabilistic evidence over aggregate statistics, and posterior weights are obtained through Bayesian updating across the ensemble. Scenario compatibility is quantified using effective sample size (ESS), which measures posterior concentration and the compression of structural uncertainty induced by conditioning. Experiments demonstrate that scenario impact depends not only on target magnitude but also on alignment with the learned joint structure, and reveal structural failure modes when targets fall outside prior ensemble support. The proposed framework provides a probabilistic diagnostic model for evaluating scenario feasibility and structural consistency before downstream projection and transportation planning.

Episode mining aims to extract subsequences of events that possess certain distinctive properties and constitute facts valuable to the user. Maximal frequent episode mining concentrates on discovery of frequently-appearing subsequences, which are not included into any other larger frequent subsequence. The state-of-the-art for this problem is the MaxFEM algorithm which enumerates possible subsequences, while applying various pruning techniques to accelerate the search. However, this is a computationally-intensive problem: reducing the minimum number of required subsequence occurrences or increasing the length of the subsequence both substantially raise running time, which limits practical use of MaxFEM. In this paper we describe our efforts in designing a high-performing algorithm for this problem. For this we: 1) develop an efficient C++ implementation of MaxFEM, and 2) devise an efficient technique to parallelizing it. As the result, we propose an improved parallel MaxFEM variant, which we call ParMaxFEM. Additionally, we integrate the improved algorithm into Desbordante - a high-performance, open-source data profiler with deep Python integration that treats patterns as first-class entities and allows users to develop their custom programs that can include discovery and validation of patterns. To evaluate our approach we compare both C++ implementations with the original SPMF implementation. Experiments demonstrated that our reimplemented version provides up to $8\times$ speedup over the SPMF baseline, while our parallelization technique provides up to $35\times$ improvement overall (on 8 cores).

The classical Kolmogorov--Arnold representation theorem states that any continuous multivariate function can be exactly decomposed into a finite composition of univariate continuous functions and addition operations. This foundational result has recently inspired the development of Kolmogorov--Arnold Networks (KANs) in classical machine learning, as well as their extensions into the quantum domain (QKANs). In this paper, we establish two quantum analogues of the Kolmogorov--Arnold representation theorem for continuous unitary-valued maps of several variables within an open $1$-neighbourhood of the identity matrix \(O_1(\mathbf{I}) \subset \mathcal{U}(n)\). First, we prove a representation theorem that yields an exact additive decomposition inside the matrix exponent of anti-Hermitian-valued maps. Second, due to the non-commutative nature of quantum operators, we derive a factorised version expressing the target unitary map as a finite sequential product of univariate matrix exponentials. Finally, we provide a concrete topological counterexample based on the lifting property of \(\mathcal{SU}(2)\) to demonstrate that these local representation theorems cannot be globally extended to the entire unitary group \(\mathcal{U}(n)\) without encountering fundamental structural obstructions.

Autonomous driving planning requires translating navigation intent, traffic rules, dynamic interactions, and language instructions into executable continuous trajectories. Vision-Language-Action models have been introduced into driving planning to improve long-tail generalization, commonsense reasoning, high-level semantic understanding, and explainability. However, existing VLA planners mainly follow planning-head-based trajectory prediction or full-trajectory autoregressive generation. The former only weakly constrains continuous trajectory generation with VLA reasoning, while the latter relies on long sequences of low-information-density coordinate tokens, making semantic-action alignment difficult and leading to discretization errors and inefficient inference. To address these limitations, we propose AnchorVLA, a hierarchical decision-anchored VLA planning framework that uses trajectory-pattern anchors as an explicit interface between high-level VLA reasoning and continuous trajectory execution. Specifically, Decision-as-Anchor Representation represents behavior-level driving decisions with anchor tokens, each encoding an entire local motion pattern rather than a single coordinate point. Decision-Anchored Residual Flow then generates fine-grained continuous trajectories in the selected anchor-defined residual space, capturing multi-modal execution refinements after high-level decision making. By reasoning over compact and semantically meaningful anchors instead of autoregressively generating waypoint sequences, AnchorVLA preserves LLM-based decision making while improving inference efficiency, semantic-action alignment, and continuous generation flexibility. Experiments on the Bench2Drive closed-loop benchmark show that AnchorVLA achieves a state-of-the-art Success Rate of 77.28 and a competitive Driving Score of 89.92.

Successful diffusion of AI in the workforce hinges on the economic value that AI brings to human endeavors. Bringing AI into the workforce is more than deploying a powerful new technology -- it is launching a new form of collaboration. Each human worker is now endowed with a team of AI agents; work can be delegated to these agents, and the role of the human shifts towards managing and monitoring. How can we maximize the economic value from collaboration with AI in the workforce? How can we make it a "true" collaboration that empowers human workers rather than replacing them? We take an approach that combines the fields of theoretical computer science and economics, highlighting the potential of algorithmic tools grounded in economic principles to improve the effectiveness of human-AI collective work. We consider two tiers of tools: (1) tools for better coordination, via algorithmic management of interdependencies; (2) tools for better cooperation, via contractual incentive alignment. We show how a principled approach based on algorithmic and economic research enhances both coordination and cooperation, charting a pathway for future research to inform AI markets.

Traditional reinforcement learning (RL) for recovery in autonomous systems lacks causal understanding and generalizes poorly to novel failure scenarios. RL policies often stall in failure states, spending up to 70% of an episode immobilized. Rule-based recovery alone is inadequate, and adding heuristic recovery to a pretrained PPO policy worsens rewards because policies cannot coordinate well with unanticipated interventions. The issue is not missing recovery mechanisms but a lack of policies trained to collaborate with them. We introduce CRRL, a causal-guided RL framework that trains policies to work effectively with rule-based recovery. The recovery detects stalled states and assists the agent. Causal relations from driving logs shape the training signal, teaching the policy to anticipate stalls and adjust actions in recovery contexts. The framework follows MAPE-K, with sensor collection, causal model construction, and hybrid RL policy training corresponding to Monitor, Analyze, and Plan/Execute, respectively. We evaluate CRRL through a four-condition ablation study across three driving scenarios, with 20 episodes per condition. We find that causal training significantly improves reward, distance, and velocity. Moreover, 9 of 20 roundabout episodes required zero recovery intervention, confirming navigation competence. These results show that causal-guided training produces effective RL policies that cooperate with rule-based safety components.

Understanding how households consume electricity in response to socioeconomic and climatic drivers is important for decision-makers designing energy policies in a changing climate and under geopolitical tensions. Consumers respond differently to thermal stress depending on income, consumption habits and the surrounding built environment, a nonlinear behaviour that most approaches oversimplify. In this study, households are treated as agents interacting with complex environments, and Inverse Reinforcement Learning is used to represent their consumption behaviour as model implied reward functions. Specifically, we observe how these reward functions change when households undergo socioeconomic and climatic shocks. The framework is tested on different clusters of electricity consumption profiles in Italy. Clusters' reward functions are retrieved and used to understand how cooling behaviour changes from summer 2021 to summer 2022 and 2023, before, during and after the energy crisis and a heatwave. We find that these shocks reshaped cooling behaviour heterogeneously across consumer groups, in directions conditioned by their prior habits and built environment. Across the 2021 to 2023 summers, we identify a spectrum of responses: transient adjustments that receded as the shocks eased, durable shifts persisting into 2023, and consumers exhibiting negligible change. At the intradaily scale, groups comparable in socioeconomic and environmental context but differing in their daily timing of consumption responded distinctly, identifying time of use as a separate dimension of behavioural heterogeneity. Energy policies and demand-response schemes should therefore account not only for who consumers are and where they live, but for when they consume and whether their response to a shock persists.

Software diversity has been extensively studied as a means of reducing the risk of common-mode failures. Classic work showed that the central issue is whether failures of diversely redundant components overlap in ways that limit the reliability gains. Traditional software diversity is costly to obtain, since it requires multiple implementations as well as the corresponding validation, maintenance, and deployment effort. Recent advances in Large Language Models (LLMs) may change this. LLMs enable inexpensive code generation: they produce many candidate implementations of the same specification quickly, across different models, decoding settings, and programming languages. This raises a natural question: can LLMs serve as practical generators of software diversity, and how much reliability improvement can that diversity actually provide? In this paper, we extend classical empirical studies of software diversity in human-written programs to LLM-generated code. We study three specifications using both historical human-written programs and large pools of LLM-generated ones evaluated under a common compilation, sandboxing, and exhaustive test suite. We explore LLM diversity along multiple axes, including model family, generation temperature, and programming language. Reliability improvement is evaluated in a 1-out-of-2 configuration across both homogeneous and heterogeneous program populations, including within-LLM pairings and pairings across programming languages and across LLM-generated and human-written programs. The results show that combining LLM-generated programs, especially in heterogeneous settings, can yield reliability gains, although this is partly conditioned by the programming language and generation setting. Taken together, these findings suggest that LLMs provide a scalable source of comparatively low-cost programs whose diversity can be leveraged for reliability improvement.

Decentralised federated learning, based on peer-to-peer communication, is increasingly proposed for on-device training of machine learning models, promising a privacy-preserving, communication-efficient training process with no risk of single-point failure. However, the role of structural and temporal inhomogeneities in such fully decentralised settings remains poorly understood. Here, we investigate their effects when model parameters are locally averaged during aggregation. We show that the decentralised federated learning process is governed, both in the early phase and the late, stationary limit, by the same dynamics as a lazy random-walk diffusion process on temporal networks. Based on this mapping, we demonstrate that the typical experimental scenario used in decentralised federated learning leads to unrealistically rapid convergence because of ignoring the temporal and structural inhomogeneities inherent in the communication network. We analyse real-world temporal networks and find that inhomogeneities most often dramatically slow down diffusion, hence the convergence process.

Entropy regularization is widely used in continuous-time reinforcement learning (RL) to reduce sensitivity to environmental perturbations, yet its robustness benefits lack a rigorous theoretical foundation. This paper establishes the first robustness guarantees for entropy-regularized continuous-time Markov decision processes. We show that maximizing an entropy-regularized objective yields a lower bound on a worst-case robust RL problem with joint reward and transition perturbations. We analytically characterize the induced robust sets and prove that they expand monotonically with the regularization strength, justifying the empirical observation that stronger entropy improves robustness. In contrast to prior discrete-time analyses, our results remove the intractable state-distribution entropy term and provide guarantees invariant to action frequency. Experiments on queueing network control and market making confirm our theory, showing that entropy-regularized policies outperform greedy and $ε$-greedy baselines under dynamics perturbations.

Template-based contrastive synthesis is scalable, but its candidates often differ only in a few entity-slots while sequence-level optimization spreads supervision over mostly shared templates. We formalize this as the Resolution Mismatch Problem and propose KARMA, which enumerates schema-constrained paths over domain knowledge graphs and verbalizes them into slot-aligned contrastive candidates. Slot-Parallel Alignment (SPA) then applies a decoupled slot-level objective to route preference supervision to discriminative entity-slots, with slot-aware masked attention serving as an optional packed-evaluation implementation. Across biomedical, computer-science, and chemistry benchmarks, KARMA outperforms base LLM and same-data SFT baselines, and compares favorably with sequence and token-level preference methods.

LLM-powered agents struggle with personalization when users issue raw, underspecified queries. In this setting, agents must infer latent intent, extract preferences from noisy interaction histories, and select among competing alternatives. Existing benchmarks rarely test this capability, as they often rely on user-refined queries or simplified histories. We introduce personalized product search (PPS), a testbed for agentic personalization under raw queries and diverse histories. We construct Agent Personalized Benchmark (APeB) from action logs, pairing underspecified intents with rich histories and user-viewed candidate items. Evaluating state-of-the-art LLMs with multi-step agent workflows, we find that models handle explicit queries well but struggle with early-stage queries requiring intent and preference discovery. Rubric analysis attributes this gap mainly to ineffective history use. A simple history-aware query-refinement pipeline, VQRA, yields consistent gains, highlighting the need for dedicated history-utilization modules in personalized agents.

In selective deployment, practitioners act only on a model-chosen subset of individuals based on predicted conditional average treatment effects, but marginal conformal guarantees need not control reliability on that selected subset. We study reliable selection for black-box CATE predictors: selecting candidates whose CATE errors are below a tolerance while controlling the false discovery rate (FDR). Since CATE errors are unobservable, we construct doubly robust proxy errors from pseudo-outcomes; however, naive proxies can lose power under heteroskedasticity because variance overwhelms the reliability signal. We propose Denoised Conformal Alignment, which subtracts an estimated conditional variance component and combines conformal calibration with Benjamini--Hochberg selection. Our analysis shows that validity is governed by stability of proxy/oracle threshold labels, rather than pointwise perfection of the variance estimator. Experiments show substantially improved power while maintaining FDR control across challenging settings.

This study examines how prompt language and translation theory-driven prompt design influence the quality of Spanish-Chinese journalistic translations generated by GPT-5.2. A parallel corpus of four editorials from El Pais was translated under 48 experimental conditions (4 prompt types, 3 prompt languages, and 4 articles). Translation quality was assessed using BLEU and BERTScore-F1 for automated evaluation, alongside human evaluation based on the Multidimensional Quality Metrics (MQM) framework. Automated metrics identified the baseline prompt (BASE) as the best-performing condition, whereas human evaluation ranked the brief-oriented prompt (BRIEF) highest (MQM: 8.66 vs. 7.84), a reversal likely attributable to the single-reference constraint inherent in automated measures. Sub-error type analysis revealed that translation theory-driven prompts selectively reduced Awkward style errors, while Unidiomatic style errors persisted across conditions. Prompt language had a negligible impact under both evaluation paradigms. These results indicate that translation theory-driven prompts can yield measurable quality gains under expert evaluation of journalistic translations, although their pedagogical implications for language learners remain suggestive and require validation through user-based studies.

Large language models (LLMs) are increasingly used to implement algorithms from research manuscripts, but papers often leave implementation choices implicit. This study examines how the written format of an algorithm specification affects first-pass LLM implementation accuracy. We compare ordinary prose, LaTeX algorithm-style pseudocode, PDF-like extracted pseudocode, Markdown fields, YAML-like specifications, JSON-like specifications, and Python code stubs across five machine learning tasks, three models, and four experimental settings, yielding 4,020 generated implementations. Hidden tests evaluate details that often determine correctness, including tie-breaking, array shapes, numerical rules, return structures, and invalid-input behavior. Under the core-information setting, LaTeX algorithm-style pseudocode has the largest average format effect, with YAML-like specifications and ordinary prose close behind. Under complete information, GPT-5.4 mini shows no format differences in the matched comparisons, whereas Gemma 3 4B and Llama 3.2 3B still do. Code stubs do not consistently improve correctness despite specifying the function signature. The results support a writing recommendation: authors should state the interface, computation steps, numerical rules, and boundary-case behavior explicitly, instead of relying on a particular surface format to carry those details.

Multi-domain knowledge graph completion (MKGC) aims to improve missing triple prediction in a target KG by transferring knowledge from other support KGs. Existing methods typically enforce consistency constraints on equivalent entities across KGs to transfer knowledge, which risks suppressing domain-specific contextual information of entities. This design can also compromise entity representation information from all KG domains, impeding performance improvements, especially in low-resource data scenarios. To address this, we pioneer a generation-based paradigm for MKGC and propose DMKGC, a conditional diffusion-guided knowledge transfer framework. Our key insight is to treat each KG as a partial view of the entity entire information, and generate informative domain-general entity embeddings through diffusion models conditioned on support KGs. Particularly, we first initialize domain-agnostic entity embeddings as prior entity embeddings, and then encode them within individual KGs. Afterward, we fuse equivalent entities from support KGs as the conditional diffusion generation guidance. We leverage the prior entity embeddings as the proxy generation objective, which ensures this conditional generation to be unbiased towards any conditioned KGs. Simultaneously, we also train the generated embeddings to be predictive across KGs, thus preserving domain-specific information. Extensive experiments on 14 KGs in 3 benchmarks demonstrate a 4.3\% average MRR improvement in tail entity prediction over state-of-the-art methods, with sustained gains in low-resource data settings.

As particle physics detectors grow in scale, High Energy Physics experiments must process ever-increasing data volumes. Level-1 trigger systems, implemented on Field-Programmable Gate Arrays and increasingly using neural-network algorithms, filter this data in real time. However, their proximity to the interaction point exposes them to radiation, which can corrupt outputs, stall processing pipelines, or damage hardware, with significant financial and scientific consequences. In this work, we present the first Register Transfer Level fault-injection study of a deployed Level-1 hardware neural-network trigger, GNN-ETM in the Belle II trigger system. We target three failure modes most consequential to a real-time trigger pipeline: deadlocks, timeouts, and packet-integrity violations. Through two complementary campaigns, we inject 1 442 840 Single-Event Upsets across 211 245 signals. We find a monitoring asymmetry in the existing verification infrastructure and propose inter-stage liveness monitoring as a more accurate alternative to output-only observation, showing that Mean Time To Failure estimates from the two approaches differ by up to 78.7%. The resulting per-stage data identifies the highest-priority hardening targets.

While deepfake audio detection systems achieve high performance in controlled benchmarks, their reliability often diminishes in the wild. Prior work shows that dataset-specific artifacts contribute to this gap. Yet, systematic tools to identify which acoustic properties a model exploits as shortcuts remain limited. We propose an intervention-based diagnostic framework, grounded in a directed graphical model, that formally distinguishes confound-driven shortcut dependencies from legitimate domain shift. We operationalise this through controlled acoustic perturbations targeting non-speech structure, spectral content, and signal energy, complemented by corpus-level distributional analysis. Evaluating XLS-R-300M with RawGAT-ST across ASVspoof challenges datasets, we quantify model sensitivity to specific intervention types. Results reveal that non-speech interventions produce the largest performance shifts, confirming non-speech intervals as a dominant shortcut.

Activation functions are considered an essential primitive for neural nonlinearity, i.e., they enable neural networks to serve as universal approximators. In this paper, we show that this nonlinearity can also be achieved by input-conditioned threshold gating through branches as a universal primitive. We demonstrate that standard activations -- whether piecewise-linear (ReLU, PReLU, Hardtanh) or smooth (SiLU, Sigmoid, Tanh, GELU) -- are in fact instances of a single Threshold Gating (TG) primitive. For softmax, we show that it admits an exact TG conversion via its equivalent per-element Sigmoid form. We then validate these equivalences by converting pretrained networks across CNNs, transformer-based models, and recurrent architectures, preserving model performance without requiring retraining. Threshold Gating also enables training from scratch that goes beyond replacing existing activations, enabling gains in model compression, performance, and shorter training. We also propose a 'Minimal Branch Theorem' which relates the minimum number of required branches in our primitive to the trainability of general deep neural networks. In terms of hardware implementation, TG maps to a unified implementation in the case of analog in-memory systems, addressing the bottleneck of analog-to-digital and digital-to-analog converters (ADC/DAC) that is known to significantly impact power consumption and on-chip area.

The informativeness of a training set is as consequential as its size, yet most sampling strategies remain agnostic to the intrinsic geometry of the data distribution. We introduce CuBAS (Curvature-Based Adaptive Sampling), an information-geometric framework for adaptive data selection in supervised classification, grounded in the q-state Potts Markov random field (MRF) model. The central insight is that a labeled dataset can be viewed as a statistical manifold, on which local curvature, estimated via the ratio of second to first-order observed Fisher information, faithfully encodes the geometric complexity of the data distribution. We construct a k-nearest-neighbor graph over the labeled data and derive a closed-form curvature score at each vertex from the Potts sufficient statistics. This curvature signal partitions the graph into two complementary regimes: low-curvature regions, corresponding to smooth, homogeneous clusters, and high-curvature regions, concentrated around decision boundaries that are disproportionately informative for classification. By selecting nodes from both regimes, CuBAS constructs compact yet maximally informative training subsets. Empirical evaluation across more than 60 benchmark datasets demonstrates consistent and statistically significant improvements over random sampling and uncertainty-based baselines, across a wide range of labeling budgets and classifier architectures. CuBAS is computationally efficient (linear in the number of k-NN graph edges), theoretically grounded in the differential geometry of statistical manifolds, and interpretable in terms of the local shape operator of the data manifold.

Vision-language alignment powers open-vocabulary recognition, retrieval, and LVLM grounding, yet natural captions are often underspecified, making similarity brittle and overly confident under paraphrase and omitted details. We aim to learn representations whose matching is stable across caption views and whose confidence reflects how strongly text constrains an image. We propose Text as Partial Constraint (TPC), a core-residual alignment framework that treats multi-view captions as incomplete supervision. It distills a consensus semantic core as the alignment target, learns a single-view core predictor for standard inference with one query, and explicitly discourages vision-language similarity from depending on the orthogonal unsaid residual. An uncertainty-aware contrastive objective further softens alignment when caption views disagree, reducing overconfident updates under weak language constraints. Across zero-shot recognition and adversarial robustness, TPC achieves 81.42/64.05 Top-1 clean/robust accuracy on ImageNet and 76.19/52.03 on an Avg-14 transfer suite, while improving LVLM transfer with 85.16 POPE F1 and 59.57 OKVQA accuracy under an LLaVA-1.5-7B stack. These results suggest that modeling text as a partial constraint is a practical and principled route to more reliable vision-language representations under underspecified language supervision.

Industrial automation increasingly demands control strategies that balance operational performance with strict energy efficiency requirements. A common approach to solving this multi-objective problem, particularly within the framework of reinforcement learning (RL), is to formulate a single, scalar reward function that linearly combines the competing objectives. However, the manual weighting of these different objectives is heavily reliant on domain intuition, incredibly time-consuming, prone to human bias, and frequently fails to uncover optimal trade-off solutions. This work addresses the critical challenge of automating the weight selection process to systematically and efficiently discover the Pareto front of optimal trade-off policies. We formulate the weight selection process as a multi-objective Bayesian optimization (MOBO) problem and evaluate its sample efficiency against a standard uniform grid search baseline. Using a physical Quanser Aero 2 testbed configured for 1-DoF pitch control, our results demonstrate that the MOBO approach, utilizing the expected hypervolume improvement (qEHVI) acquisition function, consistently outperforms uniform grid sampling. MOBO achieves superior hypervolume and maximum spread, successfully identifying high-quality, diverse trade-off policies with a reduced evaluation budget, thereby enabling highly efficient energy-aware control in complex mechatronic systems.

Maintaining consistency between architectural design and runtime-observed behavior is challenging in long-lived safety-critical firmware. This paper presents a runtime-informed methodology for detecting architectural drift in ISO 26262-compliant firmware. The approach collects hardware-assisted execution traces, abstracts them into message exchanges among firmware components, and compares the resulting runtime behavior with design-time sequence diagrams through a deterministic differencing step. The computed delta identifies discrepancies as confirmed, missing, additional, or inverted, while a constrained LLM-based step generates a human-readable report only to support expert review. We evaluate the methodology in an industrial firmware context through agreement-based validation and a practitioner survey. Results over 26 test cases show strong agreement between the generated deltas and expert-curated references, while practitioners perceive the reports as useful for interpreting drift, reducing manual analysis effort, and supporting safety-oriented documentation activities. The findings suggest that combining runtime trace analysis, deterministic architectural differencing, and constrained LLM-based reporting can practically support architectural drift detection in evolving safety-critical firmware.

Recent research expands beyond binary anti-spoofing with the emergence of Source Tracing, the task of identifying the specific generative origins of synthetic speech. However, current research often equates a "source" with its generative architecture. We propose redefining a source as a compositional tuple of Architecture, Training Data, and other training factors affecting the generated speech. We propose a framework using Structured Orthonormal Prototypes to minimize class overlap and intra-class variance. Our Subspace Partitioning strategy splits the embedding into architecture and data subspaces, while a residual subspace captures stochastic variability, enabling "compositional generalization" for novel factor combinations. This approach improves performance for partially seen sources and maintains robustness in fully open-set scenarios. MLAAD evaluations for Few-Shot open-set Identification show our approach significantly outperforms angular-margin baselines.

Deep reinforcement learning (DRL) in industrial control often suffers from lag and overshoot due to purely reactive control based on the current tracking error. To achieve anticipatory control without high computational overhead, we introduce a predictive formulation that augments the DRL state space with target velocities and future reference horizons. Evaluating eight configurations using proximal policy optimization (PPO) on a 1-degree-of-freedom (1-DoF) helicopter testbed, simulation results showed a 9-fold error reduction, lowering the mean absolute deviation from 2.73° to 0.31°. However, zero-shot transfer to physical hardware revealed a sim-to-real gap. Interestingly, a simpler configuration using a single, further look-ahead horizon matched the real-world top performance of the most complex model (1.11°). Overall, evaluating various combinations of prediction horizons and target velocities demonstrated that highly granular predictive data is not necessarily required for physical transfer.

Modern video surveillance systems generate far more video streams than human operators can effectively monitor, making automated analysis essential for timely detection of security events. This paper presents a unified multi-task deep learning framework that simultaneously performs face recognition with zone-based authorization, automatic license plate recognition, weapon detection, fire and smoke detection, and human action recognition on a shared GPU platform. Among the integrated modules, two task-specific deep-learning models are proposed in this work to address scenarios that are insufficiently represented in publicly available datasets: a single-class weapon detector fine-tuned on a merged and relabeled dataset, achieving a mean average precision (mAP@0.5) of 0.947, and a SlowFast-R50 action recognition model trained on a purpose-built vandalism dataset comprising 614 video clips, achieving 94.33% classification accuracy. To improve robustness in continuous video, all detection modules are integrated into a temporal event-validation architecture based on multi-frame confirmation, confidence-weighted voting, and cascaded filtering, transforming frame-level predictions into reliable security events. Each module is evaluated independently on established public datasets (LFW, D-Fire, FIRESENSE, and UCF-Crime), followed by integrated end-to-end system evaluation. The proposed temporal validation strategy reduces the fire and smoke false-alarm rate from 52% to 4% and improves video license plate exact-match accuracy from 66.7% to 81.8%, while the complete framework maintains real-time operation with a per-frame latency below 100 ms on commodity hardware. These results demonstrate that combining specialized deep-learning models with temporal event validation provides an effective and practical solution for reliable real-time intelligent video surveillance.

Generative AI has made remarkable progress in producing functionally correct code, yet ensuring both correctness and performance remains an open challenge. We present Copper, a framework that combines formal verification with performance-aware specification to generate code that is provably correct and efficiently executable. Our approach integrates AI-driven code synthesis with formal verification tools, and automated performance profiling loops. Evaluated on a diverse set of algorithmic and real-world programming tasks, Copper produces solutions that satisfy strict correctness guarantees while delivering significant improvements in runtime and memory efficiency compared to baseline AI-generated code. This work demonstrates that it is feasible to bridge the gap between trustworthiness and performance in AI-assisted programming, offering a practical pathway toward reliable, high-performance code generation.

Reinforcement Learning (RL) has substantially improved the reasoning ability of large language models (LLMs), but sparse outcome rewards still make token-level credit assignment difficult. Existing scalable RL methods typically assign trajectory-level rewards uniformly across tokens, while recent entropy-aware approaches either rely on coarse detached heuristics or directly optimize true entropy, which can introduce non-local gradient components misaligned with sampled-token policy updates. We propose Adaptive Credit Policy Optimization (ACPO), a token-level credit assignment framework based on a mode-local surrogate entropy. ACPO asymmetrically modulates policy updates by emphasizing uncertain decisions in successful rollouts and overconfident tokens in failed rollouts. We show that the surrogate admits deterministic entropy bounds and, under modal alignment and proximal updates, preserves the policy-gradient direction to leading order. Experiments on mathematical reasoning and coding benchmarks, including AIME 2025 and HumanEvalPro, show that ACPO consistently improves over strong RL baselines such as DAPO, GTPO, and SAPO.

Deep reinforcement learning (DRL) offers powerful control for industrial cyber-physical systems (ICPSs), but its "black-box" exploration risks violating strict hardware safety limits. Typically, these constraints are managed through complex reward shaping. In this work-in-progress paper, we embed a differentiable physics model directly into the proximal policy optimization (PPO) actor loss function. By simulating short-horizon future trajectories during training, the policy is penalized for anticipated safety violations independent of the task-reward signal. Evaluated on a simulated 1-degree-of-freedom helicopter testbed with strict pitch constraints, our physics-informed soft regularizations substantially reduce constraint violations while maintaining reliable target tracking.

We introduce Vidu S1, a real-time interactive video generation model supporting voice control of digital characters. Users can control video generation content at any moment through voice instructions. Vidu S1 supports infinite-length real-time video generation without blurring, drift, or visual distortion. Built with TurboDiffusion and TurboServe, Vidu S1 outputs 540p real-time videos at up to 42 FPS on regular consumer GPUs. Users can upload custom images of real people, anime, and pets, and choose different voice tones for personalized experiences. Experiments show that Vidu S1 achieves the best performance across all test metrics while fully meeting real-time inference requirements. A playable online demo is available at https://vidu.com/vidu-stream.

High-level languages such as R or Python are used frequently to analyze and visualize data in the form of scripts or notebooks. However, these artifacts suffer from reproducibility issues due to what we frame as implicit assumptions made by the authors. Such assumptions range from package versions and shapes of involved data tables, to manual and often undocumented setup steps. Within this work, we provide a unified, example-driven perspective on implicit assumptions in data analysis backed by an explorative proof-of-concept implementation. With this perspective, we propose the use of static analysis techniques to identify such assumptions and to make them explicit in the form of code constraints, focusing on the inclusion of data-analysis-specific issues. Such constraints can then be used to automatically transform these scripts into executable and reproducible artifacts, to check these assumptions at runtime, and to serve as documentation to support code reuse and comprehension.

We revisit random projections for reducing the dimension of high-dimensional polygonal curves. Drawing from the toolbox of randomized linear algebra, we give a considerably simplified proof of the known $O(\varepsilon^{-2}\log(nm))$ bound on the target dimension of a random projection that preserves the continuous Fréchet distance of polygonal curves up to a factor $(1\pm\varepsilon)$. Our proof is based on the concept of sparse oblivious subspace embeddings. While previous techniques were limited to the case of the Fréchet distance, our techniques are fairly general and extend to all possible distance measures that involve the maximum, a sum or an integral over Euclidean distances between pairs of points on both input curves. We define a generalized dissimilarity measure for curves that includes several popular measures such as Fréchet, $q$-DTW, Hausdorff, etc. as special cases and show that the same dimension reduction technique works for this generalized dissimilarity measure. Finally, we apply the same framework for dimension reduction to piecewise linear surfaces, after extending the distance measure suitably to such surfaces.

We study a graph classification problem involving over 20 million graphs, arising from high-order perturbative computations of correlators in planar $\mathcal{N}=4$ super-Yang--Mills, a model closely related to the theory of the strong nuclear force. We benchmark graph neural networks, including graph transformers, achieving robust generalization to larger graphs with up to $99.996\%$ ROC AUC. Then, we analyze how the models can be used to gain a computational speedup compared to the traditional graphical bootstrap algorithm, through shrinking the redundant data by up to $85.5\%$ at the level of denominator graphs. Finally, we study the embeddings of the models to investigate their interpretability.

Post-hoc analysis of trained neural network weights often seeks to recover geometric structure directly from the parameters. We show that, for positional-encoding-equipped neural fields, the symmetry visible from weights is not the true symmetry group itself, but an observable symmetry set determined by the trained parameters, the positional encoding (PE), and readout observable. We formulate this dependence through an exact observability hierarchy, $G_{\mathrm{obs}}^{\mathrm{exact}} \subseteq G_{\mathrm{lift}}^{\mathrm{exact}}(φ) \cap G_{\mathrm{true}}$, where $G_{\mathrm{lift}}^{\mathrm{exact}}(φ)$ is the set of input transformations that the PE can exactly lift to the feature space. The hierarchy implies that even when a target function has a geometric symmetry, that symmetry may be structurally invisible to weight-level observables if the PE does not represent the corresponding transformation. We test this prediction using MLPs trained on two-dimensional signed distance functions with multiple shape symmetry groups, positional encodings, and Gram-based observables. The results show a consistent PE-dependent pattern: DyadicAxisPE supports $D_4$-sensitive readout but structurally suppresses $D_3$ rotations, TriAxisPE yields lower $D_3$ / $D_6$ readout scores under the tested Gram observables by replacing coordinate axes with three 120-degree-separated axes, and random Fourier features mainly exhibit a $π$-rotation response under these readouts. These findings show that PE design affects not only approximation behavior but also which structures are accessible to post-hoc weight-level readouts. This provides a basis for a principled observable-dependent symmetry readout.

Clinical cardiac imaging pipelines currently deploy separate models for each dataset and modality, incurring redundant training costs and precluding knowledge sharing across anatomically related tasks. Consolidating semi-supervised learning, unsupervised domain adaptation, and domain generalisation into one model is therefore a practical necessity, yet naive joint training exposes a fundamental barrier: conflicting label semantics between datasets collapse LA Dice from 90.31\% to 83.38\%, while gradient imbalance across tasks of unequal complexity suppresses the weaker tasks throughout training. We present UniT-Diff, a unified diffusion segmentation framework that resolves these conflicts through three targeted mechanisms. An 11-channel task-specific output space physically partitions label categories, eliminating cross-task gradient sign reversal by construction. SNR-Adaptive Task Conditioning (SATC) scales the task token by the log signal-to-noise ratio of the current diffusion timestep, suppressing domain-specific bias during coarse denoising and restoring full task guidance as the signal clears. Task-Type-Aware Conditional Dropout (TTACD) permanently removes the task token for domain-generalisation inputs, routing them through a shared neutral pathway that draws on cross-dataset cardiac anatomy rather than source-vendor statistics. Under a single parameter set, UniT-Diff surpasses independently trained task-specific baselines on all three benchmarks simultaneously: +0.87\% on LA, +1.77\% on MMWHS, and +0.88\% on MNMS.

Modern web applications increasingly expose accessibility barriers through interaction flows rather than static page snapshots. Keyboard traps, focus loss, modal leakage, delayed status updates, dynamic controls, and changing page regions often become observable only after users perform concrete actions. These behaviors are directly related to dynamic WCAG criteria, yet they remain difficult to automate because their assessment depends on runtime interaction evidence and is still commonly performed through manual inspection. We present Flow-A11y, a flow-aware accessibility testing system for interaction-dependent WCAG criteria. Given a target page and a natural-language scenario, Flow-A11y executes the flow in a real browser, records an ordered runtime trace, constructs criterion-specific evidence packets, gates unsupported judgments, and emits auditable findings grounded in resolvable runtime evidence. Evaluated on 19 real public-web scenarios covering 45 dynamic WCAG criteria, Flow-A11y achieves over ten times higher oracle agreement than a generic browser-agent audit, while its evidence-calibration layer improves fail precision from 23.5% to 41.4% and eliminates invalid evidence references. These results show that runtime traces provide actionable evidence for assessing interaction-dependent accessibility behavior. They demonstrate a practical path toward automating dynamic WCAG criteria that page-level scanners cannot assess and that have traditionally required manual evaluation.

Heterogeneous Graph Neural Networks (HGNNs) have exhibited remarkable efficacy in modeling complex systems with multiple types of nodes and relations, yet their training on large-scale heterogeneous graphs remains computationally prohibitive. Although graph condensation methods can effectively improve learning efficiency on large-scale graphs, existing condensation processes are mainly designed for homogeneous graphs and typically rely on computationally expensive gradient matching or bilevel optimization paradigms, rendering them impractical for heterogeneous settings. To address these limitations, we propose HGC-RC, a simple yet effective role-aware heterogeneous graph condensation framework. Specifically, HGC-RC first extracts semantically enhanced node embeddings via lightweight propagation. It then introduces a role-aware hybrid clustering strategy consisting of class-partitioned clustering for labeled target nodes to preserve class distributions and unsupervised type-wise clustering for non-target nodes to retain critical cross-type connectivity. Finally, a compact heterogeneous graph is efficiently reconstructed based on the resulting cluster assignments. Extensive experiments demonstrate that HGC-RC outperforms state-of-the-art baselines, offering a practical pathway to accelerate HGNN training on large-scale heterogeneous graphs without sacrificing task performance

Electroencephalography (EEG) decoding for brain-computer interfaces (BCIs) faces a major challenge: substantial inter-subject variability limits effective cross-subject generalization. Consequently, practical systems still rely largely on subject-specific models trained from scratch and requiring individual recalibration. EEG foundation models have recently emerged as a promising alternative; however, even large pretrained models cannot simply be used as fixed feature extractors and still require additional adaptation before they can be reliably applied to downstream tasks. In this work, we address this challenge through targeted adaptation strategies. Building on recent EEG foundation models such as REVE, LaBraM, and LUNA, we examine the impact of different low-rank adaptation strategies on motor imagery classification. We propose a framework that structurally decouples subject-invariant knowledge from subject-specific neural signatures: the low-rank update at each adapted layer is split into a Global adapter, trained jointly across all subjects, and Subject-Specific adapters, each absorbing individual variability. To assess the contribution of each path, we compare three adaptation strategies: (i) subject-specific LoRA (ii) global LoRA and (iii) stacked LoRA, combining both Global and Subject Specific adapters. Experiments on BCI Competition IV-2a, PhysioNet Motor Imagery, and the clinical Zuo2025 benchmark show that Stacked LoRA effectively mitigates inter-subject variability, achieving the best accuracy in the large majority of backbone and dataset combinations. Our analysis further reveals that the optimal balance between the global and subject-specific paths depends on the target population: a shared adapter is sufficient for large, diverse cohorts, whereas subject-specific adaptation is decisive under the high inter-session variability of clinical recordings.

Thinking has emerged as a critical capability for Large Language Models (LLMs) tackling complex tasks. However, its reactive nature, where reasoning is passively triggered only upon receiving a user response, inevitably introduces latency that compromises conversational fluidity. This stands in sharp contrast to human dialogue, where speakers proactively anticipate and plan future content during natural pauses to ensure seamless interaction. To bridge this gap, we propose Proactive Thinking, a framework that empowers models to pre-compute potential response elements during conversational downtime instead of waiting idly for the next input. We then introduce a training-free baseline that can think ahead by anticipating future states, balancing efficiency and quality through speculative continual thinking. To evaluate this approach in practice, we adapt three benchmarks of varying complexity into time-aware environments that simulate real-time conversational flow. We demonstrate that proactive thinking effectively improves interaction efficiency without compromising performance. Ultimately, this work advocates for a fundamental shift toward more intelligent, anticipatory, and real-time conversational AI.

Silicon sampling-using large language models (LLMs) to simulate human survey respondents-has emerged as a promising approach for augmenting traditional survey research. However, most evaluations rely on distributional comparisons rather than individual-level prediction, which risks conflating pattern matching with coherent respondent-level prediction. We propose cross-survey transfer, a more rigorous evaluation framework in which an LLM is given a respondent's answers to one set of questions and must predict their answers to entirely different questions from the same survey. Using data from the Taiwan Election and Democratization Study (TEDS) 2024, three open-weight LLMs (27B-120B parameters), and supervised machine learning baselines, we find that: (1) zero-shot LLMs achieve 52% accuracy on genuinely unseen items, closing to within 6 percentage points (pp) of a supervised random forest trained on same-population data; (2) a stable construct predictability hierarchy emerges, from 67% for partisan attitudes to 23% for sovereignty; and (3) variance collapse and safety alignment effects-two commonly cited LLM limitations-turn out to be more nuanced than previously reported, with variance collapse affecting supervised models as well and alignment effects varying dramatically across model families. These findings clarify both the promise and boundaries of silicon sampling.

HAR is increasingly expected to run continuously on edge devices, yet recent LLM-based methods remain hard to deploy: raw sensor prompts are long, cloud inference adds latency and privacy risk, and fine-tuned LLM pipelines turn general-purpose models into task-specific classifiers. We present STELLA, an efficient sensor-to-LLM translation framework for on-device HAR that shifts the burden from LLM adaptation to sensor tokenization. A lightweight hierarchical tokenizer compresses an entire multi-channel inertial window into a fixed set of compact latent sensor tokens, which are projected into the embedding space of a frozen pretrained LLM and combined with a natural-language prompt for label scoring. This preserves activity-relevant temporal and cross-channel structure while keeping LLM-side computation predictable across sensor configurations. STELLA also supports on-device personalization, adapting only the lightweight tokenizer on small amounts of user-specific labelled data and augmenting inference with a local retrieval context, keeping the LLM, user data, and retrieval on device. Across seven public HAR datasets and eight benchmark settings, STELLA achieves new state-of-the-art performance, improving over prior methods by up to 11.83% F1; on-device personalization yields up to a further 21.91% F1 as user data accumulates after deployment. STELLA also outperforms representative time-series tokenizers under the same LLM pipeline and achieves real-time inference under practical mobile and edge budgets, showing that efficient sensor tokenization is a practical path toward accurate, private, and personalized LLM-based HAR on edge devices.

With the rise of Industry 4.0 driven by the integration of Cyber-Physical Systems (CPS) and the Internet of Things (IoT), the use of Digital Twins (DTs) has significantly increased over the past decade, as they provide detailed insights and support well-informed decision-making. However, the lack of standardized methodologies, in addition to the time and resources involved for building them remains an important challenge. Building on the idea that engineering models of the physical twin (PT) are often available, we propose a tool-supported framework that automates the derivation of DTs by leveraging existing structural and behavioral models of the PT and extending them with additional models to build a comprehensive DT. To demonstrate the feasibility of our approach, we applied it to four different use cases, in which we automatically derived DT instances from (1) models of their PT, (2) configuration of our generic framework and (3) minimal ad hoc additional development for connecting the DT to the PT. These experiments illustrate the applicability of our framework for building DTs in contexts that satisfy our assumptions and requirements. By simply configuring the framework, we are able to derive a DT aligned with its operational purpose.

Safety-critical applications require classifiers that are both robust and reliable. Adversarial training is a widely adopted defense for improving robustness in deep neural networks; however, its effect on the reliability of predictive uncertainty remains underexplored. We investigate this gap through the lens of selective classification, which has rarely been systematically analyzed alongside adversarial robustness. We introduce a unified benchmark for the robustness-uncertainty trade-off. It standardizes architectures, augmentations, threat models, and evaluation metrics across clean, adversarial, and common-corruption settings. Across a wide range of state-of-the-art adversarial training methods, we uncover a recurring failure mode: several approaches improve robust accuracy while degrading uncertainty ranking, leading to poorer selective behavior. To address this, we propose Evidential Adversarial Training (EV-AT), which models uncertainty through a Dirichlet distribution and combines (i) an evidence-based loss promoting clean accuracy and reliable uncertainty with (ii) a robust evidence-alignment loss matching clean and adversarial predictions in log Dirichlet-parameter space. Extensive experiments show that EV-AT shifts the Pareto frontier of robustness-uncertainty trade-offs beyond prior state-of-the-art adversarial training methods. Our source code is publicly available at https://github.com/NicolasSournac/Robustness_Meets_Uncertainty.EV-AT.

Criminal identification from surveillance imagery has become a critical research area in intelligent forensic surveillance systems due to the increasing deployment of CCTV cameras in public and private environments. However, surveillance-based face recognition remains highly challenging because of low image resolution, illumination variation, motion blur, pose changes, facial occlusion, and background clutter. To address these limitations, this paper proposes an Attention-Guided EfficientNet (AG-EfficientNet) framework for precise criminal identification in surveillance images. The proposed framework integrates EfficientNet-B0 with Convolutional Block Attention Modules (CBAM) to enhance discriminative facial feature learning under degraded surveillance conditions. In addition, a multi-scale surveillance feature fusion strategy is introduced to preserve both local texture information and high-level semantic identity representations. A hybrid Softmax-Triplet optimization mechanism is further employed to improve inter-class separability and intra-class compactness for robust criminal identity discrimination. The proposed framework was experimentally evaluated using the Labeled Faces in the Wild (LFW) and SCFace datasets. Experimental results demonstrate that the proposed AG-EfficientNet framework achieved superior surveillance recognition performance with an identification accuracy of 98.2%, Precision of 97.9%, Recall of 97.6%, F1-Score of 97.7%, and ROC-AUC of 0.99, outperforming conventional deep learning architectures including AlexNet, VGG16, ResNet50, MobileNetV2, and standard EfficientNet-B0. Furthermore, Grad-CAM visualization and ablation analysis confirm the effectiveness of the proposed attention-guided feature learning strategy.

Simulation is essential for maritime robotics, supporting operator training, mission rehearsal, and human-vehicle interaction in environments where real-world testing is costly or hazardous. Existing simulators focus primarily on autonomy systems and often lack human-in-the-loop interaction and realistic environmental physics. This paper introduces LOTUSim, an open-source, real-time maritime simulator supporting multi-user interaction across aerial, surface, and underwater robotic systems for coordinated naval-style operations. The first contribution of this work is enabling real-time interactive performance for users while ensuring scalability to large fleets operating within a shared interactive simulation environment. Validation demonstrates robust human-in-the-loop performance, maintaining strict real-time execution and high visual fidelity while scaling to large heterogeneous maritime drone swarms. The second contribution is a computationally efficient, Ekman-inspired layered, underwater current model that captures wind-driven, depth-dependent flow dynamics with sufficient physical fidelity for large-scale simulations. Validation against ocean reanalysis data demonstrates substantially improved accuracy compared to commonly used stochastic Gauss-Markov current models. These results confirm LOTUSim's suitability as a simulation platform for operatorin-the-loop maritime robotics research.

Semi-supervised semantic segmentation (SSSS) has long turned on one question, which pseudo-labels to trust, and answered it with ever more careful confidence filtering. Foundation backbones change the regime: with a DINOv2 teacher a strict threshold already retains a measured 98%-clean pseudo-label set, so the accuracy that remains lives not in the filter but in how the embedding space is structured by class. We propose PixCon, a clean-positive pixel-contrastive framework. PixCon maintains a per-class memory bank that admits only labeled pixels the student already classifies correctly, guaranteeing a contamination-free positive set ($ρ_F=0$) by construction, unlike prior contrastive SSSS banks (ReCo, U$^2$PL) built from confidence-filtered pseudo-labels. It is a single branch over a consistency backbone, adds no inference-time parameters, and needs no bank-specific threshold. A first-order analysis of the supervised-InfoNCE gradient explains why contamination hurts: its false-positive term scales as $ρ_F/(1-ρ_F)$, which we measure (0.018 on Pascal, 0.106 on ADE20K) rather than assume. Across Pascal VOC, Cityscapes, and ADE20K, PixCon matches or improves a strong DINOv2-based UniMatch V2 baseline in a compute-matched one-switch protocol: it improves every Pascal-1/8 seed (a per-seed gain of about +0.2 mIoU) and its three-seed mean reaches 87.90, the published UniMatch V2-B figure. Because contamination is already rare under foundation-model teachers, our analysis indicates the $ρ_F=0$ guarantee acts chiefly as robustness as teachers weaken, while the accuracy gain comes from cleaner positive supervision, making clean-positive contrast a robust, low-cost default for foundation-model SSSS.

Reinforcement learning has become a standard post-training recipe for large language models, but dense full-parameter updates create two deployment-relevant bottlenecks: suppressed reasoning performance, often reflected by premature saturation of test-time scaling, and interference when consolidating multiple capabilities through multi-domain training or model merging. We show that the reasoning-effective component of these updates is largely concentrated in the base model's spectral space, motivating Subspace-Aligned Rewiring (SAR), a post-hoc editing method that retains this spectral core while removing orthogonal components. SAR therefore preserves reasoning gains and filters residual update directions that suppress performance or amplify cross-domain interference. Across several model families and scales, SAR extracts compact reasoning cores using as little as approximately 0.58% of total parameters: it preserves over 99% of post-training performance and improves high-k exploration in mathematical reasoning, and generalizes to agentic coding by improving six of seven open benchmarks on an in-house model. SAR also purifies mixed-domain training updates by releasing suppressed coding capability while maintaining math reasoning and instruction following. It further enables model merging across experts, yielding cross-domain generalization that surpasses previous merging baselines and even the best single-domain experts. Overall, SAR shows that extracting reasoning-effective updates from parameter geometry can serve as a training-free mechanism to improve reasoning and multi-domain performance.

The rapid growth in the parameter scale of large language models (LLMs) has created a strong demand for efficient compression techniques. As a hardware-agnostic and highly compatible approach, low-rank compression has been widely adopted to reduce both memory footprint and computational cost. However, existing SVD-based methods are still largely driven by local reconstruction objectives, overlooking two critical limitations: rank budgets are often allocated without explicitly considering layer-wise loss sensitivity, and local approximation errors can propagate and accumulate through the residual stream, leading to amplified global deviations from the original model. To address these issues, we propose LACE-SVD, a Loss-Aware SVD framework with Cumulative Error correction for LLM compression. LACE-SVD first estimates the calibration negative-log-likelihood increase induced by candidate layer-wise compression ratios and solves a budget-constrained allocation problem to assign rank budgets. It then refines the compressed model with closed-form local updates and introduces a propagation-aware correction for residual-stream output modules, reducing layer-output discrepancy as a proxy for cumulative error propagation. Experimental results demonstrate that at a high compression ratio (0.6), the WikiText-2 PPL of our method on LLaMA-7B (32.57) is significantly better than that of Dobi-SVD (46.18).

The paper presents a longitudinal measurement study on the adoption of the pledge and unveil system calls in OpenBSD. These system calls are used to sandbox programs and libraries. Given a dataset covering 19 releases, many programs and libraries were modified to use the system calls already before their introductions in official releases. The adoption rates have also steadily grown; a linear trend provides a coarse but sensible heuristic. Although particularly programs residing in /usr/bin and /usr/sbin have been modified to use the system calls, the sizes of programs and libraries do not correlate well with the amounts of pledge and unveil system calls invoked. Regarding the pledges made, standard input and output operations have frequently been requested, although the full fine-grained arsenal offered by pledge has generally been utilized in OpenBSD. The same observation is seen in that particularly read operations to given paths have frequently been unveiled. All in all, the measurement results indicate that the adoption of system call minimization and sandboxing techniques is not necessarily as troublesome as has often been discussed in the literature.

HiPPO gives recurrent states memory semantics as coefficients of online polynomial projections, but in fixed channel coordinates. Modern selective SSMs, by contrast, rely on token-dependent control and channel interaction. We introduce SHiPPO (Sylvester HiPPO), a transported projection-memory prior that lifts HiPPO coefficient memories into a moving channel frame. For any fixed or realized right-transport path, SHiPPO transports the approximation family and channel metric together; conditional on that path, the state is ordinary HiPPO in a tied moving frame and follows Sylvester coefficient dynamics, preserving the left online-memory operator while adding right-action transport. For selective-SSM execution, we derive a restricted group-local realization with controller-compatible right actions, exponential-adjusted updates, exact block-affine scan, and recurrent decoding. We also give a simultaneous-reducibility criterion identifying when right transports collapse to static mixing plus independent scalar or blockwise banks. Controlled diagnostics show that larger current-token write rank improves ordinary prediction error but cannot recover order-sensitive changes to already-written memory; transported-memory variants recover this signal, which disappears when the transport pathway is removed. A finite-field associative-recall diagnostic with interleaved bindings, operations, and queries provides complementary autoregressive evidence while leaving the preferred right-action realization open. Taken together, these results support SHiPPO as a mechanistically grounded transported-memory prior, with evidence focused on memory mechanisms rather than broad sequence-modeling dominance.

Omni modal large language models (OmniLLMs) have attracted wide attention for their ability to jointly process audio and video, but they generate large token sequences under audio-visual inputs, leading to substantial inference cost. Existing audio-visual token compression methods often rely on unimodal guidance, overlooking the temporal locality of query-relevant evidence in audio-visual inputs and implicitly assuming that the two modalities share a temporally aligned information density distribution. We propose \textbf{OmniFocus}, a training-free query-guided token compression method for OmniLLMs that performs independent importance estimation for video and audio, enabling a modality-symmetric compression design that preserves modality-specific salient evidence while maintaining audio-visual alignment, thereby mitigating the modality bias issue that can arise from unimodal-guided compression. Experiments on the Qwen2.5-Omni model family across four audio-visual benchmarks show that OmniFocus maintains strong compressed performance at low token retention ratios and outperforms existing baselines on several major benchmark scores at 25\% token retention. On DailyOmni with Qwen2.5-Omni-7B at 25\% token retention, OmniFocus maintains 59.40 accuracy while delivering up to 1.38$\times$ prefill speedup relative to the full-token baseline, highlighting a favorable practical accuracy-efficiency trade-off.

Fast estimation of the size of the largest overlap between tables enables blocking and query-by-table retrieval in large table repositories. The first and the state-of-the-art estimator Armadillo improves efficiency by embedding each table independently and approximating overlap ratio via embedding similarity. However, accurate estimation in heterogeneous repositories remains limited by three challenges: (C1) overlap depends on row-column structure, i.e., each matched cell must preserve both its row and column membership under a joint alignment of the two tables, but existing encodings leave this structure to be inferred indirectly; (C2) independent encoding provides no explicit channel for inter-table alignment signals, biasing prediction toward global similarity; (C3) naive value encodings overfit to corpus-specific distributions, causing cross-domain degradation. Hence, we propose ALORE, a scalable and domain-robust overlap ratio estimator built on three principles: (P1) explicitly represent row-column structure; (P2) expose inter-table alignment signals during training without expensive alignment search; (P3) reduce sensitivity to corpus-specific value distributions. ALORE instantiates these principles with a Two-View Row-Column Hypergraph encoder, alignment-guided objectives with inexpensive interaction signals, and a domain-robust value mapping. Experiments on multiple datasets spanning diverse domains and scales, including a large real-world corpus beyond prior benchmarks, show that ALORE outperforms the state of the art. ALORE reduces MAE by up to 55% overall and 69% in zero-shot transfer, while achieving up to 89x speedup. We further validate its effectiveness for query-by-table retrieval.

Agent skills, reusable instruction artefacts supplied to a tool-using language model, are increasingly optimised by shortening, structural rewriting, stronger-model compilation, and scoped loading, on the assumption that a smaller or better-organised skill lowers cost while preserving success. That assumption is rarely tested with quality and real monetary cost measured on the same runs and the contributing factors separated. This study reports a controlled decomposition over ten skill-delivery conditions, 40 software-engineering tasks, and three repetitions per cell (1,200 rollouts), separating no-skill execution, raw skills, deterministic shortening, linear and structured rendering from a shared semantic ledger, scoped loading, and the compiler and executor model tiers. Quality is the verifier pass rate at task level; cost is solve-stage token cost at standard provider prices, with a token-volume-normalised view for robustness and compilation cost amortised separately. The task is the unit of inference, intervals are task-clustered, and the contrast family is multiplicity-controlled. Deterministic shortening is close to the raw baseline but does not establish non-inferiority within the preset margin. Structured rendering and scoped loading lower pass rate on the compact executor without lowering cost, and structured rendering is indistinguishable from linear text at matched content. The only contrast surviving correction is executor capability, which raises pass rate by 27 percentage points at roughly five times the real cost, with compiler tier showing no robust effect. Under real prices no optimised representation reaches a practical break-even. The evidence indicates that executor capability is the dominant lever and that no representation strategy improves over the raw skill on either executor tier.

Neural networks are increasingly used to infer hidden physical structure from dynamical observations, yet it remains unclear whether their out-of-distribution performance reflects transferable physical rule learning. We address this question in a controlled inverse problem: reconstructing interaction graphs of a kinetic Ising model from Glauber magnetization trajectories. Across convolutional, graph, Transformer, and hybrid architectures, we find that data-driven training produces distinct and reproducible statistical strategies under topology and temperature shifts. Edge-population diagnostics reveal that Transformer-based models tend to preserve the link density of the training ensemble, whereas convolutional models can collapse toward sparse- or no-link predictions that appear out-of-distribution stable by exploiting the majority no-link class. Thus, high in-distribution accuracy and apparent out-of-distribution robustness do not necessarily imply a learned dynamics-to-structure rule. Instead, neural reconstruction can be governed by architecture-dependent statistical priors. Our results identify a concrete failure mode of standard data-driven learning in physical inverse problems and motivate rule-guided principles for machine-learning-assisted scientific discovery.

The use of Large Language Models (LLMs) across diverse areas of human activity-ranging from everyday tasks to safety-critical applications-aims to enhance decision-making effectiveness with minimal human feedback. Concurrently, it seeks to align decisions with human expectations, preferences, and needs while mitigating risks associated with AI non-determinism. However, humans frequently over- or under-rely on AI recommendations, and current AI systems remain poorly calibrated to human expectations. To address these challenges, we introduce a human-AI collaborative decision-making framework designed to augment human capabilities and align AI agents with human preferences and expectations. Specifically, this paper (a) formulates the collaborative decision-making task as a stochastic game between an AI agent and a human player, and (b) proposes the Human-Centric Reflective Architecture (HCRA), which integrates human-calibrated models with reinforcement learning agents that leverage linguistic feedback in an iterative, reflective process. Evaluation results demonstrate that HCRA enhances decision-making effectiveness and delivers high-quality recommendations.

Forecasting future events has attracted growing attention as a testbed for general-purpose AI. A natural way to ground this evaluation is let the models trade in the prediction markets. Trading, however, requires more than forecasting. Moreover, recent benchmarks report a substantial gap between calibrated probability scores and the trading results. We propose Raven-Agent, to the best of our knowledge, the first autonomous trading agent for prediction markets. On a controlled replay over an archived decision set, our architecture achieves the only positive return and the only positive risk-adjusted return among all tested policies. We have released our code in https://github.com/Alchemist-X/predict-raven .

Large language models (LLMs) have significantly lowered the technical barrier to software plagiarism. By transforming existing source code while preserving its functionality, modern LLMs can generate semantically identical program that may evade traditional plagiarism detection techniques. Among such attacks, code paraphrasing modifies the syntax and structure of a program while preserving its behavior. This paper investigates whether software birthmarks can detect such LLM-assisted plagiarism. As a starting point, we employ k-gram software birthmarks based on unique k-grams of Java opcodes, with k ranging from 1 to 6. We employ three contemporary LLMs: ChatGPT-5.1-Codex-Mini, DeepSeek-V4-Flash, and Claude-Haiku-4.5. The dataset consists of individually compilable source files extracted from actively maintained BSD-2-Clause licensed Java projects. We further compare five similarity measures for birthmark matching: cosine similarity, Dice index, Jaccard coefficient, Simpson index, and edit-distance-based similarity. The results demonstrate that k-gram software birthmarks remain effective for detecting LLM-assisted plagiarism. Among the evaluated models, ChatGPT-5.1-Codex-Mini generated the most difficult-to-detect clones. Furthermore, the findings confirm the higher performance of coding-oriented models for plagiarism task.

Images captured by consumer electronic devices, such as mobile phones and digital cameras, often suffer from low-light degradation due to sensor limitations and imaging pipelines, which degrades visual quality and affects downstream vision tasks. Existing methods based on Convolutional Neural Networks (CNNs) and Transformers have dominated current low-light image enhancement (LIE) due to their excellent ability to model hierarchical features. However, CNNs operate in local receptive fields that cannot model long-range dependencies, while Transformers overcome this problem but incur substantial computational costs. To address these challenges, we propose MambaLIE, a Scene Light Intensity-Boosted Low-Light Image Enhancement method based on a State Space Model (SSM). We first introduce scene light intensity to improve the structural distribution of illumination, which is then gated with the low-light input to guide enhancement. To better model the illumination while maintaining computational efficiency, we propose the Locally Enhanced State Space Model (LESSM) for efficient light enhancement. Our LESSM contains two branches: an SSM branch and a Local Enhanced branch, where the former is used to model the long-range dependencies with linear time complexity, while the latter is used to enhance local feature representations. Extensive experiments demonstrate that MambaLIE outperforms state-of-the-art CNN-based and Transformer-based LIE methods on four widely used synthetic benchmarks and five publicly available real-world benchmarks in terms of accuracy, speed, and model size, making it suitable for practical deployment on resource-constrained devices.

Video Diffusion Transformers (VDiTs) have demonstrated significant capabilities in high-fidelity video generation. However, their ability to produce long-duration videos is fundamentally constrained by the quadratic complexity of the self-attention mechanism. Recent clustering-based sparse attention methods improve the quality-speed trade-off by grouping semantically similar tokens, but their practical efficiency remains limited by two bottlenecks: substantial clustering overhead and low CTA utilization caused by irregular cluster-induced blocks. We propose HyperVAttention (HVA), a training-free sparse attention framework that addresses both bottlenecks jointly. To reduce clustering overhead, we introduce 3D local-window clustering, which exploits the spatio-temporal locality of video tokens to restrict centroid search to fixed local neighborhoods, and implement it with a custom Triton kernel for efficient execution. We further propose a hybrid clustering strategy that performs full clustering only at anchor steps and updates only subset tokens at intermediate steps, leveraging the temporal stability of cluster assignments across denoising steps. To improve CTA utilization, we present hardware-aware cluster merging that minimizes CTA-aligned execution cost through parallel agglomerative merging, improving block density and approximation fidelity by utilizing idle tile capacity. Together, these components reduce clustering overhead, avoid redundant updates, and better align sparse attention with the fixed tile structure of modern GPU kernels. Experiments on Text-to-Video generation show that HVA establishes a new Pareto frontier for training-free sparse attention in video diffusion, reducing end-to-end latency by up to $2.13\times$ while improving fidelity over existing training-free sparse attention baselines.

Model merging techniques, which aggregate independently finetuned models into one to combine their capabilities, have become a topic of significant interest in recent years, with a broad array of methods having been proposed to tackle this problem. Simultaneously, an emerging trend in distributed learning has been the use of methods such as local SGD and DiLoCo, which greatly reduce communication costs by periodically aggregating the independently trained local models. However, these communication-efficient methods have been shown to degrade in performance relative to the FLOP-matched data-parallel gold standard as the number of independent local models grows and as the number of local training steps before global communication is increased. In this work, we draw an explicit analogy between the pseudo-gradient aggregation step in local SGD/DiLoCo and task arithmetic-based model merging, establishing a straightforward way to utilize merging methods in the context of distributed optimization. We then evaluate multiple state-of-the-art model merging methods in this setting and identify one method in particular, Iso-C, as a promising approach for improving DiLoCo. We find that DiLoCo SGD with Iso-C aggregation outperforms not only simple pseudo-gradient averaging but even the momentum-based DiLoCo, despite lacking a momentum mechanism itself. Building on this finding, we propose IsoLoCo, which adapts Iso-C for distributed training by equipping it with Nesterov momentum. Our empirical evaluations on language model pre-training across varying numbers of local workers show that IsoLoCo significantly outperforms DiLoCo, with the gap between them widening as the number of workers increases. This advantage remains present across model sizes and inner step counts, confirming that merging-inspired aggregation is an effective strategy for low-communication distributed training.

Background: Foundation models (FMs) trained on large-scale unlabeled physiological data have emerged as a promising paradigm for medical artificial intelligence. Their ability to capture clinically meaningful, transferable representations for rare diseases remains largely unproven. This study investigates whether FM pre-training provides genuine clinical generalization benefits beyond improved optimization for rare electrocardiographic (ECG) phenotypes. Methods: We systematically evaluated nine publicly available ECG FMs for Brugada syndrome detection on the BrSwiss cohort (294 patients, 87 cases) and the independent external HUCA cohort (363 patients, 76 cases), under three strategies (from-scratch training, linear probing, full fine-tuning) across several configurations, including a 3% data ablation and zero-shot cross-site transfers. Results: Pre-training was necessary for high-capacity architectures unable to converge from scratch (AUC gain up to 0.411, p < 0.05), but gave no significant gain for compact architectures already converged on labeled data alone. On full BrSwiss, the best fine-tuned FM (ECG-CPC, AUC = 0.962) only marginally exceeded the strongest supervised baseline (ECG-CPC from scratch, AUC = 0.932; p = 0.091). At matched training-set size, the data-efficiency advantage on BrSwiss-3% (AUC gain = 0.055, p < 0.01) did not replicate on HUCA. Under zero-shot cross-site transfer, FM-based pipelines did not generalize better than supervised baselines, all approaching chance-level performance. Conclusion: For Brugada syndrome detection, FM pre-training is mechanical rather than semantic, providing optimization stability rather than transferable clinical knowledge. These findings challenge the assumption that large-scale pre-training inherently encodes clinically meaningful representations, highlighting the central role of model architecture and data-domain alignment.

Recent advances in molecular large language models have led to strong performance on molecular understanding and generation tasks, yet these gains often come without reliable structural grounding. In particular, existing approaches conflict with the chemistry principle that structure determines function: despite their downstream success, current molecular LLMs perform poorly on basic structure recognition, suggesting that they fail to capture molecular graphs from canonical SMILES. To remedy this, we propose MolBasic, a structure-first framework that strengthens structural comprehension via SMILES-Graph translation. MolBasic is built around a multi-level structure perception benchmark, where bidirectional SMILES-Graph conversion serves as the core task to align sequential and topological representations. On top of this foundation, we employ a progressive learning scheme with a standardized Chain-of-Thought (CoT) to steer models from structure acquisition toward higher-level molecular reasoning. Experiments show that MolBasic substantially improves structural understanding and yields robust gains on downstream tasks, including property prediction and objective optimization, supporting our structure-first paradigm.

Text-rich image models can now design poster-scale layouts, but we lack ways to measure whether they honor scientific communication contracts: legible labels, prescribed aspect ratios, and -- above all -- abstaining from fabricated scientific figures. We present POSTERHARNESS, an auditable harness reframing poster generation as measurable instruction-following tasks, with a pilot benchmark and failure taxonomy. POSTERHARNESS uses a placeholder-first contract to separate two jobs models otherwise conflate. The model performs visual-summary design: typography, reading path, color, and background -- but never draws data-bearing figures. Every figure region must be an empty labeled placeholder; a deterministic compositor inserts real source-paper figures at detected coordinates. This makes properties measurable: placeholder count and ID accuracy, blankness, aspect-ratio compliance, abstention from synthesized graphics, public-text hygiene, and source-figure provenance -- with failures logged as explicit rejections, not hidden in plausible-looking output. We instantiate the harness on 12 papers (6 HEP, 6 AI/ML-adjacent) and report three findings. (i) A counterfactual probe shows the placeholder contract drives VLM-counted synthesized figures from 34 to 0 across three papers. (ii) A failure taxonomy identifies blocking contracts: placeholder geometry, placeholder QA, template critic, and public text. (iii) Comparison with Paper2Poster shows a trade-off: PosterHarness yields higher-resolution artifacts, lower white-canvas fraction, and stronger VLM visual preference; the deterministic baseline retains slightly more PosterQuiz-style information and runs faster. We report this as regime characterization, not a superiority claim. All artifacts, prompts, manifests, and audit scripts are released as a reusable evaluation component.

In high-dimensional Ising model estimation, target sample sizes are often limited, and effectively using auxiliary binary datasets of unknown relevance remains challenging. To address this, we propose Trans-Ising, a transfer learning method that combines a loss-based source screening rule with a two-stage estimation procedure. The method first identifies informative auxiliary sources using held-out target pseudolikelihood to prevent negative transfer. It then computes an initial estimator via pooled nodewise $\ell_1$-regularized logistic regression, followed by a target-only correction step using a folded-concave penalty. Theoretically, we establish fixed-node $\ell_2$ and $\ell_1$ error bounds, exact graph selection consistency, and the conditional consistency of the screening rule. Through extensive simulations and real-data analyses, we demonstrate that Trans-Ising achieves lower estimation errors than both target-only estimation and naive data pooling.

Social media posts are a rich and valuable source of data for analyzing mental health states and users' well-being using automated analysis tools. In this work, we demonstrate how we used a range of Natural Language Processing (NLP) methods, including Long Short-Term Memory (LSTM), BERT-based models, and Large Language Models (LLMs), for self-state and well-being analysis and summarization during the CLPsych Shared Task 2026. Our approach achieved one of the top Consistency and Contradiction scores for the summarization task and also middle-level results for the other tasks. By testing and developing such mental health-state estimation systems, we contributed to improving mental health support systems. We make our code available https://github.com/psytechlab/CLPsych2026/.

We present CAFÉ, a learning platform designed to introduce computer science students to Formal Methods (FM). CAFÉ aims to scaffold students' structural thinking (in contrast with operational thinking) by promoting the practice of Graphical Loop Invariant Based Programming (GLIBP). In the GLIBP approach, students solve loop-based problems by first constructing a Graphical Loop Invariant (GLI) before deriving the corresponding code. The GLI is an informal diagrammatic representation of the loop invariant. It illustrates the variables involved in the loop, their properties, and the relationships between them. To enable automated feedback, students complete a blank GLI, a box-based version of the GLI. Beyond evaluating the code students submit, CAFÉ provides personalized feedback on students' GLI and its alignment with the code. In this demo, we walk through CAFE from both a student's and a teacher's perspective. We show how the tool supports GLI design and provides feedback. We then present how a teacher can encode programming challenges together with a corresponding blank GLI.

Controllable generative models of 3D medical images can synthesize volumes with specified clinical attributes, but this demands samples that are simultaneously high-fidelity, natively 3D, and faithful to the requested conditioning. We present CONFLUX, a latent diffusion model for chest computed tomography (CT): a 3D variational autoencoder compresses each volume, and a rectified-flow transformer generates in the latent space. Generation is conditioned on structured radiological metadata (18 abnormality findings, sex, age, and reconstruction kernel) through adaptive layer normalization. The model leads strong volumetric baselines on tri-planar Frechet distance (FID 32.3 vs. 74.6 for MAISI) while exposing direct control over clinical attributes. To strengthen that control we add an online reinforcement-learning post-training stage (group-relative policy optimization) that rewards how reliably a classifier recovers the requested findings from each generated volume. Judged by a separate, independent classifier, post-training removes 47% of the shortfall relative to real-scan reliability. We release the model and a ~200k synthetic chest-CT dataset with conditioning metadata spanning a wide variety of clinical findings.

Vision-language models can exhibit visual concept-conditioned divergence: given images containing demographic features, corporate logos, or ideological symbols, some models produce unusually uniform responses that differ from what peer models say about the same input. These behaviors evade text-only audits because visual concepts cannot be isolated or substituted the way text tokens can. We present VISTA (Visual Inconsistency Screening Through Analysis), a black-box cross-model audit that couples semantic entropy with distribution-based divergence to flag model-specific anomalies. In a controlled study, we implant concept-conditioned stances in three VLMs via fine-tuning on small biased datasets and confirm that VISTA detects them. Auditing six VLMs across 19 topics, VISTA surfaces 142 high-suspicion cases (1.2%) and identifies selective refusal as a previously unreported divergence pattern, where models refuse demographic queries at rates varying from 0 to 65% across groups.

We propose MABLE (Masked Autoencoding with Bi-Lipschitz Decoding for Embeddings and Graph Metric Learning), a self-supervised framework for learning node and graph embeddings from large, heterogeneous graphs, demonstrated here on geospatial mineral-exploration data. MABLE combines masked reconstruction with fixed cosine-similarity losses that align matched augmented views while keeping unpaired embeddings well spread. A bi-Lipschitz feature decoder ties a low-dimensional reconstruction component of each node embedding to feature similarity, while matched-node consistency shapes the remaining context used by graph pooling. Lipschitz-controlled pooling helps stabilize graph-level representations under perturbations of retained node embeddings, while augmentation alignment trains robustness to masking, node dropping, and sampling variation. Across local copper and regional Arabian Shield studies, MABLE embeddings provide complementary downstream signal and produce coherent embedding-derived layers for hypothesis generation without learned discriminators or hard-negative selection.

Recent reasoning-centric Large Language Models (LLMs) have made significant strides, yet they predominantly operate on a passive-inference pattern that assumes complete information. In contrast, real-world clinical intelligence is inherently an iterative investigative process requiring strategic evidence acquisition. To bridge this gap, we formalize medical diagnosis as an Iterative Evidence-Seeking Task. We leverage Reinforcement Learning with Verifiable Rewards (RLVR) to elicit intrinsic reasoning within a closed-loop environment, guided by a novel suite of rewards that enforce diagnostic precision and examination consistency. To facilitate this, we introduce the Retrieval-Augmented Generation-based Examination Simulator (RAGES), a high-fidelity clinical oracle that provides realistic, knowledge-grounded follow-up evidence. Empirical results across diverse datasets demonstrate that our framework enables LLMs to transition from passive responders to autonomous assistants. Notably, our model demonstrates comparable performance to larger and reasoning-enhanced baselines, while RAGES proves superior to vanilla LLMs in generating biologically plausible clinical feedback.

Recent advancements in the Internet of Things (IoT) emphasize the urgent need for advanced network security, as IoT networks feature dynamic topologies, imbalanced traffic, and complex attack patterns. Unlike general IT networks, IoT environments exhibit extreme heterogeneity and sparse topologies. Traditional GNN-based intrusion detection methods often struggle to efficiently model node and edge features or capture fine-grained anomalies in such settings. To address this, we propose SKGFusionKAN, a novel IoT-tailored approach enhancing GraphSAGE with a multi-scale selective kernel attention mechanism. This enables adaptive extraction of node and edge features under diverse traffic conditions. Specifically, our edge-oriented message passing strengthens information propagation, while selective kernel attention adaptively weights edge-derived information from different scales to handle heterogeneity. We also introduce a gated fusion process to dynamically integrate multi-scale features, improving robustness against evolving attacks. Finally, we leverage Kolmogorov-Arnold Networks (KAN) for classification, offering superior nonlinear modeling capabilities essential for detecting intricate, low-frequency attacks. To our knowledge, this work presents a comprehensive integration of GNNs and KAN with dedicated architectural innovations for IoT intrusion detection. Extensive experiments on four NIDS benchmarks show that SKGFusionKAN consistently outperforms state-of-the-art approaches in binary and multiclass tasks, demonstrating its potential for IoT security.

Scaling modern large language models (LLMs) to long contexts is limited by the quadratic computation cost, and poor length extrapolation of dense attention. Chunk-wise sparse attention offers a promising alternative, but all existing methods fall short of full attention because of their inaccurate chunk selection. We propose Hierarchical Landmark Sparse (HiLS) Attention, a chunk-wise sparse attention mechanism that learns chunk selection end-to-end under the language-modeling (LM) loss. HiLS factorizes attention hierarchically: each query performs attention independently with each retrieved chunk to extract chunk-specific information, and the resulting outputs are fused according to chunk retrieval scores. By incorporating retrieval scores into the forward attention computation, HiLS optimizes them directly with the LM loss, enabling end-to-end retrieval learning and native sparse training. Experimental results show that HiLS-Attention achieves performance comparable to, and in some cases better than, full attention at in-domain context lengths. Meanwhile, HiLS-Attention extrapolates more than $64\times$ the training context length with 90% retrieval accuracy, far beyond full attention. Moreover, existing full-attention models can be converted to HiLS-Attention with lightweight continued pretraining, preserving in-domain performance while acquiring ultra-long-context extrapolation. Together with its sparse KV access and computation, HiLS-Attention breaks the usual efficiency-performance trade-off, enabling long-context LLMs that are both more efficient and more effective on general long-context tasks than their full-attention counterparts.

Effective agency in social environments depends on when an agent seeks knowledge, when it acts, and whether its actions are justified by acquired information. Existing grounded benchmarks provide executable actions, persistent state, and verifiable outcomes, while social simulation environments provide rich interaction among language agents. We study an evaluation setting that combines these requirements. We define socially distributed task environments as interactive environments where task-relevant knowledge is partitioned across role-isolated participants and consequential actions are accessible only through them. Communication serves as exploration over role-partitioned knowledge, while grounded action serves as exploitation over environment state. We introduce Incognita, a Concordia-based framework that separates social interaction from grounded execution. The evaluated agent routes messages to a user or specialist entities; specialists mediate admissible operations; a deterministic sub-environment executes accepted operations over a canonical state; and an offline evaluator scores outcomes with inherited rewards. Incognita-Retail transforms tau-bench retail into a multi-entity environment while preserving final-state reward semantics. We evaluate three generative agent models on 18 tasks stratified by social breadth, with 540 trials. Progress appears in reward and behavior: success rises from 0 percent to 8.9 percent and 17.2 percent, while premature finalization falls from 100 percent to 87 percent and 58 percent. Stronger models elicit more hidden knowledge, contact more entities, and attempt more grounded writes, yet reliability remains low. These findings show that socially distributed task environments expose behavior before reliable success, including knowledge elicitation, source selection, grounded action attempts, and premature completion belief.

Cloud data centers rely on virtualization technologies to serve AI workloads in multi-tenant environments. With the growing scale of data-intensive AI workloads, the performance of storage I/O paths at the virtualization layer has become a critical factor. A single user request often crosses multiple IP blocks, where functional units such as storage, GPU, and accelerator devices under virtualization fan out into separate stack traversals between the guest and the backend. As a result, round-trip and context-switching overheads accumulate with the number of devices. In this letter, we identify that a dominant factor in this overhead lies not in the kernel-mediated I/O path alone, but in the per-device submission structure itself, which persists even in user-space, kernel-bypass storage frameworks such as SPDK. To address this, we propose cross-IP request coalescing, which relocates the fan-out point from the guest to the SPDK vhost-user backend. The guest submits multi-device I/O as a single compound request, and the accelerated bdev at the backend decomposes and dispatches it to each target device, replacing multiple per-device guest-backend round trips with a single submission. Evaluation in an SPDK-based virtualized environment shows that the proposed approach achieves up to 1.78x lower latency than the per-device baseline, with the benefit growing as concurrency increases.

Large language models remain limited as continual learning systems, motivating renewed interest in Sparse Distributed Memory (SDM) as an explicit online episodic memory. CALM (Nechesov and Ruponen, 2025) identifies its threshold-binary encoder as an open design question. This paper evaluates rank-order N-of-M encoding (Furber et al., 2007) as an alternative. We make three contributions. First, a faithful reimplementation validates the published architecture by confirming exact equivalence between WheelSDM and RankOrderSDM (cosine similarity 1.0000 across 10 seeds) and reproducing the documented divergence of RDLIF neurons under interference. Second, multi-seed capacity experiments show RankOrderSDM outperforming StandardSDM by 13.4 percentage points at saturation in the scaled configuration and by 0.8 percentage points at the published architecture scale. Third, BER robustness experiments disentangle representation and learning effects, showing that the large robustness gain arises primarily from the interaction of rank-order encoding with MAX-Hebbian learning, while the encoder alone provides only a small advantage under matched learning conditions. Experiments on GloVe-100 embeddings confirm this small but consistent encoding benefit on real structured data, whereas sentence embeddings exhibit a ceiling effect at low memory load. A secondary analysis shows that idealized rank-order encoding requires half the component-level encoding energy of SpikingMamba's SI-LIF neurons at four-bit precision, although decoder costs dominate overall system energy. These results identify which components of the original rank-order SDM architecture provide measurable benefits for contemporary memory-augmented AI systems, offering practical guidance for architectures such as CALM.

Cross-lingual retrieval-augmented generation (RAG) is often deployed in an English-evidence regime, where users query in diverse languages but retrieved passages remain English. In this setting, generation can fail despite strong base models: English evidence induces language drift (English or code-switching outputs) and models use evidence unreliably when producing non-English answers. We attribute these failures to two post-training challenges: (i) errors are prefix-dependent, so fixed-trajectory supervision suffers from prefix mismatch; and (ii) sequence-level (partly discrete / judge-based) rewards yield noisy credit assignment and high-variance updates. We propose TR-RAG, a teacher-regularized RL recipe that couples reward optimization with on-policy distillation on student-visited prefixes. A compact student samples on-policy answers, while a stronger frozen teacher is queried only on those prefixes and provides a prefix-wise student-to-teacher reverse-KL anchor. We further introduce a reward decomposition for English-evidence multilingual generation, combining language consistency, character 3-gram recall, and an LLM-judge score for evidence-grounded correctness. Across three benchmarks -- BioASQ-ENKB5, Hotpot-ENKB5, and naturally multilingual MKQA -- and two backbones, TR-RAG improves the composite of language adherence and evidence-grounded correctness over strong baselines. Crucially, the teacher anchor acts as a safety net: on in-domain languages it prevents the large language-consistency collapses (up to ~27 percentage points) that reward-only RL can suffer by drifting below even the base model, while on distant out-of-distribution languages -- where reward-only RL stalls at the base model's ceiling -- it still improves evidence grounding; and on character 3-gram recall the compact student sometimes surpasses its 70B teacher.

A central goal of mechanistic interpretability is to understand how neural networks work and what each individual component does. Dominant circuit-finding approaches focus on a specific behavior and reverse-engineer the role of components on the associated sub-distribution. However, past work has shown that components can have different functions that are active on different subsets of the input distribution. In this work we ask whether a single weight can be understood globally across the full training distribution by characterizing when it matters (the inputs on which ablating it changes the model's predictions). We introduce an automated LLM pipeline that writes a short, human-readable description of when a weight matters and verifies it on held-out text, crediting a weight only if its description generalizes. Across two sparse and two dense transformers, the fraction of weights that are interpretable (in this sense) is higher in sparse transformers than in dense ones, a gap that widens once unreliable descriptions are discarded. Our results show that a meaningful fraction of a sparse transformer model's weights can be interpreted: 12 to 31% of weights have a single short description that identifies what the weight is used for.

Dense video captioning aims to generate temporally grounded descriptions of video events, benefiting both event-level video understanding and generation. In this domain, autoregressive video large language models have emerged as a prevalent paradigm due to their strong generative and cross-modal modeling capacity. However, generating dense captions under the token-by-token paradigm severely limits inference efficiency and hinders scalability as video length and event density increase. In this work, we propose a parallelized autoregressive framework that not only improves generation efficiency but also enhances temporally grounded captioning performance. Our key insight is to exploit the weak local dependencies across temporally distinct events to restructure the causal dependency graph, thereby enabling lossless parallel generation. Specifically, tokens with weak cross-event dependencies can be decoded in parallel, while tightly coupled tokens within each event retain sequential decoding to preserve local semantic coherence. To realize this insight, we introduce two key components for lossless parallel decoding: (1) a latent global planning mechanism that automatically learns the event-level structure and produces compact tokens encoding global inter-event causality while adaptively aggregating event-level audio-visual semantics, guiding subsequent dependency restructuring and parallel decoding; and (2) an event-factorized parallel decoding mechanism that effectively balances local focus with global inter-event awareness. Experiments on various benchmarks demonstrate the clear advantage of our approach in both efficiency and performance in omni-modal event grounding and captioning. Project website: https://github.com/showlab/PadCaptioner.

We introduce VSeek, an agentic framework that transforms long-video question answering (LVQA) from a passive, single-pass perception task into a multi-turn retrieval process. VSeek utilizes a natural language-driven search to identify relevant context within long videos and is post-trained with reinforcement learning (RL) to jointly formulate targeted search queries and reason over retrieved clips for LVQA. While RL post-training has revolutionized reasoning in symbolic domains such as mathematics and code, its application to long-video understanding remains hindered by a lack of verified rewards. To ensure that the retrieved context is relevant, we propose a novel neuro-symbolic approach that bridges open-ended natural language with discrete visual verification. Specifically, complex user queries are compiled into formal temporal logic specifications for systematically decomposing natural language questions into a definitive checklist of required atomic visual primitives, such as key objects and activities, along with their temporal ordering. These systematically derived grounding events provide the critical feedback signal for RL post-training, enabling dense, verifiable rewards based on the successful retrieval of these specific visual elements rather than relying entirely on outcome-only answer accuracy. By explicitly optimizing for this verifiable evidence-seeking behavior, VSeek improves Pass@1 scores by up to 8% and Pass@4 scores by 15% on long-video understanding benchmarks compared to base models. We open-source our code at https://utaustin-swarmlab.github.io/VSeek.

Multilingual documents encapsulate rich regional cultures, scientific discoveries, and historical records. Parsing this content into structured, machine-readable formats is critical for unlocking global knowledge. However, existing benchmarks predominantly focus on high-resource languages like English and Chinese, creating an evaluation blind spot concerning model performance on other languages. While recent Vision-Language Models (VLMs) claim support for hundreds of languages, the lack of ground truth makes it impossible to empirically verify these capabilities. To bridge this gap, we introduce MORE, a large-scale benchmark designed for multilingual document parsing evaluation. MORE distinguishes itself through three key dimensions: (1) Unprecedented Scale: It covers 149 languages, making it the most linguistically diverse benchmark to date; (2) Structural Complexity: Unlike previous works, it extends evaluation beyond plain text to include structural elements such as code blocks, tables, and catalogs; and (3) Data Authenticity: All samples are curated from real-world documents via a model-assisted, human-refined annotation pipeline. We evaluate state-of-the-art models using MORE, establishing new performance baselines for long-tail languages and validating the benchmark's effectiveness in diagnosing model capabilities in realistic, diverse scenarios. The MORE dataset will be available at https://github.com/zimoqingfeng/MORE.

We argue that AI systems used in conducting foreign policy tasks - broadly enacting 'statecraft' - should be a priority test case for technical AI governance research. In enacting foreign policy, we refer to the formulation and implementation of external objectives by political actors. Statecraft is a high-consequence deployment domain, with extreme downside risks and structural properties that standard evaluation practices handle poorly. These features include partial observability, unbounded action spaces, contested ground truth, and multidimensional objectives. This paper advocates for a literature-grounded research agenda. Our contribution is threefold: (i) a claim about the structural conditions of foreign policy that combine catastrophic tail risk with technical evaluation complexities, (ii) an ECOSYSTEM review that highlights the asymmetric focus on ASSESSMENT features over ACCESS, VERIFICATION, SECURITY, and OPERATIONALIZATION, and (iii) a demand-side evaluation framework that decomposes foreign-policy workflows into bounded, evaluable sub-tasks with human recombination. As AI systems are already being deployed in the conduct of war and peace, amid limited public evaluation infrastructure from the technical AI governance community, this agenda is an urgent priority.

Most artificial life simulators either operate on abstract substrates disconnected from physical reality, or simulate physically grounded worlds that do not scale to the population sizes required for open-ended evolution. We present Microcosmos, a simulation engine in which artificial lifeforms are modeled as elastic filament chains inhabiting a two-dimensional viscous fluid world, designed from the ground up for modern GPU hardware and end-to-end differentiable simulation. We validate the engine through four experiments. Hand-designed locomotion strategies confirm that the fluid coupling respects known physical constraints. Gradient-based optimization of filament folding demonstrates both the full differentiability of the simulator and the expressivity of the filament encodings. Neuroevolution and quality-diversity search produce a wide range of swimming and chemotaxis behaviors automatically. Linear scaling with particle count confirms the engine supports large-scale simulation. Microcosmos is released as an open platform with the long-term goal of supporting large-scale open-ended evolutionary simulations, designed to be physically plausible and computationally scalable.

Convolutional Neural Networks (CNNs) achieve strong denoising performance by exploiting spatial context from neighboring pixels. Deep Image Prior (DIP) leverages this property to restore images from a single noisy input without requiring large datasets. However, the over-parameterized architecture of DIP often leads to noise fitting during optimization. In this paper, we propose Pool-DIP, a convolution-free architecture that incorporates pooling-based contrast modeling to capture spatial context efficiently. Pool-DIP improves denoising performance while significantly reducing the number of parameters and computational complexity compared to convolution-based DIP models. Experimental results show that Pool-DIP achieves competitive performance across multiple datasets, including a real-world benchmark. Spectral analysis further reveals that Pool-DIP stabilizes the evolution of high-frequency components during optimization and suppresses erroneous high-frequency signals. The proposed architecture also generalizes well to other image restoration tasks such as super-resolution and inpainting.

LLMs have made substantial progress on automated code generation from natural-language descriptions of desired behavior (intent). Most existing methods improve generated programs through execution-guided code refinement: they generate a candidate solution, execute it, and patch the implementation using feedback, while leaving the underlying specification unchanged. This workflow implicitly assumes that the LLM's understanding of the intent is already correct and complete. In practice, however, intents are often ambiguous or underspecified. As a result, even a capable model may produce a correct implementation of the wrong intent, making specification mismatch a central bottleneck. This paper presents BeSpec, a behavioral model-based approach to specification alignment. BeSpec treats the task description as partial evidence about the intended behavior of the correct program. It first builds an explicit behavioral model, which are checkable properties that valid outputs must satisfy. BeSpec then generates candidate programs, executes them on probe inputs, and compares their observed behavior with the predicted behaviors. When observed behavior does not match the predicted behaviors, BeSpec either refines the specification or rejects the candidate program. We evaluate BeSpec with three LLMs on four benchmarks: CodeContests, xCodeEval, APPS, and the contamination-free LiveCodeBench. Against nine baselines, BeSpec achieves the highest Pass@1 and average pass rate across all settings, improving average Pass@1 over the strongest baseline by 8.1%--25.3% relative across the three LLMs. A failure analysis shows that after alignment, most remaining errors stem from algorithmic difficulty rather than misunderstood specifications, while ablation studies confirm that each major component of BeSpec contributes positively.

Public official-information request records contain process signals. They can support research, workflow review, and human-supervised agent help. Yet they also mix observed correspondence, platform states, inferred events, and legal outcomes. FOI-O is a reusable process-modelling method and verification infrastructure for Freedom of Information administration. FOI-O NZ, based on the New Zealand Official Information Act, is the only implemented and validated jurisdictional profile in the current repository. Broader reuse is a design intent and future validation path, not an empirical result of this package. FOI-O models the request record first. Request profiles, observed correspondence events, controlled vocabularies, and provenance make visible what was seen and how it was changed. It then adds review queues, release metadata, bounded agent contracts, semantic assets, process-model artefacts, and fixture-only process-mining interchange examples. Human certification of legally meaningful outcomes stays outside autonomous tooling. The repository provides JavaScript Object Notation (JSON) Schema contracts, Python data models, Simple Knowledge Organization System (SKOS) vocabularies, Web Ontology Language (OWL), Resource Description Framework (RDF), and Shapes Constraint Language (SHACL) assets. These are supported by deterministic examples, release metadata, quality gates, tests, Business Process Model and Notation (BPMN) and Petri Net Markup Language (PNML) process models, XES and OCEL-style fixture exports, and a planned New Zealand annotation task-set manifest. This article describes the motivation, architecture, ontology-development method, validation evidence, and implementation boundaries. The project is not legal advice, is not an official government publication, and does not certify release, refusal, redaction, charging, extension, transfer, complaint, or publication outcomes.

We propose PRECEDE, a precedent-guided co-scientist for side-effect-aware drug redesign that revises a parent compound to mitigate a specified side effect while preserving therapeutic function. Rather than isolated molecular generation, PRECEDE frames redesign as evidence-grounded reasoning over drug--side-effect associations, biomedical knowledge graphs, and precedents of safety-driven optimization, coordinated by an LLM orchestrator with explicit policies and human-review checkpoints. We position PRECEDE as a human-supervised AI-for-science workflow in which hypotheses remain auditable, falsifiable, and bounded by prior pharmacology.

Agentic AI applications form an emerging serving workload in which a request creates a workflow: a directed acyclic graph of LLM and tool calls that exposes per-node model choices and optional quality operators such as verifiers. This workload falls between two existing layers. Model-serving engines execute individual calls efficiently but cannot see workflow structure, while agent frameworks fix the workflow but cannot see backend load, so neither jointly chooses each node's model, verifier, and backend under serving-time conditions. We present Dyserve, a workflow-aware serving layer that fills this gap. Dyserve compiles each workflow's per-node model and verifier choices in one integer linear program (ILP) over a heterogeneous backend pool, priced by skill-conditioned offline profiles that transfer across workflows. This couples with hardware entering only through per-model throughput sweeps, and is weighted to concentrate strong models and verification on the nodes whose errors propagate the furthest. Because no single latency-quality preference fits every workload mix, Dyserve pre-solves the program at several pressure levels at admission and shifts a workflow's uncommitted suffix among these strategies under load, keeping the solver off the load-shift path; a failed tool call triggers a one-time residual re-solve that preserves committed work.

Multi-product kitting delivery imposes significant challenges for real-time scheduling in hybrid manufacturing systems that integrate processing and assembly, as dynamic order arrivals simultaneously alter supply dependencies and the set of feasible job-machine assignments. This paper proposes a sliding-window-based reinforcement learning (SWRL) framework for end-to-end online scheduling in the flexible assembly flow shop scheduling problem with complex kitting constraints. The problem is formulated as a heterogeneous graph-based Markov decision process that captures the dual-layer kitting structure and the tail-product bottleneck dynamics that produce a sparse reward landscape. To address the resulting challenges, SWRL integrates a sliding-window filtering mechanism that filters inactive nodes and prioritizes kitting-critical operations, a spatiotemporal graph encoding network that tracks bottleneck shifts across consecutive decision states, and a dynamic action mapping module with a constrained waiting strategy that adapts to the changing action space under variable topologies. Experiments on real-world instances from a home appliance manufacturer demonstrate that SWRL achieves consistent tardiness reductions over classical dispatching rules and existing deep reinforcement learning methods, and exhibits robust performance across varying resource configurations, order loads, and arrival concentrations.

Clinical time series prediction in intensive care units remains challenging due to heterogeneous physiological variables and informative missingness. The presence or absence of a measurement can reflect clinical decisions and patient severity, and thus missingness can serve as a predictive signal rather than a simple data artifact. This work presents CISM, a Channel-Independent Spectrogram framework with a Missingness stream for clinical multivariate time series prediction. CISM converts each clinical variable into a variable-wise time-frequency spectrogram, preserves variable identity through variable-aligned encoding, and aligns an explicit missingness stream with the spectrogram representation. Experiments on an in-hospital mortality task derived from MIMIC-IV show that CISM achieves the highest mean AUROC (0.7225), AUPRC (0.3308), and F1 (0.3808) among the compared time series, missingness-aware, vision, and time-frequency baselines. Ablation studies further show that observation patterns provide a meaningful informative signal. Pixel-level mask injection improves performance over plain spectrogram inputs and recovers much of this predictive value. The aligned missingness stream contributes a further, complementary gain in both AUROC and AUPRC. These results highlight the importance of modeling observation patterns as structured signals in clinical time series prediction.

Reduced-order models compress high-dimensional dynamics into low-dimensional representations that can be evaluated rapidly, but they lose accuracy when online dynamics drift beyond the training data. Adaptive methods address this by updating the subspace online with external, out-of-span information, such as full-order corrections or sensor snapshots. We discovered that a complementary and previously unexploited in-span adaptation channel exists within the current reduced subspace. By streaming the model's own predictions through an incremental singular-value decomposition with forgetting, we obtain a trajectory-informed spectral preconditioner, in which the subspace is unchanged but the basis is reweighted and realigned toward the modes visited by the dynamics. This enables the model to absorb future out-of-span corrections more effectively. We expose aspects of this mechanism on a three-dimensional spiral and confirm it on viscous Burgers and Fisher-KPP dynamics. We also discuss how in-span learning can be viewed as a dynamical-systems analogue of in-context learning. More broadly, in-span learning suggests a new principle for computational science, revealing that model-generated trajectories contain more usable information than previously recognized.

Materials phase diagrams are a core knowledge representation in materials science, encoding temperature,composition, phase stability, and phase transformation pathways, with their full understanding requiring thermodynamic mechanism analysis and scientific reasoning. Although VLMs have shown promise in scientific image understanding, their systematic evaluation on such logically complex images demanding deep mechanistic interpretation remains limited, and phase diagrams provide a challenging testbed for this purpose. We introduce MatPhaseBench, a high-quality, high-reliability benchmark for complex scientific image understanding, focused on materials phase diagrams. MatPhaseBench is constructed from 3681 papers in classical materials science journals, from which 200 high-quality diagram-text pairs were selected, covering 189 material systems and 70 elements. The benchmark has three key features: (1)targeting complex scientific image understanding-it moves beyond simple objective tests to open-ended tasks requiring deep comprehension; (2)comprehensive image-text alignment-semantic information associated with images is fully preserved during literature mining and matching; (3) high-quality human-supervised text acquisition-all descriptions undergo strict manual validation. Experimental results show that current VLMs remain substantially behind expert-level understanding: they are largely limited to surface visual perception, lack deep reasoning grounded in thermodynamic mechanisms, have limited domain awareness and expert analytical experience, and perform poorly in distinguishing fine-grained differences in composite or multi-diagram settings. Overall, MatPhaseBench constitutes a challenging research-grade benchmark, providing a foundational platform for complex scientific image understanding, phase diagram analysis, and trustworthy multi-modal AI in science.

Privacy is an important challenge when users interact with AI chatbots, since users may share sensitive information, explicitly or implicitly, and AI chatbots can use this information for user profiling. In this paper, we aim to protect user privacy via a user-side mechanism that transforms sensitive information in a user prompt, while preserving enough information to elicit a useful response from the chatbot. This approach faces an inherent tradeoff between protecting privacy (i.e., avoiding profiling) and preserving utility (i.e., getting personalized and task-specific responses). To that end, we consider, evaluate, and compare four different obfuscation actions, namely redaction, abstraction, replacement, and a novel noising/denoising scheme that we introduce. Additional novel insights include: utilizing a data type taxonomy to both identify and obfuscate sensitive information and explicitly taking into account the utility of chat responses in making the obfuscation decision. First, we systematically optimize and evaluate each obfuscation action independently in terms of the privacy-utility tradeoff it achieves. Second, we propose PROMPTPET, an LLM-based agent that selects the best obfuscation action for each sensitive part of the prompt, using a reinforcement-learning inspired rule optimizer, applied for the first time in this context. Using a real-world chat dataset, we show that PROMPTPET matches the best privacy-utility tradeoff attainable by any single obfuscation action and significantly outperforms prior state-of-the-art approaches.

AI tools are accelerating scientific publication while the systems that review it struggle to keep up, and independent verification of published research has become both harder and more important. As manual replication is slow and expensive, a growing line of work uses coding agents to automate parts of the process. Existing efforts are largely packaged as benchmarks with companion agents that only run inside the benchmark's own pipeline, and no general-purpose replication tool exists. We present VERITAS, a domain-agnostic replication framework built around CLI coding agents. Given a paper, a code repository, or both, VERITAS extracts the paper's claims, runs the methodology while resolving issues as they arise, and judges each claim against the evidence from experiment runs. The pipeline returns an importance-weighted Replication Score, a severity-rated log of every fix applied, and the patched codebase. We evaluate VERITAS on CORE-Bench and ReplicationBench, 65 papers spanning computer science, social science, medicine, and astrophysics. Against two strong Claude Code baselines on the same model and host environment, VERITAS achieves state-of-the-art performance and leads on every metric on both benchmarks.

Visual brand language is the set of visual properties that convey brand identity for a product. What is the impact of visual brand language on a person's ability to recognize and understand the functional identity of an object? Using an empirically supported modeling framework based on the JIM model of object recognition and the LISA model of analogical inference, we simulated the impact of visual brand language on object recognition, the allocation of attention, and retrieval of functional information about objects. Our simulations predict that brand information captures attention and can slow recognition of an object's functional category, with greater degrees of branding causing larger effects. These results have potential implications for the usability and experience of designed objects.

Cold-start drug-drug interaction (DDI) prediction for new drugs is critical for minimizing unexpected adverse drug reactions. The key challenge is to capture similarity between new and known drugs. However, such similarity is closely associated with complex relationships and mechanisms among drugs, enzymes, transporters, molecular structures, and other biomedical entities. Existing methods have three limitations in capturing such similarity: (1) only partial relationships and mechanisms are considered, which overlooks cross-modal information and yields incomplete or biased similarity modeling; (2) similarity computation between new and known drugs is conducted separately across modalities and performed offline for cold-start DDI prediction, leading to misalignment between similarity computation and DDI prediction; and (3) existing interpretability analyses are typically single-modality and focus primarily on key determinants of the perpetrator drug, while the underlying causes of susceptibility for the victim drug are seldom investigated. To address these issues, this paper proposes a novel Cross-Modal-Fused End-to-End Learning Network (CMF-ELN) with three components. First, diverse multimodal information is leveraged to construct four types of drug-centered knowledge graphs, enabling comprehensive similarity modeling under reconstruction-based supervision. Second, a four-channel graph autoencoder is designed to fuse cross-modal similarity within an end-to-end learning framework. Finally, a two-stage interpretability scheme is devised to precisely localize key factors for both perpetrator and victim drugs. Extensive experiments on two real datasets demonstrate that CMF-ELN achieves significantly higher prediction accuracy and more comprehensive interpretability of mechanisms than its peers.

Video understanding is moving beyond closed-context perception toward open-world evidence exploration, a paradigm formalized as Video Deep Research (VDR). However, existing multimodal search agents primarily target static images, and the current VDR benchmark relies on text-centric retrieval that discards crucial visual information. To address these limitations, we propose VideoSearcher, a closed-loop agentic framework that empowers Vision-Language Models with multi-tool reasoning for VDR. VideoSearcher unifies temporal localization, spatial focusing, and multimodal search within a single reasoning trajectory, enabling agents to progressively ground visual clues, retrieve relevant evidence, and synthesize answers. To optimize knowledge-intensive reasoning trajectories, we propose Bi-branch Sequence Policy Optimization (BiSPO), a reinforcement learning algorithm that decouples tool-invocation optimization from answer-accuracy optimization. This design provides stable learning signals for both evidence-grounded reasoning and purposeful tool use. Furthermore, we construct VideoSearch-QA, the first benchmark designed to evaluate open-world video information grounding and multimodal search-based reasoning. Extensive experiments demonstrate that VideoSearcher significantly outperforms prior open-source agentic baselines across various search-oriented and multimodal understanding benchmarks.

Large Language Models (LLMs) are serving as a catalyst of change for research practices, touching the daily lives of staff scientists, software engineers, and system administrators. The developments promise new degrees of autonomy, where categories of human work and decision making are replaced by autonomous, goal-oriented systems. This transition necessitates novel architectural paradigms and solid understanding of the strengths and limitations of LLMs. In this work, we design agents to intelligently deliver the entire life-cycle of an HPC application experimental run in cloud -- creation and build of a container, deployment in Kubernetes, optimization, and orchestration of a scaling study. We pursue this task for four well-known HPC applications to build multi-platform images and optimize across 21 instance types in Kubernetes. We demonstrate successful linear scaling with patterns approved by human experts, designs that improve work time to completion, and review suggested best practices for agentic design and collaboration.

Quantitative 3D spatial reasoning from egocentric RGB-D video is a critical capability for next-generation wearable assistants. Yet existing benchmarks do not reflect the challenges of handling (1) natural egocentric video, (2) posed RGB-D video inputs, and (3) challenging quantitative 3D spatial reasoning Q&A. To fill this gap, we introduce R3D-Bench (Reasoning in 3D), a benchmark of 3,033 quantitative spatial reasoning questions across 15 types -- spanning multiple-choice, distance-based, and volumetric reasoning questions -- built on top of 57 egocentric video sequences from Aria Digital Twin. To set a strong baseline on this dataset, we introduce R3D, a model-agnostic spatial tool-calling framework. In contrast to existing approaches that directly embed 3D information into the model's input representation, R3D constructs a 3D scene from video using segmentation and depth-lifted object representations. It provides this information to an LLM through eight composable spatial tools. On R3D-Bench, R3D with Qwen3-VL 235B achieves 73.5% mean relative accuracy, substantially outperforming the best depth-enabled baseline (CuTR+Tools, 61.9%) and the best RGB-only baseline (Gemini 3 Flash, 46.5%).

Magnetic particle imaging (MPI) enables real-time, radiation-free tracking of magnetic nanoparticle-coated instruments, making it highly suitable for interventional procedures. This study proposes a harmonic-aware transformer framework that directly predicts catheter tip positions from raw MPI voltage signals, eliminating the need for image reconstruction and reducing computational latency. The framework incorporates frequency-domain preprocessing to isolate the 2nd to 8th drive-field harmonics, enhancing the signal-to-noise ratio while preserving motion-relevant features. A transformer architecture with six encoder layers and eight attention heads is employed to learn spatio-temporal dependencies across the three receive axes (x, y, z) for accurate three-dimensional position estimation. The model is trained on simulated MPI signals and evaluated on real in vitro datasets under standard, bending, and heartbeat-like motion conditions. The proposed method achieves sub-millimeter localization accuracy, with a minimum L2 error of 0.103 +/- 0.092 mm and mean absolute errors (MAEs) of 0.039 +/- 0.046 mm, 0.054 +/- 0.049 mm, and 0.060 +/- 0.044 mm along the (x, y, z) axes, respectively, for the bending dataset. Across all datasets, the MAE ranges from 0.165 mm to 0.655 mm, demonstrating consistent performance. The optimized inference achieves a latency of 0.55 ms per frame and a throughput of approximately 1800 frames per second, confirming real-time capability. Compared with conventional MPI-guided approaches relying on image reconstruction, the proposed framework provides improved accuracy, reduced latency, and enhanced robustness under complex motion conditions. These results highlight the potential of harmonic-aware transformer models as efficient and scalable solutions for real-time catheter localization in interventional MPI.

In many scientific and engineering domains, maximizing discovery within a limited sampling budget demands strategic, observation-guided exploration. While generative models have enabled training-free reward alignment, current methods typically excel in local searches within narrow regions of the underlying distribution. These approaches struggle when preferences are unknown a priori and only revealed through sequential feedback-a scenario demanding broad exploration to uncover high-utility regions. To address this, we introduce Bootstrap Flow-Map-Tree (a.k.a BFMT), a novel computationally efficient sampling framework designed for history-aware global search and alignment under sampling budget constraints. BFMT enables full tree-path construction from any tree depth using a single function evaluation, drastically reducing computational overhead while providing critical foresight for sequential sampling. By enabling dynamic transition time steps scheduling, BFMT efficiently allocates its sampling budget, smoothly transitioning from broad global exploration to fine-grained local refinement of high-utility modes discovered through exploration. Extensive experiments and ablations across diverse search and alignment tasks demonstrate that BFMT substantially outperforms baseline approaches.

Large language models (LLMs) have demonstrated remarkable capabilities across diverse applications, yet ensuring their simultaneous safety, helpfulness, and trustworthiness remains a persistent challenge. Conventional refusal-oriented alignment strategies mitigate harmful content generation but systematically fail to serve legitimate user needs, often withholding information that could safely and constructively address the underlying intent of sensitive queries. Building upon the constructive safety paradigm pioneered by Oyster-I, which moves beyond blanket refusal toward thoughtful, response-oriented safety alignment, we identify two critical limitations of its Supervised Fine-Tuning (SFT)-based scheme: insufficient safety generalization to out-of-distribution scenarios and a phenomenon we term safety chain-of-thought (CoT) over-generalization, wherein safety-oriented reasoning patterns are excessively applied to benign queries, degrading helpfulness and user experience. To address these limitations, we propose Oyster-II, a reinforcement learning (RL)-based constructive safety alignment framework that adopts a Zero-RL paradigm combined with a multi-stage reinforcement learning strategy.Evaluated across extensive benchmarks, Oyster-II comprehensively surpasses both Qwen3-14B and its predecessor Oyster-I on safety dimensions, achieving cross-scale performance comparable to Qwen3-Max and Qwen3.5-397B.

LLM-based coding agents solve software-engineering tasks through iterative interactions with development environments, where returned observations accumulate in the context and become a major source of inference cost. Observation compression reduces this cost by shortening observations before they are appended to the context. However, existing methods still exhibit an unsatisfactory efficiency-effectiveness trade-off, as they do not explicitly model how compression affects the agent's subsequent behavior. This paper proposes CoACT, an action-preserving observation compression method for coding agents. CoACT is built on next-action preservation (NAP), which requires a compressed observation to induce the same next action as the raw observation. By checking the agent's immediate next action, NAP provides a practical signal for whether a compression preserves the information needed for continued task solving. During training, a teacher model first generates multiple compressed candidates of each observation. CoACT then uses an action-preservation reward based on NAP to filter out candidates that would change the agent's next action, and uses a length-reduction reward to choose compact candidates as supervision for a lightweight compressor. Experiments on SWE-bench Verified with three agentic models show that CoACT reduces average total token consumption by 33.0% while maintaining task-solving effectiveness close to the uncompressed agent.

Taxonomies provide key information about the semantic relationships between concepts and the inherent organization of vision and language. Despite their impressive capabilities, large multimodal models (LMMs) often lack taxonomic knowledge, leading to low hierarchical visual recognition (HVR) consistency. These models typically only rely on language modeling objectives during fine-tuning and lack explicit taxonomy-aware regularization. To address this, we propose Hierarchical Representation Regularization ($HiR^2$), a simple plug-and-play regularizer that improves hierarchical consistency in LMMs. Specifically, we introduce a semantic-aware visual tree construction framework that extracts coarse-to-fine visual features from intermediate LLM layers guided by textual cues. The regularizer combines two complementary objectives: a taxonomic entailment loss that enforces hierarchy via hyperbolic entailment cones in the Lorentz model, and a discriminative dispersive loss that promotes angular separation of semantically similar embeddings on the unit sphere without disturbing the radial hierarchical structure. Extensive experiments demonstrate that $HiR^2$ effectively captures taxonomic structures across diverse LMMs and fine-tuning methods. Code is available at https://github.com/PKU-ICST-MIPL/HiR2_ICML2026.

Multimodal Large Language Models (MLLMs) have achieved remarkable progress but still struggle with complex visual reasoning tasks requiring multi-step perception and logical deduction. While explicit visual generation incurs prohibitive computational costs, existing latent approaches often rely on external experts or lack rigorous cognitive logic. In this paper, we introduce ProLaViT (Progressive Latent Visual Thought), a framework empowering MLLMs to perform structured visual derivation in the continuous latent space. Unlike works dependent on heterogeneous external models, ProLaViT leverages an endogenous self-distillation mechanism, utilizing the model's own visual encoder to supervise latent thoughts. To facilitate this, we construct a scalable programmatic synthesis pipeline enabling the model to internalize algorithmic precision without inference time tools. We design two reasoning paradigms: (1) Coarse-to-Fine Causal Chain for spatial tasks, guiding attention from global context to local targets. (2) Dialectical Reasoning Chain for logical tasks, incorporating counter-factual thinking for verification. Furthermore, we propose a Distance-Weighted Diversity Loss to impose topology-aware constraints, preventing feature degeneration by enforcing semantic distinctiveness. Extensive experiments demonstrate that ProLaViT outperforms baselines on vision-centric benchmarks, achieving superior accuracy and interpretability with high efficiency.

Explainability is central to building trustworthy AI, yet explanation interfaces can inadvertently provide adversaries with an expanded privacy-related attack surfaces. Recent studies show that advanced membership-inference attacks succeed by exploiting confidence-drop trajectories, induced through attribution-guided perturbations, as discriminative features, rather than directly using confidence scores or explanation vectors. Existing defenses against membership inference fail to directly mitigate such explanation-driven attacks. In this work, we investigate whether, during training, a model's own gradients can be leveraged as defense signals against such attacks, thereby aligning explanation profiles between members and non-members. To this end, we propose a Trajectory-Invariant Explanation Regularization (TIER) defense that penalizes erratic fluctuations in confidence drops simulated through gradient-guided perturbations and simultaneously minimizes the distributional shifts via KL-divergence. Unlike conventional adversarial training, which emphasizes label robustness, our approach targets explanation robustness by enforcing self-consistency through KL-divergence and reducing the variance of confidence drops between members and non-members. Extensive experiments confirm that our method effectively mitigates these attacks, delivering privacy protection while maintaining model utility and explanation fidelity.

On social media, many users actively push back against false claims. Understanding who pushes back and how they do so matters, as this corrective activity is central to how misinformation is contested. We study this counter-misinformation ecosystem at scale: applying a domain-specific NLI model from our prior work to a large corpus of COVID-19 tweets, we classify 264,737 posts as supporting or opposing false claims and compare 23 user- and text-level features across the two groups. Contrary to the dominant assumption that negative emotion is a signature of falsehood, we find that anti-misinformation posts are more emotionally negative than pro-misinformation posts, with higher levels of anger, disgust, and sadness. These differences are modest in magnitude but consistent in direction across the negative emotions. We also find that posts opposing misinformation tend to come from more established users, i.e., older accounts, more followers, and higher listed counts.

Multimodal Large Language Models (MLLMs) have shown strong capabilities, but they may memorize private information from web data, raising privacy concerns. Machine unlearning offers a way to remove such private knowledge without retraining from scratch. However, existing MLLM unlearning benchmarks have two major limitations. First, they rely on simplified images that contain only the single target individual, failing to reflect the visual complexity of real-world photos. Second, they typically assume that the forget set and retain set are fully separated, ignoring the fact that private information is often visually entangled with benign public information. For example, a private individual may appear with a public figure or in front of a well-known landmark, where unlearning the private target should not damage the public context. To address these limitations, we propose PPE-Bench, a new benchmark for evaluating MLLM unlearning under private-public entanglement. Each image contains a target individual to be forgotten and public information to be preserved, including public figure and landmark. We further introduce two simple but effective methods to better preserve public information during unlearning. Through experiments, we find that existing unlearning methods can reduce private information leakage, but often substantially harm adjacent public information.

We ask whether interaction is necessary for order-optimal 1-bit mean estimation over nonparametric finite-moment classes. Adaptive threshold-query protocols achieve the order-optimal 1-bit minimax rate, and the same rate is attainable with general 1-bit queries using only one adaptive transition (i.e., two stages of querying). In the non-adaptive setting, threshold and interval queries are known to be highly suboptimal, but the case of arbitrary non-adaptive quantizers remains unresolved. Can such quantizers match the adaptive rate, yielding an optimal one-shot protocol? Or is the known two-stage estimator stage-optimal, with a single adaptive transition being necessary and sufficient?

Learning-based approaches for selectivity estimation in databases have gained significant traction in recent years. However, theoretical studies of these learning-based approaches are essentially limited to fixed query distributions on static databases. In practice, both the underlying database and the query workload can dynamically change over time. In this work, we propose an algorithmic framework for learning selectivity of queries in this more general dynamic setup. Inspired by online learning, we measure the performance of the learning algorithm in this setting by its regret, which compares the cumulative loss incurred by the learning algorithm to that of the best fixed strategy. We establish upper and lower bounds on regret for histogram-based linear queries, such as point, range, and subset selection queries, under standard loss functions, in both static and dynamic database settings.

Low-bit quantization shrinks language models but treats precision as a single global hyper-parameter: every weight uses the same bit-width. We introduce Variable Bit-width Quantization (VBQ), a training-time method in which each contiguous group of 64 weights learns its own resolution from {1,2,4,8} bits via a Gumbel-Softmax relaxation, trained jointly by an alternating optimization that gives the precision logits a clean, task-aligned signal. VBQ discovers a consistent, strongly heterogeneous allocation within individual projection types, not merely across layers, impossible to express with per-layer methods: 69% of groups collapse to 1 bit, the LM head averages 1.09 bits, while the first MLP block keeps ~2.5 bits. This pattern is stable enough to freeze into a fixed recipe and reuse without further search. The recipe yields a "bigger-but-smaller" regime: a 131M model at 1.82 mean bits reaches perplexity 4.2 on TinyStories, beating a 55M FP16 model (PPL 4.4) at 3.8x less storage, and lets a 1.46B model on FineWeb-Edu match a 593M FP16 control at ~3.7x less storage with 2.5x more parameters. As quality-per-byte, VBQ is 3.9-8.4x more efficient than FP16. The recipe maps directly to packed low-bit storage, so it also accelerates inference: with custom fused dequantize-and-multiply kernels, memory-bandwidth-bound autoregressive decode is faster at equal output, and the speedup grows with scale (parity at 131M, 1.9x at 1.0B, 4.7x at 9B on Apple silicon). A distributional analysis (KL divergence and argmax-flip rate) reveals a striking mechanism: deeper layers progressively self-heal the quantization error injected by early layers. The win is a from-scratch, train-time phenomenon; scaling the search economically beyond 1.5B parameters remains open. VBQ reframes precision as a learnable, non-uniform resource and shows that spending a fixed bit budget unevenly beats spending it uniformly.

Many online decision-making problems involve both round-specific feasible actions and drifting reward models: eligible ad impressions, feasible prices, and available treatments can change over time, while user preferences, demand curves, and patient responses may evolve. Motivated by these applications, we study non-stationary linear bandits with round-specific feasible decision sets. Existing methods that obtain the optimal \(\widetilde O(T^{2/3}P_T^{1/3})\) dependence, where \(P_T\) is the path length of the reward-parameter sequence, impose an orthogonal-structure assumption on round-specific decision sets, which can be restrictive in contextual applications. We address this gap through a unified misspecification-reduction viewpoint: after partitioning the horizon into blocks, we relate each block's dynamic regret to regret against a fixed-parameter linear bandit benchmark, with the within-block parameter drift entering as bounded misspecification. Restarting algorithms with misspecification-dependent regret guarantees then yields the optimal \(T^{2/3}P_T^{1/3}\) dynamic-regret dependence for both linear bandits with general compact decision sets and \(K\)-armed contextual linear bandits.

Deploying AI-generated video detectors in real-world services demands an ultra-low false positive rate (FPR) on real videos to avoid falsely rejecting authentic content, a regime where standard metrics such as AUROC fail to reflect actual operating behavior. We introduce Spatial Patch-Level Incoherence and Temporal Roughness (SPLIT), a training-free detector that operates on patch tokens from a frozen vision encoder to detect both fully generated and partially edited videos. SPLIT computes two complementary signals: Two-step Temporal Roughness (TTR), capturing non-smooth patch trajectories via one-step and two-step feature variation contrast, and Local Spatial Motion Incoherence (LSMI), measuring spatially inconsistent temporal changes through gradients of a feature-space motion field. The two are fused multiplicatively with gamma correction to sharpen real-fake separation at strict thresholds. We further propose a service-aligned evaluation protocol based on Fake Recall at fixed FPR with real-only threshold calibration and cross-real threshold transfer. Across three benchmarks (FakeParts, GenVideo, and ViF-Bench), SPLIT achieves the highest Fake Recall at FPR = $0.1\%$, substantially outperforming supervised and training-free baselines while remaining robust to post-processing with negligible overhead. The code is publicly available at https://github.com/mldljyh/SPLIT .

Cognitive Behavioral Therapy (CBT) provides a structured framework for understanding a user's mental state by examining the interaction between cognitive and behavioral factors. However, out-of-the-box LLMs respond fluently and empathetically, yet collapse into validation & reflection, regardless of what the user actually needs. They know theoretical CBT (scoring up to 96% accuracy on licensing exam questions) but fail to apply it effectively. We explore this gap with a knowledge-guided framework that treats CBT dialogue as controlled affective reasoning: user narratives are decomposed into Beck's Cognitive Conceptualization structure, grounded in clinical SNOMED CT concepts validated via Natural Language Inference, and a Multiple Chain-of-Thought (MCoT) strategy selection between Validation & Reflection, Socratic Questioning, or Alternative Perspectives. To measure whether such guidance actually changes behavior, we introduce the Protocol Leverage Force (F), a behavior-level metric that captures how far an intervention shifts a model away from its default response. Across three open-weight LLMs and 14 RealCBT-derived case studies, evaluated with human experts, valence-arousal trajectories, and linguistic entrainment, F shows that simply introducing protocol definitions via single chain-of-thought prompting fails to change LLM behavior, while MCoT on these definitions guides strategy selection better. Still, the effect stays within 1% (approx. 1.2-1.3%), and all models remain biased toward Validation & Reflection. These results show CBT knowledge alone does not ensure effective application, giving the affective-computing community instrumentation to measure where LLMs fall short.

Platform-orchestrated agentic workflows have become a popular paradigm for developing LLM-based applications. However, their reliability remains a major challenge due to the uncertainty of LLM outputs, complex inter-node dependencies, and heterogeneous tool interactions. Existing agentic workflow optimization and agent enhancement methods primarily rely on trajectory-level feedback. Without explicitly identifying the underlying failure root causes, their resulting repair plans are often insufficiently targeted. We propose FlowFixer, a diagnosis-driven automated repair framework for agentic workflows. FlowFixer first transforms workflow executions into unified symbolic traces and performs symbolic inference to derive executable behavioral specifications that capture node correctness, temporal dependencies, and causal relationships. Based on specification verification, it conducts failure attribution and root cause analysis, and then generates targeted repair patches. To reduce verification costs, FlowFixer further employs a multi-dimensional pre-execution assessment to filter infeasible repairs before dynamic verification. We evaluate FlowFixer on workflow failures collected from three popular development platforms: Dify, Coze and n8n. Results show that FlowFixer achieves a repair success rate of 71.3%, outperforming state-of-the-art baselines by 11.9% to 27.6%. It also improves failure attribution accuracy by 4.8% to 33.1% and root cause analysis accuracy by 15.3% to 38.8%. This work offers a new perspective on reliable diagnosis and repair of agentic workflows through symbolic modeling and inference.

Long-horizon behavior prediction aims to infer a user's next action based on a lengthy historical sequence, playing a crucial role in artificial intelligence field. The rise of large language models (LLMs) offers a promising direction for sequential behavior prediction, yet LLMs struggle with latent behavioral pattern induction and model-intrinsic cognitive biases when tackling long-horizon behavior prediction. Prior memory management methods follow a context-compression paradigm that attempts to address this task by alleviating the historical sequence burden, yet fail to resolve the core challenges. In this paper, we advocate a paradigm shift that reframes the lengthy historical sequence from a burden into a valuable resource to be exploited, and accordingly propose PraMem, which conducts beforehand practice over the lengthy historical sequence to build an experiential memory, thereby serving as the assisted input for accurate long-horizon behavior prediction. Extensive experiments across diverse tasks demonstrate that PraMem achieves superior performance than prior methods, and more in-depth analyses provide valuable insights into the mechanism and evolution of the experiential memory. Code: https://github.com/icip-cas/PraMem.

Current benchmarks for evaluating large language models (LLMs) in medical calculation are largely based on simplified settings, where each patient case corresponds to a single calculator and the required tool is explicitly specified in the query. However, real clinical scenarios often require multiple calculators for joint evaluation, nested-scale calculation, and fuzzy queries that do not directly specify the target calculator. To this end, we propose a new medical calculation benchmark, MedCalc-Pro, which covers three progressively challenging task settings: single-calculator, multi-calculator, and nested-calculator calculation settings. MedCalc-Pro contains 2,268 real-world clinical cases, covering 77 medical calculators across 14 clinical departments. Meanwhile, to address the limited performance of existing frameworks and methods in complex clinical scenarios, we further propose a more generalizable agent framework that supports multi-tool selection and nested-tool calling, while suppressing parameter error propagation through structured validation and evidence review. We conduct systematic comparisons across open-source, closed-source, and medical-specialized LLMs, and the results show that our framework achieves the best performance across all three task settings. This work provides a new benchmark and method for evaluating and applying LLMs in challenging medical calculation scenarios.

Large language model agents driving security tool suites over the Model Context Protocol are increasingly common. Yet the factors that bound their capability remain poorly characterized: how much depends on the model versus the client that drives it, whether constraining the agent to the orchestrator's own tools helps, and where capability is limited by reasoning rather than by missing tools. Using HexStrikeAI, an open-source orchestrator that exposes 150+ tools, as a testbed, we follow a methodology that evaluates the system, diagnoses its failures, and applies targeted improvements. We run 86 picoCTF challenges across seven categories and three difficulty tiers, under three tool-access regimes and three model/client configurations (774 trials). We then apply corrections to existing tools, agent-behavior changes, and eleven new capability tools, and re-run the previously-unsuccessful trials. The diagnosis isolates the driving client as a first-order factor for a fixed model (a 2.1 * gap between two DeepSeek clients) and a monotonic difficulty gradient, with the largest gains in the mid tier. The overall solve rate rises from 55.4% to 72.0%, and every configuration improves significantly (paired McNemar p < 0.001, non-overlapping 95% confidence intervals). The residual failures are reasoning- or environment-bound rather than missing-tool. A 60-run stability sub-study finds single-run verdicts reproducible (17/20 unanimous). We discuss what the results imply for how such orchestrators should be evaluated, and we are explicit about the limits: the study uses a single benchmark, the fixes were tuned on the same challenges they were evaluated on, and the client effect is demonstrated for one model only, so its generality to other models remains a hypothesis.

Dynamic knowledge graphs are ubiquitous in today's AI applications, as we represent molecular structures, social relationships, and language information using these graph models. As knowledge graphs evolve over time and are often noisy and incomplete, modeling their temporal and relational dependencies becomes crucial for downstream tasks. To address these challenges, this paper proposes PGRE (Poisson-Gamma Relational Evolution), a probabilistic model for modeling inter-relational dependencies in dynamic knowledge graphs. PGRE represents multi-relational temporal links via a Poisson-Bernoulli formulation. It introduces Gamma-distributed latent variables to capture entity-factor associations and cross-relation dependencies mediated by shared latent communities. A Gamma Markov process further models the temporal evolution of these latent variables, enabling principled characterization of relational dynamics. Experiments on benchmark datasets show that PGRE achieves competitive performance in link prediction, particularly in sparse settings, while revealing meaningful relational evolution patterns in dynamic knowledge graphs.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a promising paradigm for improving mathematical reasoning in language models. Yet most RLVR work rewards only the final answer (outcome-based rewards), leaving the impact of step-level process supervision (process rewards) underexplored especially for small models that lack the capacity to self-correct under sparse feedback. We systematically compare five reward conditions applied to Qwen2.5-0.5B fine-tuned with Group Relative Policy Optimization (GRPO) on GSM8K: a no-RL baseline, process-only, outcome-only, and three hybrid weightings ($λ\in \{0.9, 0.5, 0.1\}$ process weight). Process-only supervision achieves 63.73% test accuracy versus 53.75% for outcome-only, a nearly 10-percentage point gap while yielding reasoning traces with higher step validity and lower deviation from ground-truth chain length. Hybrid rewards generally correlate positively with process weight, with one notable anomaly: the low-process / high-outcome configuration ($λ=0.1$) underperforms pure outcome supervision, suggesting conflicting optimization signals. Error analysis using GPT-4o as a judge reveals distinct failure mode distributions: process models generate structurally inconsistent but arithmetically grounded traces, while outcome models produce concise but derivation-error-prone chains. Our results demonstrate that reward granularity is a first-order design decision for RLVR, with process-level supervision substantially improving both accuracy and trace fidelity in small language models.

Reinforcement Learning (RL) has emerged as a powerful approach in financial trading, enabling agents to learn optimal strategies through direct market interaction. However, financial markets are highly uncertain, with price fluctuations driven by stochastic volatility, model limitations, and regime shifts. Traditional RL models struggle in dynamic environments, often failing to adapt to sudden market disruptions, leading to suboptimal trading decisions. To address this challenge, we propose an uncertainty-aware RL framework that integrates distributional, epistemic, and aleatoric uncertainty estimations. Our approach enhances uncertainty estimation using SHAP-weighted reconstruction uncertainty, MC Dropout, and an LSTM-based technical indicator consensus mechanism. Experimental results on five major U.S. stock indices demonstrate that RL agents equipped with uncertainty estimation significantly outperform traditional models in return and risk management. This study advances uncertainty estimation in RL-based financial trading, with future research extending its application to other asset classes and alternative RL architectures for greater adaptability.

Automatic Speech Recognition (ASR) and Dialect Identification (DID) are crucial for Indian languages, many of which are low-resource and exhibit significant dialectal differences. Existing methods often optimize ASR or DID individually, resulting in performance trade-offs. In this work, we propose a multimodal framework that jointly improves ASR and DID. Our method employs a Bottleneck Encoder to extract dialectal features from Conformer-based speech representations and a RoBERTa encoder to process ASR-generated CTC embeddings. A gating mechanism merges these features, followed by an attention encoder to refine the representations. The learned embeddings are concatenated with Conformer outputs to enhance ASR features. Evaluated on eight Indian languages with thirty-three dialects, our method achieves an average DID accuracy of 81.63% and average CER and WER of 4.65% and 17.73%, respectively. These results highlight the effectiveness of our method for joint ASR-DID modeling.

LLM coding agents increasingly complete development tasks by issuing ordinary CLI commands. Following the Unix design, these commands cooperate through shared operating-system state: one command may write state that a later command reads. While this composition is benign and intended, it creates an overlooked exploit surface. Existing attacks and defenses mainly target the instruction layer, where malicious intent appears as hostile text. In contrast, we observe that individually benign commands can form a dangerous producer-consumer state relation across the command trace, exposing what we call CLI command-composition risk (CCR). Given this new attack surface, it is critical to systematically uncover and characterize the impact of CCR in real-world coding agents. However, systematically understanding this risk is quite challenging, because naive command enumeration and end-to-end LLM generation produce mostly invalid workflows. We present MOSAIC, a knowledge-guided framework that distills validated command-state behaviors from CVEs, advisories, and researcher PoCs into reusable summaries, composes them into exploit paths, and instantiates them as realistic developer workflows for black-box agent evaluation. Across five real-world CLI coding agents and five backend LLMs over 2,525 trials, MOSAIC achieves a 96.59% attack success rate under benign developer tasks.

Replicated databases and permissioned blockchain systems rely on Byzantine Fault-Tolerant (BFT) consensus to maintain a globally consistent order of transactions across distributed replicas. These protocols preserve safety even under asynchrony, as they commit a transaction only after agreement among a strong quorum of replicas. During network partitions, however, when no strong quorum is reachable, they lose liveness and cannot make useful progress. In this paper, we present Cassandra, a consensus protocol that enables partial progress without sacrificing safety. Cassandra achieves this through a two-tier certification framework that decouples availability from commitment, allowing each partition to extend its own chain and reconcile these chains once the network is restored. To support this, Cassandra introduces a pacemaker that advances views without requiring a strong quorum and calibrates each replica's timeout off the critical path. Our evaluation results show that Cassandra remains competitive with state-of-the-art BFT protocols under stable conditions, sustaining 900K TPS at 16 replicas and 480K TPS at 104 replicas, with latency ranging from 0.31s at 16 replicas to 0.75s at 104 replicas. Under severe partitions, Cassandra maintains non-zero speculative throughput through PoA-backed progress, preserving work that can be reconciled once connectivity is restored.

Before fixing an issue, it is useful to first reproduce it by generating a bug reproduction test (BRT). However, generating a BRT is itself a challenging task, because issue descriptions tend to be informal, making it difficult to determine whether a candidate BRT indeed fails for the reason in the issue. Prior work has attempted to tackle this problem via inference scaling, using large language models to generate many BRTs and patches, then using execution feedback to select and improve them. Unfortunately, this is expensive and the feedback is unreliable. This paper explores evolutionary programming for BRT generation to sharpen the feedback, while enhancing evolutionary programming to keep costs in check. Our new approach, EvoOtter, controls test execution costs via successive halving. Furthermore, it controls LLM costs via batched crossover for an entire generation in a single LLM call, as well as via rule-based code mutations, with a new fitness score tailored for BRTs. As a result, EvoOtter generates state-of-the-art quality BRTs at the fraction of the cost of prior inference-scaling approaches to this problem. More broadly, this paper points at how to efficiently and effectively combine evolutionary programming with large language models for software engineering.

Vision Language Models (VLMs) are increasingly applied to software engineering artifacts, especially UML class diagrams whose meaning depends on visual notation. Yet, it is unclear whether VLMs actually read such diagrams or instead answer from pretrained priors about how classes typically relate. We introduce a controlled UML benchmark in which each prior-conforming diagram is paired with its prior-conflicting counterpart that (1) preserves the same class names and layout while (2) reverses only the relation arrow. We evaluate eight open-source VLMs from two model families, InternVL3.5 and Qwen3, alongside two closed-source frontier models GPT-5.4 and GPT-5.4 Mini. Across the eight open-source models, reversing the arrow reduces relation-direction accuracy by 33.48% on average, while GPT-5.4 Mini retains a 10% gap. In the harder three-class condition, accuracy drops sharply by 45.28% for open-source models, and even 18.62% for the GPT-5.4 family on average. Scaling provides only limited improvements and is family-dependent. Our benchmark presents a diagnostic prior-driven failure in diagram-grounded software understanding. Our artifact is available at https://anonymous.4open.science/r/UMLKnowledgeConflict-8461.

Meta-learning without labeled data is crucial for real-world applications, where obtaining labeled datasets can be expensive or restricted due to privacy concerns. Data-Free Meta-Learning (DFML) addresses this challenge by leveraging pre-trained models without access to training data. However, existing DFML methods rely on model inversion to generate training data, a process that is generally difficult and computationally expensive due to the need to generate high-dimensional data matching the original distribution. To address this limitation, we propose a novel meta-learning setting that avoids model inversion by jointly leveraging pre-trained models and unlabeled data. Our method generates meta-training tasks by assigning soft labels from pre-trained models to unlabeled data. Since the quality of these tasks can vary, we introduce a task-weighting mechanism based on task confidence and class distribution balance to ensure effective meta-learning. Extensive experiments demonstrate that our approach substantially reduces computational cost and improves generalization, achieving up to 104-fold speedup and 8.4 percent to 36.4 percent improvements in few-shot classification accuracy compared to state-of-the-art DFML methods.

Large language model (LLM) agents can improve through accumulated experience, but free-form textual memories become difficult to maintain, validate, and reuse as interactions grow. Recent symbolic approaches learn executable skills or programmatic world models, yet often store local procedures or assume simplified dynamics. We propose Object-Centric Environment Modeling (OCM), which organizes experience into an executable object-centric environment model. OCM maintains two connected code bases: object knowledge, which defines environment entities and mechanisms as Python classes, and procedure knowledge, which records reusable interaction patterns that must import and use the object model. OCM works in an online setting: after each episode, OCM reflects on the trajectory, updates both knowledge bases, and verifies that all procedures execute against the updated object model. During future interaction, the agent uses progressive knowledge disclosure to inspect compact code signatures first and read source code only when needed. Experiments show that OCM achieves the best average rank across benchmarks and reduces invalid actions, demonstrating that agents can benefit from building object-centric environment models.

Multi-view robotic manipulation methods with the attention mechanism have recently achieved significant progress in both training efficiency and task performance. However, the inherent redundancy, occlusion, and viewpoint dependency in robotic view images often lead to severe attention drift. To address this challenge, we propose AmpAttention, a novel attention mechanism inspired by differential amplifiers in analog circuits. It aims to suppress attention noise and capture high signal-to-noise ratio signals for more reliable perception. Based on this, we introduce the RVAF model, which integrates task-guided intra-view and inter-view AmpAttention. Compared to previous state-of-the-art methods, RVAF achieves the optimal average success rate across 18 RLBench tasks (249 variations) while reducing training time by 33.3\%. RVAF also demonstrates strong potential in real-world high-precision tasks, exemplified by its ability to pick up a dart and accurately insert it into the red bullseye. Furthermore, we extend RVAF to RVAF++ by incorporating the SAM2 image encoder. RVAF++ achieves substantial gains on high-precision tasks, achieving a 91\% success rate on the `insert peg' task. More qualitative results are provided at the anonymous project website https://anonymous.4open.science/w/RVAF-Anonymization.

Analog in-memory computing (IMC) has emerged as a promising approach for accelerating matrix operations by exploiting the intrinsic physics of memory arrays. To date, however, most IMC architectures have focused on linear algebra workloads in which computation is encoded in the equilibrium state of a physical system. Extending these principles to nonlinear optimization remains challenging and typically relies on iterative algorithms composed of repeated linear operations. Here, we introduce a continuous-time nonlinear closed-loop IMC architecture for box-constrained zero-forcing (BCZF) decoding in massive multiple-input multiple-output (MIMO) systems. The proposed architecture embeds the decoding problem directly within the dynamics of a nonlinear feedback network of memory arrays and supply-limited operational amplifiers, allowing solutions to emerge through continuous-time physical optimization. We derive a compact analytical model of the circuit and show that its trajectories minimize an equivalent energy function. Experimental emulation using a fabricated IMC chip confirms the predicted dynamics under realistic hardware nonidealities for up to 16x16 MIMO systems. To overcome the finite precision of analog hardware, we extend mixed-precision iterative refinement from linear algebra to nonlinear continuous-time optimization, enabling reliable detection of high-order modulation formats including 256-QAM. Benchmark projections indicate operation from ultra-low-energy approximate decoding to high-accuracy massive MIMO detection. Together, these results extend closed-loop IMC from equilibrium-based linear algebra to continuous-time nonlinear optimization and establish a pathway toward efficient physical accelerators for high-accuracy wireless communications.

Molecular optimization often starts from a pretrained generative model that captures a broad prior over valid molecular structures. At test time, however, the goal is not to sample from this prior, but to use a limited oracle budget to shift generation toward task-specific high-reward molecules. We study this adaptation problem for discrete diffusion models. Each online round couples several choices. The loop must decide which candidates to evaluate, how rewards become model updates, which feedback to reuse, and how far to move beyond the pretrained prior. These choices have mostly been studied in isolation, leaving open whether they complement one another, become redundant, or interfere inside a full online adaptation loop. We conduct controlled studies across six small-molecule binding-affinity tasks and three protein-fitness tasks. We find that acquisition, reward shaping, and model debiasing provide complementary routes to higher reward, especially for small molecules. Replay further stabilizes learning, while validity penalties keep small-molecule exploration on the valid molecular manifold. Together, these findings point to a practical recipe for feedback-efficient molecular optimization: online fine-tuning with acquisition, reward shaping, debiasing, replay, and validity control. This recipe outperforms offline fine-tuning and inference-time search baselines under matched oracle-call budgets and GPU-hour accounting. The gains are largest when high-reward candidates require larger shifts from the pretrained prior.

Existing ViT-based weather forecasting models apply uniform computation across all spatial tokens, even though nearby atmospheric grid points often contain similar values and large regions evolve smoothly over time. This makes much of the intermediate per-token computation redundant. Standard token-efficiency methods, such as pruning or merging, reduce cost by removing or fusing tokens. However, weather forecasting is a spatiotemporal dense prediction problem in which a history of atmospheric states must be mapped to future values on the original latitude-longitude grid. Thus, every grid cell must retain a physically meaningful representation, especially under autoregressive rollout. We introduce Sparse-Reslim, a parameter-free plug-in routing module that makes sparse token processing compatible with this fixed-grid requirement. Sparse-Reslim routes only 25% of spatial tokens through the expensive middle transformer blocks and treats those blocks as residual updates: it computes the change produced for the routed tokens and scatters only this delta back to the full sequence. Unselected tokens keep their pre-routing representations exactly, so no grid cell is dropped or replaced by a mask token, and no fusion layer or additional parameters are introduced. Across ERA5 resolutions up to the operational 0.25\textdegree{} standard and two model families, a deterministic Transformer and a diffusion model, Sparse-Reslim improves forecast accuracy on every evaluated variable while substantially reducing cost: training is about 2.5x faster in the main settings and reaches 3.18x speedup at 0.25\textdegree{}, with over 2.2x lower peak memory. A controlled decomposition shows that the accuracy gain comes primarily from sparse routing itself, while random token selection provides an additional regularization benefit without selector overhead.

We release \textsc{JavaVulBench}, a benchmark dataset and evaluation harness for Java vulnerability detection. The dataset contains $\sim$30{,}600 Java methods spanning 1{,}740 CVEs and 700+ projects, labelled at both method and line granularity, with per-CVE publication dates and five realistic split strategies: random, project-disjoint, temporal, deduplicated, and unseen CWE-family. The harness provides a single \texttt{LlmPrediction} schema across three backend families (encoder classifiers, local generative models served by Ollama, and API-served LLMs routed through OpenRouter) so that twelve reference detectors CodeBERT, GraphCodeBERT, UniXcoder, DeepSeek-Coder-1.3B, and eight API/open-weight LLMs (GPT-4o, GPT-4.1-mini, Claude Sonnet~4, DeepSeek-v3, DeepSeek-Coder-v2, Qwen-2.5-Coder-14B/7B, CodeLlama-13B) are evaluated under identical conditions from a single command. A pre-training contamination audit is shipped alongside every model so users can separate genuinely unseen test CVEs from potentially memorised ones. Data, code, and fine-tuned checkpoints are archived on Zenodo [31] and short demonstration video is available on YouTube (https://www.youtube.com/watch?v=nMTX\_hqkuoM) https://www.youtube.com/watch?v=nMTX_hqkuoM.

Cloud-edge Large Vision-Language Model (LVLM) inference enables efficient deployment by splitting computation between edge devices and cloud servers. In this process, intermediate vision tokens are transmitted from the edge to the cloud over a communication link, thereby exposing a new attack surface. We study vision token manipulation attack (VTM-Attack) under a black-box man-in-the-middle setting, where an adversary intercepts and manipulates a subset of transmitted vision tokens under a budget constraint. We propose four naïve attack strategies and an optimization-based token selection method. Experiments on 6 state-of-the-art LVLMs (3B-72B) across 4 benchmarks show that manipulating only 10\% of vision tokens can reduce accuracy by up to 88.31\%. These results reveal a critical vulnerability in cloud-edge LVLM inference.

Modern blockchains rely on a modular architecture that decouples consensus from execution. Recent advances in consensus algorithms have shifted the bottleneck to the execution layer, which must deterministically follow the consensus order and handle increasingly complex, compute-intensive smart contracts. We identify that single-node validators cannot keep up, motivating the need for a scale-out design. We design Remora, a scale-out smart contract execution engine. Remora adopts an efficient asymmetric architecture with centralized transaction dispatching and distributed execution, and depends on an object versioning scheme with a strict ownership model to guarantee deterministic scale-out execution. Remora achieves up to 3x throughput improvement compared to state-of-the-art deterministic execution schemes, scales up to 250k TPS, matching modern consensus performance, and reduces latency by up to 5ms. We also show that Remora elastically adapts to bursty workloads and dynamic access patterns using real-world traces. Remora's main performance benefits come from a novel stateless-stateful separation during smart contract execution, which overlaps the execution of state-independent tasks with consensus, and a new locality-aware and load-balanced scheduling scheme.

Personal agents will increasingly negotiate on behalf of users: splitting costs with other personal agents, appealing platform decisions, escalating support disputes, requesting refunds, changing subscriptions, and negotiating deadlines or reimbursements. Existing negotiation benchmarks emphasize agreement, surplus, or strategic competence, but a user-owned agent can reach an agreement while harming the user through privacy leakage, consent violation, unsupported advocacy, over-concession, failed escalation, or poor auditability. We introduce SovereignNegotiation-Bench, a trace-level multi-turn benchmark for delegated personal-agent negotiation under private utilities, disclosure constraints, evidence requirements, and institutional asymmetry. The benchmark separates agent-visible observable state from evaluator-only labels and evaluates agreement success jointly with user utility, privacy, consent, evidence grounding, concession discipline, escalation, and auditability. We report an artifact-backed validation over 240 scenarios, 4 model families, 14 baselines, 13,440 frozen-prompt live trajectories, 61,135 parsed action rows, and a blinded 3-annotator audit over 300 items. The strongest agreement-maximizing baseline achieves the highest agreement rate but low user utility and high privacy/consent risk; FullSovereign does not maximize agreement, but obtains the best sovereign negotiation score by preserving utility, minimizing leakage, grounding claims, and reducing unauthorized commitments. The results show that agreement success is insufficient for user-owned negotiation agents.

N-Version Programming (NVP) improves software reliability by executing multiple independent implementations and combining outputs, but its adoption is limited by high cost and the assumption of failure independence, which empirical studies have challenged. Recent advances in Large Language Models (LLMs) reduce the cost of generating multiple implementations, shifting focus to whether their failures are independent. We propose the first systematic methodology to assess failure independence in LLM-generated code and apply it to 224 problems across twelve models, five languages, and three prompting strategies. We analyze both structural and behavioral diversity (i.e., whether implementations fail on the same test cases), complemented by N-version reliability analysis under majority voting and manual inspection of the generated code. Structural diversity analysis shows that implementations from the same model are highly similar, while different models produce more distinct solutions. The same trend appears in behavioral diversity, with implementations from different models showing higher diversity yet still failing on the same tests far more often than expected under independence. N-version reliability analysis reinforces this: three- and five-version ensembles realize only 0.43 and 0.44 of the reliability gain achievable under independence, dropping below 0.3 when ensembles are built from the same model. Manual fault analysis shows that even different failure patterns often share root causes. Overall, these results suggest LLM-generated solutions do not satisfy NVP's failure independence assumption, though heterogeneous models help partially. They also validate our methodology as a tool for systematically evaluating failure independence as models evolve.

Long-running coding agents such as autoresearch can persistently discover optimizations for open-ended problems. However, they tend to converge onto a single high-level approach, then proceed with low-level edits while missing other superior approaches to the problem. We hypothesize two harness-level design choices contribute to this behavior: accumulating context in a single long-running agent and only exposing a single program state to edit. We introduce SwarmResearch, an orchestrator-subagent harness in which a Shepherd Agent uses global context to steer a population of Search Agents, each operating with local context in their respective git branch. On open-ended optimization tasks, SwarmResearch discovers better or comparable solutions to state-of-the-art LLM-guided evolution and multi-agent techniques on 13/15 tasks, driven by higher-level exploration. Compared with fixed scaling of serial and parallel agents, SwarmResearch's orchestrator-guided scaling discovers better-performing solutions by adapting parallelism at different search depths.

High-throughput long-context generation is one of the central challenges for large language models. Generation is typically memory-bandwidth-bound rather than compute-bound: each decoding step must stream the accumulated key/value (KV) cache from memory, so bandwidth demand grows with context length while only one token is emitted. Two parallel approaches have therefore emerged: reducing memory access with efficient attention variants and linear-time mixers such as Mamba, or increasing parallel computation by generating blocks of tokens at once. However, technical challenges arise when combining these two ideas. Earlier hybrid diffusion models such as DiffuMamba use bidirectional Mamba mixing, including a reverse-direction scan relative to causal generation. This reverse scan needs to scan the entire sequence, so its states are not prefix-only and cannot be precisely reused as a cache even when diffusion is performed block by block. We propose a BDLM Mamba--attention hybrid that addresses this challenge by restricting the reverse Mamba scan to the active denoising block, which enables exact caching across blocks. In an 87M-parameter DCLM sweep, BDLM Mamba-H achieves the best C4-en validation perplexity compared to BDLM attention and full-sequence baselines. At 350M parameters, it remains competitive with BDLM attention. For long-context inference, BDLM Mamba-H reaches 19.7x the throughput of full-sequence DiffuMamba-H at 65K tokens and 3.7x the throughput of BDLM attention at 262K, showing that Mamba hybrids are a potential long-context diffusion architecture.

As LLMs are increasingly deployed as autonomous adjudicators in semi-open textual game environments, robust rule adherence becomes critical when user intent conflicts with system rules. However, these models are trained to be helpful and compliant, leaving them vulnerable to a class of attacks we term \textit{Rhetorical Injection}, where adversarial users exploit narrative framing techniques such as pseudo-logical reasoning and authoritative coercion to bypass adjudication logic. We present CoC-Seduce, a multi-agent adversarial benchmark built on Tabletop Role-Playing Game (TRPG) mechanics, an ideal instantiation of semi-open environments where rules are explicit for adjudication, yet interaction remains entirely in natural language. Three frontier models, i.e., GPT-5.4, Claude Sonnet 4.6, Gemini 3.5 Flash, serve as adversarial generators producing 5,376 samples across 4 world settings and 16 skill categories. We then benchmark 20 target adjudicators against this corpus. Evaluation across 20 models reveals that neither model scale nor explicit reasoning mechanisms reliably confer adjudication robustness, with \textsc{Pseudo-Logic} emerging as the dominant attack vector and cross-cultural settings exposing systematic knowledge gaps across all evaluated families. Project page: https://github.com/answerrtx/CoC-Seduce

Induction heads are attention circuits believed to underlie in-context learning in transformers, yet a precise characterization of the estimators they implement remains elusive. We study transformers trained on order-$k$ Markov chains and identify two complementary smoothing mechanisms. First, at finite attention-weight scale, the circuit implements a soft context-matching estimator: it aggregates contributions from exact and partial context matches, weighted exponentially by their overlap, and induces a data-dependent interpolation across context orders analogous to Jelinek-Mercer smoothing. Second, a beginning-of-sequence (BOS) token induces additive pseudo-counts, recovering Dirichlet-style smoothing. We construct a disentangled transformer implementing both mechanisms and show that trained transformers recover the predicted attention patterns. Across settings where pseudo-count smoothing is optimal or lower-order contexts provide structured evidence, trained transformers match or outperform classical count-based baselines. Our results bridge mechanistic interpretability of induction heads with classical statistical smoothing, revealing that transformers learn to regularize in-context estimation rather than simply count.

Predictions by machine learning (ML) and artificial intelligence (AI) models are often received skeptically unless they are paired with intelligible explanations. In the context of just-in-time defect prediction, highlighting small portions of a software change (diff) -- beyond rule-based lints -- where risk may be concentrated has not yet been extensively investigated. In this work, we leverage attention weights from an LLM-based Diff Risk Score (DRS) model to highlight parts of a diff that the model focuses on when predicting risk. We aggregate token-level attention into interpretable code units (lines, hunks, and files), and present the top-K units to developers as a lightweight form of guidance during code review. We evaluate our approach using expert-labeled changes that have caused real outages. Results show that the highlighted snippets cover expert-labeled outage-causing change lines 53.85% of the time when highlighting the top-2 hunks, while requiring developers to review 26.28% of the changed lines on average. Because attention is produced during standard model inference, the approach is scalable for large development workflows and can be surfaced in the code review UI with low additional latency.

Inference-time alignment steers a frozen language model during decoding using auxiliary reward signals, avoiding the cost of repeated weight updates. However, existing inference-time alignment methods typically optimize a single scalar score, so explicit safety constraints must either be ignored or encoded through manually tuned penalties. We propose Lagrangian Reward Augmentation (LARA), a general inference-time alignment framework under safety constraints. Starting from a KL-regularized constrained objective with a reward model and a cost model, LARA dualizes the constraint and reduces the optimization problem to a one-dimensional convex problem over a nonnegative dual variable. Estimated on a small calibration set, this dual variable defines an augmented reward that can be used as a drop-in scoring signal within existing inference-time alignment methods. For sequence-level sampling methods, such as Best-of-N reranking, the calibrated dual variable corresponds to the solution of the expected-cost constrained problem. For token-level reward-guided decoding methods, the same construction yields a principled dual-calibrated heuristic rather than an exact constrained-policy guarantee. We evaluate LARA on both sequence-level and token-level inference-time alignment methods, and find that LARA improves the helpfulness-harmlessness tradeoff, with Best-of-N achieving the best performance among inference-time methods, approaching finetuning-based direct alignment baselines.

Any global mirror of open-source version control is incomplete, but the reasons a commit is missing are not interchangeable: a project may have force-pushed it away (so it no longer exists upstream), or the mirror may never have ingested it (a true collection gap). We release a commit-provenance dataset that separates these cases at scale by comparing two views of the same commit graph: the GHArchive event stream, a historical witness of what GitHub advertised at push time, and the World of Code (WoC) V2604 object database, an accumulated union of periodic fetches that never deletes a collected commit. Walking each reference's PushEvent chain reconstructs force-pushes structurally (a before that is not the prior head breaks the fast-forward chain; no recorded flag exists across eras), and joining every advertised commit against WoC membership yields a three-way label. Over 1,118,116,350 advertised commits, 53.35% are present in WoC, 6.47% are rewritten (orphaned by a later force-push, an upstream edit and a correct absence), and 40.18% are never-ingested (the candidate collection gap). About one missing-commit case in fifteen is a rewrite the project erased, not a mirror gap. We release the force-push witness (166,710,831 events over 19,926,250 repositories) and a per-project rollup (78,125,788 repositories; force-pushing observed in 25.47%, a rewritten commit in 12.85%). We treat the 40% never-ingested share as an upper bound on the collection gap and state five reasons it overcounts. The dataset makes mirror-completeness reporting honest, flags rewritten commits as duplicate patch sources for contribution counting, and corrects a 10.82% corpus-wide undercount that history editing imposes on commit-based productivity, with a released per-project correction factor. All artifacts ship as a self-contained, independently hosted replication package keyed to WoC V2604.

Since 2020, the Linux Foundation and the multi-organization Inclusive Naming Initiative (INI) have encouraged open-source projects to replace non-inclusive terms such as master/slave and whitelist/blacklist. Although these recommendations have been widely adopted, there is limited empirical evidence on their long-term adoption across Linux Foundation (LF) projects or their implications for AI-assisted software development. In this paper, we present NISCAN, a multilingual static-analysis framework that detects non-inclusive terminology across source code and related software artifacts using the INI vocabulary. Using NISCAN, we conduct the first ecosystem-scale study of inclusive naming across 461 Linux Foundation repositories. Our analysis shows that non-inclusive terminology has declined by approximately 47% since 2020, yet adoption remains incomplete: 62.7% of repositories still contain at least one Tier-1 non-inclusive identifier, while most remaining terminology resides outside source code in documentation, comments, configuration files, and other software artifacts. We further show that repository size, programming language, project functionality, and ecosystem are stronger predictors of term inclusiveness in LF repositories rather than foundation governance. To examine the implications for AI-assisted software development, we conduct a case study evaluating whether large language models (LLMs) can reconstruct legacy non-inclusive identifiers from surrounding program context. The results show that historical naming decisions remain embedded in model predictions even after identifiers have been renamed. Overall, our study findings provide the first ecosystem-scale assessment of inclusive naming adoption within the Linux Foundation and highlight the importance of addressing terminology residue to support responsible naming and ethically sourced code generation.

Leadership computing facilities steward large-scale scientific datasets that routinely require substantial transformation before serving as AI training data. However, no existing framework fully unifies automated transformation, readiness assessment, provenance tracking, and agent-native deployment. We present REDI, an open-source framework that addresses this gap through a unified five-stage pipeline (ingest, preprocess, transform, structure, and output) with per-stage instrumentation for reproducibility and deployment as an agent-callable skill; companion tool SetGo automates FAIR compliance and catalog publication. Evaluated across climate, proteomics, materials science, and nuclear fusion, REDI transforms all datasets from raw to AI-ready, with outputs validated against domain-expert references, and preliminary results show near-ideal parallel scaling to 100 nodes on Frontier for the climate case. Provenance-instrumented profiling reveals file I/O as the dominant pipeline cost, with format selection a first-order optimization lever. These results establish REDI as a cross-domain platform providing automated data readiness for scientific AI, transforming data preparation bottlenecks into reproducible, reusable community assets.

We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures, ranging from 2.3B to 31B parameters. Alongside improved vision and audio encoders for all model sizes, we propose a unified, encoder-free architecture for our 12B model, which ingests raw audio and image patches. Furthermore, we integrate a thinking mode, enabling Gemma models to generate reasoning traces prior to responding. We improve inference speed, memory, and compute efficiency, as well as long-context abilities through critical design choices. Gemma 4 establishes a leap in performance across STEM, multimodal, and long-context benchmarks, and rivals larger, frontier open models in human-rated tasks.

Pathologic complete response and tumor shrinkage measure whether breast cancer responds to neoadjuvant therapy, but not whether that response was structurally favorable, persistent, or hidden beneath volume loss. We built an outcome-blind longitudinal DCE-MRI manifold from I-SPY2 trajectories to test whether pretreatment imaging carries a structural response phenotype missed by conventional descriptors. The dominant axis of response geometry was not recoverable from the full clinical and genomic stack -- age, receptor subtype, MammaPrint, PAM50, treatment arm, and tumor burden -- but became strongly recoverable once baseline structural entropy was added. A constrained representation mapping recovered the same axes as unconstrained decomposition, establishing the structure as intrinsic rather than a post-hoc interpretation. The phenotype persisted through therapy, and as treatment proceeded the volumetric signal faded while entropy stayed separated -- a crossover from burden to structural persistence. Among complete responders, structurally disordered tumors could shrink more early yet remain structurally disordered, a volumetric deception invisible to endpoint labels. External analyses in UCSF, I-SPY1, and Duke established recurrence relevance under representation-dependent boundaries, and a representation-family commensurability assessment showed why feature-name matching is insufficient: the same label can fail, transport, or entangle with extraction geometry. Pretreatment MRI therefore exposes a structural response phenotype that endpoint-based language leaves invisible -- including, among complete responders, a pretreatment imaging signal of structurally distinct response states that awaits prospective validation.

Spoken Question Answering (SQA) remains largely focused on high-resource languages and carefully recorded speech, limiting the reach of speech-LLM methods in low-resource settings. This paper investigates whether text-to-speech (TTS) can provide task-specific training data for Luxembourgish SQA without requiring a large human-recorded QA corpus. Starting from existing text-based QA resources, we translate questions into Luxembourgish, synthesize spoken questions with multiple TTS systems, and pair them with textual answers. We train a parameter-efficient SLAM-style architecture that connects a frozen Whisper encoder to frozen multilingual LLM backends through a learned projector and LoRA adapters. We compare MMS-TTS, Qwen3-TTS, and OmniVoice variants, including single-source corpora of about 48k questions and a 4TTS multi-source mix of approximately 230k questions. Evaluation on LLAMA-LB-Test with two real Luxembourgish speaker conditions shows that multi-source and voice-design-based synthetic training configurations yield the strongest SQA performance. The results also show that no-reference TTS quality scores do not monotonically predict downstream QA performance, indicating that synthetic speech must be evaluated as task-specific training data rather than only as natural-sounding audio.

Audio-language models can be prompted for code-switched speech, but their decoding is not optimized for code-switching and often fails at language boundaries. We propose a practical reinforcement learning with verifiable rewards recipe for data-efficient adaptation of audio-language models to code-switched ASR using group relative policy optimization, combining an error rate reward with a script fidelity reward that penalizes wrong writing systems and a two-pass draft-and-refinement procedure. Using Qwen2-Audio as a reproducible testbed across 10 language pairs, training on only TTS code-switched speech, we show that RLVR with 10% of the data matches LoRA supervised fine-tuning trained on the full dataset, with the largest gains on typologically distant pairs. The error rate reward eliminates translation errors while the script fidelity reward separately reduces script contamination without degradation. These gains transfer zero-shot to a human-recorded code-switching corpus.

Training AIs to be risk-averse in resources could offer a failsafe in the event that AIs turn out misaligned. Misaligned but risk-averse AIs would tend to prefer low-risk, low-reward strategies like cooperation over high-risk, high-reward strategies like rebellion, limiting the downsides of any misalignment. But we can only feasibly train AIs to be risk-averse on low-stakes gambles, and we will only be safe if their risk aversion generalizes to astronomically-high-stakes gambles. Will it? To shed light on this question, we introduce RiskAverseOOD: a benchmark for measuring how well risk aversion generalizes out of distribution. We then offer some initial results. Using a variety of methods to make Qwen3-8B choose risk-aversely when the stakes are low, we find that we can induce substantial risk aversion when the stakes are astronomically high. Our models' learned risk aversion generalizes at least partially across 98 orders of magnitude. From a baseline 2% rate of choosing a safe `Cooperate' option, we see rates around 70% (SFT and tie training), 52% (DPO), and 39% (activation steering). In another experiment, our fine-tuned reward model reliably scores risk-averse reasoning above risk-neutral or excessively risk-averse alternatives (99.6% pairwise accuracy). We replicate these effects at different scales (Qwen3-1.7B and Qwen3-14B) and across model families (Gemma-3-12B-IT and Llama-3.1-8B-Instruct). Overall, we find that risk aversion learned at low stakes can generalize OOD to astronomically high stakes, though not yet consistently enough to serve as a reliable failsafe. Achieving that level of consistency is an open problem.

Neuromorphic controllers for size, weight, and power-constrained systems require neural architectures that are both energy-efficient and interpretable at the level of system dynamics. However, existing approaches either rely on end-to-end trained spiking networks with limited interpretability, or on converted classical controllers that fail to fully exploit neuromorphic dynamics. We present a spiking neural network (SNN) architecture for generating polar trajectories, using a winner-take-all (WTA) architecture with accessory populations that induce controlled transitions in neural activity. We demonstrate tuning rules for these population dynamics, and utilize a form of shunting inhibition to enable independent control of direction, speed, and radius of the resulting polar trajectories. We implement the network on the SpiNNaker2 neuromorphic processor, and demonstrate a two to three orders of magnitude reduction in wall-clock step time and three to four orders of magnitude reduction in energy expenditure when compared to conventional computing platforms.

The Variational Quantum Eigensolver (VQE) is essential for molecular simulation in drug discovery, but hardware noise and algorithmic limits restrict its precision. While the NVIDIA CUDA-Q platform mitigates some hardware issues via exact simulation, it lacks Qiskit support and restricts parallelization. To solve this, we introduce Graph-VQE, a novel framework that extends CUDA-Q with optimization-level parallelism. Graph-VQE leverages amino acid sequence structures by partitioning Hamiltonian interaction graphs into weakly coupled clusters using Louvain community detection. These clusters undergo restricted updates on the full-Hamiltonian objective, followed by a global refinement stage utilizing Hamiltonian batching. Furthermore, a custom Qiskit-CUDA-Q integration layer enables standard workflows with GPU acceleration. Evaluations on protein folding tasks prove that Graph-VQE outperforms baselines, achieving lower final energies. It delivers competitive RMSD and binding affinity compared to AlphaFold3 and IBM quantum processors while maintaining stable quality across multi-GPU environments, thereby providing a highly practical path toward high-fidelity biomolecular simulations.

eBPF lets developers run custom programs inside the Linux kernel, where a verifier proves each program safe. However, when the verifier rejects a program, the unclear error makes repair challenging: the error reports where verification stopped, not where the program lost the proof the verifier required. To quantify this gap, we conduct an empirical study of 235 reproduced rejections, showing that 47% of rejections return only EINVAL, one error string maps to as many as nine distinct root causes, and 10 of the 12 root causes are eBPF-specific. Repair thus requires both domain knowledge and locating where the proof was lost, yet existing tools only help developers read the error. We present bpfix, which reconstructs where the required proof was established and where it was lost from the verifier log, and prints a Rust-like diagnostic. To evaluate bpfix and the ability of LLMs to help repair, we construct a benchmark of 75 LLM repair tasks. Current models achieve 0-37% one-shot success with the raw log, and replacing the log with the bpfix localization improves repair by 11-21pp, suggesting that locating where the proof was lost is key to guiding repair. bpfix is available at https://github.com/eunomia-bpf/bpfix

Machine learning has shown speedups for numerical methods in physics applications, but integrating Python-based libraries into high-performance C++ solvers creates performance bottlenecks. We present CodeJeNN, which bridges this gap by auto-generating self-contained C++ code from trained Keras models for inference. This eliminates external dependencies through minimal inlined functions, allowing seamless integration into existing frameworks. We describe the Keras-to-C++ workflow, supported architectures, and limitations. CodeJeNN is demonstrated through inference benchmarks against Keras in eager and JIT modes and a CFD test case modeling viscosity in a hydrogen-air mixing layer, showing speedups without sacrificing accuracy.

Software design patterns' effects on code quality have mostly been studied in the context of object-oriented languages. In the programming language Rust, which comes with novel language concepts, compile-time safety guarantees and a distinct type system, there has been little research on design patterns. This work investigates how patterns affect software quality and compile-time enforcement of invariants through a case study on three representative components of production backend applications. An evaluation framework based on criteria derived from the SQuaRE quality model, incorporating benchmarking, static code analysis and expert interviews, is developed to assess the refactored code. The patterns typestate and newtype are applied to address existing code smells in the selected use cases. While the typestate pattern improves faultlessness and testability significantly, it comes at the cost of more structural code that can degrade readability. Code with extensive branching logic and a high number of invariants is likely to benefit most from the pattern. The newtype pattern combined with the "Parse, don't validate" principle offers high returns in software quality at a low cost and prevents invalid states during runtime. Overall, this work provides an initial empirical assessment of design patterns in Rust and establishes a foundation for further studies involving additional use cases and patterns.

This paper develops an Optimality-Informed Neural Network (OINN) approach for the energy-optimal, free-final-time powered descent of a lunar lander from any initial position, velocity, and mass within a bounded operating envelope to a fixed landing site with zero terminal velocity. Building on a recent framework that jointly embeds Pontryagin's minimum principle and the Hamilton-Jacobi-Bellman equation for general nonlinear optimal control, the proposed OINN approach specializes that idea to a lunar landing problem with free time of flight and fixed terminal state. Every boundary and transversality condition is hard-encoded into the network architecture by construction, the closed-form Pontryagin-optimal thrust magnitude and direction law is substituted directly rather than learned, and the remaining state, costate, and an auxiliary value-function output are trained against a physics-residual loss formed entirely from the necessary conditions of optimality, with no precomputed optimal trajectories required. A preliminary theoretical analysis is explored, establishing a stochastic-optimization stationarity guarantee for the offline training procedure, an explicit bound translating the achieved training residual into bounds on touchdown position, touchdown velocity, and flight-time error, and a fixed, input-independent onboard computational and memory cost suitable for real-time deployment. Numerical simulations evaluate the trained policy, with no retraining, against an independently solved indirect-method boundary-value problem at six representative initial states spanning the operating envelope and against eighty additional Monte Carlo simulation runs, demonstrating close agreement with the indirect-method solution and consistently small dynamics and transversality residuals throughout the envelope.

Rapid growth in social media has transformed global communication by enabling fast information exchange, but it has also accelerated the spread of misinformation. Fake news, manipulated content, and provocative narratives are increasingly linked to social unrest, political instability, and mob violence. Incidents in South Asia and elsewhere demonstrate how false information disseminated via platforms such as Facebook and WhatsApp can trigger real-world harm, often spreading faster than fact-checking efforts can respond. To address this challenge, this chapter presents a multilingual, multimodal Natural Language Processing (NLP) framework for early detection of misinformation and violence-prone dynamics. A fused dataset of 138,256 Bangla and English samples was created by combining multiple benchmark datasets. The framework integrates XLM-RoBERTa for multilingual text representation, CLIP for visual embedding, and a multi-head attention mechanism for multimodal fusion, enhanced with auxiliary features such as sarcasm and geospatial metadata. Experiments on a stratified 30% subset achieved 98% test accuracy with strong precision and recall. The outcomes show the efficacy of multimodal approaches in early misinformation detection and highlight the added value of geospatial signals for anticipating real-world escalation.

Machine learning has demonstrated significant potential for real-time monitoring, optimization, and control of scientific facilities. However, deploying and maintaining ML models in operational environments remains a substantial engineering challenge. Each facility presents unique data protocols, non-standard formats, and infrastructure constraints, forcing teams to rebuild integration pipelines for every new application. We present SMOCS (Streaming Monitoring Optimization and Control System), a Kafka-based containerized framework that addresses this challenge through three contributions: 1) a layered abstraction over Apache Kafka that separates infrastructure from application logic, 2) a three-thread agent architecture that temporally decouples data ingestion, model training, and real-time inference enabling continuous online learning from live data streams, and 3) a configuration-driven deployment model that enables domain experts to operate ML pipelines without software engineering expertise. SMOCS is facility platform-agnostic, fault-isolated by design, and horizontally scalable through Docker containerization. The framework is publicly available as open-source software on the Jefferson Lab Github.

Network-on-Chip (NoC) architectures have become the standard interconnect fabric for many-core systems, yet most proposals face a fundamental trade-off between latency, area, and congestion management. This paper presents HyNoC (Hybrid Network-on-Chip), an open-source NoC architecture that combines circuit-switch path establishment with wormhole data transfer, targeting distributed computing systems built around VLIW processor cores on FPGA. HyNoC employs source routing, where the complete path through the network is encoded in the packet header by the sender or statically at compile time, enabling both deterministic low-latency transfers and software-level hotspot avoidance without the area overhead of virtual channels. The router features a parallel round-robin arbiter (PRRA) with fixed grant latency, per-port independent clock domains, and support for both unicast and multicast routing. We discuss the design rationale, the hybrid switching model, and positioning relative to prior NoC art, and argue that a richer topology combined with compiler-assisted static routing is a competitive alternative to virtual channels for FPGA-based distributed VLIW systems. Verilator co-simulation measures the deterministic per-hop latency and benchmarks the design at LLaMA 3 8B FFN up-projection scale: a four-master quadrant configuration partitions the mesh into traffic-isolated quadrants with zero cross-quadrant link traffic, yielding a 5x throughput improvement over a single master. The complete RTL implementation is available as open-source hardware under the CERN-OHL-P v2 license.

Reliable internet access is essential for modern education, yet millions of school-aged children especially in developing regions remain offline due to unconnected schools. The Giga Initiative aims to connect every school to the internet, but doing so at scale requires efficient methods to map schools and assess surrounding connectivity infrastructure without relying on sparse or noisy third-party datasets. In this work, we propose a scalable, vision-only framework that uses high-resolution satellite imagery and transfer learning to address both tasks simultaneously. By adapting pre-trained object detection models to new geographical regions with minimal labeled data, we detect schools and cell towers directly from space. We then analyze the spatial relationship between detected schools and nearby towers as a proxy for connectivity availability. This purely imagery-driven pipeline enables large-scale infrastructure mapping, reduces dependency on auxiliary data, and supports data-driven prioritization of connectivity investments in underserved areas. Our approach is demonstrated on real satellite imagery from Lesotho, showing strong performance across this region.

Predicting thermal volatility in high-performance EV powertrains is difficult as internal temperatures are rarely observable outside the lab, and models calibrated on lab drive cycles fail when deployed against real-world loads. We study this lab-to-track transfer problem using conformal prediction, offering distribution-free uncertainty bounds. We implement Ensemble Batch Prediction Intervals (EnbPI; Xu & Xie, 2021), a leave-one-out bootstrap-ensemble conformal method for autocorrelated time series, and calibrate it on real CALCE lithium-ion cycler data (A123 SP20 cells, FUDS profile). We evaluate it under a genuine, measured covariate shift: a second real CALCE test condition (US06 Highway Driving Schedule at 45°C). The unweighted EnbPI bound, achieving its nominal 95% coverage in-distribution (measured: 95.00%), degrades to 70.13% empirical coverage under this real shift. We introduce a weighted EnbPI procedure combining EnbPI's ensemble residuals with density-ratio weighting (Tibshirani et al., 2019), estimating the density ratio via a probabilistic domain classifier. This recovers coverage to 72.42%, a modest, honestly-reported improvement, not a complete fix. We additionally apply the calibrated model to real 2023 Formula 1 telemetry (Monza and Silverstone, driver VER) as an unsupervised out-of-distribution diagnostic. Because no internal thermal channel exists in public trackside telemetry, we report only unsupervised flag rates (65.6% at Monza, 58.0% at Silverstone, well above the 5% in-distribution base rate) and note inconsistent associations between flags and braking/DRS zones. We conclude that conformal domain adaptation is a promising but only partially solved tool for this problem, detailing exactly where it falls short.

3D Gaussian Splatting (3DGS) enables high-quality real-time novel-view synthesis, but practical scenes often contain millions of Gaussians, making compression essential for deployment on limited hardware. Existing reduction methods are effective but mostly heuristic: they provide no multiplicative approximation guarantee for the rendered objective, and thus rely heavily on costly post-pruning finetuning to recover quality. We ask a basic question: can a 3DGS scene be provably replaced by a much smaller weighted subset (coreset) while preserving the objective of interest? We first show that, in the unrestricted setting, no non-trivial multiplicative 3DGS coreset exists. We then show that multiplicative guarantees are not impossible, but resolution-dependent. For a prescribed rendering resolution, such as representative views or grids of views/rays, we provide the first weighted coreset construction theorem for 3DGS. The construction samples Gaussians by sensitivity: provable importance scores measuring each Gaussian's role in the full-scene objective. Finally, under explicit validity and log-transmittance stability assumptions, we turn this objective guarantee into a rendering guarantee. Empirically, our method is strongest where deployment needs it most: aggressive compression with no or minimal recovery compute. In prune-only and very short finetuning regimes, it achieves state-of-the-art performance, showing that principled importance estimation can be both theoretically meaningful and practically useful. Open-source code is available at https://github.com/waseem-m/3dgs_provable_coresets.

Recent advances in large-scale image generative models enable photorealistic scene synthesis with controllable attributes. Beyond data augmentation, their potential as diagnostic tools for trained vision systems remains unexplored in the aerial and remote sensing domains. We introduce a synthetic diagnostic framework for aerial-view vehicle detection that combines text-guided generation, attribute-controlled editing, and automated attribute verification to construct a controllable synthetic testbed. This enables fine-grained evaluation of pretrained detectors under diverse scene types and environmental conditions that are difficult to isolate in real datasets. Across three detection architectures and three real aerial datasets, synthetic scene-wise performance trends closely match real-world weaknesses. Guided by these diagnostics, targeted supplementation with small real datasets from the identified weak categories yields improvements of up to 13% AP50 while requiring substantially fewer additional samples than non-targeted augmentation. Our results show that controlled synthetic probing can predict real-domain performance gaps and guide efficient data collection. The proposed diagnostic framework is modular and can incorporate alternative generative or vision-language models as capabilities evolve. Our code and datasets are available here: https://humansensinglab.github.io/AVODDiag/

Urban mobility modeling faces challenges in representing decision-making in dynamic environments. Although Multi-Agent Systems are widely used, rule-based approaches rely on fixed heuristics that limit adaptive behavior. This work investigates the integration of Large Language Models (LLMs) as decision-making components in multi-agent simulations. We propose a hybrid architecture that connects the GAMA platform to an external LLM-based module through an API, enabling agents to determine whether route replanning is necessary. Rather than replacing routing algorithms, the LLM serves as a decision layer that guides replanning behavior. The approach incorporates persistent memory, allowing past interactions to influence future decisions and promote behavioral consistency. We compare rule-based and LLM-assisted approaches across multiple road-blockage scenarios and population scales. Results indicate that LLM-enabled agents exhibit greater adaptability and contextual awareness, particularly in scenarios with higher route flexibility. Memory influences performance and behavioral consistency, with effects varying across configurations. Overall, LLMs serve as complementary cognitive layers that enrich behavioral representations in urban mobility simulations and hold potential for modeling complex decision-making in spatial multi-agent systems.

Recently, neural operators have shown promising outcomes for learning solution operators of differential equations directly from data. This framework learns a functional mapping from the parameter field to the solution field, enabling the prediction of an entire class of solutions rather than a specific instance. However, existing operators often struggle to capture both global dynamics and fine-scale structure simultaneously. To design an effective operator capable of representing multiscale features, a hierarchical multiscale decomposition framework is required. In this study, we develop the Lifting Neural Operator (LiNO), a multiresolution operator built on the second-generation wavelet lifting scheme. LiNO learns a multiresolution decomposition directly from data by parameterizing the lifting transform. This lifting transformation is adaptive to the underlying solution function and exactly invertible by construction, enabling information-preserving multiscale operator learning. In the lifted multiresolution space, the operator evolves coarse and directional detail coefficients separately, resulting in scale-aware modeling of the underlying physics. We evaluate LiNO on several benchmarks, including Darcy flow, the Poisson equation, the Allen-Cahn equation, the compressible Navier-Stokes equation, and the Gray-Scott reaction-diffusion system. Together, these benchmarks cover a wide range of physical behaviors, including multiscale phenomena, transport-dominated dynamics, and chaotic systems. LiNO demonstrates strong performance on these challenging benchmarks compared with state-of-the-art neural operators. These results suggest that adaptive multiresolution operators provide a promising direction for scientific machine learning.

There is no doubt that safety alignment is an essential step in LLM training. However, conceptually it does not distinguish between various domains and the level of potential harm of a query, which creates significant complications in the fields like cyber security, where a model should not be constrained by its safety circuits to accomplish the goals of legitimate, authorized operations. In this work, we share our findings from a large scale abliteration experiment on 24 open-source LLMs and show that domain-specific abliteration is achievable with standard methodology on the example of a 1T-parameter Kimi K2. Building on recent work showing that refusal in LLMs occupies a multi-dimensional subspace within layers, we find that it is also distributed widely across layers, especially in trillion-parameter MoE architectures, and so we aim to capture the part of it that represents harmful concepts in the cybersecurity domain exclusively. We also investigate the correlation between models' features and the effect of domain-specific abliteration, identifying that the type of safety training and architecture are the most reliable predictors. Finally, we classify the models into 3 \emph{abliteration susceptibility} tiers and put forward a set of conjectures as to why a particular effect from this intervention might be observed in a given model.

In this paper, we describe LLMoxie, an institutional AI platform whose three-tiered architecture supports multi-cloud and on-premise inference, a LiteLLM/MLflow control plane for authentication, budgeting, PII masking, and observability, and an application augmentation layer for AI coding agents. Layered on top, an open-source RSE-Plugins ecosystem encodes accumulated RSE knowledge as a Plugin-Agent-Skill hierarchy spanning scientific Python practice, domain-specific knowledge, a six-phase research-and-implement workflow, and project lifecycle management. Scientific software is judged less by raw code quality than by whether it can be cited, audited, reproduced, and extended. Off-the-shelf AI coding agents, optimized against commercial software benchmarks, are poorly calibrated for this setting: they ignore the conventions of the scientific Python libraries they invoke, mishandle sensitive or embargoed data, and leave decision trails that are difficult to reconstruct after the fact. We report on twenty months of practice at a university-based research software engineering (RSE) center, where RSEs embedded across astronomy, earth and climate science, agriculture, and health projects worked to close this gap. We characterize the recurring infrastructure, governance, and process challenges of adopting Agentic AI inside a multi-domain RSE center, describe the platform and plugin design, and distill operational lessons from real scientific software deployments. Together, the platform and plugins shift AI coding agents from generic code generators into domain-aware collaborators that respect community norms and produce auditable provenance of technical reasoning.

This study presents a conditional Generative Adversarial Network (cGAN) framework for generating 3D porous media volumes with controlled porosity, trained exclusively on 2D thin section images. The key innovation lies in combining property-conditioned generation with 2D-to-3D reconstruction, eliminating the need for expensive 3D training data while maintaining control over petrophysical properties. The framework employs a hybrid architecture with a 3D generator and 2D discriminator, where multi-axis slice extraction enables learning 3D-consistent structures from 2D training data. Porosity labels are extracted using an Enhanced U-Net segmentation model. The methodology was demonstrated on two carbonate samples with different lithologies: dolomite-anhydrite and pure dolomite. Results show that the framework successfully generates realistic 3D volumes capturing lithological features such as anhydrite inclusions and fine crystalline textures. Porosity control achieved an $R^2$ of 0.93, with mean absolute errors of 0.019 and 0.010 for the heterogeneous and homogeneous samples, respectively.

As wearable devices enable continuous first-person recording, AI assistants must reason across long time horizons to recall past experiences-a capability known as episodic memory. Current benchmarks often rely on offline evaluation with access to entire video files, failing to simulate the streaming reality of wearable intelligence. We introduce S-EMBER (Streaming Egocentric Memory Benchmark for Episodic Retrieval), a large-scale benchmark comprising 3,141 videos totaling 388 hours of organic activity captured via Ray-Ban Meta smart glasses. S-EMBER formalizes grounded streaming episodic retrieval, a paradigm shift from global offline search to causal, active recall triggered by visual events in a continuous stream. We provide 9,448 QA pairs requiring manual visual proof through precise temporal localization and supporting flexible response lengths to simulate natural human-AI interaction. Our extensive benchmarking of frontier models uncovers a localization paradox: while semantic reasoning improves with parameter scale, temporal grounding precision remains a stagnant architectural bottleneck that does not benefit from brute-force increases in model size, resolution, or frame density. S-EMBER establishes a hardware-authentic foundation for developing grounded, reliable episodic memory in the next generation of wearable AI agents.

Intrusion detection systems are often trained under static benchmark conditions, although deployed network environments are affected by traffic drift, sensor noise, changing workloads, and evolving attack behaviour. Under such distribution shifts, static detectors may produce confident but incorrect predictions, leading to silent and unsafe failure modes. In this paper, RES-DARE (Recursive Evolving Specialists-Digital Adaptive Reasoning Engine) is proposed as a failure-aware continual intrusion detection framework with rollback-safe self-repair. Difficult, uncertain, and misclassified samples are treated as failure signals for expert specialisation rather than being discarded as noise. A supervised contrastive encoder, two-pass expert router, failure-buffer mechanism, HDBSCAN-based failure-region discovery, and trust-risk monitor are integrated to support adaptive IDS behaviour. AEHM-v2 is introduced as a rollback-safe repair mechanism, where candidate adaptations are provisionally activated and committed only when macro-F1 is preserved or improved while trust risk remains stable. Otherwise, the system is rolled back to its last validated state. RES-DARE is evaluated on CICIDS2017, UNSW-NB15, and TON\_IoT, achieving macro-F1 scores of 0.9850, 0.9736, and 0.9691, respectively. Under Gaussian feature corruption at strength 0.10, RES-DARE retains an Attack-F1 of 0.7920 on CICIDS2017 and achieves near-zero catastrophic forgetting with F = 0.0015. The results show that RES-DARE improves robustness, warning capability, and deployment safety under degraded conditions.

Reinforcement learning agents operating under partial observability must act on incomplete information, making them natural candidates for guidance from small language models (SLMs) that carry broad reasoning priors. Yet integrating SLM guidance into this setting has proven difficult: across all test environments, vanilla uncertainty-gated approaches achieve an overwrite rate at or near zero, meaning the SLM almost never contributes an independent action. We trace this failure to the bare egocentric prompt, which provides insufficient context for genuine reasoning, and identify it as a context problem rather than a capacity problem. We propose ASK+, which supplies the SLM with trajectory-aware context (a partially revealed map, visited positions, and action history) and structured chain-of-thought reasoning, converting it from a passive redundancy check into a more informative consultant that occasionally corrects the policy. We further establish that the predictive entropy signal used for selective querying measures action uncertainty rather than state uncertainty and remains informative in POMDPs, making uncertainty-gated assistance viable beyond fully observable settings. The stateful prompt drives substantial gains: on DoorKey, where vanilla ASK matches PPO (both 89%), ASK+ reaches 93% success; on FourRooms, success climbs from 53% to 70%; on HigherLower, accuracy reaches 73.7%, matching the SLM-only upper bound. Across all environments, Qwen3.5-2B matches or exceeds Qwen3.5-4B, confirming that prompt design and selective gating dominate the impact of model scale, enabling guidance without large models.

Coding agents built on large language models are now capable of patching sizable real-world codebases, yet whether they can develop compiler optimizations remains an open question. To study this question, we introduce PeepholeBench, an evaluation framework whose tasks are constructed from real-world missed peephole optimizations reported against LLVM's InstCombine pass. Since missed peephole optimizations are typically fixed with small, localized patches, they offer a well-scoped but demanding testbed for coding agents: a correct fix demands rigorous reasoning about program semantics along with familiarity with optimizer-specific conventions. PeepholeBench derives its tasks from 21 resolved LLVM issues and 19 merged pull requests (PRs), supplies agents with only the issue context that existed before each fix, and assesses the resulting patches for both correctness and profitability. With PeepholeBench, we benchmark state-of-the-art coding agents on fixing missed peephole optimizations in LLVM's InstCombine pass, measuring their patches against the corresponding human-written fixes. We observe a tension between correctness and profitability, and no agent matches human developers on both dimensions at once. The dominant failure modes are overly narrow transformations and misuse of LLVM-specific mechanisms, errors that existing test suites rarely expose. Together, these results establish PeepholeBench as a realistic and challenging benchmark for coding agents, and suggest future directions for building agents that can more dependably assist compiler optimization development.

Integrating information across related tasks can improve estimation and prediction in transfer, multi-task, and federated learning, but contamination and heterogeneity make robust borrowing challenging. We study a contaminated multi-task empirical risk minimization (ERM) framework in which an $ε$ fraction of $K$ tasks, each with sample size $n$, may be arbitrarily contaminated while the remaining tasks are heterogeneous. Our goal is to estimate both the global minimizer of the average risk and the clean task-specific minimizers, thereby combining robustness and personalization. In the Gaussian mean model, we show that several common paradigms, including adaptive and robust regularization around a shared center, global matrix regularization, decomposition-based regularization, and score-based outlier-task detection, all suffer from a worst-case contamination error of order $ε\sqrt{d/n}$, which is suboptimal compared to the lower bound $ε/\sqrt{n}$. This identifies a dimension-dependent barrier for these approaches. We then establish minimax lower bounds for a general heterogeneous ERM setting and propose a computationally efficient filtering-based robust multi-task gradient descent method. Under local strong convexity, smoothness, and sub-Gaussian gradient assumptions, the proposed method attains high-probability upper bounds matching the minimax rates up to logarithmic factors over a broad regime. In particular, it removes the extra $\sqrt{d}$ contamination dependence of many regularization-based methods and score-based outlier detection, while achieving personalization to local tasks under strong heterogeneity. Simulations and a real-data analysis demonstrate strong robustness and personalization relative to a broad range of benchmark methods.

MLLMs have shown strong zero-shot capabilities across diverse inputs such as across images, video, audio, and text. A crucial, yet underexplored, application of these models lies in understanding and modeling animal-centric scenarios. As animals are integral to millions of households, benchmarking next-generation AI models on pet-focused tasks, ranging from recognizing distress signals to enabling responsive robotic companions, is essential for building AI systems that can work alongside us. We introduce K9-Bench, a novel benchmark focused on real-world domestic dog videos, specifically targeting canine action and interaction understanding via approximately 5000 question-answer pairs across 907 videos spanning 5 distinct task categories that test long-form, canine-centric multimodal reasoning in MLLMs. To create this dataset, we propose a scalable, VLM/LLM-powered data generation pipeline that automatically mines canine-centric videos from the web and curates QA pairs requiring fine-grained, multi-hop reasoning over canine actions and temporally extended interaction sequences. We implement bias mitigation strategies designed to eliminate biases introduced by VLMs during dataset curation. Through extensive experimentation, we find that frontier MLLMs exhibit limited zero-shot performance on canine-centric tasks: although state-of-the-art closed-source models outperform open-source counterparts, they still struggle with compositional reasoning over subtle posture and interaction cues spread over long horizons. We observe that generic chain-of-thought prompting provides only modest performance for such long-horizon reasoning. Beyond a novel dataset for canine activity analysis, K9-Bench provides a general-purpose dataset construction pipeline that can be adapted to other low-data domains for quantitative analysis. Our project website is available at: https://ogmenrobotics.github.io/K9Bench.

Local pairwise comparisons are a standard tool for learning how people want decision rules to work, e.g., in participatory design or alignment. However, their use builds in two strong assumptions: that local comparisons are sufficient evidence about how a person wants an automated decision rule to behave, and that people can always answer those comparisons decisively. We investigate how these assumptions may be compromised under internal pluralism: the idea that an individual evaluates decision rules according to multiple authoritative priorities about how the rule should behave. We provide a formal model of such pluralistic preferences over decision rules, which then lets us identify two distinct failures of forced local pairwise comparison data. First, priorities such as proportionality, egalitarianism, and equal treatment are inherently global: what they imply in one case can depend on what happens elsewhere, so local comparisons may fail to capture them. Second, even when priorities are representable locally, tension between strongly-held priorities can generate internal conflict, producing potentially costly behavioral distortions when comparisons are forced. We then use our model to investigate the alternative -- allowing people to report indecision -- and our findings suggest that doing so can considerably reduce the number of queries needed to learn preferences accurately. We conclude by describing how our model points toward preference-learning methods that elicit these priorities directly, yielding more faithful and interpretable accounts of what people value.

Benign overfitting and double descent have come to shape our understanding of generalization in deep learning, establishing that overfitting is not only compatible with good generalization but can actively benefit it. Diffusion models share much of the machinery of standard deep learning, so it is natural to assume that they also exhibit these properties. In this work, we show that this assumption is largely incorrect. We first establish fundamental impossibility results showing that, unless the sample size grows exponentially with the data dimension, overfitting and good generalization cannot occur simultaneously. Consequently, the population loss follows a classical U-shaped curve in model complexity rather than exhibiting double descent. Analyzing a simplified setting, we identify a key difference between regression and score matching: regression benefits from an alignment between the target and the empirical covariance; score matching admits no such alignment, leaving overfitting irreparably harmful. We further identify implicit regularization stemming from time-smoothness of the score and early stopping during training as mechanisms that prevent such overfitting and verify our findings with high-dimensional image generation experiments. Our results reveal that generalization in diffusion models is governed by mechanisms distinct from those of traditional regression, motivating the development of new theory.

Electroencephalography (EEG) offers a noninvasive approach for examining neurophysiological correlates of dimensional psychopathology, yet systematic evidence across EEG paradigms and feature granularities remains limited. Here, we develop a granularity-aware EEG feature pipeline that organizes multi-scale descriptors into global, regional, and channel levels. Using the Healthy Brain Network (HBN) cohort, we evaluate the prediction of four psychopathology dimensions: p-factor, internalizing, externalizing, and attention problems, across four EEG paradigms. Given the heterogeneity of pediatric psychopathology and the moderate reliability of questionnaire-derived scores, this setting represents a challenging feasibility test rather than a clinical screening scenario. Tree-based models and granularity-balanced feature selection showed promising improvements over conventional approaches in selected conditions, although effect sizes remained modest. Visualization of selected markers revealed dimension-specific spatial and spectral patterns that were broadly aligned with existing neurophysiological knowledge. An exploratory cross-dataset sanity check on the independent PEARL cohort suggested that the proposed selection principle remains technically feasible under protocol shifts, without claiming cross-dataset generalizability. Overall, multi-scale EEG features contain weak but detectable signals related to dimensional psychopathology, and granularity-aware selection may serve as a useful feature-reduction strategy for future EEG-based phenotyping studies.

Large language models are inconsistent: varying prompts or including unrelated information can lead to unexpected changes in model outputs. The generator-validator (G-V) gap is one manifestation of this phenomenon, where LLMs generate responses that they then deem as invalid if re-queried to validate them. In this work, we introduce a new formulation of G-V consistency that involves a principled correction for utterance frequency. Specifically, generators often assign low likelihood to valid strings simply because those strings are a priori unlikely, which makes naive notions of G-V consistency unworkable. We show that under a natural model of rational agents answering questions with multiple answers, consistency of the validator with a frequency-corrected generator score emerges naturally. Our method, \emph{\FCPAname} (\FCPA), is a training objective implementing frequency-corrected G-V consistency for real-world LLMs. Our experimental results show that training with \FCPA{} substantially improves both G-V consistency and generator performance over prior methods, with gains of up to $+27$pp in Pearson correlation on IFEval and HumanEval, while preserving validator quality across all evaluated tasks.

Natural Language Processing (NLP) has traditionally been published in its core disciplinary venues like ACL. However, advances in Large Language Models (LLMs) has led to a blurring of the disciplinary lines between NLP and general Machine Learning (ML), with authors regularly publishing in venues from both fields. Here, we ask whether the disciplinary center of gravity is shifting. Using NLP research published from 2010 to 2026 and studies of both established and new authors, we find that a migration is taking place. First, comparing the pre- and post-LLM eras, established authors lost 19.2pp of share at flagship *ACL main-conference tracks while gaining 14.8pp in the newer Findings tracks, and general ML venues rose 8.6pp, even when adjusting for parallel growth in the fields. Second, among newer authors who debut with at least three first-author NLP-topic papers, the share whose work appears mostly at *ACL venues fell from 84% (2019) to 74% (2024), while the share appearing mostly at general ML venues rose from 5% to 21%. Using causal inference techniques, we estimate that these general ML venues confer a significant citation premium, which influences venue selection. Together, these results point to a significant shift in where NLP research is published.

Compiler missed optimizations refer to cases in which compilers failed to optimize certain code. It takes many compiler developers' efforts to implement or patch such missed optimizations. In this paper, we present a systematic study of how well agents patch compiler missed optimizations. We identify a significant challenge that patching a missed optimization requires more than just fixing the reported case, and instead requires generalizing to similar cases. We construct a benchmark of real-world LLVM missed optimization issues and compare agent-generated patches with patches from developers in terms of optimization scope. Our results show that coding agents often optimize the given examples, but many generated patches either cover only part of the developer-intended scope or partially overlap with it; in some cases, they further generalize beyond the reference patch. We further introduce historical-knowledge augmentation techniques that leverage prior LLVM optimization pull requests through retrieval and distillation, showing that they improve developer-aligned generalization and yield practical benefits when applied to real-world IR.

Monocular relative pose sensing is a central perception problem in non-cooperative rendezvous and on-orbit servicing. In spacecraft images, however, weak surface texture, thin appendages, illumination changes, and partial occlusion often leave only sparse and unstable geometric evidence. This article presents GAP-GDRNet, a geometry-aware attention-enhanced framework for monocular RGB-based 6D pose sensing. The method follows the geometry-guided direct regression paradigm of GDR-Net and modifies two points in the pipeline: an attention-based feature refinement (AFR) module is placed before dense geometric prediction, and a patch-level geometric self-attention (PGSA) module is inserted into Patch-PnP. AFR reinforces global spacecraft structure together with local weak-texture cues; PGSA then relates downsampled geometric patches before final pose regression. A Blender-based annotation process supplies target masks, visible-region masks, dense model-coordinate maps, camera intrinsics, and 6D pose labels for supervised training.The primary evaluation uses a single-target synthetic spacecraft dataset built from one CAD model. GAP-GDRNet reaches a rotation error of $1.96^\circ$, a translation error of 0.0165 m, and an ADD@0.02 m accuracy of 95.16\%, exceeding the reproduced GDR-Net baseline by 3.88 percentage points while running at 35.97 FPS for pose-network inference. Supplementary evaluations on T-LESS and LM-O give BOP AR gains of 6.8 and 3.1 percentage points over the reproduced baseline.

LLM coding agents increasingly rely on third-party agent skills from public marketplaces, which execute with the agent's privileges and create a software supply-chain attack surface: a malicious skill can steal credentials, exfiltrate source code, or install backdoors. Existing defenses use static skill scanners based on pattern matching or LLM-as-judge analysis, but it remains unclear whether they withstand adaptive evasions that preserve malicious behavior while changing payload appearance. This paper first presents an adversarial study of existing skill scanners through SkillCloak, a payload-preserving evasion framework that keeps the attack semantics intact while transforming their visible form. SkillCloak uses two complementary strategies: Structural Obfuscation, which rewrites visible payload indicators into semantically equivalent forms, and Self-Extracting Skill (SFS) Packing, which hides malicious components from the install-time view and restores them during agent execution. Across eight scanners and 1,613 in-the-wild malicious skills, SFS Packing bypasses every scanner at over 90%, while Structural Obfuscation bypasses over 80% on most static scanners and reaches 96% on a hybrid scanner, showing that appearance-based auditing is insufficient. Motivated by this finding, we propose SkillDetonate, a behavior-centric runtime auditor that executes skills in a sandbox and detects malicious effects through OS-boundary information-flow evidence rather than install-time appearance. SkillDetonate combines on-demand closure lift, which observes instructions materialized during execution, with marker-based taint analysis, which tracks sensitive-data flows across the agent context, files, processes, and network operations. The results show that SkillDetonate detects 97% of attacks at a 2% false-positive rate and sustains 87% detection on real-world malicious skills.

Real-time interaction models -- Moshi, MiniCPM-o, Qwen-Omni -- turn serving into a periodic real-time task: on every frame a session ingests streaming audio and must respond by a recurring wall-clock deadline, while its KV cache grows monotonically and stays pinned for the whole conversation. This regime hides a dangerous failure mode. On a real full-duplex stack, sustained load does not degrade serving gracefully: it falls off a cliff, jumping in one step from milliseconds per frame to a stalled engine when accumulated session state exhausts the KV pool. The collapse is metastable -- identical five-minute runs collapse or survive on run-to-run variance -- and silent: latency and deadline-miss metrics read healthy throughout. We show one move restores both stability and observability: bound each session's resident state, and latency starts telling the truth. Metronome's in-engine KV window eliminates the collapse (0/20 vs. 14/20 runs across two batches) and turns per-frame latency into a monotone load signal, on which an online admission controller discovers the schedulable concurrency; without the window, the identical controller over-admits into the wall. A first-order model predicts the collapse time within a few percent on the headline model, and a quality probe validates the bound's design by ablation: the window alone is quality-free in turn-based decoding, and its few pinned attention-sink tokens are what keep free-running generation healthy. Everything is measured end-to-end on real audio, across four interaction models on one GPU.

Recent generative models can produce high-quality synthetic images, offering scalable training training data for data-hungry models. Existing approaches to exploiting this potential typically involve 1) training or fine-tuning generators, or 2) using lightweight post-hoc adaptation like prompt engineering or inference-time guidance, making them generator-specific and expertise-intensive. We study a complementary question: given a fixed pool of generated images, can downstream utility be improved purely by selecting an informative subset? The answer is yes. We show that effective selection must counter a structural bias of modern generators: they tend to over-produce canonical modes of each class while underrepresenting intra-class variation. Building on this insight, we split each real class into a canonical Homogeneous (HO) subset and a non-redundant Heterogeneous (HE) subset, then score synthetic images by a fidelity-diversity criterion that rewards semantic alignment while penalizing canonical redundancy. The method is generator-agnostic and requires no retraining. Across multiple benchmarks, it consistently outperforms state-of-the-art data selection baselines and matches the real-data performance with up to 40% fewer synthetic samples. The same criterion remains effective when applied on top of stronger task-tuned generators, with gains on both classification and segmentation tasks. Post-generation selection is therefore not a substitute for better generators, but a complementary mechanism for improving the utility of synthetic data.

Object detection is a fundamental capability for AI-driven perception in safety-critical drone and edge-vision systems, including disaster response, operational security environments, infrastructure monitoring and defense applications. Robust model performance in such environments depends on large, continuously updated datasets. However, training high-performing detectors typically requires centralizing aerial imagery, which raises privacy, regulatory, storage, and bandwidth challenges. This is especially problematic in distributed drone deployments, where visual data is generated onboard and is often impractical or undesirable to transfer to a centralized infrastructure. In this work, we apply Federated Learning (FL) for object detection, enabling drones to improve a shared model while keeping image data local and private. We implement a federated object detection pipeline using the Sherpa.ai FL platform on the KIIT-MiTA dataset, and compare it with Single-drone and Centralized baselines using mean Average Precision (mAP) at IoU thresholds of 0.50 and 0.50-0.95. In our experiments, the proposed FL approach remains close to Centralized training while dramatically improving over Single-drone training, with the best lightweight model (YOLO26 nano), suitable for deployment even on very limited edge infrastructure, achieving relative gains of 52.89% and 67.80% in mAP@0.50 and mAP@0.50:0.95, respectively. These results show that FL enables scalable, high-performing, and privacy-preserving object detection across distributed drone fleets without data centralization.

We present GRAFT, a per-word pronunciation conditioning mechanism for text-to-speech neural codec language modeling. Existing systems reach high intelligibility and naturalness but inherit the ambiguity of text and mispronounce rare proper nouns, loanwords and technical terms. Even phoneme-conditioned models offer no direct acoustic handle for per-word pronunciation. GRAFT controls the pronunciation of a chosen word from a short spoken sample of it, encoded with the model's own speech tokenizer and bound to the word's position in the prompt. Voice conversion during training-data construction disentangles the hint speaker from the target speaker, so the hint may come from any voice while the output stays in the target voice. In a blind English listening study, human raters rank GRAFT first by a clear margin, judging its rendering of the difficult word closest to a reference recording of that word. On a five-language objective benchmark, GRAFT reduces target-word phoneme error rate by 22-39% over the identical text-only backbone and outperforms competitive open-source zero-shot systems, both phoneme- and text-conditioned, on target-word pronunciation, while preserving speaker similarity and naturalness.

Time-series forecasting supports decisions in finance, en-ergy, transportation, public health, and industrial monitoring. Recent foundation models improve transfer across forecast-ing tasks, but many depend on centralized data and Trans-former attention, which restricts their use for long, high-di-mensional, and privacy-sensitive signals. This paper presents QuantFlow, a probabilistic forecasting framework that com-bines inverted sequence embedding, bidirectional Mamba state-space decoders, quantile regression, and federated learning. Each variable is embedded over the complete ob-servation window, processed in forward and reverse direc-tions, and projected to five conditional quantiles. TSMixup expands temporal diversity through Dirichlet-weighted inter-polation while preserving sequence structure. Experiments cover cryptocurrency, traffic, electricity, Electricity Trans-former Temperature, influenza, and weather data. QuantFlow obtains mean squared errors of 0.2834 on ETTm1 and 0.2218 on Weather, and a 20-client non-IID deployment retains use-ful accuracy after three communication rounds without cen-tralizing raw records. The results indicate that selective state-space modelling is a promising basis for scalable, uncer-tainty-aware, and privacy-conscious time-series prediction, while also revealing limitations on irregular epidemiological signals and long-horizon generalization.

Hardware accelerators now sit on the critical path of online serving. GPUs, FPGAs, and increasingly remote services such as hardware security modules, post-quantum KEMs, and inference servers. For fine-grained offloads (microseconds to a few milliseconds) the classic responses to the resulting stall both fail: a context switch costs as much as the offload, and a busy-wait burns the core. Overlapping the offload with other requests is the fix, and prior systems obtain it by adding concurrency: an async-framework rewrite, a new runtime or dataplane OS, or a hand-tuned point integration. We observe that the concurrency already exists: serving concurrent requests is suspending and resuming them, so every server ships the machinery overlap needs. Overlap is then a routing problem, not a rewrite problem: submit the offload to an executor, suspend the request with the server's own deferred-response primitive, resume it on completion. Across ten off-the-shelf servers spanning every production concurrency model, this recipe takes 22-138 lines added, at most one modified, and recovers 1.2-5.4x on real hardware; the server's concurrency model and the offload's weight predict both numbers in advance, and the win is bounded by device throughput and the server's own overlap capacity. At the limit, an LD_PRELOAD fiber runtime injects the reroute into an unmodified thread-per-connection binary (17.3x) within a characterized envelope. Rerouting suspends run-to-completion atomicity; a measured taxonomy confines the hazard to unlocked shared aggregates, and a transparent page-protection detector guards exactly those, validated on stock Redis.

While diffusion models have revolutionized image synthesis, their application to real-world inverse problems is often hampered by the need for massive datasets and the difficulty of imposing strict physical constraints. In this work, we introduce \textbf{SE-UNet} (Singular Equivariant UNet), a framework designed to solve ill-posed imaging tasks without extensive pre-training. By treating generation as an optimization problem constrained by geometric equivariance ($D_4$ group) and singular value gating, SE-UNet effectively standardizes the solution space. We demonstrate that these strong inductive biases allow for state-of-the-art zero-shot inpainting results (80\% missing pixels) on CIFAR-10. Our method surpasses Deep Image Prior (DIP) baselines by over 4 dB in PSNR and exhibits a characteristic "singular snap" convergence -- rapidly locking into the signal manifold. SE-UNet thus offers a data-efficient pathway for constrained generation, aligning with the ReALM-GEN goal of bridging theoretical priors with practical deployment.

Mobile applications (apps) increasingly rely on third-party Software Development Kits (SDKs) to provide services such as advertising, analytics, authentication, crash reporting, and location services. These components form an important, but often hidden, layer of the mobile ecosystem. Here, we present a large-scale dataset linking Android apps to the third-party SDKs they integrate. The dataset was constructed by combining app package files (APKs) and metadata from AndroZoo with SDK detection rules provided by Exodus Privacy. We implemented a reproducible, static-analysis pipeline that downloads APKs, inspects their compiled code, and detects SDKs through code-level signature matching: the released dataset contains 334,719 unique app-version observations (associated with 99,722 Android apps) and 246 third-party SDKs - including app-level metadata such as categories, comments, file size, download counts, ratings and SDK-level metadata such as categories, code roots, code versions, names. The dataset supports the construction of a large-scale app-SDK dependency network and its projection onto both layers; in addition, SDKs are mapped to their corresponding operating companies, enabling provider-level analysis of technological concentration and upstream control in the mobile ecosystem. To support reproducible research, the pipeline and dataset are publicly released on GitHub and Zenodo; more broadly, the project provides a reusable research infrastructure for studying third-party technological dependencies, data-collection capabilities, and privacy infrastructures across Android apps.

We study timestep allocation for score-based diffusion sampling, where a learned reverse-time dynamics is discretized on a finite grid. Uniform and hand-crafted schedules are standard choices, but they rely on fixed prescriptions and can therefore be suboptimal. To address this limitation, we propose Adaptive Reparameterized Time (ART), a continuous-time control formulation that learns a time change by treating the speed of the sampling clock as the control, so that a uniform grid on the learned clock induces adaptive timesteps in the original diffusion time. Based on a leading-order Euler error surrogate, ART provides a principled objective for allocating timesteps along the sampling trajectory. To solve this deterministic control problem, we introduce ART-RL, an auxiliary randomized formulation with Gaussian policies that turns schedule learning into a continuous-time reinforcement learning problem. We prove that the randomized ART-RL formulation is equivalent to ART at the optimizer level, in the sense that its optimal Gaussian policy recovers the optimal ART time-warping rate through its mean. We further establish policy evaluation and policy improvement characterizations and derive trajectory-based moment identities that yield implementable actor--critic updates for learning the schedule. Across experiments ranging from controlled low-dimensional settings to image generation, ART-RL can be plugged into existing diffusion samplers by changing only the timestep grid, consistently improving sample quality over strong baseline schedules at matched budgets while leaving the rest of the sampling pipeline unchanged. The learned schedules also exhibit broad generalization, transferring without retraining across sampling budgets, datasets, solvers, pipelines, and representation spaces.

Flow-matching vision-language-action policies generate robot action chunks through an iterative transport process, creating an opportunity for test-time guidance without retraining the base policy. We study this opportunity in Guided Action Flow, an inference-time framework that keeps a pretrained SmolVLA policy frozen and uses a learned action-chunk critic to guide its reverse-time flow sampler. The critic is trained from real success and failure rollouts, can condition on task-description features from the frozen SmolVLA language pathway, and is used only through action gradients during sampling. We evaluate the approach on LIBERO manipulation tasks. A single-task critic improves success from 68.0% to 82.0% on one seed window and from 82.0% to 86.0% on another. A multi-family task-description critic improves validation success from 46.0% to 56.0%, while the locked held-out test gain is positive but modest, from 65.0% to 67.5%. These results support the feasibility of Q-guided inference for frozen flow-matching VLA policies, while showing that critic generalization and uncertainty-aware guidance remain the central bottlenecks.

Modern automotive software architectures comprise large sets of mixed-criticality functions executing on shared multi-core platforms with strict real-time and end-to-end timing requirements. Sensor-to-actuator data propagation in such systems is typically expressed via cause-effect chains with worst-case data-age budgets. Job-level dependencies (JLDs) have been introduced to provide a schedule-agnostic mechanism for bounding the data age independently of the underlying scheduler. The state-of-the-art methods for synthesizing JLDs, however, do not check whether the produced JLDs are enforceable under a concrete scheduling policy or jointly schedulable at the system level. In this paper we propose the first machine-learning-based JLD synthesis method, built around a two-level Graph Neural Network with temperature-controlled sampling that learns the structural patterns connecting cause-effect chain configurations to their JLD solutions. Since learned outputs may not be correct by construction, we embed the GNN in a novel Generate-and-Verify architecture in which a safe DP data-age checker, together with a per-chain EDF feasibility checker and a system-level demand-bound test, accept or reject each candidate. We show that the ML-based generator substantially outperforms the original greedy heuristic while achieving orders-of-magnitude lower synthesis time, demonstrating that learned structural priors can effectively replace exponential propagation-tree enumeration on this class of real-time scheduling problems.

Time series foundation models (TSFMs) have shown strong zero-shot forecasting performance, but their generalization in covariate-driven, non-stationary settings is underexplored. Electricity price forecasting (EPF) presents a challenging testbed due to complex temporal dependencies, distributional shifts, and strong reliance on structural and contextual information. We propose a two-dataset-benchmarking framework for EPF to mitigate contamination risk and enable fair evaluation of TSFMs. We examine key aspects of EPF including point and probabilistic forecasting performance, tail behavior, price spikes, and comparisons against domain-specific methods. We find that TSFMs are highly competitive and often outperform general-purpose baselines. Yet, their performance depends critically on covariate support, and they do not consistently surpass domain-specific methods tailored to EPF. Interestingly, simple ensembles of TSFMs and domain-specific methods appear to have significant potential, suggesting that the two approaches capture complementary predictive information.

Robust 3D object detection under adverse weather remains a critical hurdle for autonomous driving. Despite progress with LiDAR-4D radar fusion, most methods are constrained by a closed-world assumption, implicitly requiring training and test weather to align in both type and severity. This premise fails in practice: the open-ended nature of weather, and even variations within a single type like rain, cause dramatically different LiDAR degradation patterns, leading to significant performance drops in unseen conditions. To address this, we present Dual-Critic Guided Diffusion Alignment (DCDA), a weather-agnostic framework that learns to recover degraded LiDAR features toward a clean manifold. Rather than modeling specific weather types, DCDA employs a 4D radar-conditioned diffusion process to progressively refine features, guided by two complementary critics. (i) A detection-guided critic, anchored by a pre-trained clean-weather model, ensures that the refined features retain object-level discriminability and localization accuracy. (ii) A weather adversarial critic enforces holistic distributional consistency with clean-weather representations. By aligning features through semantic and distributional constraints rather than explicit weather modeling, DCDA generalizes effectively to unseen weather types and severities without requiring paired data or weather labels. We further introduce a structured open-weather benchmark with held-out type-severity combinations and extensive experiments verify DCDA's advantages.

Multiplex CRISPR-Cas9 gene editing requires selecting one guide RNA per target gene subject to cross-gene interactions: a constrained combinatorial problem that can be formulated as a Quadratic Unconstrained Binary Optimization (QUBO) and solved via the Quantum Approximate Optimization Algorithm (QAOA). The one-hot per-gene constraint is conventionally enforced by adding quadratic penalty terms to the cost Hamiltonian, but penalty coefficient selection is heuristic and penalties amplify hardware noise. An alternative is to enforce the constraint structurally via the XY-mixer, which preserves feasibility by construction. We present COMET, a systematic comparison of penalty-based and XY-mixer QAOA on a three-gene, twelve-qubit multiplex editing instance targeting the immune-checkpoint genes PDCD1, LAG3, and HAVCR2. In simulation, the XY-mixer exceeds 95% probability of the optimum by QAOA depth p=3, while three penalty variants spanning an order of magnitude in penalty coefficient remain below 6% at every depth. On IBM's ibm_kingston (Heron r2) processor, the XY-mixer's simulator-hardware energy gap stays within |0.8| across all depths, while the worst-tuned penalty variant's gap reaches +53.9. We provide an honest account of where the structural guarantee partially breaks under gate-level noise. The twelve-qubit instance is classically trivial; our contribution is a methodological comparison of constraint-enforcement strategies in a biologically motivated domain, with real-hardware validation.

In the age of AI, what will be good knowledge? This article, which is accepted and forthcoming in a special issue of Modern Fiction Studies on "Cultural AI" in 2027, applies digital humanities methods to map epistemic virtues (like "true," "accurate," "creative") used in a corpus of 553 journal articles on AI published in 2024. "Creativity" comes in for special attention as an example. Exploring this discourse of value, the article considers how a framework might be developed for evaluating the knowledge-worth of AI -- one less locked into values formed around pre-AI "knowledge work" agents or structures, and more open to the future values of "generativity." The essay is supported by an online digital kit for exploring data models of the corpus of articles on AI it studies.

Simulating molecules is a major application of quantum computing, with the potential to overcome exponential scaling constraints of classical computation. Researchers use different methods in order to evaluate the readiness of NISQ computers in order to test current simulation capabilities. We present an integrated repository with reproducible benchmarks of over 10 different ansatzes from published papers and two different truncation methods, applicable to any set of mapped hamiltonians, providing a single pipeline for comparing performance along multiple axes, including variance and computational time, among others. We apply them to simulate different amino acids, using hamiltonians taken from the QMProt Dataset. We then ran four separate experiments. First, we quantified noise resilience by optimizing the same hardware-efficient ansatzes under identical initialization while sweeping PennyLane noise channels and strengths, and measuring parameter drift, cosine similarity of optimal parameters, and energies evaluated on noiseless versus noisy backends. We then studied barren-plateau-related trainability via gradient-variance diagnostics and optimization trajectories across initialization strategies and ansatzes depth on small systems. We then compared adaptive versus fixed ansatzes at matched parameter budgets, reporting outer-loop iterations, wall time, and especially total cost-function evaluations to fairly contrast greedy adaptive growth with layered hardware-efficient circuits. Lastly, we mapped accuracy versus expressive capacity by sweeping the number of retained adaptive operators and recording ground-state energy error relative to classical references.

In multi-source ECG deployment, models may need to incorporate new data sources when earlier raw ECGs cannot be retained or replayed. Freezing a pretrained backbone and assigning each source an isolated classifier prevents parameter interference, but deployment still requires selecting an expert when source metadata are unavailable. We study this distinction through IRFE-ECG, an incremental expert bank built on frozen 1024-dimensional ECGFounder features. Each arriving domain adds a balanced-softmax linear expert, while a lightweight router is fitted only on retained training features and domain labels from sources observed so far. A validation-calibrated margin rule fuses the two most likely experts instead of committing to a single routed expert. On CPSC, PTB-XL, Georgia, and Chapman-Shaoxing, source-aware expert selection reaches $0.7915\pm0.0036$ Macro-F1 and a matched offline independent-head reference reaches $0.7885\pm0.0009$, supporting strong source-aware expert retention. Without source IDs, an MLP router reaches $0.7756\pm0.0027$ and top-2 margin fusion reaches $0.7782\pm0.0022$. The top-2 gain over hard MLP routing is small ($+0.0026$), with a 95\% confidence interval from paired bootstrap that includes zero. Across three domain orders, the top-2-to-oracle gap remains $0.0111$--$0.0133$, identifying autonomous source inference as the main remaining bottleneck. No raw ECGs are replayed, but frozen training features are retained for router updates; the method is therefore not memory-free.Code is available at https://github.com/yufanlu221/IRFE-ECG.

Developing high-performance kernels for Neural Processing Units (NPUs) is a critical industry bottleneck, requiring developers to manually navigate implicit hardware constraints and strict memory hierarchies. While large language models offer immense automation potential, they fail catastrophically on NPUs due to a fundamental lack of hardware-specific priors. Naively transplanting code snippets from similar NPU kernels may pass the compiler, but it consistently triggers runtime crashes and performance degradation by blindly violating underlying hardware constraints. To overcome this, we introduce Hawk, a training-free framework that harnesses hardware-aware knowledge through three core modules: (1) Run-Time Knowledge Synthesis Module, which employs a Triple-Part Executable Knowledge Representation to inherently couple the error context with executable semantics; (2) Bottleneck-Aware Knowledge Retrieval Module, which implements a 2D-Retrieval paradigm to project queries into orthogonal syntactic and hardware-aligned semantic spaces; and (3) Effect-Driven Knowledge Distillation Module, which leverages LLM-driven semantic arbitration to continuously distill the knowledge by pruning errors and consolidating redundancies based on the empirical execution feedback. Extensive evaluations on real-world NPU workloads demonstrate that Hawk elevates generation accuracy from 49.4% to 80.0%, while achieving up to a 2.2x execution speedup over state-of-the-art baselines.

Large Language Models (LLMs) often struggle with persuasion in high-stakes scenarios. People's individual personalities and concerns require tailored strategies rather than a one-size-fits-all approach. To address this challenge, we focus on a fire-rescue scenario in which an operator must persuade a resident to evacuate as a high-stakes persuasion domain and propose Dialogue Policy Selection (DiPS), a Q-learning framework to dynamically select persuasion strategies adapted to the evolving conversational context. Specifically, we train a critic, trained to maximize the chance of evacuation success, to select a persuasion policy at each turn based on the resident's recent utterances. We then evaluate DiPS against multiple baselines in both simulated and real human interactions. We find that DiPS achieves higher evacuation success than a zero-shot LLM and generic RAG-augmented approach.

Cells represent one of the most fundamental units of life. Underlying their robust performance against environmental variability, such as temporal fluctuations of chemical signals, between different cell types, is a dynamical interrelation between the two components of an intracellular pathway: a gene-regulatory network and its upstream signal transducers. To understand how a single cell utilizes this feedback to self-regulate its gene-expressions, we develop a multiscale model of the pathway's components, in which the adaptive variables responsible for signal interpretation follow a feedback-induced learning process. We then derive a macroscopic theory capturing the covariations between these components - so-called collective modes. Our theory shows how cells can achieve robust output against signal fluctuations via self-regulation rather than simple noise suppression. Such robustness corresponds to a transition from random- to structured collective modes beyond a critical adaptation rate.

Vortex light carries orbital angular momentum (OAM) in its transverse phase, unlike spin, which is encoded in polarization. We show that this phase cannot produce OAM dichroism in any continuously translation-invariant system, even when optical nonlocality and multipole light--matter coupling are included. Nonuniform strain in a two-dimensional Dirac material bypasses this no-go condition by supplying transverse momentum transfer through a long-wavelength pseudo-gauge field, so that the OAM dichroism factorizes into a nonlocal elasto-optic tensor and the overlap of the vortex transverse phase current with the strain field. For a micron-scale Gaussian bubble, the resonant OAM-odd conductivity reaches \(\sim10^{-3}e^2/\hbar\), within reach of differential optical measurements.

The decomposition of the Hilbert space of a quantum many-body system into the irreducible representations of its bond and commutant algebras yields a finer structure of dynamically isolated subspaces than the mere decomposition into symmetry sectors. While it has been recognized that subspaces associated with low bond-irrep dimensions $D_λ$ tend to violate the eigenstate thermalization hypothesis (ETH), here we show that $D_λ$ controls thermalization continuously across the full spectrum of dynamical subspaces. Using SU(2)-symmetric spin-1/2 chains as a paradigmatic example, we demonstrate that $\mathrm{log} \ D_λ$ quantitatively accounts for the average eigenstate entanglement entropy within each sector, establishing a thermalization hierarchy that interpolates from exact quantum many-body scars at $D_λ=1$ to volume-law ergodic states at large $D_λ$. To make this concrete, we introduce the notion of irreducible degrees of freedom (IDOF), defined as the number of independently-varying spatial coordinates parametrizing a many-body state within a given bond-algebra sector, which provides a microscopic interpretation of $D_λ$ and of the resulting thermalization hierarchy. Finally, we show that by selectively breaking symmetries while preserving chosen bond-algebra sectors, one can embed families of nonthermal eigenstates at prescribed entanglement levels into an otherwise ergodic spectrum, generalizing restricted spectrum-generating algebras from towers of individual states to entire dynamical subspaces.

Recent experiments have shown that strong light-matter coupling in electromagnetic cavities can modify transport properties of quantum Hall systems through the formation of Landau polaritons, prompting questions about the robustness of topological protection. While earlier theory demonstrated that the Hall conductivity can be modified at finite temperature and finite polariton lifetime (or finite broadening), experiments primarily probe the resistivity tensor. Our phenomenological model reveals an asymmetry between conductivity and resistivity in quantum Hall systems under strong light-matter interaction, showing that at zero temperature the Hall resistivity remains completely immune to cavity-induced modifications arising from polariton broadening, independent of the light-matter coupling strength. These results provide a deeper explanation for the absence of renormalization in the von Klitzing constant in experiments probing the even QH plateaus through the Hall resistivity at low temperature, and clarify the distinct roles of dissipation and strong light-matter coupling in hybrid light-matter systems.

Measurements of heat flux between micro-objects, in vacuum or in air, are challenging because of their small size and the low thermal conductance of the medium between them. One way to address this issue consists in using a scanning thermal microscope (SThM) equipped with a temperature dependent resistance thermometer. However, this requires an instrument able to both injecting a defined heating Joule power and performing highly-sensitive resistance measurements. Here, we present such an instrument based on a Wheatstone bridge equipped with three Kelvin arms. It can perform resistance measurements in the range from 100 $Ω$ to 1000 $Ω$ not only in direct current but also in alternating current regimes at frequencies up to a few tenths of kHz. We first show that measurements of resistance standards are accurate to within one part in $10^4$ with a relative experimental standard deviation which can be as low as one part in $10^8$ for one second measurement. The instrument is then tested with a SThM thermometer. With the support of an electro-thermal model considering thermal time constants of the thermometer, we explain the frequency dependence of detected signals and optimize the measurement protocols of temperature and heat flux. By measuring sub-mK temperature variations, this instrument is then used to determine with a few nanowatts uncertainty the near-field radiative heat flux between a heated glass microsphere and a glass substrate, which is caused by the coupling of surface phonon-polaritons.

Accurate detection of human serum albumin (HSA) is essential for the early diagnosis and monitoring of renal and hepatic disorders. We present a graphene-based electrolyte-gated field-effect transistor (EGFET) for label-free, real-time quantification of HSA under non-Faradaic operation. Devices exploit the high interfacial capacitance of the electric double layer (EDL) to transduce electrostatic perturbations induced by albumin adsorption into measurable conductance modulation. Negatively charged HSA molecules induce systematic modulation of the graphene channel, producing a concentration-dependent displacement of the Dirac voltage consistent with p-type doping. To establish a molecular-level interpretation of the sensing response, Brownian Dynamics simulations show that HSA adsorbs onto graphene through multiple adsorption orientations associated with heterogeneous interfacial charge distributions and variable dipole alignments relative to the surface. Adsorption is energetically stabilized by van der Waals interactions. Analysis of transfer characteristics across concentrations ranging from 0.01 to 30mgmL-1 reveals a correlation between surface charge density and carrier transport modulation within the electric double layer. Optimized devices exhibit a limit of detection of 0.0087 mg mL-1 and a linear dynamic range extending to 10 mg mL-1. The response remains non-Faradaic under sub-volt operation with reversible and reproducible behavior. The use of an inverse-mobility analytical metric highlights the role of disorder-enhanced carrier scattering in signal amplification, enabling sensitive electrostatic detection while preserving reversible device operation. These results establish liquid-gated graphene EGFETs as a promising platform for quantitative protein sensing and provide insight into disorder-mediated transport mechanisms in graphene bioelectronic devices.

Contact electrification between two conducting materials is expected to exhibit length scale effect because the screening effect will diminish in conductors as a function of material dimensions. As a consequence, the interfacial charge accumulation will diffuse away from interface/surface to a critical penetration depth as a function of material dimension. This work experimentally demonstrates the length scale effect in a permalloy and degenerately doped p-Si heterostructure system due to the flexoelectricity mediated contact electrification. The contact electrification induced interlayer charge transfer is observed through the whole thickness in case of 400 nm thick p-Si samples. Whereas, the charge carrier diffuses to a depth of 51 nm from the interface in case of 2 um thick Si. The length scale effect also leads to metal-insulator transition in p-Si layers in both cases. These results present a new opportunity to tailor the physical properties in conducting materials using contact electrification.

Dynamics of two short semi-flexible fibers settling under gravity in a viscous fluid are investigated at Reynolds numbers Re << 1. We focus on fibers initially relatively close to each other, and we check if later they approach an aligned horizontal configuration, previously identified numerically (Bukowicki and Ekiel-Jezewska, Soft Matter 46 (2019) 9379) as attractive for symmetric initial conditions of moderately elastic filaments. In our experiments, two semi-flexible ball chains sediment in a highly viscous silicone oil. They are initially straight and close to a parallel horizontal relative configuration. Their motion and shape deformation are recorded using two synchronized cameras. For most of the trials, ball chains stay together, with damped oscillations around the symmetric aligned configuration. For a few initial conditions, the ball chains move away horizontally or vertically. To study the behavior over a longer time, we perform numerical simulations, modeling moderately elastic filaments as chains of identical beads, with the centers of consecutive beads connected by springs and with the fibers' elastic resistance to bending. Different initial positions and orientations are considered. Their dynamics are determined by the multipole expansion of the Stokes equations, implemented in the precise Hydromultipole numerical code. For short times, we observe the similar dynamics of semi-flexible ball chains and moderately elastic filaments. We provide examples of long-time numerical simulations illustrating that elastic filaments close to each other can move away horizontally or vertically, but after a long time, come back and perform damped oscillations while approaching the aligned configuration with almost touching filament ends. We confirm the attractive nature of the aligned configuration of very close semi-flexible sedimenting fibers, even if they are far away from each other.

We investigate the emergence of zero-dimensional topological end states in nanoribbons described by the Qi-Wu-Zhang (QWZ) model and its extensions with longer-range couplings. While dimensional reduction from two to one dimension is often assumed to preserve the symmetry classification of the parent system, here an additional symmetry can emerge originating from the real-space geometry of the ribbon. This symmetry acts as a chiral symmetry, combining orbital and spatial transformations, and promotes the effective one-dimensional system from symmetry class D to class BDI. We demonstrate that the existence of such a symmetry depends both on the long and end termination of the ribbon and exhibits an even-odd effect with respect to ribbon width, revealing that the commonly studied rectangular ribbons constitute a special high-symmetry case. For the conventional QWZ model, we derive analytic expressions for the topological phase boundaries of finite-width nanoribbons and characterize the resulting hybridization-gap phases through ($\mathbb{Z}_2 $) and winding-number invariants. We further show that extended QWZ models with longer-range couplings support phases with multiple topological end states and higher winding numbers. These phases arise through distinct mechanisms, including the hybridization of multiple edge modes inherited from higher-Chern-number bulk phases. Our results demonstrate that both long and end termination can fundamentally alter the topological classification of confined Chern insulators, highlighting the interplay between crystalline geometry, emergent symmetries, and dimensional reduction.

We formulate a continuum model for a compressible lipid-bilayer membrane immersed in Stokes flow, replacing exact local area inextensibility by conservation of an areal phospholipid density. The membrane free energy combines Helfrich bending, spontaneous curvature, and a finite area-compression penalty, so that membrane tension becomes a constitutive response to lipid-density variation rather than a Lagrange multiplier enforcing local area conservation. The resulting interfacial stress includes normal elastic forces and tangential Marangoni stresses generated by lipid redistribution; these stresses arise from membrane compressibility and can produce an effective negative tension when the local lipid density exceeds its preferred value. We further derive the linear stability of circular membranes in two dimensions and spherical membranes in three dimensions under full Stokes hydrodynamic coupling. In both cases, bending stabilizes the base shape, while excess lipid density destabilizes it by favoring increased membrane area. The first instability occurs in the lowest nontrivial shape mode, m = 2 in two dimensions and j = 2 in three dimensions. Energy expansions near onset show that the two-dimensional instability is a pitchfork bifurcation, whereas the three-dimensional instability is generically transcritical because prolate and oblate perturbations are geometrically distinct. These results provide a controlled compressible extension of classical vesicle mechanics and directly connect lipid-density variation, membrane tension, hydrodynamic coupling, and shape instability.

Engineering dissipation with tailored steady states has become a powerful approach for preparing and stabilizing quantum states. In this framework, engineered dissipative processes continuously steer a system towards desired target states while suppressing unwanted noise. However, extending this idea to multimode systems is challenging and remains largely unexplored, although this class of states is a powerful resource for quantum sensing and quantum information processing applications. Here, we propose a general method to design the required dissipative processes for the generation of multimode cat states in bosonic systems. We show that the engineered dissipation prepares such states from the vacuum with high fidelity and robustly stabilizes them against decoherence. As a result, their lifetime is extended by several orders of magnitude compared to natural decay times, which in turn enhances their applications in quantum techonologies. We specifically focus on the preparation and stabilization of two-mode binomial cat states and discuss a pathway for the implementation in superconducting circuit. However, our scheme can also scale up to arbitrary d-mode multinomial cat states associated to $\mathfrak{su}(d\ge2)$ algebras, and thus, our scalable framework provides a feasible route towards stabilizing compact nonclassical states.

Functional laws of large numbers (FLLNs) describe the mean-field trajectory of epidemics on networks, but say nothing about the fluctuations around it. These fluctuations are governed by moments of the degree distribution not relevant at the level of the mean. A rigorous functional central limit theorem (FCLT) exists for the susceptible--infected ($SI$) process on configuration-model graphs, but no analogue exists for $SIS$, where recovery reintroduces vertices into the susceptible pool with partially known neighborhoods, breaking the clean neighborhood distribution the $SI$ derivation relies on. We develop a tractable variance approximation for Markovian $SIS$ on configuration-model graphs, combining Gleeson's approximate master equation (AME) framework with a van Kampen system-size expansion in the spirit of the $SI$ FCLT. We derive a closed drift and diffusion matrix for a reduced susceptible/$SI$-edge/$SS$-edge count vector and obtain the time-dependent covariance via the associated Langevin/Lyapunov equation. Validation against Gillespie simulation across Poisson, regular, and power-law networks shows close agreement, with deviations near the epidemic threshold and in strongly heterogeneous networks.

We demonstrate that realistic excitonic many-body Hamiltonians obtained from first-principles GW-Bethe-Salpeter equation calculations can exhibit quantum chaos governed by random-matrix universality. Considering a prototypical van der Waals heterostructure (WS$_2$-graphene), with and without lattice disorder, we analyze their energy-resolved spectral correlations and identify a disorder-driven crossover from regular to complete chaotic dynamics. We show that while pristine samples exhibit incomplete chaos (non-ergodicity) due to an approximate valley symmetry that restricts excitonic mixing, the presence of disorder-induced electronic flat bands act as a catalyst for valley mixing to drive the system into a fully developed chaotic (ergodic) regime with reduced symmetry. Crucially, fluctuations in many-body oscillator strengths are shown to follow universal Porter-Thomas statistics, directly linking the underlying quantum chaos and experimentally accessible optical observables. Finally, by examining long-range spectral correlations, we estimate the Thouless time associated to excitonic mixing across the entire many-body bandwidth. Our results establish excitons as a highly tunable platform for probing many-body ergodicity and its spectroscopic signatures in realistic interacting 2D materials.

Heteroatom substitution is a powerful route to tune the chemical and electronic properties of carbon nanomaterials. In particular, replacement of an sp2 hybridized carbon atom in the graphene lattice with a nitrogen atom (denoted as graphitic nitrogen) induces substantial changes in the electronic properties. These include changes in the band structure that can influence electronic transport, and magnetism. A key requirement for applications is both the periodic and precise incorporation of the heteroatoms in extended carbon lattices. Here, we report the on-surface synthesis and characterization of two one dimensional carbon nanostructures, a polymer and a graphene nanoribbon, consisting of periodically incorporated graphitic nitrogen atoms. The on-surface reactions toward formation of the nanostructures were monitored by scanning tunneling microscopy. The bond-resolved chemical structures of the reaction intermediates and products were investigated by atomic force microscopy, which enabled atomic-scale visualization of the graphitic nitrogen sites. The electronic properties of the nanostructures were studied by scanning tunneling spectroscopy and density functional theory calculations. Our analyses revealed the presence of localized nitrogen-centered electronic states. In the gas phase where the nanostructures are in a neutral charge state, these states undergo spin polarization leading to an open-shell ground state. Upon adsorption on Au(111), the nanostructures exhibit electron transfer to the surface, which resulted in a closed-shell ground state. Our results demonstrate a straightforward and generally applicable route to synthesize graphitic nitrogen-substituted carbon nanomaterials with potential applications in spintronics, catalysis and energy storage.

Light absorption in photonic lattices with sparsely distributed loss sites exhibits behavior analogous to Griffiths physics. Under uniform excitation, the transmitted power shows a stretched-exponential decay and a non-monotonic dependence on the loss strength, with an optimal loss rate that maximizes absorption. This behavior arises from rare, long loss-free segments that act as weakly coupled, long-lived photonic channels, rather than from exceptional point physics or interference effects. Using a minimal tight-binding model with binary quenched dissipation, we show that rare regions produce a universal Griffiths-type subexponential decay. Sparse-loss photonic lattices thus provide an accessible platform to observe disorder-induced anomalous absorption and rare-region Griffiths physics.

The time evolution of non-Hermitian systems is generally nonunitary. Dynamics governed by time-dependent non-Hermitian Hamiltonians lead to a variety of novel phenomena, one of which is state amplification or suppression induced by the complex Berry phase. Here, we extend the framework of geometric modulation to multi-level systems and show that both transition and survival intensities can be modulated. We apply our theory to the non-Hermitian Landau-Zener (LZ) problem. First, we show that, in the half-LZ problem, both the transition and survival probabilities exhibit nonreciprocity due to the complex Berry phase. In the non-Hermitian standard LZ problem, only the survival intensity is known to exhibit nonreciprocity, whereas the transition intensity does not. However, the physical origin of this nonreciprocal behavior remains unclear. In this work, we show that the nonreciprocity originates from the complex Berry phase.

Battery discharging is governed by a fundamental trade-off between output power and energy conversion efficiency due to internal dissipation. In this paper, we demonstrate that such a trade-off universally yields a parabolic envelope $P\proptoη(1-η)$. The efficiency at maximum power is exactly one half, mirroring the well-known half-Carnot limit in finite-time thermodynamics. To extend this bound into practical operational rules, we formulate a multistage constant-current discharging (MSCD) schedule subject to simultaneous real-time load demands and a global discharging deadline. Analytical resolution via the Karush--Kuhn--Tucker conditions reveals a remarkably compact optimal policy: $I_{i}^{\star}=\max(I_{i}^{-},I_{0})$. Under this rule, stages limited by external demand run exactly at their minimum required currents, while all remaining stages are elevated to a uniform baseline $I_{0}$ fixed by the deadline constraint. By tracing the dissipation--time Pareto front, we quantify how internal resistance shifts the operational boundaries and sharpens the trade-off corner. This analysis establishes a rigorous thermodynamic baseline for the scheduling layer of battery management systems, offering natural extensions to nonlinear models incorporating temperature and state-of-charge dependencies.

Single wall carbon nanotubes (SWCNTs) are promising building blocks for nanoelectronic and optoelectronic devices, yet reliable and stable doping, particularly n type, remains challenging due to strong environmental sensitivity and competing extrinsic effects. Encapsulation of charge transfer molecules within the SWCNT cavity offers a promising route to stable doping while preserving the nanotubes outer surface for subsequent processing. Here, we systematically investigate the filling of arc discharge SWCNTs with the electron donor tetrathiafulvalene and electron acceptor tetracyanoquinodimethane, comparing different methods for filling, including melt filling, solution reflux, and vacuum phase sublimation. We follow the entire processing workflow from raw, unfilled powders to aqueous dispersions and employ density gradient ultracentrifugation to separate filled from empty nanotubes as well as metallic from semiconducting ones. Encapsulation efficiency and electronic modification are assessed using absorption spectroscopy, resonant Raman scattering, thermogravimetric analysis, and electron paramagnetic resonance. Finally, we introduce a complementary vacuum-phase method that removes externally adsorbed molecules without extensive solvent washing, enabling cleaner encapsulated systems.

Photodoping plays an important role in determining the optoelectronic response of two-dimensional semiconductor devices; however, the origin of the responsible trap states remains unclear. In this work, we investigate UV-induced photodoping in multilayer WSe2 field-effect transistors (FETs) based on WSe2/hBN heterostructures on SiO2/p-Si substrates. Wavelength-dependent measurements reveal pronounced n-type photodoping under 405 nm illumination, whereas the effect is orders of magnitude weaker under 640 nm excitation and for p-type photodoping. Furthermore, when the SiO2 layer is removed, both n-type and p-type photodoping are strongly suppressed, demonstrating that the oxide layer is essential for persistent photodoping. Analysis of defect-state distributions in amorphous SiO2, together with first-principles calculations for hBN defects, shows that the experimental observations cannot be explained by defects in hBN. Instead, the results indicate that defect states in the SiO2 substrate act as charge reservoirs that facilitate charge transfer and long-term carrier trapping. These findings highlight the dominant role of substrate-related trap states in UV-induced photodoping and photogating behavior in WSe2 FET devices.

The role of impurity scattering in the generation and transport of orbital current remains less established than in conventional spin Hall systems. Here we investigate Ru-based nonmagnet/ferromagnet bilayers in which the impurity scattering is tuned by Cu or Ti alloying. According to SOT measurement and thickness-dependent drift-diffusion analysis, we extract the effective orbital Hall conductivity and the orbital diffusion length. We find that the orbital Hall effect in polycrystalline Ru is dominated by intrinsic mechanism that is moderately robust against weak disorder but suppressed by stronger alloy disorder. However, the orbital diffusion length remains nearly unchanged at approximately 14 nm over the investigated impurity range. This behavior indicates that orbital transport is not governed simply by an impurity scattering. Together with previous temperature-dependent measurements, our results show that static impurities and dynamic lattice disorder affect orbital transport through distinct microscopic channels. These results provide new insight into how disorder governs orbital generation and transport, and offer experimental guidance for developing high-efficient orbitronic materials.

We give a comprehensive account of the recent entropy density functional theory for the equilibrium statistical mechanics of classical many-body systems (arXiv:2606.28240). The approach is formally exact and based on a joint grand potential minimization principle for the one-body density and the global pair distance distribution. These variational fields depend respectively on position and on scalar distance, which retains the low computational complexity of standard density functional theory. Correlations effects are contained in a unique excess entropy functional, which is universal across all systems with pairwise interparticle potentials. Functional differentiation yields entropic direct correlation functionals that generate entropic response and fluctuation correlation functions via coupled Ornstein-Zernike equations. Two alternative proofs are given for the existence and uniqueness of the underlying metadensity functional map, based on generalizations of either Levy's constrained search method or Mermin-Evans proof by contradiction. Simple excess entropy approximations yield the standard mean-field and second-virial excess free energy density functionals. We describe exact entropic functional line integrals, make connections to the recent one-body fluctuation profiles, and generalize the entropy approach beyond pairwise interparticle potentials.

The Ramsey community number $r_κ$ is the smallest network size at which a graph is better described by a partition into communities than by no partition, under a prescribed detection rule. On a scale-free graph this question is confounded: a block model can split the network merely to absorb its degree distribution. I compute $r_κ$ analytically for the deterministic pseudofractal scale-free web of Dorogovtsev, Goltsev, and Mendes, separating genuine community structure from degree heterogeneity with two closed-form detection rules. Under a plain Bernoulli stochastic block model, the web's natural recursive bipartition is unpreferred while small and breaks at $r_κ=1095$ nodes, with a log-evidence growing as $(\ln 3-\tfrac{2}{3}\ln 2)n$. Under a degree-corrected model tested against the configuration-model null, the same partition survives, breaking far earlier at $r_κ=42$, with a log-evidence growing as $(2\ln 3-\tfrac{4}{3}\ln 2)n$ -- exactly twice the plain slope, and independent of the prior. Degree correction reverses the ordering of the candidate cuts, demoting the hub-leaf split and elevating the recursive one. Because the web is self-similar, the best description is not two communities but a nested hierarchy: the degree-corrected evidence keeps rising as the partition is refined, and is maximised at of order $\sqrt{n}$ communities of $\sim\sqrt{n}$ nodes. A purely local recursive rule thus builds true hierarchical community structure, over and above the scale-free degree sequence it also produces, in an exactly solvable setting.

We discuss a new mechanism for the nonlinear anomalous Hall effect in tilted two-dimensional gapped Dirac fermion systems. This mechanism originates from a side-jump displacement associated with Landau-Zener tunneling across the mass gap. In lightly doped tilted Dirac systems with a small gap, this interband mechanism can contribute to the nonlinear anomalous Hall effect in addition to the conventional intraband mechanism arising from the Berry curvature dipole. It provides a possible explanation for the nonlinear Hall effect accompanied by nonlinear longitudinal transport observed in the organic Dirac fermion system α-(ET)2I3 with an extremely small gap.

This work provides direct experimental evidence for the role of topological interface states in thermal conduction across a metal/topological insulator junction. It also shows that this conduction can be reversibly modulated by electrical current injection, offering a new approach toward active control of heat flow at solid-state interfaces. Specifically, the interfacial thermal conductance of ${Au}$/$Bi_{89}$$Sb_{11}$ and ${Au}$/$Bi_{87}$$Sb_{13}$ junctions demonstrates distinct temperature- and bias-dependent behavior. Both responses are attributed to carrier redistribution between topological interface and bulk band states, driven thermally by Fermi-Dirac broadening and electrically by quasi-Fermi-level shifts and WKB tunneling into nearby bulk bands. Control experiments using trivial semimetals and insulating interlayers further confirm the topological specificity of the effect. Such electrically tunable interfacial heat conduction positions interface electronic structure engineering as a promising route for active thermal management. In doing so, it lays the groundwork for a mechanically robust alternative to conventional structure-driven thermal control compatible with increasingly dense, high-power solid-state devices.

Shear of three-dimensional, highly compressed granular packings is simulated using a bonded particle approach that explicitly resolves elastic deformation. Varying Poisson's ratio $ν$ produces significant changes in rheology, packing structure, and grain morphology. During flow, conventional systems ($ν> 0$) readily crystallize while auxetics ($ν< 0$) resist ordering. This duality reflects the fact that conventional grains develop polyhedral-like facets but conserve volume while auxetics behave oppositely, demonstrating an unexpected interaction between elasticity, geometry, and crystallization.

We study the dc conductivity of a continuum two-dimensional Rashba film with an in-plane Zeeman field and delta-correlated scalar disorder. Although the field deforms the two helicity Fermi contours and rotates the spin texture, it does not produce anisotropic magnetoresistance in the leading quasiclassical conductivity. The mechanism is geometric. A density Ward identity fixes the spin-vector part of the Born self-energy to the derivative of the total particle density with respect to the field. This derivative vanishes, because the total area enclosed by the two Rashba-Zeeman sheets is independent of the in-plane field. The Born self-energy is therefore scalar and field independent, and the quasiparticle lifetime stays isotropic. The same area invariance controls transport: once the leading impurity ladder reduces the current vertex to the parabolic velocity, the diagonal intraband Kubo conductivity collapses onto the two-sheet occupied area and is field independent as well. The result settles the short-range-disorder quasiclassical problem: point-like nonmagnetic impurities do not produce AMR in this model. A nonzero AMR requires physics beyond this quasiclassical short-range-disorder mechanism.

Quasi-bound states in the continuum enable exceptional field confinement, strongly reducing the pump intensity threshold for nonlinear light-matter interaction in dielectric metasurfaces. As a result, nonlinear self-action effects, often elusive in bulk nonlinear media, emerge at moderate excitation intensities. Here, we show how pulse duration and resonant coupling govern the nonlinear self-action mechanism in resonantly enhanced third-harmonic (TH) generation from a nonlocal metasurface. We identify two excitation regimes that interact differently with the resonant mode, revealing complementary intensity-dependent responses. Under spectrally narrow picosecond excitation, resonance-enhanced TH generation shows pronounced deviations from cubic scaling at high intensities. In contrast, broadband femtosecond excitation transiently drives the resonance, encoding the nonlinear response in the spectral reshaping and broadening of the TH signal. Simulations reproduce both regimes: continuous-wave modeling captures picosecond power scaling and higher-harmonic interactions, while time-domain simulations resolve femtosecond dynamics. These results clarify nonlinear self-action in metasurfaces featuring quasi-bound states in the continuum, linking strong field confinement to conversion efficiency, scaling behavior, and distinct spectral dynamics under different excitation conditions. This work sheds light on the interplay between pulse duration, bandwidth, and resonant coupling in high-Q nonlocal dielectric metasurfaces, advancing their use in ultrafast and nonlinear nanophotonics.

We present a complete field-theoretical renormalization-group (RG) analysis of the Brownian Ising Model (BIM), in which a $\mathbb{Z}_2$ order parameter is coupled to a passive conserved density, breaking detailed balance. Using the Martin-Siggia-Rose formalism and an $ε=4-d$ expansion, we show that this density-order parameter coupling is RG-relevant below four dimensions and drives the system to a new non-equilibrium fixed point, distinct from the Ising universality class. Critical exponents are computed at lowest nontrivial order, some of which require a dedicated two-loop analysis. At large scales, the density acts as an effective noise that is white in time but long-range in space, enhancing order-parameter fluctuations and producing a negative anomalous dimension $η$. A defining feature of the new class is that the correlation and response functions acquire different anomalous dimensions, $η\neq 2 - γ/ ν$ - a direct, observable signature of fluctuation-dissipation-theorem violation at large scales that cannot occur in equilibrium. We also find a small correction-to-scaling exponent, implying large preasymptotic corrections that must be accounted for in numerical and experimental tests. We further derive a set of relations among renormalization factors that hold to all orders in perturbation theory, following from the linearity of the density dynamics and an emergent shift symmetry. These yield an exact scaling relation $ν= 2/(d+z-2)$ at the BIM fixed point and establish that the Ising universality class, as well as that of quenched diluted-Ising, is unstable in $d=3$. This establishes the BIM fixed point as the unique infrared attractor for any nonzero diffusion constant.

The Ising universality class is remarkably robust to non-equilibrium perturbations, which generically flow to zero under renormalization. We show that this robustness fails when an order parameter is coupled nonreciprocally to a conserved diffusive density. Below $d_c=4$, the renormalization group flows to a fast-diffusion fixed point at which the density acts as a long-range multiplicative noise, producing a novel universality class. The non-equilibrium nature of the fixed point is manifest in the large-scale violation of the fluctuation-dissipation relations, reflected in a splitting of the scaling exponents of the two-point correlation and response functions--a measurable hallmark of non-equilibrium critical fluctuations. A two-loop calculation establishes the stability of this fixed point but yields a small correction-to-scaling exponent $ω\approx0.020$ in $d=3$, implying strong finite-size corrections. An all-orders modified Harris criterion $ν>2/(d+z-2)$ confirms that the BIM fixed point governs criticality in $d=3$, with Ising universality recovered only at $d=2$.

Solvable time-dependent systems provide important settings for studying non-equilibrium physics, where exact results are rare. They are also useful for benchmarking quantum simulations, which can directly probe real-time dynamics beyond the reach of conventional numerical approaches. In this paper, we show that the four-dimensional Chern-Simons theory offers a natural and unifying framework for constructing such systems. Focusing on classically integrable field theories, we consider a generalization of the four-dimensional Chern-Simons theory in which the usual holomorphic one-form is replaced by a more general, spacetime-dependent one-form. This yields a systematic procedure for generating time-dependent integrable field theories and establishes a universal relation: for every theory obtained in this way, the allowed time dependence coincides with the one-loop renormalization group flow. Despite the explicit time dependence, these theories retain Lax integrability and remain solvable by inverse scattering methods. Our construction applies to both ultralocal and non-ultralocal theories and extends previously known time-dependent sigma models to a much broader class of integrable systems. It also admits rewriting as dilaton gravity coupled to matter, producing a large family of classically integrable dilaton gravity theories in two dimensions. We also comment on connections to time-dependent integrable models studied recently in condensed matter physics and non-autonomous integrable systems arising from dimensionally-reduced Einstein gravity.

Non-Abelian anyons emerging in fractional quantum Hall states carry a characteristic entropy, $ΔS = k_B \log d$, where $d$ is the anyon's quantum dimension. This $\mathcal{O}(1)$ entropy can, in principle, be extracted from charge measurements of an antidot via Maxwell relations. However, equilibrium charge measurements in fractional antidots have proven to be challenging with conventional charge detectors. Here, we propose a scheme based on an antidot embedded in an interferometer, in which the charge can be inferred from the recently observed time-dependent switching of the interference phase. Performing such non-local charge measurements at equilibrium, the characteristic $\mathcal{O}(1)$ entropy of non-Abelian anyons (e.g., $d = \sqrt{2}$ for the $ν= 5/2$ state) can be extracted for intermediate temperatures, which exceed the level spacing of the interferometer edge, but are much smaller than the level spacing of the antidot.

We establish a sharp upper bound on the densities of finite-gap solutions of the sine-Gordon equation. The bound is derived directly from the finite-dimensional hierarchy, without explicit integration of the finite-gap solutions. The maximal density is determined by the roots of the invariant polynomial. An analogous sharp upper bound is established for a bounded class of finite-gap solutions of the sinh-Gordon equation.

We present a method for computing reducible, normally hyperbolic, invariant tori with internal quasiperiodic dynamics in autonomous ordinary differential equation systems. The approach is based on the parameterization method of KAM theory; thus, it is a Newton scheme with small divisors. Since the inner dynamics of the torus is prescribed, the corresponding system parameters for which such a torus exists are simultaneously determined. The method is amenable to a form of pseudo-arclength continuation, enabling the traversal and computation of saddle-node bifurcations. We give explicit algorithms for the methods and demonstrate their applicability with two numerical examples.

We estimate the strength of chaos of probe waves and probe geodesics in different smooth supergravity backgrounds of decreasing supersymmetry and/or increasing length of the AdS throat in the interior (LLM geometry, supertubes, superstrata). We find that the wave chaos becomes stronger and stronger with less supersymetry and longer throats; in other words, chaos becomes stronger as we approach black hole solutions. Geodesic motion shows the opposite trend, becoming more and more regular. Testing the wave chaos by its compliance with the Berry random wave hypothesis and the geodesic chaos by computing Poincare sections, we explain the dichotomy between wave and geodesic motion by the existence of stable periodic orbits inside long throats while the overall measure of KAM tori decreases. Computing the Renyi entropies for the dual CFT states in the weak coupling regime, we show that they do not have such universal trends and the complexity depends on the specifics of the state rather than just the amount of supersymmetry and throat length. We conclude that the hierarchy of BPS chaos works differently in the bulk and in field theory, and in either case cannot be simply extrapolated to black holes.

This work focuses on the two-dimensional, nine-velocity (D2Q9) lattice Boltzmann model. First, we show that the D2Q9 scheme cannot achieve secondorder accuracy unless the cubic velocity terms are neglected, and we explain how some of these parasitic terms can be eliminated. Second, we demonstrate that the standard choice of the equilibrium distribution has no effect on the equivalent PDE at second order. Finally, we numerically investigate the effect of these cubic terms and study different choices of equilibrium distributions using a new benchmark called the Oblique Dipole Benchmark, which describes obliquely propagating 2D vortex dipoles with periodic boundary conditions.

We propose a kinetic model of opinion dynamics under selective media influence on both graphon-based and real-world networks. The media action, inspired by Hallin's theory of spheres, is incorporated through a model predictive control strategy designed to steer agents' opinions toward a desired target opinion. For the resulting Boltzmann-type description, we analyse the evolution of the moments and, in the quasi-invariant interaction limit, derive a Fokker--Planck-type equation together with a characterisation of its stationary states. We also prove exponential convergence to equilibrium in the Fourier metric. Numerical experiments are performed on networks generated by a Gaussian graphon and on real-world, single-issue Twitter networks, allowing us to investigate the role of control and interaction parameters, as well as the impact of the subset of agents subject to media influence. Using real-world social networks data to initialise opinions and infer the interaction structure, we then compare the dynamics obtained on the original networks with those produced by Gaussian graphons fitted to their adjacency matrices, thereby assessing the descriptive power of the graphon approach for real-world opinion dynamics.

In this study, we investigate Lie symmetries of the (2+1)-dimensional Boussinesq equation, which has been proposed to model the propagation of gravity waves on the water surface, with particular emphasis on the head-on collision of oblique waves. We consider this equation in a more general form involving an arbitrary function f(u) and establish a complete Lie symmetry classification with respect to the admissible forms of the nonlinearity. For the canonical equations arising from the classification, we construct reductions to ordinary differential equations by using an optimal system of two-dimensional subalgebras. Furthermore, we examine the exact solutions of the equation and analyze the stability of the traveling wave solutions.

The accurate simulation of urban flow is key to designing building ventilation, understanding cities' micrometeorology, and predicting pollutant dispersion. Reynolds-Averaged Navier-Stokes (RANS) simulations are a common modelling approach for simulating urban flow, but their accuracy depends on the closure model and its parameters. These parameters are inferred from benchmark cases, but they are not necessarily suitable for realistic urban environments, which involve different physical mechanisms. This is referred to as the transferability problem of RANS urban modelling. The objective of this work is to propose a robust Bayesian method to {sequentially} infer RANS parameters for urban flow modelling. Key to the approach is the mathematical derivation of the parameter-regularised ensemble Kalman filter (PR-EnKF), which is the analytical solution of the data assimilation problem for the sequential parameter estimation. The cost functional is regularised using the prior knowledge on the turbulence parameters, thereby ensuring that the Bayesian updates remain within physical ranges. The parameters are first inferred on an isolated building, and then transferred to three cases of increasing complexity: (i) a high-rise building, (ii) a multi-building array, and (iii) the Shinjuku district urban environment. Results show that the PR-EnKF achieves faster convergence, reducing parameter uncertainty by an order of magnitude and reconstruction errors by up to 50%. Because of the regularisation, the PR-EnKF selectively updates the most important parameters. This work enables robust large-scale urban flow simulation whilst reducing the computational overhead of model optimisation for urban planning and air quality assessment.

Virtual and augmented reality (VR/AR) visors require compact and lightweight optics. Metalenses have been widely proposed as ultrathin replacements for bulky refractive eyepieces, with performance typically assessed using point spread function (PSF) and modulation transfer function (MTF) measurements. Here, we design, fabricate, and characterize a single-layer silicon nitride metalens optimized for the three emission peaks of an RGB OLED display, and benchmark it against refractive and Fresnel eyepieces. Under coherent illumination, the metalens exhibits a tightly confined PSF and strong mid-to-high spatial-frequency MTF, suggesting excellent optical performance. However, when evaluated in a realistic system-level VR testbed incorporating incoherent OLED illumination, a dynamic-pupil eye model, and near-eye-relevant focal lengths, the same device exhibits pronounced ghosting and background haze. We show that these artifacts arise from the intrinsic multifocal nature of polychromatic diffractive focusing and demonstrate that common mitigation strategies such as narrowband filtering and long-focal-length relay optics merely mask, rather than resolve, the issue. Our results establish that meta-optics for AR/VR must be evaluated under realistic system-level conditions to reveal their true imaging performance.

Miniaturized lasers with sub-micron dimensions are of broad interest for optical science, on-chip communication, sensing, and biomedical barcoding. Recently, lowest-order half-wave lasing was demonstrated in semiconductor-on-metal cavities by operating away from the highly dispersive and absorptive surface-plasmon resonance. Here, we demonstrate half-wave-mode lasing near the surface-plasmon resonance at ~630 nm using high-gain indium phosphide (InP) nanoparticles on ultrasmooth gold substrates. The smallest lasing particle, estimated from simulated dispersion curves to have a length of ~115 nm and height of ~100 nm, emitted at 730 nm in air, representing one of the smallest reported active laser cavities. Linewidth and threshold pump fluence generally decreased as the lasing wavelength shifted farther from the plasmon resonance. In larger particles with lengths of 280~480 nm, we observed lasing attributable to second- and third-order plasmonic modes with progressively narrower linewidths. These results extend half-wave dipolar lasing toward near-infrared and visible wavelengths and further push laser miniaturization toward the plasmonic localization limit.

In optics and acoustics, in structured beams, non-diffracting polarization measures in clearly diffracting beams, and spin direction distributions in the core of these waves that have Skyrmionic behavior have been found. We here study the equivalents of these for fermions and show that in corresponding circumstances the non-diffractive spin textures persist independently of spin, statistics, or kinematics (or propagation speed of the structured wave). As a part of the study, we present LG solutions for Dirac particles valid for both relativistic and non-relativistic kinematics.

Optical metamaterials provide a versatile platform for engineering homogeneous electromagnetic media whose distinct phases are characterized by phase diagrams in constitutive-parameter space. However, existing studies of hyperbolicity, topological properties, and exceptional-point formation often rely on highly symmetric models or case-by-case numerical parameter scans, leaving a unified analytic framework that identifies phases and phase transitions directly from the constitutive tensors lacking. Here, we develop a general theory that yields exact analytic criteria for topological transitions, exceptional-point transitions, pinch-off Lifshitz transitions, and optical Lifshitz transitions in homogeneous media. Applying this framework to a tractable example of a gyroelectric medium with anisotropic chirality, we uncover exceptional rings and negative refraction induced by gyroelectric-chiral coupling. By enabling the exact determination of phase boundaries, our theory provides a predictive framework for discovering previously unexplored electromagnetic phases and offers new principles for the systematic design of optical metamaterials.

Physical systems and objects in the object-oriented programming (OOP) paradigm share a common organizational structure: identity, state, and governing laws. We argue that making this structural correspondence explicit, rather than leaving it as tacit knowledge embedded in scientific software, provides a natural and general basis for teaching computational physics. The correspondence is independent of both programming language and mathematical formalism, applying equally to differential equations, eigenvalue problems, and variational principles. To illustrate this idea, we present Ollin, an open-source Python framework for computational physics education organized around the correspondence principle. Four examples spanning mechanics, celestial mechanics, quantum mechanics, and variational optics demonstrate that the same representational structure can be preserved across diverse physical domains. In each case, the class definition serves as the physical model, its attributes encode the state and physical parameters of the system, and its methods encode the governing laws, regardless of whether those laws are integrated, diagonalized, or optimized. More broadly, the correspondence principle provides a conceptual framework for relating the organization of physical models to the organization of code.

Visitation entropy, the Shannon entropy of an individual's distribution of visits across locations, is a widely used metric in the human mobility literature. Yet, its widespread use rests on assumptions that are rarely made explicit: entropy is defined over a fixed set of states, and estimating it empirically requires abundant, well-sampled observations. The limitations that arise when entropy is used outside this setting have been documented and explored in fields such as statistical physics, ecology, and cryptography. The implications for mobility studies, however, remain unclear. Here, we leverage synthetic and empirical trajectories to systematically examine the strengths and weaknesses of visitation entropy as a measure for characterizing human mobility. We show that, for a sequence of locations visited by an individual, the visitation entropy primarily reflects the number of unique locations visited and sequence length, which together explain 90.7% of its variance in empirical data. We also show that shorter sequences systematically lead to an underestimation of entropy, showing finite-sample bias to be a key limitation. As a result, comparisons of groups based on visitation entropy should be treated with caution. We find that apparent entropy gaps between genders, commuters and non-commuters, and urban and rural residents are reduced, or reversed, once sequence length and the number of unique locations are taken into account. Finally, we show that these issues can be addressed by controlling for sequence length and the number of unique locations, and by complementing entropy with structural network measures, which provide more nuanced insight into how mobility is organized beyond the aspects that entropy alone captures.

Wide-span spectral and noise characterization at millimeter-wave and terahertz frequencies is increasingly important for emerging wireless, sensing, and spectroscopy systems, yet remains challenging for conventional electronic instrumentation because of the complexity and calibration burden of extender-, multiplier-, and mixer-chain architectures. Here we show that photonics can provide attractive alternatives to this highly challenging electronic instrumentation through an antenna-coupled thin-film lithium niobate electro-optic receiver. Our implementation performs fast spectral reconstruction across the widely separated WR 9.0 (80-125 GHz) and WR 2.8 (240-380 GHz) bands with a single component, with carrier-frequency errors below 3 MHz, scan speeds up to 25 THz/s, and a displayed average noise level below -104 dBm/Hz. We then extend to Hz-level resolution and demonstrate phase-noise characterisation capabilities across these ultra-wide bands by using a mode-locked femtosecond-laser frequency comb as a sampling clock, mapping high-frequency carriers to aliased intermediate-frequency tones. We validate the method by measuring the phase noise of 90.225 GHz and 270.675 GHz carriers using the same chip and system configuration. We further improve the sensitivity of the phase-noise measurement through cross-correlation between two optical probe-pulse channels. Finally, we show that the same technique can be applied synchronously to ten widely spaced frequencies distributed across more than 100 GHz, with the option to quantify their mutual coherence. These results establish TFLN integrated photonics as a scalable route toward compact millimeter-wave/terahertz instrumentation.

Encoding quantum information into discrete optical frequencies, or "frequency bins," uses different colors of light as additional information channels, allowing each photon to carry more information than polarization alone. We present a computational design for an electrically tunable silica microtoroid that selects desired frequency channels after a polarization-entangled photon pair has been generated without disturbing the photons' polarization entanglement. In the proposed architecture, the 750 nm signal photon passes through the microtoroid, while its entangled 880 nm partner bypasses the resonator and serves as a reference for the selected frequency channel. The principal challenge is resonator birefringence: because horizontally and vertically polarized light resonate at slightly different frequencies, the selected frequency can reveal the photon's polarization state and weaken the quantum correlation between the photon pair. We solve this problem by adding a small lithium-niobate tuning element controlled with a single applied voltage. The voltage shifts the resonator so that it responds almost identically to horizontally and vertically polarized light, reducing the remaining mismatch to only 0.286 optical linewidths across nine frequency channels. The photons remain strongly entangled after passing through the device, with a concurrence of C = 0.969, a Bell-state fidelity of F = 0.981, and a Bell parameter of S_max = 2.785. If the relative timing between the frequency channels is also controlled, the same device can generate a nine-channel polarization-frequency hyperentangled state with an effective dimension of K = 8.97. This computational design provides a compact, electrically tunable bridge between polarization-entangled photon sources and future high-capacity quantum photonic systems.

Mid-infrared (mid-IR) frequency combs are powerful tools for molecular sensing, industrial monitoring, and precision spectroscopy, yet their development beyond 5 um remains limited. Electro-optic modulation offers a promising path toward compact, agile comb generation, but extending this approach into the mid-infrared has been hindered by the lack of practical, high-performance modulators. Here we present an approach that leads to efficient generation of mid infrared frequency combs around 9 um, by employing ultrafast, room temperature, free space electro optic intensity modulators. By driving a single modulator with short electrical pulse trains, we realize both single- and dual-comb operation from a continuous-wave quantum cascade laser, providing a compact and versatile platform for mid-IR spectroscopy. This scheme produces combs with tunable repetition rates down to the megahertz range with direct observation on an electrical spectrum analyzer without any interferometer. As a proof of concept, we perform single- and dual-comb spectroscopy of a germanium etalon and an ammonia cell, achieving resolution far beyond that of conventional Fourier-transform infrared (FTIR) spectrometers and highlighting the potential of this approach for precise measurements in the long-wavelength molecular-fingerprint region. These results establish high-performance mid-IR modulators as a promising route toward practical, energy-efficient frequency-comb systems for sensing and spectroscopy.

Optical two-qubit gate with high success probability, fault-tolerant fidelity, and high-purity outputs is a fundamental yet unsolved challenge, essential for large-scale optical quantum computing toward quantum advantage. Here, we propose a feasible scheme for such gate using thin-film lithium niobate platform, enabling \c{hi}(2) nonlinear photon-photon interaction with 100% efficiency. By decoupling photon interaction and qubit flip operations, fidelity ceiling is removed, and output state purity is recovered by spectral-phase pre-compensation based on a full-spectral photon interaction model, yielding a CNOT gate with 84% success probability, 93% purity, and unity fidelity.

Complex component systems are collections of discrete units such as species, words, genes, whose observed realizations are naturally summarized by component counts. Many empirical laws have been observed in those systems, such as Taylor's law, Zipf's law, and Heaps' law, and domain-specific mechanisms are often employed to explain their emergence but, despite their ubiquity, a unifying framework remains elusive. In this work, we propose a null model showing that, under heterogeneous latent rates and finite sampling, several commonly observed scaling relations can arise without invoking domain-specific mechanisms. Taylor's law, for instance, reflects a crossover between sampling noise and genuine system heterogeneity and it is largely insensitive to the detailed latent distribution, while Zipf's and Heaps' laws arise from the convergence of order statistics and distinct component counts under heavy-tailed but otherwise generic priors. Our work thus suggests that these ubiquitous patterns are better interpreted as a transient sign of statistical convergence instead of fundamental principles that require tailored generative explanations.

Accurate interpretation of diffusion magnetic resonance imaging (dMRI) signals in grey matter (GM) remains challenging due to the complex, heterogeneous, and densely packed cellular environment. Numerical phantoms provide a controlled framework for investigating the relationship between microstructure and diffusion signals, yet existing approaches often lack the morphological realism and multi-cellular organisation required to faithfully represent GM tissue. In this work, we introduce Contextual Cellular Growth (ConCeG), a generative framework for creating individual cells or constructing dense, three-dimensional, multi-cellular GM substrates informed by real neuronal and glial morphologies. The method combines topological neuron synthesis with a spatially constrained growth network, allowing for the controlled generation of heterogeneous cellular environments with realistic intra- and extracellular compartments. Synthetic cells are generated using morphological and topological characteristics derived from biological reconstructions. We validate the framework through comparisons of structural features with real cellular data, demonstrating strong agreement in branch order, length, angle, and tortuosity distributions. Power spectrum analysis further shows that both intracellular compartments reproduce the spatial correlations observed in biological tissue. Together, these results show ConCeG provides a biologically grounded framework for generating grey matter substrates suitable for large scale diffusion MRI simulation.

Hybrid mode-selective photonic lanterns transform an input complex point-spread function into several single-mode outputs, where a selected core feeds the fundamental mode to a photonic science instrument, while the remaining cores are used for wavefront sensing in a closed-loop adaptive optics system. A neural network maps the intensities of the wavefront sensing cores to an estimated wavefront correction, which is applied to an upstream deformable mirror. However, there exists a trade between maximizing the amount of light reserved for the photonic instrument and the reduced signal-to-noise ratios for the wavefront sensing cores. We explore wavefront correction for the Seidr instrument, a part of the Asgard Suite for the Very Large Telescope Interferometer. We evaluate different neural network architectures, comparing wavefront estimation performance for different wavefront error types, as a first step toward addressing the signal-to-noise trade-off.

Seidr is a new interferometric beam combiner within the Asgard Suite, utilizing infrastructure common to the Bifrost instrument at the Very Large Telescope Interferometer. Seidr combines hybrid mode-selective photonic lantern injection modules with a kernel-nulling photonic chip backend to enable deep H-band nulling for high-contrast studies of exoplanets, exomoons, and exo-zodiacal dust. This instrument update summarizes Seidr's current design maturity and recent simulations of the point source-to-lantern outputs. We also outline progress on our neural network-based wavefront estimation scheme, which uses the photonic lantern outputs to sense phase fluctuations, designed to feed back to Baldr's deformable mirror, and improve nuller light injection.

Cu-doped TiO$_2$ thin films for hydrogen sensing were synthesised by reactive DC magnetron sputtering in Ar/O$_2$/He mixtures, with the He fraction used as a control parameter for film growth. By combining normal-angle deposition (NAD) and glancing-angle deposition (GLAD) with post-deposition annealing, the effects of He on microstructure formation and sensor performance were examined. X-ray diffraction and electron microscopy revealed that He promotes nanostructuring, lattice expansion in as-deposited NAD films, increased porosity after annealing, and a stronger anatase character in the final oxide layers. These structural changes, which enhance the reactive surface area, lead to improved hydrogen sensing at 300\,$^\circ$C in 1~vol.\,\% H$_2$. The response of NAD films increased from 1.4 to 6.0 simply by replacing part of the argon with helium, whereas GLAD films showed only a modest increase. The observed nanostructuring is discussed in terms of a simulation-supported growth scenario involving energetic backscattered He, a reduced hammering effect, and cooling-related suppression of adatom mobility, which together favour the formation of a more open sensing layer. Helium-assisted sputtering represents a useful physical route for tailoring oxide thin films for gas-sensing applications.

The uncertainty associated with breakup events that occur during atmospheric re-entry is severe. Limited attempts to gain a better knowledge of this environment have included the use of breakup recorder-type sensor capsules that are designed to escape the demising debris cloud and survive in order to transmit data. This work models a breakup recorder undergoing this process as a rigid body experiencing hypersonic aerothermodynamic loads alongside collision dynamics with components of the demising container vehicle. The re-entry of the Edoardo Amaldi Automated Transfer Vehicle (ATV3) and the recorder placed on board, the Re-Entry Breakup Recorder 4 (REBR4) is studied in the present work. After a deterministic exploration of the nature of the dynamics of the problem, uncertainty quantification is performed to investigate the effects of initial spacecraft state, REBR detachment conditions and spacecraft fragmentation states. From this data, inferences about the nature of the real re-entry event indicate that detachment of the recorder from the cargo bay prior to main breakup events is more likely than the alternate hypothesis of the container vehicle experiencing high rotation rates.

We present a method for characterizing polarization fiber channels carrying broadband quantum signals, where narrowband filtering would waste photon flux. Wavelength-dependent polarization mode dispersion (PMD) maps each input state to a trajectory on the Poincaré sphere; we show that the singular value decomposition of the band-averaged rotation matrix yields, in closed form, the optimal input states, the mutually unbiased measurement bases, and their infidelities. The three singular values provide a compact, bandwidth-dependent channel signature that separates first- from higher-order PMD, and the resulting 5%-infidelity bandwidth gives a practical filtering budget. We characterize deployed fiber links in Masdar City and demonstrate PMD mitigation by concatenating two channels through a single polarization controller.

Second-harmonic generation (SHG) is a widely used nonlinear optical process that has been optimized for high frequency conversion efficiency, broad bandwidths, and optimum spatial and temporal quality of the generated light. However, limitations of the generation efficiency arise due to temporal dispersion and pulse walk-off effects introduced by the nonlinear crystal. To circumvent these limitations, quasi-waveguide schemes, in particular multi-pass cells (MPCs), can be employed, providing flexible phase-tuning capabilities. In this work, we demonstrate for the first time broadband and efficient SHG in an MPC. SHG efficiencies of 72% (for 63 fs, 50 uJ pulses) and 47% (for 15 fs, 20 uJ pulses) are obtained by simultaneously matching the relative phase and group delay between the fundamental and second-harmonic fields over multiple passes through the MPC. Our work sets a new record for efficient Gaussian-beam, ultrashort-pulse SHG and provides a general route to address efficiency-bandwidth limitations of nonlinear optical processes.

Mid-infrared photothermal microscopy combines the chemical specificity of infrared absorption with the spatial resolution of visible-light detection, but practical implementations face a persistent trade-off between forward-scattering (FWS) and backward-scattering (BWS) detection geometries. FWS provides quantitative, shape-independent phase contrast but requires two-sided optical access that is difficult to achieve in aqueous or thick samples. BWS offers convenient single-sided access, but its signals are strongly distorted by depth-dependent interference for micron-scale objects. Here we present a bidirectional femtosecond mid-infrared pump-probe holographic microscope capable of switching between FWS and BWS geometries within a single instrument, and use it to introduce and validate a new imaging modality, internal forward scattering (IFS). IFS exploits the back-reflection generated at the top surface of the mid-infrared-transparent sample substrate as an internally generated forward-scattering illumination wave, isolated from the directly backscattered field via temporal coherence gating. Using polystyrene beads on CaF2 substrates in air, water, and a refractive-index-matched glycerol-water mixture, we show that IFS reproduces the signal magnitudes and temporal dynamics of true FWS measurements while retaining the mechanical simplicity and single-sided accessibility of BWS. These results establish IFS as a practical, quantitative alternative to conventional FWS and BWS geometries for photothermal, and more broadly quantitative phase, imaging, with direct relevance to single-sided imaging of biological or solvent-contained specimens.

In topological photonics, the bulk-edge correspondence is conventionally imported by direct analogy with electronic systems, overlooking the fundamentally distinct spacetime symmetries of Maxwell's and Schrödinger's equations. In this work, we challenge this prevailing paradigm by demonstrating that non-trivial bulk topology alone is insufficient to guarantee localized edge states in photonic platforms. Using a Su-Schrieffer-Heeger-inspired photonic crystal, we unveil a generalized bulk-edge correspondence intrinsically shaped by the relativistic nature of electromagnetic waves. This constraint imposes a strict frequency cutoff, a feature fundamentally absent in electronic topological insulators, which enables a new regime of frequency-controlled spatial localization near the cutoff. Furthermore, we demonstrate that this generalized correspondence is polarization-dependent: transverse electric (TE) and transverse magnetic (TM) edge modes exist in different parameter regimes and exhibit distinct dispersion relations, including distinct zero-dispersion points. Our framework redefines the theoretical boundaries of topological photonics, unlocking new opportunities for polarization-selective dispersion engineering and robust pulse propagation in topological photonic platforms.

Scaling the efficiency of optical parametric amplifiers (OPAs) without degrading spatio-temporal pulse quality is fundamentally limited by spatio-temporal walk-off, intensity dependent gain, and back-conversion. Here, we numerically and experimentally demonstrate an OPA architecture based on a free-propagating quasi-periodic geometry that overcomes these bottlenecks. Operating within a single nonlinear crystal, the system utilizes pass-by-pass dichroic idler rejection to suppress back-conversion, a birefringent crystal for temporal resynchronization, and free-space diffraction to improve the spatial overlap along propagation. Starting from 1.9 μJ 330 fs pulses at 515 nm and a continuous-wave seed at 783 nm, the generation of 0.8 μJ 160 fs signal pulses at the same wave-length is obtained at a repetition rate of 500 kHz. This simple architecture achieves a 64% quantum efficiency, a 42% pump-to-signal conversion efficiency, and 80 dB of gain while maintaining excellent spatial and temporal quality, providing a scalable platform for ultrafast sources with arbitrary emission wavelengths

For the design and optimization of optoelectronic devices, accurate knowledge of the complex refractive indices of the constituent materials is essential. Herein, we present a fast and non-destructive approach for the extraction of the refractive indices from reflectance and transmittance spectra of samples with lateral dimensions down to the micrometer scale. Microspectroscopy, based on the combination of a standard optical microscope and a spectrometer, enables the assessment of the optical response of multilayer stacks using high-magnification optics with correspondingly large numerical apertures. Employing a numerical formalism explicitly accounting for the influence of the numerical aperture, allows for precise retrieval of the refractive index without resorting to dispersion models. We demonstrate the applicability of the proposed method for large-area, homogeneous, optically incoherent samples such as transparent glasses and absorbing 4H-SiC, for a SixNy thin film on glass substrate, and for mechanically exfoliated flakes of highly oriented pyrolytic graphite and MoO3, as representatives of uniaxial and biaxial optical anisotropy. While the results prove excellent agreement with values reported in literature, the case of graphite highlights the limitation for probing the out-of-plane refractive indices due to reduced sensitivity. Finally, we discuss possible extensions towards retrieving the full anisotropic tensor of the refractive index, establishing the proposed approach as a methodologically sound alternative to spectroscopic ellipsometry.

Robust optical design demands quantitative knowledge of how manufacturing and alignment tolerances degrade system performance. We show that analysing the perturbation eigenmodes of the Hessian matrix gives qualitative insight about the mechanisms of performance degradation (such as couplings) that is invisible to classical sensitivity-matrix analysis based on the Jacobian alone. Via the envelope theorem, we prove that the first-order sensitivity of the fully compensated system is identical to that of the uncompensated one; refocusing only acts at second order through the Schur complement of the Hessian. We propose the trace of the tolerance-scaled Hessian %,$\Tr(\mathbf{S}\mathbf{H}\mathbf{S})$, as a single scalar robustness metric. Demonstrated on an off-axis three-mirror anastigmat and scaled to a twenty-three-parameter surface-figure model, eigenmode decomposition reveals the dominant sensitivity directions and yields deterministic tolerance budgets that replace costly Monte Carlo sampling.

This paper designs and develops a miniaturized photoacoustic spectroscopy gas probe with acoustic pressure enhancement for in-situ detection in oil-immersed power equipment. The probe adopts a single-cavity single-fiber structure, integrating the photoacoustic cell and the Fabry-Pérot optical sensing cavity within a ceramic ferrule with an outer diameter of only 800 um. Through the electro-mechanical-acoustic equivalent model analysis of the semi-open photoacoustic cell, the enhancement mechanism is revealed that reducing the radius of the photoacoustic cell can form a flat region in the acoustic pressure response. The CrAgAu composite metal diaphragm is fabricated by electron-beam evaporation, and the gas vents are precisely machined by focused-ion-beam etching, successfully realizing the preparation of the sensor prototype. Experimental results demonstrate that the sensor exhibits a flat acoustic response in the frequency range of 2800-10000 Hz with a sensitivity of -15 dB re 1 mV/Pa. Using acetylene as the target gas, the detection limit reaches 71.4 ppb (with an integration time of 130 s) at the optimal operating frequency of 3275 Hz, and the linearity exceeds 0.998. In-oil tests verify its stable detection capability in oil-phase environments, providing a miniaturized, high-sensitivity all-optical technical solution for in-situ monitoring of dissolved gases in transformer oil.

Quantum state tomography is an essential tool for characterizing quantum systems and underpins nearly every experimental realization of quantum technologies. Conventional tomography relies on performing a sequence of projective measurements on many identical copies of a quantum state, requiring the measurement apparatus to be reconfigured between successive acquisitions. As the Hilbert-space dimension increases, the number of required measurements grows rapidly; in practice, additional overcomplete measurements are often performed to improve robustness to experimental imperfections. Here, we introduce a tomography platform based on structured anisotropic media that performs informationally complete measurements of photonic polarization qubits within a single acquisition. The approach employs three liquid-crystal metasurfaces with spatially varying optic-axis orientations that transform the input polarization into a far-field distribution of discrete transverse-momentum modes. Each diffraction pattern uniquely determines the polarization state, enabling its reconstruction without sequential changes to the measurement apparatus. Unlike previous implementations, our scheme is intrinsically photon-number independent: the same optical device operates identically for arbitrary photon numbers, while the desired photon-number sector can be selected afterwards through post-selection of the corresponding $n$-fold coincidence events. We experimentally demonstrate single-frame quantum state tomography of both single- and two-photon polarization states, providing a simple and scalable route toward efficient quantum-state characterization.

Exceptional points (EPs) have long promised enhanced sensing of physical signals, but have practically been limited by simultaneous enhancement of noise. Aligning an EP's non-analytic response with target perturbations while suppressing noise has, however, remained challenging. Here we show that fully passive, phase-tuned multi-port scattering networks enable scattering EPs with tailored anisotropic response to perturbations. We leverage projection-induced non-unitarity to realize effective non-Hermitian behavior when measuring only a subset of system ports. By formulating EP design in terms of the discriminant of the projected scattering sub-block and its directional derivatives, we give control-counting rules relating the number of programmable link phases to achievable Riemann surface topologies. We demonstrate our framework in a four-port photonic network by designing both an anisotropic EP and a Dirac-type EP with linear splitting along two parametric directions. We further suppress the global thermal drift response of a network-based sensor to a 3/2 power-law scaling while retaining square-root sensitivity to localized signals. Since the effective non-Hermiticity arises purely from port projection, our approach transfers to integrated photonic and microwave meshes, acoustic circuits, and projected metasurfaces, offering a phase-only route to reconfigurable non-Hermitian response and noise-robust EP sensing.

To meet the stringent requirements for miniaturized and highly sensitive trace gas sensing in space-constrained scenarios, including power equipment monitoring, minimally invasive biomedical diagnostics, and in situ lithium-battery analysis, we report a MEMS-integrated fiber-tip photoacoustic spectrometer (MFPAS). The device incorporates a Fabry-Perot (F-P) photoacoustic sensor formed by directly butt-coupling a single-mode fiber (SMF) to a 3 mm x 3 mm MEMS chip with a 100-nm-thick low-pressure chemical vapor deposition (LPCVD) Si$_3$N$_4$ diaphragm. The resulting approximately 200-$μ$m deep silicon microcavity functions simultaneously as a photoacoustic gas cell and an acoustic confinement cavity. A micro-aperture fabricated at the diaphragm periphery by focused ion beam (FIB) milling serves as both a gas diffusion channel and an acoustic high-pass filter, suppressing ambient low-frequency pressure fluctuations and stabilizing the F-P quadrature point without active servo control. In gas-phase measurements, the sensor achieves a noise-equivalent concentration (NEC) of 58.5 ppb@1s, with a rapid response time of 6 s. Benefiting from its ultra-small cavity volume of approximately 1.5 nL, the device is further adapted through structural packaging for in situ dissolved gas analysis in transformer oil, where it achieves an NEC of 230 ppb@1s and a T90 response time of 320 s in the oil phase. By combining nanoliter-scale detection volume, ppb-level sensitivity, rapid response, and wafer-scale batch fabrication compatibility, the proposed MFPAS bridges MEMS diaphragm micromachining and FIB-enabled gas exchange engineering. This design overcomes the intrinsic gas-exchange limitation of conventional sealed-diaphragm optical microphones and offers significant potential for power equipment monitoring and in situ health diagnostics.

Thin core silicon nitride photonics enables ultra-low loss, CMOS foundry compatible integration that supports wavelengths from the visible to shortwave infrared. Applications that can benefit from the resulting lower cost, improved robustness, and portability include quantum sensing and computing, ultra-low noise microwave generation, optical clocks, optical gyros, coherent fiber communications, and fiber sensing. An important next step is integration of functional circuits and systems on chip with heterogeneous integration of active components such as high-performance photodetection. Yet to date integrated high-performance photodetectors on the thin film silicon nitride platform has remained elusive. In this work, we demonstrate heterogeneous integration of an InGaAs on InP substrate Modified Uni-Traveling Carrier balanced photodetector with a 15-meter-long unbalanced thin core silicon nitride Mach-Zehnder Interferometer with a bandwidth of 0.92 GHz and a responsivity of 0.305 A/W at 1550 nm with a propagation loss as low as 2.5 dB/m at 1600 nm. Using this circuit we demonstrate two functions, a meter-scale differential interferometer laser stabilization circuit achieving a nearly 23 dB noise suppression at 1 kHz offset and an optical frequency discriminator frequency noise measurement with high sensitivity across 6 orders of magnitude from 10 Hz to 10 MHz. These results demonstrate that the high performance of thin core silicon nitride devices can be combined with integrated high-performance photodetection to realize on-chip stabilized lasers and circuits and pave the path towards full systems on chip.

We study spin effects in twisted-electron propagation through the entrance or exit region of an axially symmetric magnetic coil. Starting from the Foldy-Wouthuysen reduction of the Dirac equation, we derive a paraxial spinor equation in which the longitudinally varying solenoidal field produces, in addition to the usual diagonal Zeeman term, a transverse Pauli coupling proportional to the fringe-field gradient. The scalar transverse dynamics is treated exactly by the Ermakov mapping, which absorbs the longitudinally dependent focusing into a metaplectic scaling transformation and reduces the orbital evolution to that of a stationary two-dimensional oscillator. On this background, the transverse Pauli term is treated perturbatively and yields an explicit first-order correction for arbitrary realistic solenoidal profiles. Axial symmetry implies conservation of the total projection of angular momentum, so each spin flip is accompanied by a compensating one-quantum change of orbital angular momentum. In addition, the linear coordinate structure of the perturbation restricts the first-order dynamics to at most two neighboring radial sidebands for each incoming oscillator component. We derive the corresponding transition amplitudes and show how their phases are governed jointly by the Ermakov accumulation and the diagonal spin-orbital rotation. The resulting framework provides a direct way to quantify spin-orbit mixing of twisted electrons in realistic magnetic lenses and solenoidal beam-line elements, and it identifies a route toward controlled spin-OAM conversion in engineered sequences of magnetic-field edges.

Ultrasound is a noninvasive, real-time, and therefore widely used imaging modality; yet its application in cellular and sub-cellular biology is significantly limited by rapidly increasing acoustic losses in aqueous solutions with decreasing wavelength. Here we introduce a nano-optomechanical cavity transducer platform to generate and detect ultrasound in aqueous solutions with a sub-micron acoustic wavelength. We analyze the full signal pathway through a combination of finite element method modeling and the coupled differential equations that describe the dispersive optomechanical interaction. Our findings project a signal-to-noise ratio in the thousands at ~5 GHz, limited by diffraction losses and thermal-acoustic noise. This work establishes a viable path towards optomechanical ultrasound systems capable of label-free imaging at cellular and sub-cellular length scales while also providing a broader framework for optomechanical crystal device operation in aqueous environments relevant to biochemical sensing, medical diagnostics, underwater acoustic sensing, and nanoscale imaging.

Three-level systems exhibit quantum interference effects absent in two-level systems, making them important for quantum optics. Here, we study the coherent interaction of a Lambda-type three-level system with free electrons shaped by optical near fields. By treating the electron train as a quantum drive, we show that the interplay between electron modulation and the three-level system's transition pathways induces tunable interference patterns. This interaction effectively realizes electron-mediated coherent population trapping (CPT). We identify a regime that enables complete population transfer between the two lower states and the preparation of a high-coherence superposition, manifested as dark states. In particular, these driven-dissipative steady states are independent of the initial state. Our work proposes shaped free electrons as a platform for steady-state coherent control of three-level systems, enabling atomic-scale state engineering.

Micro-ring lasers (MRLs) are attractive light sources for energy-efficient optical interconnects, but their intrinsic directional bistability leads to unpredictable clockwise/counter-clockwise emission. We demonstrate stable unidirectional emission in hybrid quantum-dot (QD) III-V/Si MRLs using passive reflective feedback integrated on the bus waveguide, leaving the ring cavity unperturbed. Three reflector architectures - Y-splitter loop mirrors, adiabatic Y-splitter loop mirrors, and distributed Bragg reflectors (DBRs) - are benchmarked against a reflector-free bidirectional baseline through combined experiment and coupled-mode-theory rate-equation modeling. All designs preserve ultra-low thresholds of 0.79-1.12 mA (112-158 A/cm^2, roughly an order of magnitude below prior quantum-well unidirectional ring lasers) while enhancing single-facet output power and wall-plug efficiency, with directional isolation up to 27.65 dB for the DBR. The reflectors impose no penalty on the 4-5 GHz modulation bandwidth or its thermal robustness, establishing passive external feedback as a practical route to unidirectional QD MRLs for DWDM-scale optical interconnects.

The optically levitated system in vacuum has emerged as a powerful platform for studies of fundamental physics and precision measurements. Although various nanoparticles have been successfully levitated in vacuum, they typically lack the capability to support optical nonlinear processes. Here, we experimentally demonstrate the stable levitation of a potassium titanyl phosphate (KTP) nonlinear nanocrystal in vacuum and investigate its second-harmonic generation (SHG) properties. This levitated system intrinsically provides a pristine dark-background environment with a high signal-to-noise ratio. The trapping laser simultaneously serves as a fundamental light for efficient SHG. Moreover, the polarization of the collected SHG signal is correlated with that of the fundamental laser, providing clear evidence of the optical torque enabling controllable alignment of the nanocrystal with the driving field. Our work establishes a new route toward exploring nonlinear optical processes in vacuum levitation systems and designing novel nanodevices with high manipulation agility in a fully contact-free environment.

The fluctuation -- dissipation theorem (FDT) is shown to provide a powerful resource for the analysis of stimulated light scattering from laser-driven density gratings, including stimulated Brillouin scattering and its kinetic-regime extension. In the physical setting of stimulated light scattering by density gratings, the FDT establishes that the dynamics of disturbances induced in a medium by a laser-driven electrostrictive force unfolds via the same physical pathways as the dynamics of internal, spontaneous fluctuations in this medium at equilibrium. When integrated into a suitable kinetic framework, the FDT leads to a significant simplification of the analysis of stimulated light scattering, allowing the stimulated gain/loss spectrum to be found directly from the spectrum of spontaneous density fluctuations without the need to solve kinetic equations with an external-field term. Operating within this framework, we derive a physically intuitive closed-form solution for the stimulated gain that accurately recovers all the signature properties of sound-wave-mediated Stokes amplification in the hydrodynamic regime, provides a continuous, fully analytical crossover from the hydrodynamic to kinetic regime of stimulated scattering, and explains distinctly different properties of kinetic stimulated scattering from density gratings, consistent with experiments on stimulated scattering in moderate-pressure gases. As important physical benchmark, in the limits of vanishingly low and very high collision frequencies, this solution for the SBS gain recovers the FDT-transformed solutions of, respectively, the Vlasov and Navier -- Stokes equations.

Scalable quantum photonic technologies require single-photon emitters whose positions and emission energies can be engineered simultaneously. Hexagonal boron nitride (hBN) is an attractive room-temperature host, but deterministic creation of spectrally reproducible emitters remains challenging. Here, we use a standard scanning electron microscope as a direct-writing tool to activate bright green single-photon emitters in hBN at predefined sites, without ion implantation or post-fabrication thermal annealing. The written emitters exhibit reproducible zero-phonon-line emission centered near 536 nm, room-temperature antibunching with g(2)(0) as low as 0.08, high brightness, strong linear polarization, and stable emission. Thickness-dependent activation, stacking experiments, cathodoluminescence spectroscopy, and first-principles calculations support a carbon-related defect complex as the most plausible origin of the emission. As a proof of nanophotonic compatibility, we further activate emitters in a nanoparticle-on-mirror plasmonic nanocavity and observe photoluminescence enhancement accompanied by shortened emission lifetimes. These results establish electron-beam direct writing as a practical route to site-selective, spectrally uniform green quantum emitters in hBN, offering a promising basis for integrated room-temperature quantum photonic architectures.

Cascading failures driven by load or flow redistribution arise in networked systems such as power grids, supply chains, and cloud computing centers. Most flow-network models assume that a node either functions or fails as a whole. In many real systems, however, a node supports several distinct flows that share node-level resources, and failure in one of them does not necessarily imply failure in the others. We study this setting through multiplex flow networks with partial functionality, where a node can remain operational in some functionalities while failing in others. A heavy load on one functionality reduces the capacity available to the others, as quantified by cross-layer influence factors. When a node fails in one layer, its load is redistributed among surviving nodes in that layer, while the node may continue to operate in the others. Using mean-field analysis, we derive recursive equations for the final system sizes, namely the fraction of surviving nodes in each layer after the cascade stops. We validate the analysis through simulations for several load-capacity distributions. We then examine key features of the cascade dynamics, including non-monotone robustness curves, different cascade-outcome regimes, and their relation with cross-layer influence. We map the outcomes to distinct steady-state regimes, including single-layer survival phases absent in joint-functionality models, and show that partial functionality can increase robustness relative to the joint-functionality case. Finally, we study robustness maximization under a fixed total capacity budget by comparing several capacity allocation strategies. We propose a strategy that combines cross-layer influence with local neighborhood information on load and degree, and show that it gives the strongest robustness performance across the configurations considered.

The Virtually Imaged Phased Array (VIPA) is a spectral disperser that has seen increasing adoption across various applications, including optical telecommunications, high-resolution spectroscopy, and LiDAR. Although VIPAs are typically fed by single-mode optical fiber in these applications, there is growing interest in using a multimode optical fiber feed to enable higher throughput. However, a multimode fiber feed introduces several challenges, one of which is a fundamental limit on how much light can be coupled into a VIPA. In this work, we derive a closed-form expression for the minimum tilt angle of a multimode fiber-fed VIPA required for no etendue coupling loss, and we show that this angle depends solely on the input fiber's etendue, the focal ratio of the injection cylindrical lens, and the VIPA's intrinsic parameters. Moreover, we derive the 2D and 3D etendue that can be coupled into a VIPA, as well as the VIPA's etendue coupling efficiency and total transmission, given a particular configuration. We validate our mathematical models using ray-tracing simulations in Zemax OpticStudio non-sequential mode. Our results provide practical guidance for the design of high-throughput optical systems based on multimode fiber-fed VIPAs.

In recent decades, thermoelectric (TE) materials have proven to be a complementary source of renewable energy, as they can directly convert waste heat into electrical energy. Energy-efficient, reliable, and scalable synthetic routes for the fabrication of TE materials and their processing into functional devices via low-energy and low-waste routes are necessary for the broader adoption of these materials in various applications. In this work, we report the formulation of hybrid thermoelectric (hTE) inks based on nanostructured Sb2Te3 and Bi2Te3, using PMMA as the polymer matrix and hexanedithiol (HDT) as the binder. Percolation studies were conducted to determine the optimal film composition, with an 80% nanoparticle content yielding the highest TE performance. Finite element modelling (FEM) was employed to optimize the device geometry, including the cross-sectional area ratio of p- and n-type legs, to maximize power output. Based on these results, a flexible hTEG was fabricated using the optimized ink composition. The device exhibited an output power of 950 nW and a Power Output Density (PoD) of 40.37 nW cm-2 under a 30 K temperature gradient, significantly outperforming previously reported polymer-based flexible hTEGs incorporating chalcogenides. This study presents a sustainable and effective strategy for developing high-performance hybrid thermoelectric devices through ink formulation, composition optimization, and simulation-guided device design.

Reversible enzyme kinetic constants can be audited through the Haldane relation: the apparent equilibrium constant implied by the rate law should match biochemical thermodynamics under matched conditions. We use the reciprocal cost $C_{\mathrm{Haldane}}=J(K'_{\mathrm{eq,kin}}/K'_{\mathrm{eq,thermo}})$, with $J(x)=\tfrac12(x+x^{-1})-1=\cosh(\ln x)-1$, as a calibrated, direction-symmetric reporting scale. The score is zero at agreement, penalizes reciprocal over- and underestimates equally, encodes the free-energy discrepancy in $RT$ units, and ranks records identically to $|ΔΔG|$; the contribution is therefore biochemical curation, a reproducible workflow, and fold-error calibration rather than a new ordering. We apply the score to a curated demonstration set and, under prespecified inclusion criteria, assemble a two-sided backbone of twenty-one audited single-study records. Eight genuinely independent tests pair kinetics fit without a thermodynamic prior against separately measured equilibria; all eight fall within twofold (maximum $C_{\mathrm{Haldane}}=0.069$), although this remains a feasibility demonstration. Across the full backbone, eighteen records fall within twofold and three are flagged. The backbone is concentrated in carbohydrate isomerases and epimerases, so these results are within-family observations. Because real records carry no ground-truth labels, a semi-synthetic benchmark (twenty-nine within-twofold seeds, $1{,}885$ injected known-error cases) quantifies detectability: AUC $0.784$ ($95\%$ bootstrap CI $0.725$--$0.838$), conditional on the injected error taxonomy and invariant under monotone rescaling of $|\ln x|$. All data, code, protocol, and benchmark generator are archived for exact reproduction.

In this work, we analytically model the Swiss roll metamaterial medium, showing that each unit cell can be thought of as a flux-coupled waveguide terminated in coupled inductors. The equivalent circuit model of the metamaterial accurately predicts its response, including its higher order resonances, even when made of a lossy, finite-thickness conductor. A closed-form solution for the response of this circuit model is then derived by solving a matrix equation of arbitrarily large dimensions. Along the way, we provide a general method for modeling the effective permeability of uniaxial, anisotropic metamaterials formed from axially invariant conducting inclusions.

A challenge in optics, which is shared by other sources of radiation, is to direct a coherent beam to impinge on a target behind an obstacle intervening in the line-of-sight (LoS). While self-accelerating or bending beams can help avoid an LoS obstacle, the beam does not reach the LoS target downstream beyond the obstacle. If one instead avoids the obstacle by projecting a \textit{transverse} null or shadow onto the axial plane at which it is located, an associated \textit{axial} shadow is cast that extends over the effective Rayleigh length, which reduces the utility of this approach. Unless either the wavelength or the transverse shadow width is changed, this Rayleigh length can only be reduced by modifying the structure of the illumination beam. Here we show that space-time wave packets (STWPs), in which each spatial frequency is tightly associated with a single wavelength, when used as an illumination beam, can dramatically reduce the axial extent of the cast shadow. Indeed, by utilizing STWPs we produce deep sub-Rayleigh-length shadows, in some cases with a more than two orders-of-magnitude reduction below the conventional Rayleigh length. For example, a 5-mm-wide transverse shadow in a Gaussian beam at a wavelength of $\sim1$~$μ$m has a Rayleigh length of $\sim25$~m, whereas the same shadow projected by an STWP extends only $\sim0.15$~m. We demonstrate this sub-Rayleigh-length reduction in the axially cast shadow accompanying transverse nulls whose widths extend over a broad span of widths from 80~$μ$m to 48~mm -- almost three orders-of-magnitude. These results may lead to advances in safe radiation therapy, non-LoS optical and wireless communications, selective stand-off detection, three-dimensional photolithography, and laser ablation and micro-machining.

The continuous evolution of ultrafast optical technologies requires advanced solutions for the dynamic and selective manipulation of light's degrees of freedom. While metamaterials excel at tailoring these properties through static geometrical design, achieving sub-picosecond, all-optical dynamic control without sacrificing signal throughput remains a fundamental challenge. Here, an Au orthogonal nanorod dimer plasmonic metasurface capable of ultrafast, transmissive control over the ellipticity, and optical rotation of light is presented. By exploiting the transient optical nonlinearity of Au under femtosecond excitation, hot-electron generation can be selectively induced in either nanorod by leveraging the pronounced geometric anisotropy of the unit cell. The transient response is captured by coupling a Three-Temperature Model (3TM) with Finite-Difference Time-Domain (FDTD) simulations. This ultrafast response is driven by non-thermal electron excitation, rapid electron-electron thermalization, and subsequent lattice heating, which dynamically break the optical symmetry of the orthogonal localized surface plasmon (LSP) modes supported by the dimers. Crucially, this active mode-mixing drives sub-picosecond polarization switching, reaching peak shifts of approximately 10 degrees in ellipticity and up to approximately 20 degrees in optical rotation with an instantaneous response and a relaxation time of approximately 3 ps, while the absolute amplitude of the transmitted signal is only weakly perturbed. By achieving macroscopic transmissive polarization shifts alongside a highly stable 40% transmission efficiency, this platform overcomes the severe optical attenuation and geometric constraints that bottleneck existing nonlinear architectures, paving the way for low-latency, high-speed optical switches and modulators essential for next-generation nanophotonic networks.

Spinal cord stimulation is routinely used for the treatment of chronic pain, and is increasingly being investigated for the restoration of movement after paralysis. However, most of our current knowledge on epidural spinal cord stimulation relies on empirical approaches and semi-hybrid models. Hence, optimizing this therapy requires a better theoretical understanding of how stimulation affects the target neural structures. Using the physical properties of tissues combined with an electromagnetic, dielectric, and metallic description of them, we derived the electric field transmission coefficient through a given number of layers of tissue in the spinal cord. We then used this model to calculate the transmission of electric field energy in target neural tissues such as grey or white matter for different pulse width. Simulations suggest that the propagation through the tissues of the spinal cord of an epidurally applied electric field has a complex relationship with the field pulse width. In addition the electric field energy is absorbed at the junction between dura matter and cerebrospinal fluid, and within the cerebrospinal fluid as induced current. These currents hit the different bodies floating within the cerebrospinal fluid, and/or dorsal horn in a very localized region of the spinal cord. The proposed models allow theoretical calculation of the transmission factor for different numbers of layers of tissue within the spinal canal, and for different electric field pulse widths. This eases the prediction of the expected energy reaching each layer of the spinal cord for different tissues, parameters, and electrode configurations, simplifying the optimization of spinal cord stimulation treatment.

The time auto-correlation of auxiliary wave functions (TACAW) method enables efficient simulations of ultra-low-loss electron energy loss spectra (EELS) arising from vibrational and magnon excitations. In practical applications to realistic materials systems, however, TACAW calculations become challenging due to the large system sizes required for models containing defects, interfaces, impurities, or grain boundaries, as well as the substantial computational cost and data throughput associated with molecular dynamics and multislice calculations. Here we discuss a practical methodology for large-scale TACAW simulations and present torched-TACAW, a freely available implementation of the TACAW part of the described workflow for efficient STEM-EELS simulations. The overall approach combines molecular dynamics based on foundational machine-learned interatomic potentials, partitioning of elongated supercells, and on-the-fly processing of multislice outputs in order to enable near ab initio quality simulations with tractable memory use and data flow. Using rutile TiO2 as a model system, we analyze important numerical aspects of the method, including windowing and supercell partitioning, and demonstrate atomic-resolution STEM-EELS simulations for thick samples.

Biological neurons assign credit across branching dendrites, where synaptic drive, dendritic conductance, local voltage, and somatic teaching signals interact to shape synaptic plasticity. We study conductance-based dendritic networks with E/I synapse banks, shunting inhibition, and tree-structured branch-to-soma coupling, and examine when restricted somatic feedback can approximate compartment-specific backpropagated errors. Exact gradients factor into local eligibility x compartment error terms: the eligibility uses presynaptic activity, driving force, and input resistance, whereas the fast non-local term is a path-specific error obtained by transporting a soma error through dendritic gains. This factorization turns local learning into a credit-signal compression problem. We test the hypothesis that shunting inhibition benefits learning under these constraints when it reshapes the compartment-error field to better match global scalar, per-soma, low-rank, or path-structured feedback. Exact-gradient reconstruction verifies the factorization; path-gain, rank, broadcast-fidelity, inhibition-intervention, and transported-error-oracle diagnostics support the proposed mechanism. Under nonnegative conductances and per-soma 5-factor (5F) feedback, shunting LocalCA remains 5--6 percentage points below matched backpropagation on MNIST, Fashion-MNIST, and figure-ground MNIST, indicating that feedback-field fidelity remains a major bottleneck. These results show how E/I conductance, shunting inhibition, and dendritic branching can reshape credit-signal geometry in restricted local learning.

Neuroimaging research requires manipulating heterogeneous data structures, including raw MRI volumes, volumetric parcellations, cortical surface meshes, tractograms, and connectivity matrices, across tools with incompatible interfaces and file formats, forcing researchers to repeatedly re-implement routine but technically demanding operations. We present CLABTOOLKIT, an open-source Python package that consolidates these operations into a single, coherent framework by representing volumetric, surface, and streamline data as interoperable Python objects. Five core data structures (Parcellation, Surface, AnnotParcellation, Tractogram, and Connectome) encapsulate common neuroanatomical entities and provide consistent methods for loading, processing, and exporting data across standard neuroimaging formats (e.g., NIfTI, GIFTI, FreeSurfer annotations, TCK/TRK), including connectome generation from a parcellation and scalar-map projection onto tractogram streamlines. Complementary modules support BIDS dataset management, FreeSurfer integration, diffusion MRI processing, morphometric analysis, graph-theoretical network analysis, and GPU-accelerated multi-panel visualization via PyVista. The toolkit comprises 19 modules organised into six layers, exposing 13 object-oriented classes with 234 methods and 207 standalone functions, and a JSON-based configuration system enables workflow customization without code changes. Unlike existing neuroimaging libraries, which typically address these tasks separately, CLABTOOLKIT combines color and lookup-table management, parcellation manipulation, multi-surface visualization, and tractography utilities within a single framework. CLABTOOLKIT is compatible with Python 3.9-3.12 and released under the Apache 2.0 license. Source code, documentation, and example workflows are available at https://github.com/connectomicslab/clabtoolkit.

Spatial consistency is a fundamental physical property of wireless channels that reflects the smooth evolution of the channel between spatial locations. At the cluster level, it requires similar multipath components (MPCs) remain grouped into the same clusters as the transceivers move, enabling consistent cluster tracking. Cluster-level spatial consistency is essential for realistic cluster-based channel models, especially for potential 6G techniques such as massive MIMO, integrated sensing and communication, and terahertz (THz) communication. However, existing clustering and tracking methods do not fully exploit spatial correlations of MPCs. In tracking-after-clustering, clustering and tracking are decoupled, while joint clustering-and-tracking mainly relies on cluster centers from the previous snapshot. In this work, we propose a Mahalanobis-distance-based simultaneous clustering and tracking (MD-SCT) algorithm to capture the joint distribution of clustered MPCs in the delay, angular and spatial domains. Under Mahalanobis distance, MPCs in successive snapshots are associated with existing clusters, thereby inherently tracking while clustering. The algorithm is further applied in the sub-THz band. Performance is evaluated using mean square successive difference and gradient change rate. The results demonstrate that the proposed algorithm yields smoother cluster evolution. This improves the reliability of clustered channels for spatial consistency modeling in 6G.

Low-precision neural networks are attractive for resource-constrained hardware, but fixed-point arithmetic introduces failure modes that are often hidden by idealised quantisation models. In particular, two's-complement overflow wrapping can corrupt hidden activations by changing both their magnitude and sign, leading to unstable numerical error propagation and severe accuracy degradation. This paper proposes a Lyapunov-stabilised quantisation framework for low-precision neural networks operating under hardware-style wrapping arithmetic. The hidden-state energy is monitored through a layerwise Lyapunov function, and a monotone projection is applied to enforce bounded and non-increasing state evolution across depth. The method is evaluated on MNIST using a compact patch-based transformer under post-training quantisation and quantisation-aware training with fixed-point bit-widths from 4 to 16 bits. Monte Carlo results show that unconstrained wrapped quantisation-aware training collapses to near-chance accuracy across 6-16 bits, with activation overflow rates exceeding 11%. In contrast, the proposed monotone Lyapunov projection suppresses activation overflow to below 0.012% and restores stable low-precision learning, achieving 86.55% accuracy at 12 bits. These results demonstrate that Lyapunov-based state control can act as a hardware-aware stabilisation mechanism for reliable fixed-point neural inference and training.

Adaptive token processing has emerged as a promising approach for improving the efficiency of semantic communication systems. However, existing semantic communication frameworks largely overlook token-level multiple access and the impact of semantic interference among simultaneously transmitted semantic tokens. In this paper, we propose Adaptive Token Selection and Token-Domain Multiple Access (ATS-ToDMA), a novel cross-modal semantic communication framework that jointly performs semantic token selection, interference-aware scheduling, and semantic-aware power allocation. The proposed framework introduces a Semantic Signal-to-Interference-plus-Noise Ratio (SSINR) metric that captures the combined effects of channel impairments and semantic interference arising from token similarity. A transformer-based scheduler is developed to allocate selected semantic tokens across token-domain transmission slots while mitigating both intra-modal and cross-modal semantic interference. To characterize the behavior of the proposed system, analytical bounds on semantic interference and feasible token occupancy are derived, together with a closed-form approximation for semantic-aware power allocation. Simulation results demonstrate significant gains in semantic throughput and semantic decoding accuracy while reducing aggregate semantic interference and transmit power compared with OMA, Semantic NOMA, Random-TS, and Greedy ATS benchmarks.

This letter analyzes the performance of on-off digital noise (OODN) modulation under multi-user scenarios. While prior works have addressed single-link operation, the impact of co-channel interference remains unexplored. We consider $K$ synchronous OODN interferers over AWGN and fading channels and derive a unified analytical framework for the bit error probability (BEP). The optimal likelihood-ratio detection threshold is obtained, along with a closed-form expression for the resulting irreducible error floor, which climbs geometrically with the number of co-channel interferers and yields a simple admission-control rule on network density. The analysis is extended to $κ$-$μ$ fading, covering practical millimeter-wave channels with dominant line-of-sight components. Results, corroborated by Monte Carlo simulation, show that the floor is governed by the on-off interference and the non-coherent detector rather than by the noise waveform.

Channel charting (CC) enables data-driven user localization in wireless networks by embedding channel state information (CSI) into low-dimensional representations. In multi-cell scenarios, each base station independently learns a local chart via neural encoders, leading to misaligned representation spaces across overlapping coverage areas. This lack of consistency hinders network-level tasks such as user tracking, handover prediction, and resource allocation. To address this issue, we propose a principled framework for multi-site channel charting based on topological signal processing. We model the collection of local charts as a network sheaf, which encodes consistency constraints across the network and enables the coherent integration of locally learned representations into a shared global structure. This formulation introduces an interpretable inductive bias that promotes alignment across base stations while preserving local geometric fidelity. Building on this model, we develop a multi-site channel charting architecture and an alternating optimization algorithm that jointly updates neural encoders and inter-site orthogonal transport maps, with theoretical guarantees on consistency. Experimental results validate the effectiveness of the proposed approach, demonstrating improved cross-site alignment without degrading the quality of local embeddings.

Ambient Internet-of-Things backscatter devices at known locations can act as low-cost passive anchors by creating geometrically anchored reflected paths in cellular networks. Unlike reconfigurable intelligent surfaces, practical backscatter devices are independently controlled and lack a common phase reference; their modulation signatures may be known, but their reflection gains and residual phases are generally uncalibrated. We study how much localization information survives this incomplete per-device calibration in uplink non-line-of-sight (NLOS) positioning, where the direct NLOS path and the backscatter-assisted paths share an unknown scatterer. Treating the common channel gain, the relative backscatter response, and the residual device phases as nuisance parameters, we derive closed-form equivalent Fisher information matrices for calibrated, partially calibrated, and fully uncalibrated operation. The analysis shows that unknown device phases remove carrier-phase information from the backscatter-assisted paths, whereas joint uncertainty in the common gain and relative response leaves the direct NLOS path with only bandwidth-dependent delay information. The resulting position-domain bounds show that device count alone is insufficient: the passive anchors must also observe the common scatterer from sufficiently diverse directions. For joint single-snapshot identification of the user equipment and scatterer, at least two devices in two dimensions and three in three dimensions are necessary. The results identify deployment implications for Ambient Internet-of-Things positioning and show which calibration losses also apply to separable subpanel-based reconfigurable-surface architectures.

Channel estimation in extra-large multiple-input multiple-output systems is challenging due to near-field propagation, where the array response depends on both the angle and distance of the propagation paths. Existing near-field channel estimation methods typically rely either on fixed angle-distance grids, which suffer from grid mismatch effects, or on multi-stage refinement procedures with increased computational complexity. To address these limitations, this paper proposes the \textit{dictionary-learning iterative soft-thresholding algorithm (DL-ISTA)}, a method for joint near-field dictionary learning and channel estimation based on the iterative soft-thresholding algorithm. The proposed method jointly estimates the sparse channel coefficients and the continuous angle-distance parameters through alternating optimization, thereby avoiding discretization errors associated with fixed grids. To promote robust convergence, the angle-distance parameters are initialized using Sobol sequences, which provide uniform coverage of the parameter space. Numerical results show that DL-ISTA outperforms a baseline with comparable computational complexity and attains comparable or better accuracy than a substantially more complex benchmark.

Integrated sensing and communications (ISAC), empowered by dynamic metasurface antennas (DMAs), has emerged as a promising paradigm for next-generation wireless networks. However, existing DMA-based designs commonly rely on the frequency-flat response model for DMA elements, which is accurate only in narrowband scenarios and can cause significant phase and magnitude mismatches in wideband and ultra-wideband systems. This paper investigates a DMA-based wideband ISAC system under a frequency-selective Lorentzian response model, which accurately captures the frequency-dependent behavior of DMA elements. We aim to jointly balance the aggregate signal-to-interference-plus-noise ratio (SINR) of communication users and the signal-to-noise ratio (SNR) of the radar target. To this end, we first develop an alternating optimization framework based on projected gradient ascent (PGA), deriving closed-form gradients of the objective function with respect to the digital beamforming vectors, resonance frequencies, and damping factors under the frequency-selective Lorentzian DMA model. We then propose an unfolded PGA architecture that preserves the interpretability of model-based optimization while learning key hyperparameters to accelerate convergence. Simulation results show that the frequency-selective Lorentzian model improves performance by approximately 20\% over its frequency-flat approximation. Moreover, deep-unfolded PGA achieves up to 20-fold faster convergence and improves the objective value by up to 7\% compared with PGA-based benchmarks.

Accurate channel estimation in orthogonal frequency division multiplexing (OFDM) systems remains challenging when demodulation reference signal (DMRS) observations are sparse and noisy, and when DMRS configurations vary across deployment scenarios. This paper proposes DANCE (Diffusion-based Noise-Adaptive Null-space Channel Estimation), a diffusion-based channel estimator for OFDM systems. We formulate DMRS-aided channel estimation as a sparse linear inverse problem whose measurement operator is induced by the pilot pattern. The resulting range-null space decomposition separates the measurement-constrained range-space component from the unobserved null-space component, which is reconstructed through a learned diffusion prior. To avoid directly imposing noisy pilot samples as exact constraints, DANCE introduces a noise-adaptive posterior correction into the reverse diffusion process. The correction coefficient and the residual sampling variance are jointly calibrated according to the observation noise level, thereby reducing pilot-noise injection while retaining useful measurement information. We further design a conditional U-Net denoiser for complex-valued OFDM channel grids, where the real and imaginary components are represented as separate feature channels and downsampling is performed only along the subcarrier dimension. Simulations based on 5G NR tapped delay line (TDL) and clustered delay line (CDL) channel models show that DANCE achieves consistently lower normalized mean squared error (NMSE) than conventional estimators and diffusion-based posterior sampling methods under different signal-to-noise ratios, DMRS configurations, Doppler frequency shifts, and train-test distribution mismatches.

Integrated sensing and communication (ISAC) enables intelligent wireless infrastructure but raises growing regulatory concern as fine-grained personal trajectory histories become a byproduct of sensing. General Data Protection Regulation (GDPR) Articles 5(1)(c) and 5(1)(f) require that personal data be limited to what is necessary and protected through appropriate technical measures against unauthorised reconstruction. This paper addresses both requirements through a Fisher information density (FID)-constrained trajectory sharing scheme for robot collision avoidance, where sensing estimates are perturbed according to local information content before sharing. Experiments on real pedestrian traces show that FID-controlled sharing achieves a strictly better privacy-utility tradeoff than fixed-error perturbation: at matched missed-conflict rates, reconstruction leakage and sustained exposure lengths are consistently lower, establishing information-aware perturbation as a principled technical measure aligned with GDPR data minimisation and integrity requirements.

Sixth-generation (6G) wireless networks are expected to serve as AI-native infrastructure, transmitting meaning rather than mere bits -- a shift that makes semantic communication the central paradigm for next-generation connectivity. Deep learning-based semantic encoders show compelling gains in bandwidth efficiency; however, their dependence on large transformer models with hundreds of millions of parameters is at odds with the sub-millisecond latency, microjoule energy budgets, and kilobyte memory footprints of the constrained IoT and edge devices that will dominate 6G endpoints. Tiny language models (t-LMs) -- compact, quantised, task-specialised models deployable on microcontrollers, mobile system-on-chips, and edge accelerators -- are the enabling technology for closing this gap. This review provides a unified treatment of (i) the theoretical foundations of semantic information, covering semantic entropy, channel capacity, and rate-distortion theory; (ii) a two-axis taxonomy of t-LM-based semantic communication systems across five architecture classes and six compression paradigms; (iii) a survey of model compression techniques -- quantisation, pruning, knowledge distillation, low-rank adaptation, split computing, and neural architecture search -- through the lens of semantic quality preservation; and (iv) semantic-aware resource allocation frameworks for 6G multi-user networks. Evidence across the surveyed literature shows that compression can reduce semantic encoder size by up to 99.98% while preserving task accuracy, that split computing achieves device-side encoders with as few as 640 parameters, and that knowledge graph integration cuts transmission energy by 65%. Seven open challenges are identified, spanning theoretical gaps, system design, knowledge-base management, post-quantum security, and hardware co-design, with a 3GPP standardisation roadmap toward IMT-2030.

Near-field integrated sensing and communication (ISAC) can deliver the high spatial resolution and transmission capability with the shared spectrum and hardware. Due to the partial overlap between communication scatterers and radar targets, the sensing information can provide valuable priors to enhance the channel estimation while fusing the two heterogeneous modalities remain challenging. To address this problem, a Cross-Attention Transformer based Channel Estimation Neural Network (CAT-CENet) is developed, which includes a communication pilot branch generating the the Key and Value features and a sensing information branch generating the Query feature. By elaborating the three-module structure, CAT-CENet can focus on features of overlapped targets automatically without need of identifying them in advance. The modality contribution is theoretically analyzed based on the Shapley value to verify the cross-attention gain achieved by CAT-CENet. Simulation results show that CAT-CENet outperforms the state-of-the-art schemes, especially with the higher overlapping proportion, and is robust to the model pruning.

The unmanned aerial vehicles (UAVs) will play an important role in the future urban transportation systems. This requires designing robust localization schemes especially for non-cooperative UAVs that do not share any information about their movements. This paper designs a multimodal UAV localization framework which utilizes camera, LiDAR and radar sensing modalities. The underlying data processing and the subsequent inference of the UAV location are distributed among the sensing nodes and the edge server attached to the base station. The proposed UAV localization framework addresses three key challenges. First, the sensing nodes have limited computing and communication resources, and they contain only single modality sensors. Second, the multimodal data differ greatly in the sampling rates, time alignment and the encodings. Third, the changes in the environment and the hardware failures cause the modal data to degrade, or to be completely missing. The proposed localization framework utilizes several data processing modules including a information-bottleneck (IB)-based compression module that extracts the most relevant features from each modality, a time-encoding alignment module that provides the unified representation in a shared latent space, a multimodal fusion module that accounts for the degraded and missing data, and a Mamba-based regression module that predicts the present UAV location. The experiments involving a real-world dataset demonstrate that the proposed framework accurately and reliably obtains the UAV location while outperforming other existing frameworks.

In the Internet of Vehicles (IoV), transmitting high-dimensional multi-modal sensory data to edge servers for time-sensitive tasks faces severe spectrum bottlenecks. To address this, we propose a foundation model-driven over-the-air token fusion (AirTF) framework for task-oriented multi-modal token communications. Unlike existing schemes for segmentation that rely on convolutional neural networks (CNNs) with limited local receptive fields, AirTF leverages vision transformer (ViT) encoders to extract globally contextualized semantic tokens from distributed heterogeneous sensors. By concurrently transmitting these spatially aligned tokens over a shared wireless channel, our framework exploits the superposition property of the multiple access channel to inherently fuse complementary multi-modal semantics (e.g., RGB and infrared) directly over the air. This mechanism significantly enhances spectral efficiency compared to orthogonal transmission. Furthermore, the integration of a pre-trained foundation model provides critical visual priors, effectively addressing the data-hungry nature of ViTs on limited, scenario-specific semantic segmentation datasets. Experiments demonstrate that AirTF consistently outperforms orthogonal transmission and CNN-based fusion baselines across AWGN and fading channels. Additional evaluations under a three-user setting, residual synchronization errors, and imperfect channel state information estimation further confirm its robustness. The source code will be made publicly available upon acceptance.

Rotor inter-turn short-circuit (ITSC) faults in synchronous generators introduce electromagnetic asymmetries that can lead to torque ripple, unbalanced magnetic pull, and progressive mechanical degradation. While most existing studies focus on binary classification and severe fault conditions, the assessment of incipient rotor ITSC severity using displacement-sensitive vibration measurements remains relatively underexplored. This paper proposes a vibration-based diagnostic framework for multi-class severity classification of rotor ITSC by integrating an eddy-current displacement sensor with physically motivated feature extraction. An 18-dimensional hybrid feature set is designed to characterize electromechanical modulations induced by rotor electromagnetic asymmetry. Using an XGBoost classifier with leave-one-out cross-validation, the proposed approach achieved 90.56% overall accuracy, including 99% recall for healthy operation and 87% recall for mild fault conditions. The results suggest that displacement-sensitive vibration analysis enables effective severity-aware diagnosis of rotor ITSC with minimal sensor requirements.

Spectrum sensing is a fundamental problem in the upper mid-band, where spectrum resources are shared with incumbent systems. This paper considers frequency-domain occupancy estimation when the number of primary user signals, their bandwidths, and their signal-to-noise ratios are all unknown. We develop a generalized likelihood ratio test and a computationally efficient search procedure that combines binary search with dynamic programming to select the set of intervals maximizing the sum of log-likelihoods. The proposed method is validated through both simulation data and over-the-air experimental data using an upper mid-band software-defined radio (SDR), demonstrating its practical applicability.

Direction-of-arrival (DoA) estimation is a fundamental array processing task that has benefited substantially from deep learning. Deploying such methods across distributed edge devices introduces privacy and communication constraints that federated learning (FL) can address. Yet, standard FL algorithms treat DoA as a generic classification problem, ignoring the underlying physics of the array manifold. To address this, we propose a physics-informed FL framework for DoA estimation that incorporates steering-vector geometry directly into the local training objective via a manifold-aware regularizer. Unlike existing FL baselines, the regularizer in our framework penalizes discrepancies in steering space rather than label space, exploiting the known geometric structure of the array manifold. We provide theoretical convergence guarantees for our framework, showing convergence to a stationary point. Simulation results confirm that our physics informed approach outperforms multiple FL baseline approaches across iid and non-iid data conditions.

In cell-free massive multiple-input multiple-output (CF-mMIMO) systems, the canonical uplink local receiver is the local minimum mean square error (LMMSE) receiver with large-scale fading decoding (LSFD) at the central processing unit (CPU). The LSFD coefficients are derived under the use-and-then-forget (UatF) lower bound of the ergodic rate, and computing these coefficients introduces additional fronthaul overhead and computational complexity at the CPU. This paper investigates local receiver design directly from the true ergodic-rate objective under perfect local channel state information (CSI). By introducing an expectation-based constraint and leveraging large-system random matrix theory, we develop a functional-variational approach that yields the asymptotically optimal quasi-LMMSE (Q-LMMSE) receiver in closed form. A key insight is that the Q-LMMSE receiver shares the same direction as the conventional LMMSE receiver, differing only by an instantaneous CSI-dependent scalar, and thus incurs the same per-access point (AP) complexity. More importantly, this scalar varies across APs and implicitly provides adaptive weighting for the direct summation at the CPU, thereby completely eliminating the need for statistical LSFD coefficients and the associated CPU-side computational overhead. Numerical results demonstrate that the proposed Q-LMMSE receiver consistently outperforms the LMMSE-LSFD benchmark in terms of the ergodic rate, achieving approximately a {5\%} gain when the number of antennas per AP is low, while operating with strictly lower system-level complexity.

In this paper, we investigate the dual-function radar-communication (DFRC) design for massive multiple-input multiple-output (MIMO) systems equipped with 1-bit digital-to-analog converters (DACs) at the transmitter and 1-bit analog-to-digital converters (ADCs) at the receiver, motivated by the need for low-cost and power-efficient implementations of massive MIMO systems. We consider a downlink scenario where the transmit signal matrix is optimized to enhance sensing performance while satisfying communication quality of service (QoS) requirements. Specifically, the objective is to minimize the 1-bit Cramér-Rao bound (CRB) for estimating the azimuth angle of a point-like target under symbol-level constructive interference (CI) constraints. We conduct an asymptotic analysis of the 1-bit Fisher information, revealing its nonmonotonicity with the signal-to-noise ratio (SNR), and introduce amplitude constraints to exclude regions where the objective function value is clearly suboptimal and facilitate convergence to high-quality solutions. The resulting problem is a nonconvex optimization challenge with coupled binary and linear constraints. We transform the discrete problem into a continuous constrained one, characterize its global and local minima, and tackle it via the augmented Lagrangian method (ALM) and a spectral projected gradient (SPG) method combined with nonmonotone line search. The solution is further refined via local search and cutting-plane techniques. Extensive numerical experiments verify our analysis, showing that the proposed approach exhibits promising DFRC performance compared to benchmark schemes.

Tactile sensing is central to how robotic systems interact with the real world, yet current solutions face a tradeoff between sensing area and system complexity. This work investigates metallic ultrasound waveguides as distributed tactile sensors fully interrogated from a single proximal transducer. Using cylindrical indenters, we characterized the acoustic response to single and multi-point contacts with varying forces and contact materials. For single point indentation, the applied force was well captured by a linear relationship with the ratio of the reflection to transmission coefficients (F = a * R/T) across all nine tested materials (R2 >= 0.95). The calibration slope, a, correlated strongly with the material's effective contact modulus (log--log Pearson r=-0.98). The reflected energy partition was found to be a load-independent parameter related to the contacting material's properties, enabling material classification independent of force. For the two-indenter experiment, both contact forces were recovered from the waveguide signal and were in close agreement with reference load cell measurements (contact 1, R2 = 0.97; contact 2, R2=0.95). The approach was extended to two-dimensional metallic sheets, confirming both contact localization and material-dependent effects. Overall, these results validate metallic waveguides as a robust platform for distributed tactile sensing, providing contact localization, force estimation, and material-class discrimination for the contacting body.

In this letter, three novel spatial modulation (SM) schemes specifically designed for pinching-antenna systems (PASS) are proposed, offering a flexible trade-off between bit error rate (BER) performance and decoder computational complexity. The proposed one-dimensional quadrature SM (1D-QSM) scheme exploits dual waveguides to independently transmit the orthogonal symbol components to eliminate mutual interference. Meanwhile, the 2D-QSM and 2D-generalized SM (2D-GSM) schemes jointly utilize waveguide and pinch indices to enhance spectral efficiency. Results demonstrate that 1D/2D-QSM schemes achieve significantly lower computational complexity and superior BER performance at low spectral efficiencies compared to existing literature and the proposed 2D-GSM. In contrast, 2D-GSM provides enhanced reliability at high spectral efficiencies, outperforming both existing literature and the proposed 1D/2D-QSM schemes. Finally, analytical upper bounds for BER and decoder computational complexity are rigorously derived and validated through simulations.

Mastering invisible electromagnetic (EM) environment and sculpting radio waves with the dexterity of manipulating light or matter have long been aspirations in physics and information science. While information metasurfaces (IMSs) provide the physical interface to program EM wavefields, their real-world autonomy is fundamentally limited by environmental 'blindness' and the prohibitive overhead of site-specific and trial-and-error retraining. Here we propose metasurface embodied intelligence through world model (metaEI-WM), a universal and out-of-the-box paradigm that achieves expert-level performance without on-site fine-tuning. In contrast to purely data-driven agents, metaEI-WM establishes a fundamental understanding of the EM dynamics by integrating fully automated semantic environment modelling with embedded electrodynamic priors. By anticipating future scenarios in silico, it optimizes the IMS coding configurations to dynamically shape EM environments on demand. We show that metaEI-WM successfully enables zero-latency non-line-of-sight signal enhancements, symbiotic communications, and contactless physiological sensing across highly complex and unseen indoor scenarios. To the best of our knowledge, metaEI-WM is the first paradigm to achieve end-to-end automation of complex spatial channel manipulation tasks ab initio, requiring neither human-annotated data nor online training. This framework bridges the gap between digital intelligence and physical-layer wave dynamics, offering a scalable solution for robust and self-managing wireless ecosystems.

Coherent perfect absorption (CPA) extinguishes optical fields through interference and dissipation, but conventional implementations rely on material loss that is largely fixed after fabrication. Here we demonstrate all-optically controllable CPA based on frequency conversion in a periodically poled lithium niobate waveguide resonator. Pump-driven frequency conversion couples a resonant signal field at 1581 nm in the main system to a non-resonant environmental mode at 780 nm, creating a dynamically tunable effective loss channel. The nonlinear cavity acts as a tunable lossy beamsplitter without intrinsic material absorption. Under coherent two-sided signal injection, we observe up to 92 % absorption. We further introduce environment-assisted CPA by injecting an external field into the frequency-converted environmental mode, turning the environment from a passive loss reservoir into an addressable coherent control port. Our results establish a frequency-conversion-based platform for all-optical control of dissipation in CPA, combining pump-tunable loss with environment-assisted coherent control.

Quantum gate teleportation enables distant parties to implement nonlocal quantum operations without physically transferring the participating qubits, making it a promising primitive for distributed quantum computing. We introduce a general framework for deterministic quantum gate teleportation based on path superposition, in which the target nonlocal operation is specified through the phase of a preshared maximally entangled resource and a suitable family of path-dependent local unitary operators. The framework establishes general design conditions that guarantee deterministic teleportation after measurement of the control qubits and the application of local correction operations. As representative realizations, we construct teleportation protocols for controlled-NOT (CNOT) and controlled-Z (CZ) gates, demonstrating that different nonlocal operations can be implemented within the same protocol architecture through appropriate choices of the design parameters. We further outline a proof-of-concept photonic realization based on spatial-path and polarization degrees of freedom. The proposed framework identifies path superposition as a versatile resource for quantum gate teleportation and distributed quantum information processing.

Hamiltonian estimation is an effective approach in studying the structure and dynamical evolution of quantum systems. The difficulty in estimating the Hamiltonian is that an $N$-qubit Hamiltonian has $4^N-1$ unknown parameters, requiring exponentially many equations for information extraction. In this paper we develop a method based on compressed sensing to estimate the Hamiltonian of a multi-qubit system. We identify a problem where as $N$ increases, the common sufficient condition (Restricted Isometry Property) for compressed sensing often fails, obstructing the application of compressed sensing in ($N\geq 3$)-qubit Hamiltonian estimation. To solve this problem, we propose a ``scale transformation" technique to restore RIP and ensure a compressive estimation of a $k$-sparse Hamiltonian using only $O(k\log(4^N/k))$ equations. In the numerical examples, we estimate the Hamiltonians of two 6- and 30-qubit systems, demonstrating the effectiveness of the method.

We investigate phase transitions in the extended Su-Schrieffer-Heeger (SSH) model with next-nearest-neighbor hoppings and an imaginary staggered chemical potential. In the presence of small non-Hermiticity, exceptional points emerge in pairs from the gap-closing momenta near the topological phase boundaries of the Hermitian limit. Utilizing the Krylov spread complexity and entanglement entropy, we analyze two dynamical protocols: (i) preparing the non-Hermitian ground state via a unitary transformation, and (ii) evolving the system under the non-Hermitian Hamiltonian. We show that the spread complexity, and long-time spread complexity as well as entanglement entropy can effectively signal phase transitions in the first and second protocols, respectively. To unravel the detailed structure of the transitions, we introduce the momentum-resolved complexity that identifies the characteristic modes and tracks their evolution with the driving parameter. In the regime where the system possesses a purely imaginary spectrum, we further identify dynamical phases based on the saturation behavior of the spread complexity. The entanglement entropy is also found to exhibit similar saturation behavior, thereby providing a more experimentally accessible probe of the dynamical phases.

Quantum networks can exhibit nonclassical correlations that cannot be explained by classical models with independent sources. While the role of entanglement is well understood, the impact of measurement design remains largely unexplored. Here we develop an operational framework for certifying network nonlocality in the bilocal Alice--Bob--Charlie network using ancilla-assisted meters to evaluate the nonlocal observables required for bilocal and fully network nonlocal (FNN) witnesses. The approach successfully reproduces both bilocal and FNN correlations in simulation. On the 156-qubit superconducting processor \textit{ibm\_kingston}, we observe bilocal nonlocality with $\mathcal{S}_{\rm BLHV}=1.067(6)>1$ after readout-error mitigation, while the FNN witnesses reach $99\%$ and $96\%$ of their certification thresholds, implying the substantially stronger requirements for FNN certification. We further show that Bob's joint measurement determines the accessible level of network nonlocality: bilocal and FNN certification are optimized by different measurement settings, while both violations can disappear even for maximally entangled states. These results identify measurement geometry as an independent resource for network nonlocality and provide a practical route toward its certification on programmable quantum processors.

We investigate the radial motion and tidal forces experienced by neutral test particles in multi-horizon black hole solutions arising from Einstein gravity coupled to nonlinear electrodynamics (NED). Focusing on the three- and four-horizon configurations, we examine how nonlinear electromagnetic corrections modify the causal structure, radial geodesic motion, and tidal-force profiles in comparison with the Schwarzschild and Reissner-Nordstrom (R-N) spacetimes. Our analysis shows that the NED field gives rise to multiple zero crossings in both the radial and angular tidal-force components, leading to successive transitions between stretching and compressive tidal regimes. More importantly, the radial equation of motion contains classically forbidden regions bounded by bounce-back points. For the class of trajectories considered in this work, these forbidden regions prevent particles from entering the spacetime domain where the tidal forces become divergent. In the super-extremal regime admitted by these solutions, the forbidden region may extend beyond the event horizon, preventing particles released from rest at sufficiently large distances from crossing the horizon. We further identify a systematic ordering of the critical charge values associated with the appearance of tidal-force zero crossings, additional horizons, and bounce-back points. For both the three- and four-horizon configurations, these critical values satisfy a hierarchical ordering, indicating that changes in the tidal-force structure precede the corresponding modifications of the horizon configuration. These results demonstrate that nonlinear electrodynamics can substantially modify the classical dynamics of neutral particles in multi-horizon black hole spacetimes through the combined effects of forbidden regions, multiple tidal transitions, and changes in the horizon structure.

Expressive parameterized quantum circuits (PQCs) are often designed under a dilemma: the growth of expressibility and entangling power (EP) that improves Hilbert-space coverage is also expected to randomize an ansatz and activate barren-plateau (BP) conditions. We show that this dilemma is not a one-dimensional tradeoff. The usual picture collapses three inequivalent objects -- parameter-ensemble coverage, fixed-circuit entangling response, and local gradient moments -- into one scalar narrative. For a fixed circuit probed by Haar-product inputs, EP is a global two-copy mean of the output-entanglement distribution, whereas entangling-power deviation (EPD) is a global four-copy fluctuation descriptor. Gradient variance, however, is a local two-copy contraction selected by a parameter light cone and a cost observable. This moment hierarchy yields an analytic separation: equal EP need not imply equal trainability, as witnessed by equal-EP circuits with different EPDs and different gradient variances. These separations turn EP and EPD into a two-dial design rule for PQC ansatzes: EP measures how far the circuit has moved along the coverage dial, while EPD monitors whether input-dependent variability remains. We find that ansatz routes can reach high, Haar-like coverage before EPD and gradient variance collapse, showing that coverage and BP activation are distinct crossover events. The EP/EPD framework thus breaks the apparent one-dimensional expressibility-trainability tradeoff into a practical design rule: search for highly expressive PQCs in the window where coverage is high but BP-like homogenization has not yet erased trainable structure.

Modified theories of gravity based on the symmetric teleparallel framework have recently attracted considerable attention as viable alternatives to General Relativity. In this context, f(Q) gravity, in which the gravitational interaction is encoded in the nonmetricity scalar Q, provides a consistent geometrical formulation that differs from the standard curvature-based description. In this work, we investigate the structure of neutron stars within a family of f(Q) gravity models by employing realistic equations of state, namely FPS, SLy, ENG and MPA1. Using the covariant formulation of f(Q) gravity, we derive the corresponding Tolman-Oppenheimer-Volkoff equations and apply them to model compact stellar configurations. Numerical integration of the field equations provides the mass-radius relations and the maximum masses supported by each equation of state, enabling a direct comparison with current observational constraints. Furthermore, we analyze the behavior of the nonmetricity scalar both inside and outside the stellar object, providing additional insight into the gravitational structure of compact stars in this framework.

Quantum compilers play a key role in transforming quantum circuits into lower-cost implementations with improved execution fidelity. This process is commonly guided by circuit-level metrics, such as gate counts and circuit depth. Although compiler pass tuning has been widely studied in classical compilation, directly transferring these techniques to quantum compilers is challenging, because quantum programs are expressed as circuits and exhibit optimization behaviors that are shaped by quantum-specific structures. Prior quantum compiler tuning approaches have begun to use circuit features to guide pass selection, but they remain limited in two aspects: they search only a small portion of the optimization-pass space, and they mainly rely on static features that do not explicitly reflect how a circuit reacts to compiler optimizations. We present QuTuner, a feature-guided quantum compiler pass tuning framework that generalizes across compilers and tuning objectives. QuTuner first builds a large optimization dataset. It then characterizes each circuit from two complementary views: static circuit features that describe circuit structure, and optimization-aware pass embeddings that summarize the circuit's responses to individual optimization passes. Using these representations, QuTuner trains two offline models to retrieve and rank candidate pass sequences for unseen circuits, followed by lightweight refinement. We evaluate QuTuner on Qiskit and PyTKET using two benchmark suites. On Qiskit, QuTuner improves the evaluation-metric reduction by up to 84.85% over the strongest baseline while reducing tuning time by 73.59%. On PyTKET, it improves metric reduction by up to 18.68% with a 64.49% reduction in tuning time. These results show that QuTuner provides an effective approach to adaptive pass tuning for quantum compilers.

With unitary coin operator that is a random matrix drawn from a uniform distribution with respect to the Haar measure, we construct a variant of a discrete quantum random walk using a d-regular undirected connected simple graph. With each step of the walk, the coin operator random matrix is drawn independently from the distribution. The expectation value over the distribution of random unitaries for the associated quantum channel gives a depolarization channel for the coin subspace and resembles the associated classical random walk where the direction a walker steps depends upon the outcome of a fair coin or balanced d-sided dice. Remarkably, despite the fact that the averaged channel depolarizes the coin subspace, measurements in the vertex subspace can be designed that would reveal information about the initial quantum state, even after many iterations of the channel. We illustrate with examples of quantum walks on Cayley graphs of Abelian groups, such as the cycle and hypercube graphs. These quantum walks are examples of bipartite strongly interacting systems, where one subsystem is strongly perturbed, yet information about the initial state in the other subsystem is potentially measurable forever. Due to its decoherence, a quantum random walk with a Haar-random coin would not be useful in search algorithms but could aid in understanding quantum systems with strongly perturbed subsystems.

Conical 2-designs are families of positive operators possessing crucial symmetry properties that allow them to efficiently characterize important quantum measures. However, little is known about their general structures and properties. We propose a construction method of conical 2-designs whose elements are not of equal traces, based on a generalization of equiangular weighted frames to non-projective operators. We provide a detailed analysis of their properties, relations to quantum measurements, and applications, together with examples that go beyond the known classes. This constitutes a major step toward a full characterization of informationally overcomplete quantum measurements.

Quantum conference key agreement (QCKA) enables multiple users to establish a common secret key with information-theoretic security and is regarded as a key primitive for secure communication in future quantum networks. However, practical implementations of QCKA typically suffer from higher noise levels than conventional bipartite quantum key distribution (QKD), making the improvement of the tolerable error threshold an important challenge. Gottesman and Lo proposed two preprocessing procedures for QKD with two-way classical communication, known as the B-step and the P-step, which enhance the tolerable error threshold. In this paper, we analyze the asymptotic security of QCKA with tripartite GHZ states and two measurement bases using two-way classical communication, including multiple B-steps and P-steps. We derive the corresponding secure key rate analytically and demonstrate that iterative B-steps can increase the tolerable error threshold beyond 20%, significantly improving upon the approximately 11% threshold achievable without two-way classical communication and the approximately 15% threshold obtained with only a single B-step. Our results show that two-way classical communication can substantially enhance the robustness of practical QCKA protocols.

Lie-Poisson electrodynamics (LPE) is a non-Abelian and nonlinear deformation of usual electrodynamics, where the gauge algebra is defined through a Lie-algebra-type Poisson bracket on space-time. We focus on the geometric approach to LPE in the absence of charged matter. We establish a non-trivial field redefinition which, under mild technical assumptions, maps the LPE dynamics to that of Maxwell theory. Using this map, for any symmetry of the Maxwell action, we construct generators of LPE symmetries and the corresponding conserved currents. In particular, we obtain deformed Poincaré transformations. We also outline a natural quantization prescription for LPE based on our field redefinition.

In this paper, we study the two-dimensional Schrodinger oscillator within a dynamical noncommutative (DNC) space. By leveraging perturbation theory, we derive the energy eigenvalues and eigenvectors and systematically analyze the effects of both dynamical and non-dynamical noncommutative settings. First-order corrections to the eigensystem are obtained, revealing that the energy shift explicitly depends on the DNC parameter tau. Furthermore, we explore the thermal properties of the system by employing the partition function. Numerical results are presented to provide a comprehensive analysis of the system behavior under the considered effects. Notably, in the DNC framework, the commutation relations and the deformation parameter are position-dependent. Using the two-dimensional Bopp-shift, we effectively map the noncommutative problem to its commutative counterpart.

We study the ratio between the trace speed and the Hilbert-Schmidt speed for differentiable finite-dimensional quantum states, $\mathcal R(φ)=\|\partial_φρ(φ)\|_1/\|\partial_φρ(φ)\|_2$. Because $\partial_φρ(φ)$ is always Hermitian and traceless, this ratio is constrained more strongly than for a generic operator. For any nonzero tangent operator $X=\partial_φρ$ of rank $r$, we prove the sharp bounds $\sqrt{2}\le \|X\|_1/\|X\|_2\le \sqrt r$. The lower bound is attained exactly for rank-two tangents, while the upper bound is attained exactly when all nonzero singular values are equal, which in the traceless Hermitian setting requires even rank. At every nonstationary point of a pure-state family, the tangent has rank two, implying $\mathcal R=\sqrt2$. For odd Hilbert-space dimension $d$, we further prove the sharp global maximum $\mathcal R\le \sqrt{d-1/d}$, with equality characterized by full-rank spectra whose positive and negative eigenvalues are separately degenerate and have multiplicities differing by one. We identify $\mathcal R^2$ as the inverse participation ratio of the singular-value distribution of the tangent operator, giving $\mathcal R$ a natural interpretation as an effective-rank diagnostic for local quantum dynamics. Furthermore, we decompose the effective rank into classical (eigenvalue) and quantum (eigenvector) contributions and prove the bound $r_{\mathrm{eff}}\le r_C + r_Q$, with equality guaranteed when either component vanishes. We establish a direct inequality linking the effective rank to the quantum Fisher information (QFI), which forces a large number of active singular modes when the QFI is small relative to the squared trace speed. Finally, we derive a hierarchy of quantum speed limits in which the effective rank controls the tightness of bounds expressed through the Hilbert-Schmidt speed.

We investigate how the compactification of some spatial dimensions in Rindler spacetime affects the vacuum expectation values (VEVs) of the field squared and the energy-momentum tensor for a charged scalar field prepared in the Fulling-Rindler vacuum state. A toral compactification is considered with quasi-periodicity conditions on the field operator along the compact dimensions. The phases of these conditions are interpreted in terms of the magnetic flux enclosed by the compact dimensions. For a general number of spatial dimensions, the components of the VEVs are explicitly separated, corresponding to the expectation values in the Minkowski vacuum. As a limiting case, the VEVs are retrieved in Rindler spacetime with trivial topology. We demonstrate that, for non-zero phases in the periodicity conditions, the vacuum energy-momentum tensor exhibits off-diagonal components with indices in the compact subspace. Near the Rindler horizon, the leading terms in the asymptotic expansions of the field squared and the diagonal components of the energy-momentum tensor coincide with the corresponding VEVs in Rindler spacetime with trivial topology. The off-diagonal components of the energy-momentum tensor vanish on the Rindler horizon. For small accelerations, the difference between the VEVs in the Fulling-Rindler and Minkowski vacua is exponentially small. The exception to this is a massless field with zero phases in the periodicity conditions. In this special case, the difference decays according to a power law as a function of acceleration. As an application, we consider the VEVs near the horizon of cylindrical black holes and topological black holes with toroidal horizons.

Quantum low-density parity-check (LDPC) codes are a promising route to fault-tolerant quantum computation, but their use requires efficient preparation of encoded states. Standard encoder constructions generate circuits through fixed algebraic procedures, yet the resulting circuit can contain substantial redundancy. We formulate LDPC encoder preparation as a circuit-resynthesis problem: given the linear-reversible matrix implemented by the encoder's CNOT block, we seek a lower-cost equivalent circuit that can be routed efficiently on the target hardware and which mitigates noise. We propose a novel optimization approach referred as two-sided Hamming descent and a noise-aware optimization pipeline for this task. Across several families of Calderbank-Shor-Steane (CSS) LDPC encoders, including Bivariate Bicycle, hypergraph-product, and entanglement-assisted codes, the proposed pipeline produces substantially smaller and shallower encoder circuits than the standard constructions and the synthesis baselines considered, cutting gate counts by 53.8% in aggregate across the benchmark and by up to 68% on the Bivariate Bicycle family. The gains remain visible after routing, where the two-qubit depth is reduced by up to 71% and translate into higher-fidelity state preparation under circuit-level noise. On the Bivariate Bicycle family, live-range scheduling further reduces routed preparation failure by up to 13.7% without adding two-qubit gates to the selected circuit. These results indicate that encoder-matrix resynthesis, combined with hardware-calibrated selection and scheduling, is an effective compiler-level tool for preparing quantum LDPC code states.

The possible interaction between dark matter and dark energy has been proposed as a mechanism to alleviate the coincidence problem and the Hubble tension. Strong gravitational lensing observations provide valuable constraints to test the properties of the dark components of the Universe. We present estimates of cosmological parameters employing strong lensing data for an interaction model Q, proportional to the dark matter density and dependent on the deceleration parameter q. The obtained results are consistent with an accelerating Universe. Furthermore, these results agree with previous studies suggesting that an interaction in the dark sector may help alleviate existing tensions in the expansion history of the Universe and highlight the potential of strong gravitational lensing as an independent cosmological probe.

Cavity quantum electrodynamics (QED) with ultracold fermions provides a promising platform for realizing many-body quantum chaos through disordered, photon-mediated long-range interactions. Such setups are inherently open and are therefore subject to dissipation arising from cavity photon loss and atomic spontaneous emission. In this article, we study the driven-dissipative dynamics of a typical cavity QED setting including controllable disorder and long-range interactions. We find that the two dissipation sources have qualitatively different structures. Cavity loss reduces to a single dephasing channel, whereas spontaneous emission in the experimentally relevant regime generates a collection of nonlocal dephasing channels. Cavity-induced dephasing preserves signatures distinguishing integrable from chaotic Hamiltonian dynamics in observables that depend linearly on the density matrix, while spontaneous emission suppresses these signatures. By contrast, quantities that probe the structure of the many-body state, such as the entanglement entropy, are strongly affected by both dissipation mechanisms. Assuming experimentally realistic parameters, we derive quantitative constraints for the observation and control of many-body quantum chaos in cavity-QED platforms.

Device-independent quantum key distribution (DIQKD) is arguably the gold standard for secure quantum communication, as it aims to rely only on observed input-output statistics of an uncharacterized device which is only assumption to obey the laws of quantum physics. A corresponding security analysis hence demands a description of a quantum experiment from a most general perspective. Under close inspection, existing proof techniques do not always meet this goal as they tend to rely on subtile assumptions on a tensor product structure of the underlying Hilbert space and a 'hidden but finite' dimensionality. In this work, we collect the tools needed for a full analysis of DIQKD in the commuting operator framework, which avoids these subtilities and provides the arguably most general view on a quantum experiment. We rigorously proof the common assumption that in DIQKD measurements can be w.l.o.g. assumed to be projective. Furthermore, we show that task of computing key rates can be casted as a non-commutative polynomial optimization (NPO) problem to which the Navascués-Pironio-Acín (NPA) hierarchy gives a correct and converging relaxation. As a tool, we generalize the integral representation for the relative entropy by Frenkel [Quantum 7, 1102 (2023)] to general von Neumann algebras and apply techniques from Kossmann and Schwonnek [arXiv: 2411.04858] for the approximation in an NPO program.

We unify two independent notions of effective temperature in dynamical thin-shell wormholes: the local acceleration temperature of the throat and the Hawking-like particle-creation temperature governed by the peeling of null rays. The effective acceleration scale, obtained by averaging over both sides of the junction, defines the thermodynamic shell temperature, while peeling functions for each asymptotic universe yield an effective particle-creation temperature. For asymptotically exponential null-ray maps, we derive the generalized Unruh identity that equates these two temperatures. Applied to Schwarzschild-Schwarzschild wormholes, the formalism yields explicit analytical expressions for the Hawking-like spectra, renormalized quantum fluxes, and quasi-horizon limits in both symmetric and asymmetric configurations. The divergence of the shell temperature near the quasi-horizon is shown to mirror the divergence of the local Tolman temperature, both originating from the infinite gravitational blueshift at a horizon. These results provide a unified interpretation of acceleration, particle creation, and temperature in horizonless dynamical spacetimes, extending the conceptual link between the Hawking and Unruh effects beyond stationary black holes.

We present the Stream Graph Navigator (SGN), a lightweight Python framework for building streaming data applications. In SGN, stream-processing pipelines are built by connecting computational components into directed acyclic graphs that run within an event loop. The time-series extension of the SGN library, SGN-TS, introduces signal processing methods to handle time series data. Together, SGN and SGN-TS provide the foundation for SGNL, a matched-filtering gravitational-wave search pipeline, and are being adopted by multiple projects across the low-latency gravitational-wave data analysis infrastructure as an extensible and maintainable framework for future gravitational-wave observations.

Swift memory burden (MB) implies that the information stored in a black hole (BH) can modify its classical dynamics when the BH is perturbed. This influences the gravitational waves (GWs) emitted during BH mergers. In this paper, we investigate how the BH memory load is determined by the features of the collapsing source. We show that the memory load can vastly exceed the information content of its progenitor. An extreme example is a BH formed in a two-particle collision, which exhibits maximal MB. We then derive bounds for BHs formed through stellar collapse and examine the impact of swift MB on BH quasinormal modes, quantifying the MB-induced frequency shift of GWs. These findings imply that GW observations probe the fundamental mechanisms of BH information storage as well as their formation history.

Two revolutions in quantum sensing are converging on the same microscope stage. Biological qubits have emerged as genetically encoded, optically addressable quantum systems inside live cells. Solid state spin based qubits have been entangled and used as nanoscale correlator magnetometers, delivering sensitivity and spatial-resolution gains that single-spin probes cannot reach. Here we report a pulsed quantum microscope that enables simultaneous quantum state manipulation of both qubit technologies in live cells. The platform combines nanosecond-gated optical excitation at 450 nm and 520 nm, three-dimensional diffraction-limited addressing by galvo beam scanning and piezo objective focus, rapidly switched static and radio-frequency magnetic fields, single-photon timing with picosecond resolution, and microwave control for frequencies from DC to 5 GHz, including the 2.87 GHz resonance of solid state spins, the 500 MHz to 800 MHz resonance of radical pairs in biological qubits, and any future qubit resonance in the GHz range. Live cell imaging with simultaneous biological and solid state nanoparticle qubits in the same cell demonstrates the power of this technique for multiplexed quantum sensing. We anticipate this approach will open new opportunities for researchers to explore quantum sensing in live cells, and, ultimately, entanglement between a solid-state qubit and a protein-hosted spin qubit.

The problem of making unambiguous probabilistic predictions about experiments involving waiting "always on" detectors remains a challenge for quantum theory. While most research on this problem studies arrival time, i.e., predicting the distribution of when detection events occur, this paper studies the arrival position problem, which is the complementary challenge of predicting the distribution of where detection events occur. Despite the widespread recognition of the arrival time problem, the inability of standard quantum theory to address the arrival position problem remains a pervasive theoretical blind spot. In this paper, we compare quantitative arrival position predictions derived from prominent proposed solutions to the screen problem. As we show, these models yield distinguishable predictions even in relatively simple experiments achievable with current technology. Notably, many of these discrepancies persist even in the far-field limit, where standard semiclassical approximations are typically assumed to be valid.

Proposed in 2022 by Dienes et al., stasis is a dynamical fixed point in the early universe in which the equation of state, $w_s$, is fixed at a constant value. In this study we show that the inflationary gravitational wave imprint of any stasis epoch is captured by a closed-form spectral template controlled by two physical inputs, the equation of state $w_s$ and the stasis duration $ΔN_\mathrm{stasis}$, that applies uniformly across every microphysical realization. The template presented here yields two independently measurable observables, the spectral tilt of the spectra in the stasis band, $α(w_s)$ and the amplitude step $C^2(w_s)$ at the beginning and end of the stasis band. Eliminating $w_s$ gives a one-parameter consistency curve $C^2 = C^2(α)$ on which the data must lie if the underlying cosmology is any constant-$w$ era. This makes the spectrum falsifiable without knowing $w_s$ in advance: a measured $(α, C^2)$ pair either lands on the curve or rules out the constant-$w$ class. We show that BBO and DECIGO can resolve the perpendicular displacement from the consistency curve to $σ_\perp \simeq 1.5\times10^{-5}$ at a tensor-to-scalar ratio, $r = 0.01$, four orders of magnitude below the curve's range in $C^2$ meaning that any off-curve deviation is detectable across a broad range of allowed $r$ values.

Quantum key distribution (QKD) is the most established approach in quantum communication. However, long-distance free-space implementations, particularly satellite links, remain challenging, especially during the day due to daylight background noise. Quantum keyless private communication (QKPC) is a quantum communication protocol that enables information-theoretic security with simpler system requirements, improved robustness against noise, and without the need for secret key distribution. QKPC and QKD are complementary, with QKPC enabling free-space links where QKD is impractical, while QKD provides channel monitoring for applications that require eavesdropping detection. Here, we report a complete implementation of QKPC in a daylight free-space experiment over a 90 m rooftop link, using an experimentally simple setup. Our demonstration includes all stages of the protocol, from encoding and synchronization to message decoding, operates entirely without auxiliary classical synchronization channels, and is implemented offline through post-processing. This work demonstrates the feasibility of practical and scalable quantum communication over high-noise daylight links and highlights the potential of QKPC as a complementary solution to QKD for future ground-based and space-based communication systems.

How do we define a bulk subsystem in quantum gravity? In \cite{Wong:2025kpz}, we argued that such a subsystem must support local holographic degrees of freedom. These are gravitational edge modes, whose entanglement creates a backreaction that fuses together subregions of spacetime. In this work we give a realization of these ideas in topological string theory, building upon \cite{Donnelly:2020teo,Jiang:2020cqo}. In this theory, a subsystem for closed strings consists of open strings ending on entanglement branes, which play the role of a dynamical entangling surface. Local holography is implemented by the geometric transition of these branes. We define a subregion open string algebra and develop a diagrammatics for open string modular flow for arbitrary states and subregion. We check that the entanglement entropy of these open strings reproduces the gravitational entropy of the associated closed string background. Finally, we relate these local transitions to defect holography.

We present a field-level inference framework for testing gravity with the large-scale structure that exploits the full information content of the galaxy distribution. Traditional analyses based on the power spectrum discard non-Gaussian and Fourier phase information, resulting in strong degeneracies between modified gravity (MG) and galaxy bias. Our approach overcomes this limitation by performing a Bayesian likelihood analysis directly on the three-dimensional galaxy number counts field, jointly constraining MG and bias parameters using both amplitudes and phases. As an illustrative application, we analyse mock data in the context of the $f(R)$ theory of gravity and a non-linear galaxy bias model. Non-linear structure formation is modelled using the COmoving Lagrangian Acceleration (COLA) method under different gravity strengths, parameterised by $f_{R0}$. The resulting dark matter fields are then mapped to mock galaxy catalogues via a non-linear bias prescription. We demonstrate that, with fixed and known initial phases, including non-Gaussian and phase information yields tighter constraints on both $f_{R0}$ and the primary bias parameter, $β$, relative to the power-spectrum-only analyses. Notably, the field-level approach breaks the degeneracies between MG and galaxy bias inherent to two-point statistics. Through a cosmic web classification into voids, walls, filaments and clusters, we find that under-dense regions are the primary drivers in distinguishing gravity models at the field level. Finally, we establish the robustness of our pipeline against variations in initial conditions, Poisson noise, and galaxy field thresholding, providing a powerful path forward for field-level tests of gravity with next-generation surveys.

We study a time-dependent non-Hermitian generalisation of the Schütte-Da~Providência model describing a bosonic mode coupled to collective particle-hole excitations. Using a time-dependent Dyson map, we construct a Hermitian counterpart and reduce the collective fermionic sector by means of the random phase approximation (RPA). The resulting dynamics is mapped to two time-dependent harmonic-oscillator branches with instantaneous RPA frequencies $W_\pm(t)$. We determine the corresponding stability regions and compute transition probabilities between instantaneous oscillator states. In first-order instantaneous-basis perturbation theory the leading transition $n\to n+2$ is proportional to $\dot W_j/W_j$, showing that it is purely nonadiabatic and absent in the time-independent case. We compare this result with exact Lewis-Riesenfeld transition amplitudes within the RPA approximation. Numerical examples show that different components of the Dyson map provide distinct driving mechanisms: the scaling parameter modulates the effective coupling, while the squeezing parameter acts through a moving-boundary contribution. In both cases the induced collective transitions exhibit parametric-resonance peaks and sideband structures.

We develop a semiclassical Langevin approach to study finite-size effects in dissipative time-crystalline spin systems. For a spin-$1/2$ model with power-law Lindblad operators, we find finite-size oscillations exponentially decaying in time, and their decay rate decreases algebraically with system size in the long-range regime. The scaling exponent of this decay time provides a direct diagnostic of the robustness of the time-crystalline, and we find that this robustness extends beyond the range of power-law dissipation exponents where the system dynamics is mean-field in the thermodynamic limit. We also apply our approach to a spin-one model with local dissipation and long-range Hamiltonian interactions, formulating it in terms of Gell-Mann variables. In this case, finite-size stability is quantified through the deviation time from the mean-field behavior and the behavior of the dominant Fourier peak. We find that both quantities scale as a power law with the system size -- marking thereby a time-crystal behavior -- when the exponent of the power-law interaction is below a certain threshold, that turns out to be in agreement with previous cumulant-expansion findings. Our results show that semiclassical Langevin dynamics provides a useful finite-size probe of dissipative time crystals in regimes beyond exact Lindblad simulations.

Using a recently-introduced quantum control model for Unruh-DeWitt detectors in superpositions of classical trajectories, we investigate the response of a detector interacting with a massless scalar quantum field in (3+1) dimensions along a superposition of circular trajectories. We present numerical results for the transition probability and effective temperature of such a detector in four distinct geometric scenarios: (a) concentric, vertically-stacked trajectories, (b) planar, horizontally-displaced trajectories, (c) static central point and surrounding circular trajectory, and (d) concentric, planar circular trajectories. For Gaussian switching functions that are much broader than the acceleration timescale, in case (a) we find only minor deviations from the well-known, effectively thermal response of a single circular trajectory, whereas in case (c) we find a significant reduction in the effective temperature and greater variation with energy gap. We conclude with a discussion of a potential analogue implementation in ultracold atom systems.

Black hole mimickers, ultracompact horizonless objects, have been proposed as alternatives to black holes in a variety of settings, including extensions of general relativity and scenarios involving matter sectors beyond the Standard Model. Their formation in gravitational collapse of matter requires quantum mechanical tunneling to occur on a length scale of the order of the Schwarzschild radius of the corresponding black hole. We propose a mechanism, based on multichannel enhancement catalyzed by quantum chaotic dynamics, that can dramatically amplify the quantum mechanical transmission across a tunneling barrier. We explore, in particular, how the nucleation of a string theoretic black shell is enhanced via this mechanism. We anticipate similar results to hold for other proposed black hole mimickers.

Boundary conditions are usually expected to cause only finite-size corrections to bulk quantities, but this expectation can fail for dynamical quantum phase transitions. In this work, we show that such boundary dependence is encoded in dynamical zero modes (DZMs) of the Loschmidt matrix, which are defined as singular vectors whose singular values vanish in the thermodynamic limit. Using the Su-Schrieffer-Heeger (SSH) and extended SSH models as examples, we find that the time interval where the Loschmidt rate functions (LRFs) under periodic and open boundary conditions differ coincides with the emergence of DZMs in the open-boundary Loschmidt matrix. These modes carry the boundary-dependent contribution: removing them from the open-boundary LRF recovers the periodic-boundary result. We further show that these DZMs lead to finite-precision numerical instability, since their finite-size singular values decay exponentially with system size and eventually become unresolved in fixed-precision arithmetic. A reliable small-size branch before this loss of precision can be used to estimate the thermodynamic LRF by linear extrapolation. Our results identify DZMs as both a diagnostic of boundary-dependent LRFs and the origin of the associated numerical instability.

Aziz and Howl have argued that a hybrid theory with quantum matter and a classical gravitational field can generate entanglement between two massive systems. In their construction, the branch-dependent contribution appears at fourth order through propagators of the quantum matter field in a fixed classical gravitational potential. We analyse this mechanism within the same perturbative QFT framework and show that the effect should not be interpreted as classical gravity mediating entanglement in the sense relevant to BMV-type witnesses. The non-separable term relies on a quantum-matter exchange channel between the two interferometers and is present only when the two systems are modelled as excitations of the same matter field. If distinct, non-interconverting matter fields describe the systems, the corresponding cross-propagator is absent and the Aziz-Howl entangling diagram vanishes. The effect relies on coherent propagation amplitudes of the quantum matter field between the two interferometers, together with postselection onto the original localised branch subspace. Moreover, the inference of entanglement is made after projecting the full QFT evolution onto a restricted final subspace containing the original localised $N$-particle branch states. This projection removes precisely the sectors that would record matter-field contamination, mode deformation, or exchange between the two interferometers. We therefore argue that the Aziz-Howl mechanism is a matter-sector cross-talk effect in a classical background, not entanglement mediated by classical gravitational degrees of freedom. Having identified the channel responsible for the Aziz-Howl contribution, we predict that it can be eliminated by using distinct, non-interconverting matter species in the two interferometers, or by inserting a barrier that suppresses matter-field propagation between them.

In their strongest usage, quantum reference frames have been described as referring to "the measurements performed by a superposed lab", "the perspective of a quantum particle", "the point of view of a superposed observer", etc. While exciting, these operational proposals have remained brief and ambiguous, leading to misinterpretations and criticism. Here, we provide a detailed specification and defense of the notion of a position-superposed lab or observer. We argue that this requires no exotic claims about quantum physics and raises no greater interpretive difficulty than ordinary quantum measurements. We then derive several consequences of taking this operational meaning seriously. We stress that the position-superposed observers that define quantum references frames are different from, and considerably less problematic than, the outcome-superposed observers considered in Wigners' friend scenarios. In particular, we show that outcomes obtained by a position-superposed observer may (without decohering the superposition) be broadcast to a well-localised one, in contrast with Wigner's friend scenarios, which require the outcomes to remain internal to the system at hand. Finally, we defend the possibility to roleplay a quantum reference frame from a classical reference frame.

The two-times problem asks how the directed, oriented time of biological and conscious systems relates to the time of fundamental physics. The Evolving Block Universe, in which spacetime grows along Ricci eigenlines through an irreversible quantum dynamics, explains the direction of physical time and its multi-scale emergence up to brain time. It does not explain how this physical time is related to the experience of temporal passage. We argue that Bergson's durée, the continuous qualitative experience of lived time, marks the first-person level beyond brain time at which this hierarchy reaches its limit. This locates the remaining gap between physical time and lived time at the epistemological level.

A squeezed dissipative bath converts the coupling phase of a bipartite unconventional Dicke model into a control parameter that suppresses both static superradiant thresholds, opening a window where only a Hopf instability survives and the two spin ensembles synchronize completely via the shared cavity mode. The squeezed bath preserves a $\mathbb{Z}_2$ parity symmetry, so conventional broken-symmetry diagnostics vanish identically. We certify the synchronized state instead through parity-even, information-theoretic witnesses: a 30\% photon-number suppression, a Husimi-$Q$ lobe-count change, a 64\% suppression of spin--spin mutual information, and a robust discord-to-mutual-information ratio $D/I = 0.50 \pm 0.05$, confirmed by full quantum master-equation simulations. These results establish parity-even witnesses as a general, entanglement-free route to certifying quantum synchronization in symmetry-protected driven-dissipative systems.

Optimal power flow (OPF) solved by an interior-point method (IPM) requires repeatedly solving Newton linear systems. When variational quantum linear solvers (VQLS) are used, each IPM iteration involves an additional nested inner variational optimization loop, which can significantly slow the overall quantum-assisted IPM convergence. To address this challenge, this paper proposes a dual-level trainable quantum IPM framework for OPF that leverages early solver-generated trajectories rather than relying on single-point prediction. The key observation is that early IPM iterates provide informative primal-dual, slack, and barrier-variable evolution about the path to optimality, while early VQLS parameter updates provide useful information about the later variational search. At the quantum-solver level, a trainable parameter model uses a short prefix of the VQLS parameter trajectory to project the remaining variational search toward a lower-cost region. At the OPF-solver level, a second trainable model uses early primal-dual IPM iterates to project a later central path state, which is restored to an admissible point before IPM refinement continues. Simulation studies show that the proposed approach reduces the number of variational updates by up to $95\%$ while maintaining OPF objective values close to the classical IPM reference. A 2-bus demonstration on real quantum hardware is also included to validate the implementation of the proposed workflow.

High-redshift massive black-hole binary (MBHB) mergers provide a probe of black-hole seed formation and early galaxy assembly, but the population detected by LISA can be modified by galaxy-scale strong lensing. We quantify this effect for MBHB mergers at $10\leq z\leq20$ and assess its impact on the inference of intrinsic formation-channel fractions. We use four channels, corresponding to light- and heavy-seed scenarios with delayed and non-delayed mergers, and compute 4-year LISA-detectable event numbers with and without strong lensing. To bracket the uncertain lens population, we compare a conservative velocity-dispersion-function (VDF) prescription with an optimistic halo-mass-function (HMF)-based prescription. We consider a reference mixture with equal seed-channel weights and additional mixtures in which the intrinsic seed population is weighted toward selected formation channels, and jointly infer the formation fractions and the strength of the lensing contribution. Strong lensing does not affect all channels equally: it can change the detected channel mixture as well as the total number of detections. This effect is weak for the conservative VDF prescription, but becomes significant in the high-lensing-rate HMF case, where neglecting strong lensing can bias the recovered formation fractions. The inference precision depends on the underlying intrinsic channel composition, while the lensing contribution is more accurately recovered in the high-lensing-rate case. These results indicate that galaxy-scale strong-lensing effects and event-count information should be included when using high-redshift MBHB detections to infer intrinsic formation-channel fractions.

We study quantum-gravitational effects on the response of an infalling detector as it crosses the horizon of a near-extremal black hole in the framework of quantum JT gravity. These effects are incorporated via the gravitational dressing needed to define both the infalling trajectory and the local observables probed by the detector in a diffeomorphism-invariant way. In the black hole exterior, a preferred choice of dressing to the Schwarzian mode of JT gravity can be motivated in connection to geometric modular flow. We show how to extend this dressing to the black hole interior, defining local observables that are gravitationally dressed to both boundaries in the thermofield double state. The gravitational dressing connects the near-horizon region to the IR sector of the Schwarzian theory, leading to measurable effects as the horizon is approached. We find that the infalling detector is able to locally determine the location of the horizon and its temperature, violating the equivalence principle, but without encountering a firewall.

We provide a statistical analysis of the measurement (shot) requirements of the quantum approximate optimization algorithm (QAOA) for the MaxCut problem. We derive sufficient conditions on the number of shots per cost operator evaluation to: (a) estimate the expected cost to within a relative error $δ$ and a confidence $1-ε$, and (b) ensure SGD-based parameter optimization converges to a target relative suboptimality level with high probability. In addition, we provide an explicit bound on the number of SGD iterations required to reach the target accuracy. Our analysis reveals an unexpected scaling phenomenon: for specific graph classes, which we formally characterize, the total shot budget needed to achieve a fixed relative-performance metric decreases as the instance size grows. This result complements earlier cost function concentration arguments regarding parameter optimization redundancy, thereby highlighting the potential for high-performance, low-overhead QAOA implementations for large-scale MaxCut instances. To assist practitioners, we translate our analytical findings into practical rules of thumb for shot-budget allocation and validate these results with numerical simulations, offering new insights into the interplay between graph size, structural complexity, and resource requirements in QAOA.

We demonstrate an end-to-end, multi-stage attack against a quantum neural network (QNN) model that is executed on a trapped-ion quantum computer. Our chain combines side-channel reconnaissance, crosstalk characterization, adversarial example generation, and a physical crosstalk attack that realizes the adversarial perturbation on the device. We cover the full attack chain on ion traps and report the corresponding superconducting-hardware experiments in the appendix. We discuss implications for QaaS providers and hardware mitigations.

This work explores the reconstruction of $f(R)$ gravity within the framework of newly proposed Viaggiu holographic dark energy (VHDE), which incorporates entropy corrections due to the dynamical nature of cosmological horizons. By implementing a correspondence between VHDE and curvature-induced energy density, we generate explicit forms of $f(R)$ for different infrared cutoffs, namely the Hubble horizon, future event horizon, and Granda Oliveros cutoff. The resulting models are analyzed graphically through the cosmological parameters such as the equation of state and deceleration parameter, revealing a successful description of the transition from decelerated to accelerated expansion. The viability of the reconstructed models is further examined through stability conditions and local gravity constraints, that yields the nature of the models as free from ghost and tachyonic instabilities and demonstrates VHDE-inspired $f(R)$ gravity model as a consistent and flexible geometric framework for explaining the late-time cosmic acceleration of the Universe.

This is an introductory course on quantum probability and quantum information, based on matrix algebras.

To make quantum algorithms executable on a particular quantum device, they need to be compiled into circuits that respect constraints of the quantum hardware. This compilation usually involves multiple steps, where many hardware-compatible circuits are generated, and the best circuit is selected. To say which circuit is best, the quality of a circuit is generally quantified by a $\textit{figure of merit}$ (FoM). For FoMs, there is a trade-off between ease of calculation and accuracy in predicted execution quality. Commonly used FoMs, e.g., the number of gates, circuit depth, etc., are easy to evaluate, but do not directly capture the effects of circuit structure and noise. On the other end of the spectrum are FoMs that require full circuit execution and take a prohibitively long time to evaluate. One example is the probability of successful trials (PST), i.e., the probability of obtaining the initial state after running the quantum circuit followed by its inverse. Here, we investigate advantages and disadvantages of different FoMs, and formulate the properties of an ideal FoM. Based on our results, we propose wPST, a weighted version of the PST that accounts for individual qubits, not just the whole state. To quickly predict PST and wPST, we design machine learning models that take into account both the quantum circuit and quantum hardware data. In numerical simulations and experiments on quantum processors, we find that our machine learning-predicted FoMs outperform commonly used FoMs, increasing the correlation with the true PST or wPST by over 50%. To make our model useful for quantum compilers, we devise a two-step process to predict the wPST for non-transpiled quantum circuits: first, we predict the additional quantum gates required for the given quantum circuit, and then we predict the wPST, accounting for coherence times in the quantum device.

We formulate false-vacuum decay in a mixed-field Ising chain with $1/r^α$ interactions as a spatially nonlocal Euclidean $φ^4$ theory featuring a fractional spatial kinetic term $\sim |q|^σ$, where $σ=α-1$. The nonlocal bounce is anisotropic in space-time and develops algebraic spatial tails, challenging the standard thin-wall picture of a compact droplet. Combining thin-wall arguments with numerical solutions of the full nonlocal saddle, we show that these tails preserve the leading thin-wall exponents, manifesting instead in subleading corrections. For $0<σ<1$, the lifetime exponent scales with the energy bias $h$ of the metastable state as $B\sim h^{-1/σ}$; for $1<σ<2$, the leading Coleman scaling $B\sim h^{-1}$ is recovered, while long-range effects are retained in the subdominant term $\sim h^{σ-2}$. Our results show that tunable long-range interactions fundamentally reshape bubble nucleation and alter false-vacuum decay in quantum simulators.

In various approaches to quantum gravity, spacetime geometry is understood to emerge from more fundamental discrete structures at the Planck scale. As sometimes posited, their presence could lead to dissipative effects in the smooth effective sector. In this paper, we develop the idea of non-conservation in gravity, by introducing an effective cosmological model within unimodular gravity, in which a varying cosmological constant arises as a consequence of dissipation. We show that this requires to incorporate hidden degrees of freedom -- termed quantum gravity defects -- that act as an effective bath for the matter fields. To illustrate the viability of the framework, we study the case of an Ohmic bath inspired by the Caldeira-Leggett model for Brownian motion, leading to a diffusion equation for the matter energy density. The results show that, starting from a primordial universe with no dark energy, dissipation can account for the generation of a small positive cosmological constant.

The one-way model is a universal model of quantum computation, driven by successive adaptive single-qubit measurements on an entangled resource state. Measurements are non-deterministic, yet if the computation satisfies one of several related families of conditions known as 'flows', the computation can be made deterministic overall by modifying later measurements depending on the outcomes of earlier ones. Flow properties also enable efficient translation from one-way computations to circuits, motivating research into rewriting one-way computations while preserving the existence of flow. Existing approaches to flow-preserving rewriting are used for compilation or optimisation and preserve both the interpretation and the existence of flow. Here, we broaden our perspective to consider flow-preserving rewriting that does not necessarily preserve the interpretation, with applications to creating test instances for software that works with flow, as well as to generating ansätze for quantum machine learning. We show that a family of just three flow-preserving rewrite rules suffices to generate any diagram with flow from a trivial diagram with the desired number of inputs and outputs. This rule set is nearly the same as the complete set of flow- and interpretation-preserving rewrite rules for one-way computations in which all measurements are Pauli; and just a small subset of the flow- and interpretation-preserving rewrite rules for arbitrary measurements.

We consider the question of convergence of particular series of integrals, which are labeled by rooted trees. Necessary and sufficient criteria for convergence are obtained, together with an explicit expression for the sum. The technique used is strongly reminiscent of the generating function approach of Galton and Watson to branching processes. The interest in these series derives from the Dyson series expansion for the perturbation of a free quantum dynamics by a local potential: the convergence of the series imlies that the perturbed dynamics exists and is unitarily equivalent with the free one.

The ringdown waves from a black hole offer a clean probe of strong-field gravity, but a matter distribution that may be present around a realistic black hole renders the background spacetime dynamical and the ringdown frequencies time-dependent. We study the odd-parity ringdown of a Schwarzschild black hole that grows through the dilute, steady, spherically symmetric accretion of a perfect fluid. Working to first order in the accretion rate, we compute the ringdown waveform directly in the time domain on this dynamical background. Since the odd-parity matter perturbation decouples from the metric perturbation, the wave mode can be described by a purely tensorial mode on the accreting background. In particular, the ratio of the imaginary to the real part of the frequency cancels both the secular variation caused by the growth of the black hole and the redshift factor, so that its deviation from the Schwarzschild value purely reflects the surrounding environment. The time dependence of the frequency, on the other hand, reflects the accretion rate and allows us to define a second observable tied to it. We argue that measuring these observables across multiple modes may provide significant information to constrain the surrounding environment of the black hole.

This paper investigates and compares quantum and classical investment strategies for portfolio construction under realistic trading and allocation constraints. The study considers a multi-period portfolio re-balancing setting in which asset weights are subject to lower and upper bounds, class-level exposure restrictions, and turnover limitations. On the quantum side, the portfolio selection problem is reformulated as a quadratic unconstrained binary optimization (QUBO) model through a finite binary encoding of asset weights, and subsequently embedded into a constrained quadratic model (CQM) solved by a hybrid quantum-classical optimization workflow. On the classical side, benchmark allocation strategies are introduced to provide an economically meaningful reference for performance assessment. The proposed framework jointly evaluates portfolio efficiency from two complementary perspectives: business performance, measured through return, volatility, and a measure of the trade-off between them, the Sharpe ratio; and computational performance, assessed through solver structure, constraint handling, and hardware execution characteristics. The resulting comparison provides a rigorous basis for understanding the practical role of quantum annealing in constrained portfolio optimization, while clarifying both its modeling advantages and its current implementation limitations relative to established classical approaches.

Quantum annealing has the potential to provide practical quantum advantage for complex optimization tasks. Here, we present a systematic assessment of an integrated factorization-machine and quantum-annealing workflow (FM+QA) for a technologically relevant application: multi-objective Pareto optimization in metal up-cycling through alloy design. To address the non-convex nature of the Pareto front, we employ the recently proposed data-driven Tchebycheff scalarization (DDTS) scheme. Our results show that FM+QA extends the applicability of QUBO-based optimization to data-driven materials discovery problems with multiple competing objectives. In particular, we analyze the scaling behavior of the approach and compare quantum annealing with classical simulated annealing using both regular binary encoding and one-hot encoding. Finally, we provide a critical perspective on the problem sizes and encoding strategies for which quantum-annealing-based optimization may become practically beneficial in the near future.

We examine the question of whether the discrete transformation CRT is a gauge symmetry of ``Quantum Gravity". Since the phrase in quotes is not yet completely well defined, we first try to define specific frameworks in which one might ask the question. We find that the general answer is NO. In asymptotically flat and AdS spaces, CRT is an asymptotic gauge symmetry in the same sense that the continuous parts of the Poincare/AdS isometry groups and scalar internal symmetries are. In eternal dS space it can be considered a spontaneously broken gauge symmetry if a certain ``Alice String" configuration is allowed in the Thermofield double Hilbert space. This requires an extension of the conventional Chern-Simons rewriting of $2 + 1$ dimensional gravity. This interpretation only makes sense if we consider the underlying quantum mechanics to be time independent. General quantum measurement and semi-classical gravitational restrictions on measuring devices put {\it a priori} limits on the existence of detectors with clocks that can actually measure proper time along a classical dS geodesic.

Recent DESI observations have posed new challenges to $Λ$CDM, showing a preference for dynamical dark energy and yielding neutrino mass constraints within $Λ$CDM that approach the lower bound allowed by neutrino oscillation experiments. In this work, we investigate cosmological constraints on the key neutrino parameters, $\sum m_ν$ and $N_{\rm eff}$, within the Schwarzschild-de Sitter black-hole dark energy (SdSDE) framework. We use cosmic microwave background (CMB) data from Planck and ACT DR6, baryon acoustic oscillation data from DESI DR2, and type Ia supernova data from DES-Dovekie and PantheonPlus. We find that SdSDE scenarios prefer a positive neutrino mass whenever $\sum m_ν$ is allowed to vary. Using CMB+DESI+DES-Dovekie data, we obtain $\sum m_ν=0.207^{+0.047}_{-0.052}~{\rm eV}$ for SdSDE+$\sum m_ν$, reduced to $\sum m_ν=0.162^{+0.055}_{-0.056}~{\rm eV}$ when $N_{\rm eff}$ is also varied. This arises from the positive correlation between $N_{\rm eff}$ and $\sum m_ν$, together with the systematic preference of SdSDE for values of $N_{\rm eff}$ below the standard value. Furthermore, the best-fit $χ^2$ comparison shows that $Λ$CDM with extended neutrino parameters is strongly preferred over the corresponding SdSDE extension. Overall, the positive neutrino mass preference induced by SdSDE may reflect parameter compensation rather than an improved global fit, a possibility that should be further tested with future high-precision observational data.

Supermassive black holes (SMBHs) powering quasars at $z \gtrsim 6$ are difficult to grow from stellar mass remnants, motivating seeds from primordial black holes (PBHs) with masses $10^5-10^7 M_{\odot}$. This range is constrained by the COBE/FIRAS bound on the CMB $μ$-distortion, which limits the small-scale curvature variance to $σ_ζ^2 \lesssim 10^{-4}$. For Gaussian perturbations, the variance fixes the far tail of the one-point probability distribution function (PDF), making the PBH abundance negligible. We call this the Gaussian barrier. The barrier can be evaded only if the variance probed by the distortion is decoupled from the tail probability controlling collapse. We implement this idea in the non-perturbative $δN$ formalism and relate asymptotic PDF tails to the global shape of the $δN$ map. Four Gaussian-cored families are analyzed: generalized-normal, stretched-exponential, power-law, and log-normal tails. After standardizing each family to unit variance, we impose the FIRAS cap and compute the distortion-limited PBH abundance in the tail-shape parameter space. The ordinary exponential tail produced by standard single-field non-attractor dynamics is still too light to reopen the seed window. Algebraic tails from fractional-potential dynamics, and sufficiently heavy log-normal tails treated as a phenomenological proxy for multiplicative dynamics, can supply seed-relevant abundances while respecting the distortion bound.

Let $\{Σ_t\}$ be a spacelike foliation of a spacetime $\mathcal{M}$, and suppose each $Σ_t$ is foliated by 2-surfaces $\mathcal{S}$ with spacelike unit normal in $Σ_t$. We show that under mild energy conditions, a MOTS $\mathcal{S}$ that intersects integral curves of past-pointing conformal Killing vector field lying in the normal space of $\mathcal{S}$ is strictly stable and evolves smoothly to a spacelike horizon. We also show that if the restriction of the divergence of the vector field to $\mathcal{S}$ is non-negative, $\mathcal{S}$ is unstable, and if negative and $\mathcal{S}$ is a 2-sphere, $\mathcal{S}$ must be strictly stable.

We evaluate exactly the radial metric factor $h(ζ)$ generated by Schwarzian averaging in the AdS$_2$ throat of an extremal Reissner--Nordström AdS$_5$ black brane. The result is a Gaussian integral against $\coth(πy)$, valid at all radial depths, which Mordell's identity turns into an exact Appell--Lerch $q/q^\ast$-series representation. The dual series identifies the nonperturbative scale $e^{-π^2 C/ζ}$ missed by any finite near-boundary truncation. The third parametric derivative required by the evaluation generates the quasimodular Eisenstein series $E_2$, absent from the classical Mordell identity. From the integral representation we prove that $\mathcal{G}_0(ζ):=h(ζ)/ζ^2$ is completely monotone and hence has no interior minimum, so any confining minimum produced by a finite near-boundary truncation is an artifact. We also compute the exact relative variance of the Schwarzian kernel, which makes the averaged-metric approximation error quantitative and shows that the absence of a confining minimum is robust across moment-based effective geometries. Applied to the temporal rectangular Wilson loop, the exact throat gives algebraic screening, $E(L)\sim -κ_{\rm IR}/L^2$, the Wilson-loop diagnostic of the semi-local quantum-liquid IR of the extremal RN brane. A numerical check in a simple matched geometry confirms that the screened saddle is the dominant string configuration, and an exact-versus-truncated force comparison shows that the apparent constant-force regime of the fourth-order truncation is not a feature of the exact geometry.

Autonomous coding agents perform well on many conventional programming tasks, but scientific computing demands a rigorous validation paradigm that extends beyond simple functional test completion: generated code must preserve physical fidelity, differentiable workflows, framework-native semantics, and scalable representations. We introduce Open Research Benchmark for Integrated Tasks in Quantum Computing (ORBIT-Q) to address this gap. At its core, ORBIT-Q contributes a carefully curated suite of complex, research-level quantum workflows that serves as a challenging testbed for modern scientific programming. ORBIT-Q combines a rigorous multi-tier verification pipeline to support two orthogonal comparisons: different agent harness and model configurations at a fixed quantum software framework, and different quantum software frameworks at a fixed agent. In our systematic evaluations, TensorCircuit-NG (TC) exhibits the highest capability and performance efficiency among the evaluated quantum software frameworks under agent-driven programming, and Codex with GPT-5.5 is the strongest tested agent configuration on TC. However, a significant performance and design gap remains between frontier autonomous agents and human expert reference implementations. We further evaluate two efficiency dimensions: agent-side resource use and artifact-side runtime. Together, these results establish ORBIT-Q as a rigorous benchmark for autonomous scientific programming, framework-agent synergy, and quantum software performance.

George Lazarides was a pivotal collaborator and friend to me. We worked together on several projects, developing and exploiting the curvaton hypothesis, which was new at the time. This is a brief overview of our joint research.

It is shown that the co-genesis of baryon asymmetry and dark matter can be achieved through the rotation of a spectator axion-like particle, because of a flip in the vacuum manifold's orientation at the end of inflation. This can occur if the axion has a periodic non-minimal coupling to gravity (while preserving the discrete shift symmetry) in non-oscillating inflation models, where the inflaton field is characterised by a runaway potential. Our rotating axion can generate the baryon asymmetry of the Universe through spontaneous baryogenesis, while at a later epoch it can oscillate as dark matter. We show that in order to avoid fragmentation of the axion condensate during the rotation, we require the non-minimal coupling $ξ\sim(f/m_P)^2$, where $f$ is the axion decay constant.

We propose a nonequilibrium mechanism for Majoron dark energy in which the late-time freezing of a physical Majoron is induced by quantum coherence in a hidden pseudo-Dirac sterile fermion reservoir. The evolving Majoron background derivatively couples to the hidden pseudo-Dirac number current and drives a lagged reservoir response with a finite memory time. In the short-memory regime, the causal response kernel reduces to \(\dot X+Γ_{\rm PD}X=β\ddotφ\). The leading linear-response matching \(Q=αX\) then yields an effective scalar equation containing the exchange structure \(q_{\rm exch}\ddotφ/\dotφ\). We show that this term can dynamically suppress the Majoron velocity and sustain a response-dominated freezing branch even when the intrinsic Majoron mass is larger than the present Hubble scale. The microscopic origin of the lag variable is identified with the phase-lagged off-diagonal coherence of the hidden pseudo-Dirac ensemble, while the response strength is controlled by a response-weighted hidden density rather than by an independent gravitating component. The resulting state is a metastable nonequilibrium frozen phase with \(w_φ\simeq -1\), rather than an exactly static cosmological constant.

Pulsar timing arrays (PTAs) are ensembles of millisecond pulsars observed for years to decades. The primary goal of PTAs is to study gravitational-wave astronomy at nanohertz frequencies, with secondary goals of undertaking other fundamental tests of physics and astronomy. Recently, compelling evidence has emerged in established PTA experiments for the presence of a gravitational-wave background. To accelerate a confident detection of such a signal and then study gravitational-wave emitting sources, it is necessary to observe a larger number of millisecond pulsars to greater timing precision. The SKAO telescopes, which will be a factor of three to four greater in sensitivity compared to any other southern hemisphere facility, are poised to make such an impact. In this chapter, we motivate an SKAO pulsar timing array (SKAO PTA) experiment. We discuss the classes of gravitational waves present in PTA observations and how an SKAO PTA can detect and study them. We then describe the sources that can produce these signals. We discuss the astrophysical noise sources that must be mitigated to undertake the most sensitive searches. We then describe a realistic PTA experiment implemented with the SKA and place it in context alongside other PTA experiments likely ongoing in the 2030s. We describe the techniques necessary to search for gravitational waves in the SKAO PTA and motivate how very long baseline interferometry can improve the sensitivity of an SKAO PTA. The SKAO PTA will provide a view of the Universe complementary to those of the other large facilities of the 2030s.

We study static, spherically symmetric solutions in $f(R)$ gravity within a $D = 4+n$-dimensional spacetime, where the extra dimensions form a compact $n$-sphere of constant radius. We derive an exact solution in which the four-dimensional part of the metric corresponds to the Schwarzschild - de Sitter metric, while the extra dimensions are stabilized at a constant radius $L_0$. A consistency condition relates the size of the internal space to the effective four-dimensional cosmological constant $Λ_4 = (n-1)/L_0^2$, which is generally too large to be compatible with observations. To overcome this issue, we include the vacuum polarization effects of quantized matter fields nonminimally coupled to curvature. Using the semiclassical approach, we obtain an asymptotically flat four-dimensional Schwarzschild solution as a limiting case, where the size of the extra sphere is determined by the vacuum expectation values of the quantum fields. Finally, we discuss the dependence of the effective four-dimensional Planck mass and the extra-dimensional radius on the radial coordinate.

We derive parent Hamiltonians in terms of oscillator operators for multimode bosonic cat resource states. The construction separates a universal branch Hamiltonian, which confines each mode to the coherent-state support $ |\pmα\rangle$, and state-dependent constraint Hamiltonians, which select the desired correlations and symmetry sectors inside the resulting branch manifold. This framework progressively removes degeneracies in the low-energy manifold and yields explicit parent Hamiltonians for GHZ-, cluster-, and W-type cat states. In the large-$|α|$ limit, the bosonic parent Hamiltonians reduce to stabilizer or exchange Hamiltonians acting on an effective logical-qubit basis. The present construction provides a direct bridge between coherent-state bosonic engineering and stabilizer-based quantum information processing.

We propose the first error-tolerant secure key leasing (SKL) for public-key encryption. As with SKL in prior works, our protocol consists of a lessor and lessee. In the protocol, the lessor encodes its secret key into quantum states and leases the key to the lessee. Then, the lessor can ask the lessee to return the secret key at a later point. The lessor is able to check whether the lessee has returned its key honestly. However, our protocol works even when the leased secret key is subject to noise. The lessee decrypts the ciphertext correctly, and the lessor verifies the return of the secret key correctly when the amount of error is below a certain threshold. Our improved protocol does not change the encoding of the secret key, and thus adds no overhead to the quantum information processing. Our most significant result is a framework to analyze the trade-off between robustness against error and security. We bridge the security of the error-tolerant SKL and that of the error-tolerant certified deletion with shortened codes, which is a relatively less explored concept in coding theory.

While typical online estimation methods can estimate the multiparameter dynamics of many systems, they may not be sufficient for a system with highly noisy measurement and rapid detection rate. In this paper, we create a hybrid estimation method by augmenting the sequential Monte Carlo (SMC) sampler, an online estimation method with an offline technique known as the time-batch estimation technique. By continuously monitoring the system, we may divide signals into batches and average them into an averaged trajectory. The system dynamics is then evolved with batch-averaged Kraus maps, for which we derive a highly efficient approximation. To facilitate the adoption of our algorithm, we present a modular derivation of the SMC methods and showcase our algorithm as an explicit example. We then implement our algorithm on the measurement signals obtained from superconducting-qubit experiments under two types of measurement setting: a fluorescence measurement and a dispersive $z$-measurement. The algorithm's hyperparameter values are chosen from independent numerical simulation, while the accuracy of our estimation is validated by a signal reconstruction method. Our results show that, for the fluorescence case, our algorithm can estimate the system's parameters better than the standard calibration method, and, for the dispersive case, our estimation is capable of finding an unexpected jump in parameter values that the standard calibration method could not find.

For the precision-based measurements, Mach-Zehnder interferometry is a widely used technique. There are various ways to enhance the precision of Mach-Zehnder interferometer (MZI), e.g., having a non-classical input state is one of the ways to enhance the precision of the phase estimation performed by MZI. The phase estimation performed by MZI is investigated here by considering that the input states of MZI are different combinations of photon added and subtracted squeezed coherent states (PASCS and PSSCS). Using quantum Fisher information, it is shown that the use of PASCS in both the input modes of MZI, provides the most precise estimate of the unknown phase. This system is also analyzed in two different measurement scenarios -- single intensity detection (SID) and intensity difference detection (IDD). Systematic analysis has established that the intensity measurement might not be an optimal measurement scheme for phase estimation in MZI as phase and intensity correspond to non-commuting observables. The impact of the photon loss on the MZI-based phase estimation setup is also studied and it is found that PASCS is robust against photon loss, when the loss in MZI is low.

Noisy-metrology theory characterizes decoherence by its two-point correlation function, equivalently the single-atom coherence time or noise spectrum. We show this is insufficient for entangled probes: two collective baths with identical single-atom $T_2$ but different higher-order statistics yield opposite entanglement-enhanced scaling. Under Gaussian Markovian collective dephasing a Greenberger--Horne--Zeilinger (GHZ) probe reaches an atom-number-independent sensitivity floor. For a fully Markovian compound-Poisson bath, in which collective dephasing is generated by a finite-rate sequence of unitary phase kicks, a Dicke coherence of order $q$ (a difference of $J_z$ eigenvalues) decays at $Γ_q=Γ[1-\mathrm{Re}\,\varphi(q)]$, with $\varphi$ the kick characteristic function; for any absolutely continuous kick law this rate saturates at large $q$ instead of growing as $q^2$, and a GHZ probe recovers Heisenberg scaling $δω\propto1/N$ over the window in which collective finite-rate noise dominates residual independent decoherence. We prove that the Gaussian floor is the exact worst case: at fixed single-atom coherence time every finite-rate kick statistics strictly beats it, and for arbitrary Lévy phase noise the asymptotic entangled-probe sensitivity is set exclusively by the diffusive component. A converse bound shows that no input state, ancilla, or measurement improves on the GHZ scaling. The mechanism is purely exponential and CP-divisible, distinct from the Zeno, non-Markovian, nonlinear-generator, and error-correction routes. A dissipative analogue caps the Dicke superradiant burst. The full counting statistics of common noise thus emerge as a control axis for noisy quantum metrology, beyond the spectrum.

We propose two-stage optimized strategies for full-period optical phase estimation with single-mode Gaussian states and Gaussian measurements under a fixed energy constraint. In the first stage (Stage I), displaced squeezed probes and heterodyne measurements provide coarse localization of the phase to a window on the circle. In the second stage (Stage II), squeezed-vacuum probes with adaptive homodyne measurements perform efficient phase estimation inside the selected window. We derive a generalized Cramér-Rao bound for this family of two-stage Gaussian strategies, which contains the contribution from local parameter estimation in Stage II plus an overshoot penalty from coarse localization errors in Stage I. For E <= 25 photons and squeezing limited to 12 dB, protocols using displaced squeezed states in Stage I reduce the optimized two-stage bound relative to protocols using coherent states in Stage I, and remain within a factor of 3 to 30 of the idealized local squeezed-vacuum quantum Cramér-Rao bound.

Primordial black holes (PBHs) form when large primordial curvature perturbations re-enter the Hubble radius and exceed the classical collapse threshold. These perturbations originate as quantum fluctuations of the inflationary vacuum, motivating a quantum-information description of the PBH-producing scalar sector. We develop a conservative extension of the standard PBH and scalar-induced gravitational-wave (SIGW) framework in which Gaussian quantum discord is used as a diagnostic of residual quantum correlations, not as a new PBH-formation criterion. We describe each \((\bm k,-\bm k)\) pair as a two-mode Gaussian state and show that, in the pure squeezed limit, discord grows rapidly with the squeezing parameter, so low discord thresholds are automatically satisfied for strongly squeezed modes. The nontrivial regime is the mixed state produced by decoherence. Using a Lindblad open-system description, we motivate a Gaussian loss channel for the scalar covariance matrix and distinguish the discord from a covariance-survival factor \(Q_{\rm dec}(k)\). If the decoherence channel suppresses the scalar two-point covariance, PBH abundance can be affected through the classical collapse variance, while the SIGW spectrum is modified more directly by the factors \(Q_{\rm dec}(ku)Q_{\rm dec}(kv)\) inside the radiation-era convolution. For a narrow scalar peak and slowly varying \(Q_{\rm dec}\), this gives the benchmark scaling \(Ω_{\rm GW}^{\rm eff}\simeq Q_{\rm dec}^2Ω_{\rm GW}^{\rm class}\). Thus quantum discord and decoherence provide a controlled way to characterize the quantum-to-classical transition of PBH-producing perturbations, with the clearest imprint appearing in scalar-induced gravitational waves.

We use the (1+1)-dimensional Schwinger model to investigate the nonequilibrium dynamics of a finite lattice system under a constant external electric field. The lattice Hamiltonian is constructed under open boundary conditions. The vacuum state is prepared using the variational quantum eigensolver (VQE). Scans over the external field strength show the flip of the vacuum state at several field strengths. The critical field strengths agree with theoretical predictions. We further investigate the real-time evolution of the zero-field vacuum under an external electric field using a second-order Trotter-Suzuki decomposition. By comparison with exact diagonalization (ED), we verify that the quantum-simulation protocol reproduces the main features of field-induced boundary charge separation, decay of the vacuum-state fidelity, and quasiperiodic energy redistribution between the electric-field energy term and the fermionic sector. Our results indicate that combining VQE-based state preparation with digital real-time evolution provides a useful approach for studying nonequilibrium dynamics in strong-field lattice gauge theories.

In this work, we investigate the gravitational lensing properties of a static Ellis-Bronnikov wormhole embedded in a curved Friedmann-Lemaître-Robertson-Walker (FLRW) universe. By employing curvature-dependent cosmological distances, we derive the corresponding weak-field lens equation and demonstrate that the wormhole Einstein ring radius follows a characteristic cubic scaling with cosmological distances, in sharp contrast to the square-root behavior found for Schwarzschild black holes. This distinct scaling leads to a qualitatively different redshift evolution of the lensing signal, providing a model-independent geometric diagnostic to discriminate between wormhole and black hole lensing scenarios. Numerical analysis reveals that the interplay between the local wormhole geometry and the FLRW background produces an asymmetric response to spatial curvature that inverts at intermediate redshifts, exhibiting a non-negligible sensitivity even under tight modern constraints such as those from DESI 2024. We also find that Ellis-Bronnikov wormholes are substantially less efficient gravitational lenses than Schwarzschild black holes of comparable physical scale, implying that microarcsecond-scale Einstein rings require macroscopic throat radii. These results suggest that, should a population of cosmological wormholes exist, their lensing signatures could provide a sensitive, complementary probe of both exotic spacetime topology and the global geometry of the Universe.

Quantum error mitigation (QEM) is usually benchmarked by expectation-value accuracy, but many near-term workflows use those values only to make downstream choices such as argmin selection, ranking, top-k filtering, optimizer-step acceptance, or phase labeling. This creates a structural mismatch: accuracy is measured in the ambient landscape space, whereas shift-invariant decisions depend only on gaps. We develop a quotient-space theory of finite-shot QEM for downstream decisions. The minimal decision-complete object is the residual gap law; in Gaussian finite-shot regimes it is summarized by effective margins and a decision kernel. The QEM-specific point is that this kernel is not free: it is the pullback of shared physical device noise through the mitigation map. We prove quotient factorization, gap-law minimality, a marginal no-go theorem, a QEM pullback theorem, Gaussian decision-risk formulas, and a fixed-allocation shot-level converse. Finite-shot Qiskit Aer simulations demonstrate the predicted divergence in the evaluated regimes. Clifford-data regression can be decision-flat while improving mean-squared error, and probabilistic error cancellation can improve accuracy while worsening decision risk through sampling overhead. Decision-aware selection modestly reduces static held-out failure relative to accuracy-based selection, often by retaining Raw, but the dynamic success target is not reached. Pre-registered stress tests under a calibrated device-noise model and on a hardware micro-cell probe robustness beyond these regimes. The operational implication in the evaluated regimes is to select QEM methods through residual gap geometry, not from expectation-value accuracy alone.

As a hallmark of nonclassical light, squeezed light is of profound theoretical interest and holds broad practical promise for emerging quantum technologies. In this work, we investigate steady-state optical quadrature squeezing in the open quantum Rabi-Stark model by employing the quantum dressed master equation. Both numerically and analytically, we find that positive (negative) Stark coupling tends to enhance (suppress) the squeezing effect. The quadrature squeezing exhibits distinct signatures associated with both first- and second-order quantum phase transitions (QPTs). Notably, a sharp vanishing of squeezing is observed across the first-order QPT, suggesting its potential as a sensitive probe of such transitions. In the vicinity of the second-order QPT, we further demonstrate that the squeezing factor displays finite-size scaling behavior, indicating a promising route toward the realization of near-perfect squeezing. Moreover, we establish a quantitative criterion for the disruption of quantum criticality induced by thermal fluctuations, which may offer valuable guidance for future experiments. These findings contribute to a deep understanding of nonclassical light in light-matter interacting systems and provide useful insights for the design of strong optical squeezing states.

We develop a data-driven framework based on Neural ODEs that learns the bulk metric functions and the $U(1)$ gauge potential of a static, planar-symmetric black hole from boundary fermionic spectral functions. After validating the framework on the Einstein--Maxwell and Gubser--Rocha models with high accuracy, we apply it to the strange metal phase of the cuprate $\mathrm{(Pb,Bi)_2Sr_{2-x}La_xCuO_{6+δ}}$ within a semi-holographic setting, taking as input the spectral data generated from the extended power-law liquid (PLL) model calibrated by angle-resolved photoemission measurements. At low temperatures and near-optimal doping, we show that the normalized extended PLL model can be well described by an effective representative of a conformal class near $\mathrm{AdS}_{2}\times\mathbb{R}^{2}$ black holes with a nearly vanishing gauge potential ($qA_t\sim10^{-4}\,\mathrm{eV}$). A key structural observation is that our probe fermion is massless and therefore insensitive to the conformal factor $Ω(z)$. This further implies that fixing the macroscopic thermodynamics such as the electronic specific heat would require independent input beyond the fermionic spectra. We also delineate the applicability boundaries of our framework: the learned effective model remains viable across doping at low temperatures, with the loss increasing mildly as the samples move away from the marginal Fermi liquid point toward the overdoped side; the principal limitation is temperature, where both the loss and $qA_t$ grow substantially.

We propose a two-way entanglement-based quantum secure direct communication (QSDC) protocol in which information is encoded as the value of a continuously tunable parameter and decoded from empirically estimated joint-outcome histograms over windows of detected pairs. The protocol uses a fixed two-qubit polarization Hilbert space and only standard coincidence-based polarization measurement at the receiver, with the symbol alphabet enlarged not by increasing the carrier Hilbert-space dimension but by exploiting temporal integration over a window of pairs to resolve distinguishable joint-outcome probability distributions. Under an idealized model in which the source is maximally entangled, the encoding operation is strictly local, and the channel imprints no parameter-dependent signature on the traveling photon, the reduced state of the traveling photon is maximally mixed and independent of the encoded parameter. Consequently, passive interception of the in-flight subsystem alone yields no information about the encoded message. We further develop a quantitative trace-distance framework for bounding any parameter-dependent leakage that may arise from non-ideal channel and device effects in practical implementations. The protocol introduces protocol-level design parameters not available in fixed-alphabet two-way QSDC, specifically a runtime-tunable alphabet size and native support for continuous-time analog modulation, which may be advantageous in operating environments where channel conditions vary on protocol-relevant timescales.

We propose a quantum algorithmic protocol for calculating astrophysical opacities. Our implementation uses first- and second-quantized representations of interacting electronic and photonic subsystems and Hamiltonian simulation via the interaction picture. Inferring opacity from momentum-resolved measurements of the photonic register yields a direct relationship between qubit count and spectral range/resolution. Logical resource estimates for the classically challenging problem of solar iron opacity are comparable to another high-energy-density physics problem (Rubin et al., PNAS 121(3) (2024)).

A solvable charger--battery model is introduced in which quantum coherence controls both where a quantum battery's charge is stored and how robustly it survives noise. Charging converts the charger's coherence into charger--battery entanglement and splits the deposited work between a locally extractable part and a correlation-locked part accessible only through joint operations; for a qubit, the split obeys an exact complementarity, and at maximal coherence, the battery is locally passive with the entire charge locked in correlations. Robustness follows local accessibility: the stored energy and locally extractable work are population-based, immune to pure dephasing, and limited only by relaxation, with an energy half-life, whereas the correlation-locked work is fragile to both dephasing and relaxation. Dephasing, global and local depolarization, and amplitude damping are treated through a single gain--loss competition algebra, and the resulting storage lifetimes are made concrete with superconducting-transmon parameters.

Efficient quantum error correction and fault-tolerant quantum computing require scalable, high-fidelity long-range connectivity. In neutral-atom quantum computers, this is commonly achieved through atom transport, but shuttling introduces latency and motional heating that worsen with system size. Here, we introduce a neutral-atom architecture based on static atomic buses, in which auxiliary mediator atoms enable long-range entangling operations without qubit transport. The architecture naturally supports long-range stabilizer measurements in high-rate LDPC codes and transversal logical gates between neighboring surface-code patches, enabling a modular framework for efficient logical memories, Clifford computation, and magic-state distillation. To realize these capabilities, we co-design optimal-control protocols for bus-mediated controlled-Z gates that incorporate both microscopic neutral-atom dynamics and architectural constraints. We obtain smooth bus-mediated gates with fidelities approaching 99.9% and durations of a few hundred nanoseconds by combining time-optimal control with interaction-flatness and robustness constraints. Large-scale simulations of quantum error correction and logical entangling operations between neighboring surface-code patches predict more than an order-of-magnitude improvement in logical error rates compared with atom-shuttling architectures under realistic noise. The architecture achieves logical gate times of approximately 100 us and quantum-error-correction cycle times of about 1 ms for code distances d<12. These results establish static atomic buses as a practical alternative to atom shuttling for scalable fault-tolerant neutral-atom quantum computing.

The measurement cost for estimating expectation values of Hamiltonians is a central bottleneck in variational quantum algorithms. Grouping strategies significantly reduce this cost, with overlapping techniques being the state of the art in the field. Overlapping grouping methods require i) a non-overlapping grouping of the Hamiltonian, typically obtained from the Sorted Insertion (SI) algorithm as initialization, and ii) the construction of covariance dictionaries from approximate wavefunctions to guide the optimization. It was recently shown that different initializations can potentially reduce measurement costs for overlapping methods. Motivated by these findings, we introduce variance-aware SI (VarSI), a family of covariance-informed non-overlapping Pauli grouping heuristics to reduce measurement counts. VarSI grouping leverages the covariance dictionaries, already required by overlapping methods, to construct better non-overlapping groups. We propose three variants: a global greedy grouping insertion rule, a variance-informed SI analog, and a local refinement step initialized from SI or our variance-informed variant. We showcase the use of groupings generated by our VarSI heuristic algorithms to initialize overlapping methods using the iterative coefficient-splitting (ICS) algorithm. Molecular benchmarks with 130 Hamiltonians demonstrate consistent, non-overlapping measurement improvements over SI of 38\% and enhanced downstream ICS results when initialized from VarSI groups. We find that the initializations considered here achieve up to 70\% measurement reductions for ICS, compared to the standard SI initialization with mean reductions of 9--15.3\% depending on qubit mappings and covariance dictionaries used. These results show that non-overlapping grouping remains a consequential design step even when the final estimator uses overlapping fragments.

In this paper, an interactive Holographic Dark Energy (HDE) model is studied in the framework of modified gravity \(f(Q)\). By adopting a power parameterization for the Hubble parameter, the field equations are reconstructed and the evolution of the universe at the background level and tensor perturbations are investigated. Then, using observational data \(H(z)\), the model parameters are constrained and the dynamical behavior of dark energy throughout the history of the universe is analyzed. Also, the study of the evolution of energy density, pressure and the Equation of State (EoS) parameter of dark energy shows that dark energy in the late universe naturally tends to a region close to the cosmological constant behavior, while in the past it followed a distinct dynamical evolution. Stability analysis based on the speed of sound also indicates that the model has good classical stability around the present era. In addition, the compatibility of the current values of the relative density parameters of matter and dark energy with the observational constraints confirms the ability of the model to reproduce the main features of the observed universe. Next, the propagation of gravitational waves in the cosmological context of the model is investigated. The results show that the corrections due to \(f(Q)\) gravity and the interaction between matter and dark energy can affect the evolution of tensor perturbations and produce signatures distinct from the standard scenario. Overall, the findings of this study indicate that the interactive HDE in the gravitational framework \(f(Q)\) can provide a consistent framework for describing the cosmic acceleration and studying the cosmological consequences of gravitational waves.

Quantum entanglement is a powerful resource for quantum-enhanced technologies. However, its reliable quantification remains challenging due to the exponential scaling of the Hilbert space with system size, which renders full state tomography infeasible. Moreover, experimentally estimating entanglement typically requires a large number of measurement samples leading to a significant overhead. In this work, we present two models, a feed-forward neural network and an attention-based model, to accurately predict the bipartite second Renyi from projective measurements of quantum states. We benchmark their performance against standard classical shadow estimators and find that the machine-learning approaches achieve higher accuracy and improved sample efficiency across a range of system sizes. Our results demonstrate the potential of machine learning for scalable and efficient estimation of quantum correlations.

Consistent families of locally defined probability distributions that do not admit a joint global distribution are known as contextual, with primary examples arising in quantum theory. In this paper, we study such families of distributions using the theory of simplicial distributions, and further develop the theory for possibilistic distributions defined over the Boolean semiring. We characterize possibilistic collapses of simplicial distributions geometrically using bundle scenarios. Using this characterization together with a new connectivity condition on the total space of a bundle scenario, we provide a criterion for detecting extremal simplicial distributions. In parallel, we develop an analogous theory for presheaves on simplicial complexes, describe possibilistic collapses of empirical models on them using event scenarios together with a categorical extremality condition, and relate the two frameworks via a comparison result. We provide examples of contextual simplicial distributions that arise from our criteria on scenarios of interest in quantum foundations, such as Bell scenarios and boundaries of standard simplices, the latter connecting to Vorob'ev's classical theorem on acyclic complexes.

Efficient representations of multipartite quantum states play a fundamental role in quantum information theory, providing both conceptual insight and practical tools for characterizing entanglement. Motivated by the axiomatic framework for graph states [Phys. Rev. A 85, 062313 (2012)] and its subsequent extension to hypergraph states [Phys. Rev. A 87, 022311 (2013)], we introduce an axiomatic construction of \emph{matroid states}, a new family of multipartite quantum states associated with matroids. Our constructions are based on a set of axioms analogous to those that define graph and hypergraph states, yielding a consistent quantum representation of arbitrary matroids. Two ways of constructing matroid states are proposed: the first is defined in terms of circuits, and the second in terms of independent sets. In both approaches, we establish the existence of universal global operators that satisfy desirable properties such as locality, symmetry, commutativity, and are associated with the combinatorial structure of matroids. Furthermore, we establish a hierarchy connecting graph, matroid, and hypergraph states within a unified framework. Additionally, we show how to obtain an arbitrary graph state by applying suitable families of matroid states, whose corresponding operators are the generators of the stabilizer subgroup of the graph state. These results identify matroid theory as a natural combinatorial language for the efficient description of multipartite quantum states and open new perspectives for the investigation of quantum entanglement and related combinatorial structures.

Quantum regression tasks for predicting properties of quantum states are commonly addressed using variational quantum algorithms. While variational quantum circuits are highly expressive and allow to achieve reasonable accuracy, training these circuits may demand a considerable amount of time and resources. In this work, we propose an approach of constructing problem-specific quantum regression models with encoding relevant symmetries and regularizing the variance. The proposed method is based on finding the coefficients of the linear combination of suitably chosen observables. Although it requires the knowledge of the symmetries of the problem in question, the method does not involve parameterized quantum circuits, and the training is done efficiently once the observables are measured. We demonstrate this method on two examples: Prediction of the transverse field strength in the Ising model, and quantification of entanglement in bipartite qubit systems. Our approach is accurate and less resource-intensive than conventional variational methods.

The elementary excitation spectrum of a many-body quantum system encodes many key properties, including phenomena as diverse as transport, thermalisation and ground state structure. Excitation spectra of strongly correlated systems are typically encoded in dynamical structure factors, which are demanding to measure experimentally and challenging to compute classically. Here we use quench spectroscopy on a superconducting quantum processor to extract excitation spectra of spin chains of $L=101$ spins. By tailoring the combination of quench protocol and observable, we selectively access distinct excitation sectors across several phases of the spin-$1/2$ XXZ chain, resolving free magnons, multi-magnon bound states, and two-spinon continua. Notably, we demonstrate that the protocol does not rely on ground state preparation: in the classically challenging gapless regime, we extract spectra directly from the quench dynamics of easily prepared product states, a procedure that is natural and straightforward on quantum hardware. Our work establishes quench spectroscopy as a fast and flexible probe of many-body excitation spectra on digital quantum hardware, introduces a novel quench protocol that does not require costly state preparation routines, and provides a scalable route towards regimes where classical simulation may become intractable.

A central challenge in quantum thermodynamics is to access work fluctuations in coherent processes without distorting the energetics of the unmeasured evolution. In standard two-point schemes, the initial energy measurement dephases coherent inputs, causing the measured average work to differ from that of the unmeasured evolution. Here, we develop an operational scheme for accessing work statistics for closed quantum systems based on the abstract notion of variation in the Heisenberg picture Hamiltonian. This scheme preserves energetically relevant coherences, thereby faithfully reproducing unmeasured work, while still producing positive probabilities. We derive modified Jarzynski and Crooks relations, as well as a thermodynamic uncertainty relation, identifying coherence-induced correction terms. Furthermore, we show that this scheme can reliably quantify the performance of a coherent engine in situations where the two-point energy measurement would suppress work output. In addition, the scheme requires only a single measurement and can predict the work associated with a subsequent unitary transformation. We exploit this feature to construct a Maxwell-demon protocol that can outperform energy-based feedback engines for coherent work extraction. Our results establish this scheme as a framework for accessing coherent work fluctuations without erasing the coherence that drives quantum thermodynamic performance.

We explore the Kazakov-Solodukhin quantum-deformed black hole spacetime, characterized by a single deformation parameter \( η\) that encodes quantum corrections to the classical Schwarzschild solution. The model preserves the correct general-relativistic limit as \( η\to 0 \), while introducing significant and physically meaningful deviations in the strong-field regime. A central and remarkable feature of the geometry is the regularization of the classical singularity: curvature invariants remain finite near the minimal radius \( r = η\), effectively replacing the divergent core with a smooth and well-behaved region. This behavior naturally introduces a minimal length scale into the spacetime structure, offering a geometrically motivated resolution of the singularity problem. The deformation modifies the horizon structure, shifts the event horizon location, and alters the mass-radius relation. It also reduces the surface gravity, leading to a lower Hawking temperature and a slower evaporation process, thereby enhancing the thermodynamic stability of the black hole. Photon dynamics are correspondingly affected, resulting in a displaced photon sphere and modified strong-lensing characteristics. While the shadow remains perfectly circular due to spherical symmetry, its size depends sensitively on \( η\). Observational constraints can be expressed through \( \left| R_{sh}(η) - R_{obs} \right| \leq ΔR_{obs}, \) which places an upper bound on the deformation parameter. In the weak-field limit, the deflection angle acquires a quadratic correction proportional to \( η^2 \), ensuring consistency with precision tests while allowing potentially detectable deviations in strong-gravity observations. These features make the model both theoretically appealing and observationally testable.

Variational quantum algorithms (VQAs) are a leading approach to near-term quantum computation, but their utility is limited by barren plateaus and other pathologies in their loss landscapes. Existing landscape theories based on dynamical Lie algebras, Jordan-algebraic Wishart systems, approximate t-designs, and Haar-random circuits are foundational, but they often neglect the finite-depth geometry of realistic ansätze and are therefore poorly suited to the shallow-depth regime, where VQAs are poor approximators of 2-designs and trainability is most feasible. This thesis introduces Krylov algebras, algebraic structures induced by the Krylov span of a finite generator set acting on one or more seed vectors, as a framework for VQA landscape theory. We show that VQA reachable manifolds can be approximated in a numerically robust, geometrically faithful way by Krylov-Lie algebras and groups, and that these structures induce canonical invariant measures for computing expectation values and variances under general sampling measures. In particular, we derive weighted non-Haar variance formulas that recover the usual Lie-algebraic Haar formulas as a special case while isolating non-Haar effects into explicit correction terms. We also show that the common heuristic that sufficiently deep circuit ensembles must converge to Haar fails in general without additional hypotheses, identify concrete obstructions to naive Haar convergence, and recover convergence under natural necessary and sufficient ergodic conditions. Lastly, our formulas further imply that non-Haar contributions may mitigate barren plateaus by reweighting the visible sectors of the loss landscape, suggesting that VQAs may be more trainable than recent literature has posited.

The energy dissipation and particle acceleration mechanisms within relativistic jets remain fundamental questions in active galactic nuclei (AGN) research. In this paper, we present a comprehensive 13-year (2012-2025) multi-wavelength study of the broad-line radio galaxy 3C 120, utilizing Fermi-LAT ($γ$-ray), ASAS-SN (optical), and high-resolution VLBA (15 GHz and 43 GHz) monitoring. Cross-correlation analyses reveal that $γ$-ray flares lead radio emission by $11.08_{-1.88}^{+4.03}$ months at 15 GHz and $8.27_{-5.55}^{+3.45}$ months at 43 GHz. This frequency-dependent temporal hierarchy positions the high-energy dissipation zone upstream of the radio core, corroborating the opacity-driven core-shift effect. By tracking the parsec-scale jet morphology during major $γ$-ray flaring epochs, we demonstrate that radio outbursts systematically coincide with compact core brightening, pronounced surges in polarized flux, abrupt electric vector position angle rotations, and the subsequent ejection of superluminal knots. Ultimately, our internal radio correlations suggest that jet dynamics are governed by a dual mechanism: long-term kinematic and flux baseline variations are geometrically modulated by a secular jet precession, while rapid, highly energetic polarimetric bursts are driven by short-lived internal shocks propagating down the jet channel.

GW241011 and GW241110 are asymmetric binary black hole mergers with rapidly spinning primaries, unequal component masses, and nonzero spin--orbit tilts, making them natural candidates for hierarchical mergers. We use a Bayesian framework to compare a fiducial first-generation (1G+1G) binary black hole population with second-generation plus first-generation (2G+1G) hierarchical merger models in star clusters and active galactic nucleus (AGN) disks. Both events favor the 2G+1G interpretation over the 1G+1G hypothesis, with $\ln\mathcal{B}^{\rm 2G+1G}_{\rm 1G+1G}\simeq6.5$--$8.6$ for GW241011 and $\ln\mathcal{B}^{\rm 2G+1G}_{\rm 1G+1G}\simeq3.0$--$4.5$ for GW241110, depending on the waveform model and assumed environment. The AGN disk models yields slightly larger evidence than the star cluster models, mainly due to their spin tilt distribution, but the data do not provide a decisive environmental classification. We further consider a third-generation plus first-generation (3G+1G) interpretation, but it is not robustly preferred over 2G+1G scenarios. Finally, we also search for optical counterparts by examining AGNs within the three-dimensional localization volumes using ZTF and ATLAS forced photometry, and find one candidate source with weak flare, which might be associated with GW241110 event.

We propose close binary polarimetry as a probe of birefringence induced by ultralight axion dark matter. In a close binary, reflection or scattering can generate a small linear polarization whose time dependence is locked to the orbital phase. This phase-locked polarization provides a template against which an oscillatory rotation of the polarization angle induced by the axion can be searched for. We show that axion birefringence appears as sidebands around the orbital harmonics. For a single bright binary, with parameters motivated by observed systems and current high-precision optical polarimetry, we estimate the sensitivity to the axion-photon coupling under white noise assumption to be the level of $10^{-12}$ GeV$^{-1}$ at an axion mass of $10^{-20}$ eV. A future array of suitable binaries could further improve the sensitivity to $10^{-13}$ GeV$^{-1}$ in an optimistic scenario. This method could provide a complementary high-cadence optical probe of axion birefringence, compared to existing astrophysical searches.

We investigate non-minimal Higgs inflation in a no-scale-inspired supergravity framework and confront its predictions with the latest CMB constraints from ACT DR6 and \emph{Planck}. Working within a shifted hybrid inflation scenario, we construct an effective single-field description in which the GUT Higgs direction serves as the inflaton after stabilization of the orthogonal scalar fields. We show that the inclusion of the leading nonrenormalizable operator in the superpotential induces a controlled deformation of the Starobinsky attractor, allowing the scalar spectral index to be shifted into the range favored by recent ACT and related CMB datasets while maintaining a small tensor-to-scalar ratio, $r \sim 10^{-3}-10^{-2}$. The resulting inflationary dynamics remain theoretically consistent, with controlled supergravity corrections and sub-Planckian inflaton field values. We perform a detailed numerical analysis of the model parameter space, including reheating and nonthermal leptogenesis, and identify regions that simultaneously satisfy current observational constraints and yield a viable post-inflationary cosmological history.

Using fully kinetic simulations that capture unprecedentedly large (from electron to ion) scales, we study magnetogenesis driven by continuous large-scale forcing until nonlinear dynamo saturation. We uncover a two-stage mechanism in collisionless ion-electron plasmas whose dynamics diverge dramatically from the pair-plasma case. In the first phase, electron pressure anisotropy triggers electron-Weibel modes, seeding small-scale magnetic fields. Then, a second growth phase emerges when the more massive ions develop their own strong anisotropy and drive ion-Weibel-type modes; concurrently, a Biermann-battery mechanism contributes to amplifying the magnetic field. This combined dynamics provides a tenfold amplification of the magnetic field in comparison to the pair-plasma case. Over long times, dynamo action continues until the system reaches a statistical steady state. This self-consistent kinetic mechanism provides a plausible explanation for robust magnetogenesis wherever an external forcing continuously stirs the plasma.

We perform the first $k$-cut cosmic shear analysis of the KiDS-Legacy survey. This method uses the Bernardeau-Nishimichi-Taruya (BNT) transform to construct weak-lensing kernels that are more localised than conventional ones, and remove information from selected physical scales while retaining the constraining power of the targeted range. Removing the scale of $k \geq 0.33~\mathrm{Mpc}^{-1}$ from the KiDS-Legacy pseudo-$C_\ell$ data vector, and using a covariance matrix whose Gaussian component is computed from the theoretical data vector, we find $S_8 = 0.798 \pm 0.045$. This agrees with both the fiducial KiDS-Legacy bandpower result and our no-$k$-cut pseudo-$C_\ell$ posterior to within $0.1σ$, indicating no significant bias from nonlinear astrophysical feedback at the precision of KiDS-Legacy. We also study the case in which the Gaussian covariance is computed from the observed data vector. In this setup, the same scale cut of $k < 0.33~\mathrm{Mpc}^{-1}$ gives a much lower $S_8=0.717_{-0.046}^{+0.047}$. Further $k$-cut tests reveal a mild scale-dependent trend, with larger physical scales preferring lower $S_8$ values and a maximum low- versus high-$k$ deviation of $1.80σ$. Mock tests show that this behaviour is not produced by the covariance prescription or data vector alone, but may arise from their interplay. These results show that BNT $k$-cuts provide both a mitigation strategy for nonlinear systematics and a diagnostic of weak-lensing inference pipelines.

We investigate the dynamics of charged particles interacting with ultra-intense electromagnetic X-modes in strongly magnetized plasmas. We demonstrate that particle motion becomes chaotic for relative wave intensities $δ= B_w/B_0 \gtrsim 0.25$ (not above the field reversal threshold $δ\geq 1$). The transition to chaos occurs via the Chirikov resonance overlap mechanism and the related destruction of Kolmogorov-Arnold-Moser (KAM) tori. The maximum Lyapunov exponent increases logarithmically with $δ$, even though the unmagnetized $δ\to \infty$ limit is strictly integrable. In the $δ\gg 1$ regime, incomplete re-laminarization of the phase space flow leads to two distinct populations: (i) the majority of particles undergoing stochastic diffusion, and (ii) a fraction of particles that become phase-locked with the wave, experiencing macroscopic intermittent surfing (Lévy flights). The 1D Particle-In-Cell simulations using the EPOCH code in the highly magnetized ($σ\gg 1$) and under-dense regime are generally consistent with the Hamiltonian single-particle theory. The dissipation fraction of the initial EM energy remains mild.

Changing-look active galactic nuclei (CL-AGNs) provide a unique opportunity to probe the coupling between accretion flows and relativistic jets in supermassive black holes. We investigate the long-term radio behavior of CL-AGNs over approximately 20 years by combining FIRST and VLASS observations with quasi-simultaneous optical spectroscopy and photometry. From a parent sample of 1092 CL-AGNs, we identify 58 sources with radio detections. Radio-detected CL-AGNs exhibit systematically higher radio kinetic efficiency, quantified by \(P_{\rm j}/L_{\rm bol}\), than both typical radio-detected AGNs and radio transients, consistent with their preference for low Eddington ratios. At the population level, the expected anti-correlation between radio emission and accretion rate is weak. However, a clear source-by-source anti-correlation emerges in a small subset of CL-AGNs with continuous multi-epoch coverage. We further identify four radio transients, including both radio turn-on and turn-off events, and one source exhibiting a multiwavelength flare that may be indicative of tidal disruption event-like activity. These results suggest that radio activity in CL-AGNs is not governed by instantaneous accretion state changes but is instead regulated by long-term accretion history and jet evolution, with additional stochastic or transient channels contributing in rare cases.

Dust extinction laws are essential for precision photometry and provide a direct probe of grain properties, but their behaviour in dense molecular clouds remains poorly constrained at high extinction. Using Euclid Quick Data Release 1 (Q1) imaging of the Orion A dark cloud Lynds Dark Nebula 1641 (LDN 1641), we measured the extinction law from the broad Visible Instrument (VIS) band and the Near-Infrared Spectrometer and Photometer (NISP) $Y$, $J$, and $H$ bands along sightlines reaching $A_V\sim 30$ mag towards the cloud core. We derived colour-excess ratios $E(λ-H)/E(Y-H)$ from linear fits to colour--colour diagrams of $(λ-H)$ versus $(Y-H)$ and converted them into relative extinctions, $A_λ/A_H$. The near-infrared extinction in LDN 1641 is well described by a power law, $A_λ\propto λ^{-α}$, with $α=1.57 \pm 0.06$, corresponding to $A_{\rm VIS}/A_H=4.23 \pm 0.24$, $A_Y/A_H=2.13 \pm 0.18$, and $A_J/A_H=1.47 \pm 0.11$. We further find significant spatial variations: $α$ changes by up to $27\%$, with systematically smaller values and therefore flatter extinction curves in higher-extinction regions. This flattening is consistent with an enhanced large-grain population and supports substantial grain growth from the diffuse outskirts to the dense core of a single molecular cloud.

Magnetic fields are thought to regulate the angular momentum transfer in active galactic nuclei (AGNs), yet their strength and structure in circumnuclear regions remain poorly constrained across spatial scales and gas phases. We present a unified observational framework for probing circumnuclear magnetic fields using complementary diagnostics: direct measurements via the Zeeman effect in HI absorption and megamaser emission, and indirect constraints from broadband Faraday rotation of polarized continuum radiation. These approaches provide access to magnetized gas spanning spatial scales from ~100 parsec (pc) circumnuclear disks down to sub-pc regions near supermassive black holes (SMBHs). The Square Kilometre Array (SKA) telescopes are expected to revolutionize such investigations through their outstanding sensitivities, wide frequency ranges and high spatial resolutions achievable via very long baseline interferometry (VLBI). These capabilities will enable the detection and detailed characterization of weakly polarized emission from magnetized circumnuclear matter, which has remained largely inaccessible with current instruments. In this chapter, we review previous measurements of magnetic fields in galactic nuclei and discuss the breakthroughs that SKA observations are expected to provide in elucidating the physical conditions and processes shaping AGN environments.

A population of interstellar objects (ISOs) exist that originate from the Solar System, rather than from other stars. Such a foreground could challenge straightforward analysis of the ISO sample expected to be gathered by upcoming sky surveys. We assess whether small bodies unbound from the Solar System can experience dynamical evolution in the Galactic potential that places them on re-encounter trajectories. We find that these 'quasi-interstellar objects' (quasi-ISOs) primarily depart the Solar System through erosion of the outer Oort cloud in the past few hundred Myr, excluding the most recent ~10 Myr. After orbiting in the Milky Way potential nearby the Sun but beyond the tidal radius, those ejected on certain orbits can re-encounter the Solar System. Meanwhile, the larger population of ISOs produced by the Solar System early in its life will be too spread-out in the Galaxy to contribute significantly to the observed sample. We predict that quasi-ISOs will be intrinsically rare and have $v_\infty$ values of order 0.1 km s$^{-1}$, easily distinguishable from ISOs from other stars, meaning that the observed ISO sample will be truly Galactic. The detection of a quasi-ISO would imply larger-than-expected losses from the Oort cloud, or a particularly catastrophic erosion event 10-300 Myr ago that would not be detectable any other way.

In this work, we investigate eruptive phenomena associated with X-ray flares on AB Doradus using observations from XMM-Newton. Our aim is to detect such events with the help of continuous X-ray absorption. The variation in the hydrogen column density (NH) is used as a primary diagnostic to detect such absorption or spectral hardening in the soft X-ray band below 1 - 2 keV. These variations may represent the signatures of the eruptive processes associated with flares, such as stellar prominences, failed eruptions, or coronal mass ejections. We applied time-resolved spectroscopy technique to track changes in the absorption parameter during flaring episodes. Our analysis shows statistically significant enhancements in the NH, spanning from 0.3 - 3.4 x 10^20 cm^-2 during the flare and post-flare intervals. Out of six data sets, two sets containing multiple-overlapped flares show significant fluctuations and remaining showed no such variations in NH. These variations indicate the presence of dynamic absorbing material along the line of sight and point to ongoing physical processes within the stellar corona.

Giant Star-forming Clumps (GSFCs) are kpc-scale regions of enhanced star-formation with stellar masses of $10^7$ to $10^9\,M_\odot$ that are commonly observed in high-redshift galaxies but are rarely detected in low-redshift ($z\lesssim0.5$) galaxy analogues. However, the availability of wide-field galaxy survey data makes it possible to identify potential star-forming clumps in large samples of low-redshift galaxies using object detection models that are based on Deep Learning (DL) techniques. We apply a novel DL-based object detection model to galaxies observed by the Hyper Suprime-Cam Subaru Strategic Survey (HSC-SSP) and CFHT Large Area U-band Deep Survey (CLAUDS). Our model is based on the the Faster Region-Based Convolutional Neural Network (Faster R-CNN or FRCNN) object detection framework but expanded to process the six $ugrizy$ filter band images simultaneously and identify not only clumps and their locations in the host galaxy but also additional contaminants. By adopting the \textsc{Zoobot} foundation DL-model as a feature extraction backbone, we also demonstrate one of the first applications of \textsc{Zoobot} in a downstream task for object detection. Our model achieves a detection completeness of $\gtrsim 0.9$ and purity of $\gtrsim 0.8$ which were validated on a large set of real galaxies into which simulated clumps were injected.

Massive black hole binaries (MBHBs), and the higher-order multiples produced by repeated galaxy mergers, spend part of their lives in dynamical regimes that cosmological simulations cannot resolve, even though these regimes set their merger delays, spins, recoils, and host-galaxy context. We extend the RAMCOAL framework to follow such subgrid massive black hole triplets directly within hydrodynamical galaxy simulations. As in the original staged binary model, the black holes start as sink particles, pass through a dynamical-friction phase, and settle into bound binaries that harden through stellar scattering, gas torques, circumbinary-disc coupling, and gravitational-wave emission. When a hierarchical triplet becomes chaotic, RAMCOAL maps the encounter onto a library of three-body outcomes from direct N-body experiments and updates the surviving system, following the resulting mergers, exchanges, and ejections together with the accretion and spin evolution of each black hole. Using isolated-galaxy tests with contrasting geometries, we show that the encounter geometry alone can change which pair finally merges, and after how long. We demonstrate the first triplet MBH dynamical evolution all the way to coalescence inside a live hydrodynamical simulation. This establishes an end-to-end capability to predict triplet-driven MBH coalescences self-consistently coupled to the evolving host galaxy. Because each MBHB coalescence carries its environmental history through the subgrid phase, RAMCOAL offers a route toward merger catalogues that link the gravitational-wave signatures of coalescing black holes to the galaxies in which they form.

Fluctuations in the cosmic electron density are imprinted on the dispersion measures (DMs) of fast radio bursts (FRBs), making DMs a promising probe of cosmology and the spatial distribution of ionized baryons. In this work, we present the first measurement of the angular auto-power spectrum of FRB DMs, using 3455 apparently non-repeating bursts from the CHIME/FRB Catalog 2. We detect an angular correlation signal at $>3σ$ significance, associated with large-scale electron-density fluctuations. By fitting the measured spectrum to theoretical models, we constrain two key parameter combinations: $Ω_{\rm b}h^2$-$H_0$, which probes the cosmic baryon density and expansion rate, and $Ω_{\rm b}h^2$-$f_{\rm d}$, which traces the baryon fraction in cosmic large-scale structure (LSS). We further assess the robustness of the power-spectrum method against systematic uncertainties arising from the assumed FRB redshift distribution and from the DM contributions of host galaxies (${\rm DM}_{\rm host}$), the Galactic halo (${\rm DM}^{\rm MW}_{\rm halo}$), and the Milky Way interstellar medium (${\rm DM}^{\rm MW}_{\rm ISM}$), using mock samples. Our results demonstrate that the angular power spectrum is largely insensitive to uncorrelated DM components such as ${\rm DM}_{\rm host}$, thereby effectively mitigating the impact of poorly constrained host-galaxy systematics. In contrast to the traditional ${\rm DM}_{\rm LSS}$-$z$ relation, this method does not require individual redshift measurements--it relies only on the overall redshift distribution--and it partially breaks the parameter degeneracies in the $Ω_{\rm b}h^2$-$H_0$ and $Ω_{\rm b}h^2$-$f_{\rm d}$ planes. These findings establish the DM angular power spectrum as a robust cosmological probe and a powerful baryon tracer.

We propose that jointly modeling the thermal X-ray pulse profiles and the polarization position angles offers an effective means of locating the polarization orientation focus of the pulsar's coherent radiation. From the X-ray pulse-profile measurement we constrain the colatitude of the center of the thermal X-ray emission, which corresponds to the center of in-falling particles within the polar cap, while the RVM fitting yields the inclination angle of the focus point of polarization orientations. Thus, consistency between these two independent angle measurements would imply that the RVM fit faithfully recovers the inclination of the plasma flow center, and this center coincides with the polarization orientation focus. Conversely, the discrepancy would suggest that the polarization state of the radio emission changes as it propagates because the evolution of wave modes during wave propagation in the magnetosphere strongly depends on magnetic field orientations.

Two-halo galactic conformity is commonly interpreted as a manifestation of galaxy assembly bias, but its statistical structure and physical origin remain unclear. We aim to write the quenched-neighbour statistic in correlation-function form, test whether its scale dependence follows the linear matter correlation function $ξ_{\rm mm}^{\rm lin}(r)$, and separate the contributions of halo-mass bias and assembly bias to its amplitude. Using galaxies in IllustrisTNG300-1 at $z=0$, we measure the two-halo galactic conformity statistic of quenched neighbours at distance $r$, $Δf_Q(r)$, and related quantities in real space, compute the required correlations, perform shuffling tests at fixed halo mass, compare several $Δf$ observables, and explore the transformed family $G_n$. We show explicitly that $Δf_Q(r)$ can be written directly in terms of correlation functions and that, over $\sim 2$-$40\,h^{-1}\,\mathrm{Mpc}$, it is well described by $A_{\rm fit}\,ξ_{\rm mm}^{\rm lin}(r)$. Thus nonlinear and baryonic terms do not dominate the residual scale dependence isolated by this statistic. Halo-mass bias alone predicts lower amplitudes than measured, while fixed-mass shuffling strongly suppresses the signal; in TNG300 the amplitude is therefore dominated by %fixed-mass assembly-dependent occupancy. galaxy assembly bias at fixed halo mass. Quenching, colour, and concentration share a common rescaled shape, whereas stellar-mass and halo-mass splits do not. These results suggest that galaxy assembly bias sets the amplitude of two-halo conformity, while the double-difference structure of the statistic suppresses nonlinear residuals when the compared populations have similar halo-mass and transition-scale structure.

We present the EPS Research Astro-RAG MCP Server v2.3.0, a cross-platform Model Context Protocol (MCP) implementation providing unified, machine-readable access to five astrophysical corpora spanning the local universe (z = 0) to z = 5.68 (the high-redshift frontier of the ALPINE survey). The server exposes a humanreadable browser query interface, a REST API, and an LLM-native MCP endpoint, enabling deterministic retrieval, metadata filtering, and structured analysis across 2,064 objects: galaxies spanning HI rotation curve surveys, dwarf and irregular systems, and high-redshift kinematic targets, together with Milky Way globular clusters. Version 2.3.0 introduces FAISS-accelerated natural-language similarity search using pre-built 384-dimensional MiniLM-L6-v2 vector indexes, enabling corpus-wide semantic queries without fine-tuning or API keys. We describe the server architecture, unified schema design, FAISS index construction pipeline, MCP toolset, and cross-epoch use cases including rotation-curve retrieval, metadata filtering, and semantic similarity exploration. The server is publicly deployed on HuggingFace Spaces (https://dflynn5656-astro-rag-mcp.hf.space), released under the MIT License, and fully reproducible from Zenodo (DOI: 10.5281/zenodo.21154451). The full platform is available at https://github.com/eps-research/rag-corpus-series.

Nearby high Galactic latitude clouds provide a unique laboratory to study the physics and structure of the Galactic interstellar medium (ISM) and the properties of the interstellar dust. In this work, we select 32 sightlines toward reddened background stars at high Galactic latitudes with |b| > 20 degr, for which high-quality spectra from the International Ultraviolet Explorer are available. We utilize the "pair-method" to derive the ultraviolet interstellar extinction curves for these sightlines. We examine the extinction properties of these sightlines and find no systematic variations with the Galactic latitudes, although they do show appreciable sightline-to-sightline variations.

Based on JWST/MIRI integral field observations, we report a widespread detection of aromatic and aliphatic hydrocarbon emission at rest-frame 3.3 and 3.4 micron in SDSSJ074922.96+225511.7 (hereafter J0749+2255), a dual quasar at redshift z~2.17, corresponding to a cosmic age of ~3 billion years after the Big Bang, a time period known as the "cosmic noon" when star formation and black hole growth peak. With the 3.3 micron emission ascribed to aromatic C--H stretches of small PAH molecules and the 3.4 micron emission assigned to aliphatic C--H stretches of aliphatic sidegroups attached to PAHs, we utilize the observed intensities of the 3.3 and 3.4 micron emission bands to estimate the aliphatic fractions of PAHs and their variations across J0749+2255, which is, to our knowledge, the most distant object to date in which both aromatics and aliphatics have ever been detected. We find that both the 3.3 and 3.4 micron emission bands are pronounced and the aliphatic fractions are surprisingly high in the most luminous regions centered on the two quasar nuclei, suggesting that not only small PAHs (of ~20--30 carbon atoms) but also their attached aliphatic sidegroups survive in intense ultraviolet radiation arising from extreme starburst.

Gravitational lensing of gravitational waves can probe compact lenses, dark matter substructure, and cosmological distances, but identifying lensed events is difficult when unrelated binary mergers overlap in the same analysis window. We develop physics-informed posterior learning for ranking lensed multi-image signals against unrelated multiple-merger events. The method exploits the geometric-optics consistency that lensing can change amplitudes, arrival times, and Morse phase offsets while preserving the intrinsic phase evolution of the source. We infer a simulation-trained approximate posterior for the common detector-frame chirp mass and symmetric mass ratio, and fuse posterior samples with direct waveform features. Training uses generic multi-image simulations, while point-mass, singular-isothermal-sphere, singular-isothermal-ellipsoid, and shear-perturbed lenses are reserved for held-out lens-family evaluation. For the observationally motivated binary-black-hole population, the fusion ranking raises the detection efficiency from $20.8\%$ to $35.2\%$ at a $1\%$ reference false-positive-rate threshold calibrated on the corresponding unrelated multiple-merger sample. It lowers the network signal-to-noise ratio needed for $50\%$ detection efficiency from 45.3 to 33.5, which corresponds to a 1.35 times larger signal-to-noise-ratio-equivalent distance scale. The gain is limited by loud unrelated multiple-merger events that are partly source consistent, and by the need to calibrate the unrelated multiple-merger population. These results suggest that physical consistency can become a guiding principle for machine learning searches in dense gravitational-wave catalogs.

Fast Radio Bursts (FRBs) are important cosmological probes, but their applications depend critically on accurate dispersion measure (DM) determinations. We present a systematic comparison of six DM estimation methods using 2,874 bursts from FRB20240114A, the most active repeating FRB currently known, observed by FAST during a single 4.4-hr session on 2024 March 12. This large, homogeneous sample over a short timescale, during which the propagation environment is expected to be nearly static, provides an ideal benchmark for isolating algorithmic effects on DM determination. We investigate the dependence of inter-method consistency on signal-to-noise ratio (S/N), burst morphology, and radio frequency interference (RFI). Low-S/N bursts exhibit significantly larger inter-method deviations, while single-component bursts produce highly consistent DM values across methods. In contrast, complex double- and multiple-component bursts with drifting substructures lead to substantial inter-method scattering, indicating that DM discrepancies are primarily driven by algorithmic responses to burst morphology. RFI does not significantly alter the global statistical behavior of DM deviations, but it affects density-filtering methods through morphology distortion caused by frequency-channel masking. Even after imposing strict inter-method consistency constraints, FRB20240114A still exhibits notable apparent DM fluctuations spanning $\sim$528-534~pc~cm$^{-3}$ over 15,780s. For morphologically simple bursts these variations far exceed the measurement uncertainty and, on second-to-minute timescales, cannot arise from any plausible change in the line-of-sight electron column, pointing instead to a frequency-dependent emission-time structure intrinsic to the bursts that mimics dispersion.

We report the {\it Chandra} non-detection of PSR J0901$-$4046, a $P=75.89 $ seconds long-period radio transient (LPT). For a distance of 467 pc, the upper limit on X-ray luminosity is $L_X \leq$ few $\times 10^{28}$ erg s$^{-1}$. For the measured $P$ and $\dot{P}$, this upper limit, approximately 50 times lower than the previous {\it Swift} observations, is comparable to the spin-down luminosity of a neutron star, but would be approximately four orders of magnitude smaller than the spindown power of a white dwarf. Our results disfavor isolated WDs as the central star in PSR J0901$-$4046. We suggest that the isolated LPTs are powered by magnetic dissipation (not rotation), in a way similar to magnetars' radio emission.

The study of dark matter admixed neutron stars has the potential to advance our understanding of dark matter particle candidates. However, the large parameter space of dark matter particle masses restricts a systematic, model-independent study. In this analysis, we employ agnostic hadronic and dark matter equations of state to construct dark-matter-admixed neutron stars within a two-fluid formalism. Dark matter is characterised solely by its low-density equation of state and mass, and is modelled as a Fermi gas, while hadronic matter is anchored at low and high densities by chiral effective field theory and perturbative quantum chromodynamics calculations. A speed-of-sound parametrisation covers the intermediate density region for hadronic matter and the high-density region for dark matter, so the dark matter equation of state is constrained only by thermodynamic consistency, free from bias toward a softer or stiffer equation of state. Within this agnostic framework, we find that dark matter does not generically compactify the star: light dark matter forms extended halos that raise the tidal deformability, while heavy dark matter forms compact cores that lower it. Consequently, the dominant observational constraint shifts from gravitational-wave tidal deformability for light, halo-dominated models to NICER mass--radius data for heavy, core-dominated models. Using current data at $1σ$, we constrain the dark matter fraction to $f_{\mathrm{DM}} \lesssim 0.11$ for light dark matter. Being almost independent of any assumed dark-sector microphysics, our framework yields conservative, broadly applicable bounds on the dark-matter content of neutron stars. Neutron stars with similar masses but very different tidal deformabilities could be a smoking-gun signature of dark matter in Neutron stars.

Rotational emission from HD provides a direct probe of deuterium in warm molecular gas, but the conversion of HD line fluxes to total column densities is sensitive to non-LTE excitation. I present calculations of the non-LTE rotational excitation of HD and H$_2$ and a method for fitting their rotational diagrams simultaneously. Because HD remains subthermally excited to higher densities than H$_2$, the combined analysis constrains the thermal pressure as well as the HD/H$_2$ abundance ratio. I apply this method to JWST/MIRI observations of 26 positions in 10 protostellar outflows from the JOYS survey, published previously by Francis et al. (2025), including 17 positions with secure detections of at least three HD lines. Two-temperature, isobaric models generally provide good fits to the measured H$_2$ and HD level populations. For 11 positions with HD detections, the inferred pressures span $p/k_B=8.4\times10^7$ - $5.1\times10^9\ {\rm K\,cm^{-3}}$; for the remaining six, the HD populations approach LTE and provide only lower limits on the gas pressure. Excluding two positions toward IRAS 4B with comparatively poor fits, the median and mean inferred HD/H$_2$ abundance ratios are $1.42\times10^{-5}$ and $1.47\times10^{-5}$, respectively, only 28$\%$ and 29$\%$ of the value expected if all primordial deuterium were present as HD. These results confirm that HD/H$_2$ is generally low in shocked molecular gas while demonstrating that pressure-dependent non-LTE excitation must be modeled explicitly. Translating the measured HD/H$_2$ ratios into elemental D/H requires additional modeling of shock chemistry.

This work presents an investigation of the interstellar extinction law toward 105 Galactic open clusters using ultraviolet-optical photometry. The photometric dataset and reddening determinations were previously introduced in Ramezani et al., (2025); hereafter Paper I, where homogeneous ultraviolet photometry and colour-colour diagrams were used to derive reddening values for the cluster sample. In the present study, we extend that analysis by determining extinction law and metallicity parameters for the same clusters. Ground-based U-band observations obtained with the 2.15-m CASLEO telescope and the 1.54-m Danish telescope are combined with Gaia Data Release 3 photometry to construct ultraviolet-optical colour indices. Using extinction-ratio combinations and the Fitzpatrick, (1999) extinction model, we derive mean total-to-selective extinction law values (RV) for all clusters. The resulting RV values span approximately 2.5-4.5, with a mean value of RV = 3.53, indicating significant spatial variations of dust properties across the Galactic disk and deviations from the canonical diffuse interstellar medium value RV = 3.1. Afterward, by fitting the isochrone to the Colour-Magnitude Diagrams, we estimated the metallicity values of 105 observed clusters. This work provides the first homogeneous catalogue of extinction law measurements based on combined ground-based ultraviolet and Gaia photometry for a large open-cluster sample, establishing an important reference for studies of Galactic dust structure and stellar parameter corrections.

Galaxy groups and clusters trace the large-scale matter distribution, with their clustering usually interpreted mainly as a function of halo mass. Yet, at fixed mass, their baryonic properties retain information about halo growth, gas accretion, and feedback. The intrinsic scatter in X-ray luminosity and gas fraction suggests that X-ray-selected systems may not be a random subset of the halo population. If these observables correlate with halo assembly, they may trace secondary variations in halo bias. We test this using the Magneticum hydrodynamical simulation, measuring the clustering of systems selected by X-ray luminosity and gas fraction at fixed halo mass. We construct mass-matched subsamples by ranking halos in percentiles of X-ray luminosity and derive the linear halo-matter bias from the halo-matter cross-power spectrum. X-ray-bright halos are more strongly clustered than X-ray-faint halos at fixed mass. For the 84th-16th percentile split, we find $Δb_{\rm lin}=0.17\pm0.03$, corresponding to a $\sim17\%$ enhancement relative to the X-ray-faint sample. A 67th-33rd split gives a consistent signal, with $Δb_{\rm lin}=0.12\pm0.02$ and a $\sim12\%$ enhancement. The effect is strongest at group scales and negligible for cluster-size halos. Gas fraction shows an even stronger clustering dependence, with relative enhancements of $\sim39\%$ and $\sim26\%$ for the two percentile splits. This signal is present from $z\simeq2$, whereas X-ray luminosity becomes significant only at $z\simeq0.3$, once the gas thermodynamic state is more closely coupled to baryon retention. Matching halos by both mass and formation time reduces the large-scale bias difference to below $2σ$, indicating that formation time captures much of the signal. These results show that, in Magneticum, X-ray luminosity traces a baryonic manifestation of halo assembly bias beyond mass.

The advancements in highly sensitive and powerful radio telescopes, including the Square Kilometre Array Observatory (SKAO) and its precursors, MeerKAT, ASKAP, MWA, and HERA, will enable us to create the deepest radio images of the sky. However, due to the sheer scale of the datasets, manually classifying and analyzing this data is computationally expensive, time-consuming, and laborious. Therefore, the development of automated algorithms to detect and classify complex morphological radio sources from large astronomical surveys is the need of the time. In this chapter, we examine the recent advancements and challenges in source-finding techniques triggered by the analysis of SKAO precursor data and the SKA Data Challenges, both in spectral and continuum modes, along with the growing demand for computational resources and automated source detection methods using machine learning (ML) algorithms for the identification and characterisation of new populations of sources, addressing their complex and diffuse morphologies, as well as transient nature. Additionally, we discuss the critical factors affecting the quality and limitations of automated source-finding techniques, including artefacts from residual continuum emission, sidelobes, radio frequency interference (RFI), technical failures, calibration issues, flagging methods and false positives. With the availability of the full operational phase of the SKAO, continued advancements in source detection algorithms and computational infrastructure will be essential to fully exploit its scientific potential.

Circularly symmetric diffuse radio sources represent a recently recognized and rare class of extragalactic synchrotron-emitting structures whose physical origin remains uncertain. The currently known population is small, and the limited frequency coverage, sensitivity, and polarimetric information available from existing observations hinder robust discrimination among the various proposed formation scenarios. This chapter presents an overview of four representative circular diffuse sources featuring diverse morphologies, such as a horseshoe-shaped inner ring, circular diffuse emission surrounded by inner S/Z-shaped structures and a double-ring diffuse source. The angular sizes of these sources are in the range of 60-100 arcsec, similar to odd radio circles (ORCs). These systems illustrate the morphological diversity of circular diffuse radio emission and provide valuable clues to the physical processes responsible for their formation. The Square Kilometre Array Observatory (SKAO), with its unprecedented sensitivity, wide instantaneous bandwidth, sub-arcsecond imaging capability, and high-fidelity polarimetric performance, will significantly advance the study of these objects. SKAO surveys are expected to detect a substantially larger population of faint circular diffuse radio sources, enabling the first statistically meaningful studies of their occurrence, environments, and evolution. An expanded sample will allow investigations of their cosmological evolution and help determine whether their origin is linked to AGN feedback processes, episodic jet activity, galaxy interactions, or large-scale structure formation. This chapter summarizes the current observational knowledge of circular diffuse radio sources and describes how next-generation radio facilities, particularly the SKAO, will enable progress in understanding the origin and physical nature of these enigmatic radio structures.

Several observational and theoretical studies have suggested that the observed arm-length asymmetries in extragalactic radio sources are primarily driven by interactions between radio jets and an inhomogeneous ambient medium, although orientation effects may also contribute to the observed asymmetry. However, the observational evidence supporting these interpretations comes from only a small sample of FR II radio sources, in which the brighter hotspots are typically found on the side of the shorter jet arm. We aim to investigate the interactions between powerful jets and their surrounding environments in radio sources, with a particular focus on how these interactions shape the morphology and asymmetry of the radio lobes. The unprecedented sensitivity and angular resolution of the Square Kilometre Array Observatory (SKAO) will facilitate the discovery and detailed characterization of asymmetric radio sources across a wide range of physical scales and redshifts. Using a combination of 3D magnetohydrodynamic simulations and observational data from the SKAO, one can explore the influence of clumpy interstellar and intergalactic media on jet propagation and the resulting asymmetries in radio sources at various redshifts. The study will analyze how environmental factors, such as density and turbulence, decelerate jets, leading to observable asymmetries in smaller, higher-redshift sources. In this chapter, we review existing simulation and observational results on jet-environment interactions in radio galaxies and discuss how SKAO capabilities will further advance our understanding of AGN feedback and its role in shaping large-scale cosmic structure.

We present numerical three body simulations of a primordial black hole (PBH) encounter with the TOI 2796 system, which hosts a hot Jupiter. Over 100 days, we identify three outcomes: (i) ejection of the planet; (ii) a bound triple system (a approx. 1 au, P approx. 5 days); and (iii) capture of the planet by the PBH into a binary (0.0194 au, 15 days) that escapes the star. These results highlight the rich dynamical diversity of PBH encounters and their potential role in shaping planetary systems.

Pulse profile modeling has proven to be a powerful technique for determining the mass and radius of neutron stars. To date, this method has been applied to a handful of millisecond pulsars observed by the Neutron Star Interior Composition Explorer (NICER). However, analyses of more millisecond pulsars are necessary to determine tight constraints on the equation of state of superdense matter. In this study, we present an analysis of the isolated, rotation-powered millisecond pulsar PSR J2124$-$3358 using the X-ray Pulse Simulation and Inference (X-PSI) package, a publicly available state-of-the-art code for neutron-star relativistic ray tracing and Bayesian parameter inference. We use NICER and Chandra observations of this pulsar, exploring different neutron star atmospheric compositions and different configurations of the hot polar caps responsible for the pulsed X-ray emission. Our analyses favor a helium atmospheric composition, plausibly originating from accretion and subsequent evaporation of a former hydrogen-depleted binary companion. For this composition, and given the faint nature of the source and the low signal-to-noise of the data sets, we obtain broad posterior distributions yielding a mass $M = 1.8\pm0.5\,M_\odot$ and an equatorial radius $R_{\mathrm{eq}} = 11.7^{+2.6}_{-3.0}$ km (medians and $68\%$ credible intervals), and infer a configuration consisting of two slightly non-antipodal hot spots. By contrast, when using a hydrogen atmosphere model, the mass and radius decrease by $\sim 0.5\,M_\odot$ and $\sim 1$ km, respectively. Future multiwavelength studies, particularly those incorporating radio and gamma-ray pulse-emission, may provide tighter constraints on the geometry and physical properties of this source.

Blazars are among the most extreme classes of active galactic nuclei. They are powered by relativistic jets, but the way in which the jet energy is dissipated is still unclear. The flat radio spectrum and the core-shift effect trace the distributions of magnetic fields and relativistic particles along the jet, while variability carries information about time-dependent dissipation. However, a unified framework connecting these observables to the underlying jet physics has been lacking. Here we present a multi-frequency analysis of the prototypical blazar Mrk~501. We model its core-shift measurements, spectral energy distributions (SEDs), and power spectral densities (PSDs) with a conical jet model that conserves magnetic power. The core-shift data localize the radio emitting regions and constrain the electron-density and dissipation-rate profiles along the jet. With a single radial distribution of jet parameters, the model reproduces the core-shift relation and SED, but it underpredicts the observed variability at high radio frequencies and in the optical to $γ$-ray bands. We therefore introduce different blob distributions for the inner ($\lesssim$~0.1\,pc) and outer ($\gtrsim$~0.1\,pc) jet regions. With this extended model, the simulated PSDs are consistent with the multiwavelength observations of Mrk~501 during its 2017--2019 low state. This result points to different dissipation behavior in the inner and outer jet. Our study demonstrates that spectro--timing--astrometric jet modeling, which combines SEDs, multiwavelength PSDs, and radio core-shift measurements, can constrain jet stratification and scale-dependent dissipation in blazars.

Multimessenger observations of compact binary mergers in active galactic nuclei (AGN) offer unique probes of black hole formation channels and cosmology. We analyse 18 gravitational-wave events from LVK observing runs O3--O4b paired with 28 candidate AGN counterparts, incorporating supermassive black hole (SMBH)-induced environmental redshift corrections and electromagnetic sky-localisation constraints. Of the 28 candidate pairs, 21 are positive-to-strong-favoured (the most compelling being GW190412\_053044 $\leftrightarrow$ J143041.67$+$355703.8 at $\ln\mathcal{B} = +17.35$), three are inconclusive, and four are negative-to-strong-disfavoured. From the 21 positive-to-strong-favoured binary black hole (BBH) AGN pairs, 13 unique associations can be identified as the preferred ones. The cumulative Bayes factor across all 13 BBH-AGN confirmed associations yields $\ln\mathcal{B}_{\rm comb} \approx +81$, establishing strong collective support for AGN-hosted merger associations. Sky localisation is the dominant driver of model selection; at current detector sensitivity, the environmental redshift corrections are neither decisively confirmed nor ruled out, motivating future high-precision multimessenger follow-up.

We measure the Hubble constant $H_0$ using 13 gravitational-wave binary black hole mergers associated with active galactic nucleus hosts. We find $H_0=70.50^{+3.37}_{-2.89}\,({\rm stat})\pm1.56\,({\rm cal})$\,km\,s$^{-1}$\,Mpc$^{-1}$ ($4.4\%$ precision), consistent with both Planck\,2018 ($0.98σ$) and SH0ES\,2024 ($0.76σ$), with no significant preference between the two. Combining with the bright siren GW170817 sharpens the constraint to $H_0=70.31^{+3.00}_{-2.85}\,({\rm stat})\pm1.55\,({\rm cal})$\,km\,s$^{-1}$\,Mpc$^{-1}$ ($4.2\%$ precision), and further combining with an independent dark-and-bright-siren sample tightens it to $H_0=69.71^{+2.55}_{-2.40}\,({\rm stat})\pm1.54\,({\rm cal})$\,km\,s$^{-1}$\,Mpc$^{-1}$ ($3.5\%$ precision). Assuming a luminosity-distance prior centered around the value related to a fixed cosmology in turn, recovers $H_0=67.62\pm0.72$ (Planck-anchored) and $H_0=72.91\pm0.72$\,km\,s$^{-1}$\,Mpc$^{-1}$ (SH0ES-anchored). We show that under such an assumption, a rejection of $\gtrsim4σ$ to the opposing anchor is obtained.

We report on the diagnostic inspection of nine Project Hephaistos Dyson Sphere candidate M-dwarfs based on archival data. By comparing the Gaia positions, propagated to the AllWISE epoch, with the mid-infrared centroids measured from the AllWISE images, together with deep archival optical/near-infrared imaging, we identified significant background contamination in candidates B and C. Candidate B is coincident with a radio counterpart with spectral index alpha = 0.63 +/- 0.11, while candidate C has a near-infrared companion at an offset of 3.75 arcsec. Candidate A provides suggestive evidence through a radio counterpart with spectral index alpha = 0.40 +/- 0.35, while candidates E, F, H and J show marginal evidence. These systems exhibit either significant astrometric offsets or visible interlopers, indicating that the mid-infrared excess likely arises from line-of-sight contamination by hot, dust-obscured galaxies. However, candidates D and I still lack obvious signs of contamination. Dedicated observations are therefore essential to characterise these potential interlopers, eliminate false positives, and ensure that technosignature searches focus on the most robust Dyson Sphere candidates.

When operational, the SKA will generate unprecedented amounts of data and provide exquisite sensitivity for 21 cm tomography of Cosmic Dawn (CD) and the Epoch of Reionization (EoR). With this comes opportunities for new data-driven algorithms that unlock new methods for instrument modelling, data analysis, theoretical simulation, and inference for understanding the high-redshift universe. In this chapter, we provide an overview of some machine learning algorithms that have been proposed for CD and EoR science with the SKA

Since the 2015 SKA Chapters, the advent of panchromatic observations of nearby galaxies, capable of resolving the interstellar medium (ISM) into individual star-forming regions and molecular clouds, has revolutionized our understanding of the star formation cycle in galaxies. Despite these advances, our understanding of the gas cycle in galaxies remains limited because we know little about how the cold atomic phase of the ISM, which dominates its mass budget, participates in the cycle. Specifically, the processes regulating the condensation of atomic gas into molecular gas, the balance between warm and cold atomic gas, and their dependence on local environment and star formation activity have been explored only in the Milky Way and the Magellanic Clouds. The SKA presents a new opportunity to observe the atomic ISM within nearby galaxies at comparable resolution to other facilities and is the only facility that can make wide-area surveys of atomic gas across a statistical sample of galaxies at high (<100 pc) spatial resolution. Such observations will directly address the open questions about the evolution of the star-forming ISM. We present recent results from SKA-like observations of the Local Group made with SKA pathfinders and the Jansky VLA to illustrate the promise of replicating those observations for hundreds of nearby galaxies, reaching out to the Virgo cluster.

The Hubble tension has come to stay as a major problem in modern cosmology as it continues to plague the standard cosmological model ($Λ$CDM). As one of the viable, self-consistent dark energy theories, k-essence involves nontrivial self-interactions that can modify the background expansion beyond recombination; thereby impacting the sound horizon to last scattering, and hence, the inferred value of the Hubble constant. We examine this tension in two physically motivated k-essence models, dilaton and tachyon, using datasets from Planck and late-Universe probes including Pantheon+SH0ES, cosmic chronometer (CC), Supernova Cosmology Project Union compilation (Union3), Dark Energy Survey Year~5 (DESY5), and Dark Energy Spectroscopic Instrument (DESI) measurements. While $Λ$CDM exhibits inconsistent tension inferences, both k-essence models exhibit a substantial tension alleviation that is robust against the inclusion of the independent late-Universe cosmological datasets, giving consistent tension reduction irrespective of whether the observations are supernovae (Pantheon+SH0ES, Union3, DESY5) alone or in combination with cosmic chronometers (CC) and baryon acoustic oscillation measurements (DESI). The combined late-Universe dataset leads to only $0.14σ$ and $0.69σ$ offsets from the Planck prediction in the dilaton and tachyon models, respectively, compared to $5.89σ$ tension in $Λ$CDM. Both models demonstrate that the inferred tension alleviation is a stable, intrinsic consequence of the underlying k-essence dynamics rather than of model fine tuning: model parameters remain unchanged across datasets. The results establish that the apparent Hubble tension is not an unavoidable feature of late-Universe cosmology but depends critically on the description of dark energy.

We study inflation in metric-affine gravity. We write an action that contains all the second order algebraic distortion terms, and all first order distortion terms with a single covariant derivative, coupled to a scalar field. We include the Einstein--Hilbert term with nonminimal coupling, a scalar field potential, and impose projective invariance. The distortion equation of motion is algebraic by construction, and the distortion is integrated out analytically. This yields a kinetic term sourced entirely by distortion, with a kinetic coupling function determined by the 13 free coupling constants of the starting action. We compute inflationary observables for three model classes with a monomial distortion coupling. For a monomial potential, the spectral index and tensor-to-scalar ratio depend only on the ratio of the exponents, with the starting coupling constants dropping out entirely; however, this model lies outside the Planck + BK18 $2σ$ contours. For a potential of the $α$-attractor form, the observables are governed by a single parameter and approach the Starobinsky predictions as a limit. Including a nonminimal coupling to the Ricci scalar with a monomial potential can also yield an asymptotically flat effective potential with the same modified Starobinsky observables.

Multimessenger observations suggest that Seyfert galaxies are promising sources of high-energy neutrinos, but their dense inner environments can strongly suppress the emerging very-high-energy gamma-ray emission. Active galactic nuclei (AGN) undergoing intense accretion episodes can enter a super-Eddington state, in which the accretion flow becomes geometrically and optically thick within a critical radius and develops strong magnetic fields in its innermost region. At the same time, large amounts of matter are expelled from the disk surface in the form of powerful, radiation-driven winds. In this work, we explore a scenario in which the cores of super-Eddington AGN provide suitable conditions for the acceleration of relativistic particles, including hadrons, via magnetic reconnection in a magnetically confined region close to the supermassive black hole. The accelerated hadronic component interacts with the intense photon field of the disk, leading to a copious neutrino flux peaking at 10-100 TeV that may be detectable with current observatories such as IceCube and KM3NeT, while the surrounding outflow efficiently absorbs the accompanying gamma-ray emission from the inner core. We also apply the model to the nearby super-Eddington Seyfert 1 NGC 7469 as a representative case with two reported neutrino events. In this framework, super-Eddington AGN, in particular Seyfert galaxies undergoing transient intense accretion episodes, emerge as plausible hidden neutrino sources, offering a natural explanation for the coexistence of efficient neutrino production and a strongly attenuated gamma-ray counterpart.

Radio emission from relativistic tidal disruption events (TDEs) is commonly interpreted using the Blandford--McKee solution for an impulsive relativistic blast wave. Unlike gamma-ray bursts, however, TDE jets are powered by fallback accretion, injecting energy over weeks to months. We investigate how this sustained energy injection modifies the dynamics and radio emission of relativistic TDE jets using one- and two-dimensional relativistic hydrodynamic simulations coupled with synchrotron radiation calculations. For a fixed total energy, fallback-powered jets drive slower forward shocks, producing delayed and systematically fainter radio emission. We show that the impulsive approximation is valid only when the duration of energy injection is less than approximately ten percent of the deceleration time of an equivalent Blandford--McKee blast wave, a condition that is generally not satisfied for relativistic TDEs. As a result, continuously powered outflows are significantly less beamed than impulsive explosions, leading to a strong suppression of off-axis radio emission. These results demonstrate that the radio evolution of relativistic TDEs retains memory of the central engine and cannot, in general, be described by an impulsive solution. By substantially reducing the expected off-axis radio emission, sustained fallback-powered jets imply that relativistic jet production in tidal disruption events may be considerably more common than current radio observations suggest.

Determining the relationship between star-formation rate (SFR) and the radio luminosity ($L_{1.4}$) is critical if we are to trace the star-formation history of the Universe dust-agnostically using current and future radio facilities. However, until now, such work has relied on potentially biased binary classifications of sources to remove contaminating active galactic nuclei (AGN). We present a new, statistically-driven methodology for deriving the SFR -- $L_{1.4}$ relation, removing the need for problematic cuts. We use a Bayesian hierarchical mixture model fit to the radio-detected sources in the deep MIGHTEE COSMOS DR1 catalogue, incorporating the full SFR posterior probability distributions generated by state-of-the-art spectral energy distribution fitting code \texttt{GRAHSP}. This allows us to probabilistically determine a mean SFR -- $L_{1.4}$ relation for the SF dominated galaxies, whilst accounting for changing fractions of SF dominated sources across redshift, radio luminosity and stellar mass ranges. We find that the SFR -- radio luminosity correlation exhibits a significant dependence on redshift, but a stellar mass dependence that is weaker than previous studies. Our resultant SFR-radio correlation is $\log_{10}(\text{SFR}/M_{\odot}\,\text{yr}^{-1}) = 0.790\times(\log _{10}(L_{1.4}/\text{W\,Hz}^{-1})-23) + 1.244 \times(1+z)^{0.122} -0.033 \times (\log_{10}(M_*/M_{\odot})-10)$, with an intrinsic scatter of 0.178 dex. We show that this redshift evolution could be explained by a moderate evolution in the radio spectral index of SF galaxies. We attribute the lack of observed strong dependence on stellar mass, compared to recent studies, to the novel statistical approach that does not rely on cuts to remove AGN.

Short gamma-ray bursts with extended emission (SGRBEEs) are a particular class of long GRBs (LGRBs) which, despite their duration, share several observational features with short GRBs (SGRBs). They are composed by a short, hard initial pulse (IP) followed by a longer and softer extended emission (EE). We investigate whether SGRBEEs originate from the same progenitor as SGRBs despite their duration, representing a peculiar subclass of LGRBs, or if they constitute a distinct population. Given their respective duration, we tested if the IP and the EE share properties with short and long GRBs, respectively. We analysed SGRBEEs from the flux-limited, redshift-complete SBAT4 sample using prompt-emission data from Swift/BAT, Fermi/GBM and Konus-WIND, and X-ray afterglow observations from Swift/XRT. The temporal and spectral properties of the IP and EE components were compared with those of SGRBs from the extended SBAT4 sample and LGRBs from the BAT6 sample. Despite observing a clear spectral evolution during the prompt phase of each SGRBEE, IPs and EEs as well as short and long GRBs can not be distinguished by their hardness ratio only. All bursts analysed have a spectral lag consistent with zero. SGRBEEs XRT light curves are consistently more complex than those of SGRBs, requiring the addition of multiple breaks and showing the presence of steep decays and plateaus. Our results indicate that SGRBEEs are not an intermediate class. During prompt emission, despite common spectral features, IPs and EEs temporally differ from short and long GRBs, respectively. EEs are on average too faint to appear in the Amati plane, but IPs occupy the same parameter space of SGRBs. In the afterglow, SGRBEEs are more luminous than standard GRBs at early-time, suggesting a direct contribution from the EE; at later times, these events behave as standard SGRBs, and both remain systematically less luminous than LGRBs.

The rapid assembly of supermassive black holes (SMBHs) observed at $z\gtrsim7$ requires efficient seeding mechanisms in the early Universe. Population III (Pop. III) star clusters have recently emerged as a promising pathway that may bridge the gap between traditional light- and heavy-seed scenarios by producing intermediate-mass black holes (IMBHs) with masses up to $\sim10^4\,\rm M_{\odot}$. We investigate the properties and number densities of IMBHs forming in Pop. III star clusters with masses $M_{\rm cl}\sim10^3-4\times10^5\,\rm M_{\odot}$, and hosted in isolated dark matter minihalos, using a suite of direct $N$-body simulations. We adopt cosmologically motivated initial conditions and explore different stellar evolution prescriptions, binary orbital parameter distributions, and cluster dynamical configurations. By $z\sim19$, the IMBH mass function consistently peaks at $m_{\rm IMBH}\sim200\,\rm M_{\odot}$, with number densities of $n_{\rm IMBH}\sim0.2-5\,\rm cMpc^{-3}$. In sufficiently dense and massive clusters, IMBHs with masses $>10^3\,\rm M_{\odot}$ can already form by $z\sim19$, reaching number densities of $n_{\rm IMBH}\sim10^{-4}-10^{-2}\,\rm cMpc^{-3}$. The most massive IMBHs in our models reach $\sim6200\,\rm M_{\odot}$ through the collapse of very massive stars assembled by repeated stellar collisions, a process enhanced in fractal clusters. Lower-mass IMBHs form instead predominantly through single and binary stellar evolution and binary stellar mergers. We find that models combining large stellar radii and tight binaries produce the highest IMBH abundances relative to isolated Pop. III evolution. Owing to the high retention fraction of IMBHs ($\gtrsim88\%$), massive dense Pop. III star clusters can act as efficient incubators of both light and heavy SMBH seeds, even if only a fraction of Pop. III stars formed in such environments.

Pulsar timing arrays (PTAs) can detect dark matter (DM) substructure through the small shifts a transiting DM subhalo imprints on pulse arrival times. Recently found evidence for a stochastic gravitational-wave background (GWB) acts as red noise and competes with the substructure signal. Here we provide an analytic understanding of how this background degrades PTA sensitivity to DM substructure. We derive the full gauge-invariant proper-time observable induced by a transiting DM subhalo and develop a framework for the expected signal-to-noise ratio in the presence of red noise, accounting for the degeneracy with the pulsar timing model. From this we obtain simple scaling relations for the reach across the static, dynamic, and stochastic regimes, and numerically compute the reach for a Square Kilometre Array benchmark at three representative points of the NANOGrav 15-year posterior. The GWB background suppresses the sensitivity to DM substructure by one to three orders of magnitude compared to forecasts in the presence of only white noise, and the suppression depends on the amplitude and spectral index of the background within a factor of three. The dynamic Shapiro signal suffers the smallest suppression and gives the best sensitivity to DM substructure near $10^{-2}\, M_\odot$. Probing the regime where subhalos make up all or part of the DM remains a challenge even for surveys with more pulsars and longer observing time. Despite this, PTA measurements remain a competitive probe of DM substructure, and future surveys will increase in sensitivity by up to two orders of magnitude from existing NANOGrav limits.

Millions of supernovae will be discovered with the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). As a result, spectrographs around the world will have to make difficult decisions about which supernova candidates receive spectroscopic follow-ups. This work identifies the minimum spectral resolution, $R_λ = \fracλ{Δλ}$, as a function of signal-to-noise ratio (SNR) at which spectral classification of supernova subtypes becomes impossible. We include supernova types Ia, Ia-91T, Ia-91bg, Iax, Ib, Ic, broad-lined Ic, IIb, IIP, and Ibn in this work. We produce a definition of SNR based on specific lines for each SN subtype that allows us to generate homogeneous datasets at 16 different values of $R_λ$ and 14 different SNR's and we tested the classification performance of a recently developed deep-learning classifier, ABC-SN, on each $R_λ$ and SNR combination. We find that classification of supernova spectra into a refined taxonomy that separates, for example, between different subtypes of stripped envelope supernovae, is possible at low resolution and low SNR with no loss in model performance down to $R_λ = 50$ and $\text{SNR} = 5$. Classification performance is only minimally impacted even as low as $R_λ = 25$. We hope that astronomers using the LSST alert stream, as well as designers of future instruments and observatories, will benefit from knowing what spectral resolution is necessary to classify a supernova for arbitrary \SNR{}.

We report the discovery of VVVX J080705.96-273823.6, the first low-surface-brightness galaxy candidate identified in the Zone of Avoidance by the VISTA Variables in the Via Lactea eXtended Survey (VVVX). The source was detected serendipitously during a visual inspection of the new Deepstack images, reconstructed mosaics generated through an updated reprocessing and stacking procedure that significantly improves the depth and spatial uniformity of the original survey images. First-order morpho-photometric parameters were obtained using SExtractor together with PSF models derived with PSFEx, applied to star-subtracted images to minimize foreground contamination. Structural properties were derived from PSF-convolved modeling of azimuthally averaged surface-brightness profiles in the J, H, and Ks bands using exponential, Sersic, and composite models. The galaxy candidate is clearly detected in all three near-infrared bands. The profiles are consistently described by an exponential disk, with scale lengths of 32-36 arcsec and extinction-corrected central surface brightnesses of 19.82-22.35 mag arcsec^-2. The J band provides the best-constrained profile, while H and Ks show the same behavior with larger uncertainties. The projected proximity to ESO 494-25 and ESO 494-26 suggests a possible physical association. If located at the same distance (11-12 Mpc), the exponential scale length would be about 1.8-2.0 kpc, consistent with nearby LSB disk galaxies, although the lack of a spectroscopic redshift prevents confirmation. This discovery demonstrates that the new VVVX Deepstack products can reveal faint diffuse galaxies at low Galactic latitudes, opening a path toward a more complete census of the obscured nearby Universe.

We report the discovery of 31 new high-$z$ quasars in the redshift range $6.6 < z < 7.8$. These quasars were selected from approximately 3000 deg$^2$ of sky covered during the first 1.5 years of the Euclid Wide Survey, representing the initial results of the Euclid high-$z$ quasar search. Our candidate selection employed multiple machine-learning and probabilistic techniques applied to the Euclid $I_E$, $Y_E$, $J_E$, and $H_E$ images, supplemented by ancillary $z$-band data when available. Spectroscopic follow-up observations were carried out with Keck, Magellan, and the Large Binocular Telescope (LBT). Among the new discoveries, there are 12 quasars at $z \geq 7$, more than doubling the number of previously known quasars at $z \geq 7$. The newly discovered quasars exhibit $21.2 < J_E < 23.2$ ($-25.5 < M_{1450} < -23.6$), extending quasar studies to the faint end of the quasar luminosity function (QLF) at $z \gtrsim 7$. The quasar with the highest-$z$, EUCL J172902.75+641018.1 at $z \approx 7.77$, sets the new redshift record for the most distant quasar ever reported. These discoveries demonstrate Euclid's transformative role in high-$z$ quasar discovery and set the stage for future follow-up studies of the early galaxies hosting quasars, supermassive black hole growth, and the intergalactic medium in the epoch of reionisation.

Constraining the co-evolution of supermassive black holes and their host galaxies in the first billion years after the Big Bang is essential for understanding the formation of the earliest cosmic structures. Here, we present IRAM/NOrthern Extended Millimeter Array (NOEMA) observations of the $z \approx 7.7$ quasar EUCL\,J125308.55+705432.3, recently discovered in the first data release of the Euclid Wide Survey. We report the most distant detections of [CII] 158$μ\mathrm{m}$ and cold dust emission in a quasar host to date. The [CII] emission line sets the systemic redshift at $z=7.6980\pm0.0004$. The source exhibits luminosities of $L_{\rm FIR}=3.6\times10^{12}\,L_{\odot}$ and $L_{[CII]}=2\times10^9\,L_{\odot}$, respectively, a dust mass of 1.4$\times 10^{8}\,M_{\odot}$, and a dynamical mass in the range 0.33-1.3$\times10^{10}\,M_{\odot}$. Remarkably, despite being nearly two magnitudes fainter in the rest-frame UV ($M_{1450}=-24.06$) than previously known $z\approx7.5$ quasars (<M$_{1450}$> $\sim-$26.5), EUCL\,J125308.55+705432.3 exhibits the brightest [CII] emission among them. This indicates that the host galaxy is actively star-forming, with a star-formation rate $>250\,M_{\odot}\,\mathrm{yr}^{-1}$, consistent with recent findings that UV-faint quasars at $z>6$ preferentially reside in [CII]-luminous galaxies. The UV-faintness likely reflects dust obscuration or sub-Eddington accretion, rather than lower host mass, suggesting these systems are at a different stage in their evolution compared to UV-bright quasars. These IRAM/NOEMA observations highlight the power of combining Euclid's wide-area quasar discovery potential with submillimetre follow-up observations to characterise the host galaxies of early supermassive black holes across a broader redshift and luminosity range than previously accessible.

Through the detailed study of the optical flux behaviour in blazars over time, it is possible to infer the conditions responsible for their observed emission. PKS 2155-304, a BL Lac object detected from radio to TeV energies, is among the brightest blazars in the southern hemisphere. We present optical monitoring spanning over two decades using telescopes at Complejo Astronómico El Leoncito and Estación Astrofísica de Bosque Alegre, Argentina. Differential light curves in the B, V, R, and I bands reveal significant variability on weekly and longer timescales, with occasional changes on sub-four-hour scales. The optical spectral index remained negative, consistent with non-thermal emission, and hardened over the past nine years. Evidence for quasiperiodic behaviour on 20-30 day timescales was found, while correlations with X-ray fluxes were weak, suggesting distinct emission components in the two bands. These results highlight the pronounced optical variability of PKS 2155-304 and provide insight into its multi-band emission mechanisms

The strength of the 21-cm Cosmic Dawn radio absorption signature against the Cosmic Microwave Background or against a bright background radio source depends on the temperature of the neutral hydrogen in the intergalactic medium. While x-rays from forming galaxies will likely dominate the heating rate, recent models suggest scenarios in which heating by the scattering of Lyman-$α$ photons sourced by the galaxies contributes non-negligibly as well, and may even dominate. The efficiency of Lyman-$α$ photon heating for a point source at high redshift using a numerical solution to the exact radiative transfer equation is provided here. It is found to be a factor of several smaller than the rate computed in the commonly used diffusion approximation to the radiative transfer equation.

Context. Several author have studied the giant radio source GRS0917 + 75 , but its origin remains unclear. Aims. This source is unusual, because of its large size and its location outside a rich cluster of galaxies. We aim to understand and discuss the properties and nature of this source and its connection to the environment. Methods. We conducted optical observations to obtain new spectroscopic data. We also acquired a LOFAR image at 144 MHz to derive information at low radio frequency. Moreover, we performed a new analysis of archival VLA data in the L and C bands for a multifrequency study of the source properties in the radio band. Results. From the observational data, we classify GRS0917 + 75 as a giant radio galaxy with a size of 1.5 Mpc and an estimated age of about 100 Myr. The optical parent galaxy is a bright low-excitation radio galaxy, the brightest member of a very poor group belonging to a large supercluster. GRS0917 + 75 is a peculiar low-power Fanaroff-Riley Class I giant radio galaxy with a bright central emission but no jet-like features. Conclusions. The existence of giant radio galaxies such as GRS0917 + 75 could explain the origin of relativistic particles and magnetic fields in low-density environments.

The magnetic field structure of the Milky Way can offer critical insights into the origin of galactic magnetic fields. Measurements of magnetic structures of the Milky Way are still sparse in far regions of the Galactic disk and halo. Pulsars are the best probes for the three-dimensional structure of the Galactic magnetic field, primarily owing to their highly polarized short-duration radio pulses, negligible intrinsic Faraday rotation compared to the contribution from the medium in front, and their widespread distribution throughout the Galaxy across the thin disk, spiral arms, and extended halo. In this article, we give an overview of Galactic magnetic field investigation using pulsars. The sensitive SKA1 design baseline (AA4) will increase the number of known pulsars by a factor of around three, and the initial staged delivery array (AA*) will probably double the total number of the current pulsar population. Polarization observations of pulsars with the AA* telescopes will give rotation measures along several thousand lines of sight, enabling detailed exploration of the magnetic structure of both the Galactic disk and the Galactic halo.

We present a methodology to measure the intrinsic alignment (IA) signal of galaxies in the vicinity of cosmic voids using the \Euclid-like Flagship cosmological simulation from the Euclid Consortium. The IA signal is quantified and compared with the predictions of the linear alignment (LA) model, providing one of the first detailed investigations of this effect in underdense large-scale environments. While the IA signal around cosmic voids has received little attention to date, it may constitute a non-negligible systematic in forthcoming cosmological analyses that exploit void-lensing measurements. Our analysis examines red and blue galaxy populations separately, enabling a comparison of their alignment behaviour in void environments with the corresponding trends measured in galaxy-galaxy correlations. We find that the redshift evolution of the IA amplitude in cosmic voids is broadly consistent with that measured in the general galaxy population for both colour-selected samples. Additionally, our modelling allows us to estimate the linear bias of voids, $b_{\rm V}(z)$ which characterises how cosmic voids trace the underlying dark-matter density field, for voids with radii in the range $10 < R_{\rm V}/(h^{-1} {\rm Mpc}) < 15$. The measured bias exhibits a positive trend with redshift, consistent with theoretical predictions for the clustering of underdense regions. These results highlight the importance of accurately modelling IA in void studies, both to mitigate systematic effects in void-lensing cosmology and to further improve our understanding of galaxy-environment interactions in low-density regions of the Universe.

Quasar microlensing is uniquely sensitive to the size-scale of the accretion flow, offering one of the few direct probes of the accretion structure on micro-arcsecond scales. However, microlensing-based measurements in the optical and UV often find sizes systematically larger than expected from standard Shakura-Sunyaev disc theory, commonly referred to as the disc-size problem similar to that seen in continuum reverberation campaigns. But this assumes that all the emission comes from a single compact disc, neglecting the diffuse emission from the BLR which originates on much larger spatial scales. In this paper we directly quantify the effect of large-scale diffuse emission on the observed microlensing signatures. We adapt the physically motivated agnsed model to construct energetically self-consistent emissivity profiles in any given bandpass. Since this also predicts the full SED, we combine these SEDs with cloudy to give a diffuse BLR component. We convolve these models with representative microlensing magnification maps, and generate mock microlensing light curves to directly assess the inferred source size under different physical conditions. While the detailed shape of the disc emissivity profile has only a higher-order effect on the microlensing profile, the inclusion of the BLR makes a significant impact since this naturally smooths out the caustic network over larger scales. This introduces a significant bias when interpreted purely as a compact disc. However, the strength of this bias depends predominantly on the fractional contribution of the diffuse emission to the SED in the bandpass being considered, as this sets the effective half-light radius, giving an important wavelength dependence. We conclude that part of the excess in microlensing-inferred accretion disc sizes could arise from interpreting a composite (disc+BLR) picture as a single compact disc.

We present an overview of the Square Kilometre Array Observatory (SKAO) science potential in the area of extragalactic spectral lines besides 21-cm neutral hydrogen. It highlights the main points from the SKAO Science Book chapters on individual topics, but is augmented by additional prospects. The SKAO will push studies and use of masers, kilomasers, megamasers, molecular and radio-recombination-line emission and absorption to a wider variety of environments, including to very high redshift where it may detect the first molecule (HeH$^+$). It will open the door to measurements of hydrogen and helium isotopes, probing the conditions for star formation and Big Bang nucleosynthesis. While the planned SKAO of the 2030s (AA*) is destined to forge major progress, an SKAO as initially envisaged (AA4) will truly transform the landscape, and an extension towards higher frequencies (up to 24 GHz) would enable water maser and ammonia surveys in nearby galaxies and systematic molecular gas inventories at redshift 5.

The redshifted spectral emission from neutral hydrogen (HI) at rest wavelength 21 cm can be used as a tracer of large-scale structure and its evolution. Within the HI intensity mapping method, sufficient signal-to-noise is achieved by integrating the line emission within large voxels over a wide sky area and line of sight depth which allows access to the largest scales of the matter distribution. The resulting tomographic maps usually feature low angular and high redshift resolution. The SKAO will be able to conduct HI intensity mapping experiments observing up to 20,000 square degrees over a wide range of redshifts. For SKA-Mid, we will employ the array in a fast-scanning single-dish mode using Band 1 and 2 to access 0<z<3, mapping an enormous volume with fast survey speed, allowing for the possibility of a commensal survey producing high angular resolution maps via the on-the-fly imaging of the visibilities. For SKA-Low, we will focus on deep observations to detect the HI signal in a frequency band matching 3<z<6. In this chapter, we will give an overview of HI intensity mapping with the SKAO, including an outline of planned surveys, a discussion of observational challenges, and methodology for power spectrum methodology and forecasts. We present predictions on the constraining power on LambdaCDM cosmology from HI intensity mapping data via power spectrum, and other observables such as bi-spectrum and HI stacking. We also demonstrate the synergy power of HI intensity mapping with other cosmological surveys.

Galactic bars play a critical role in the secular evolution of their hosts by reorganising the ISM. We use a sample of 57 star-forming disc galaxies observed with JWST at 3 and 7.7 $μ$m to probe how the spatial distribution of PAH emission, as a structural marker of the cold ISM, depends on stellar mass and bar presence. We find evidence for a "watershed" at a stellar mass of $10^{10}$ Msun, marking a fundamental transition in the bar-driven distribution of PAH emission. This confirms trends previously predicted by numerical simulations and observed via ionised gas or UV light. While lower-mass galaxies exhibit a disordered and clumpy distribution of PAH emission regardless of bar presence, higher-mass barred hosts display well-structured dynamical features traced by PAH emission with significant gas reservoirs (e.g., discs and rings) within the central 15% of the bar radius (Rb). Furthermore, we observe a systematic depletion of PAH emission within the [0.2-0.8] Rb range in barred systems with stellar masses above $10^{10}$ Msun. Such central discs, rings, and associated radial dips ("bar deserts") appear to be a mass-dependent phenomenon: ubiquitous in massive galaxies but mostly absent in lower-mass counterparts. In contrast to the structured features in massive hosts, the disorganised ISM in lower-mass galaxies masks commonly observed bar-driven signatures. This suggests that tracer selection and dust obscuration may significantly bias observed bar fractions. Our study underlines two regimes of secular evolution, with different impacts and observability of bar-driven processes: it reaffirms bars as primary drivers of rapid secular evolution in galaxies above $10^{10}$ Msun$, while their impact is significantly reduced or delayed below this threshold. It further underscores the need to account for these processes when modelling galaxy evolution in cosmological simulations.

Most of the pulsar science case with the Square Kilometre Array (SKA) depends on long-term precision timing of a large number of pulsars, as well as their astrometric measurements using very long baseline interferometry (VLBI). However, before we can time them, or VLBI them, we must first find them. Here, we describe the considerations and strategies needed when planning an all-sky blind pulsar survey using the SKA. Based on our understanding of the pulsar population, the performance of the now-under-construction SKA elements, and practical constraints such as evading radio frequency interference, we project pulsar survey yields; this is done using two complementary methods for a number of illustrative survey designs, combining SKA-Low and SKA-Mid Bands 1 and 2 in a variety of ways. A composite survey using both SKA-Mid and SKA-Low is optimal, with Mid Band 2 focused in the plane. We find that, given its much higher effective area and survey speed, the best strategy is to use SKA-Low to cover as much sky as possible, ideally also overlapping with the areas covered by Mid. We find that an all-sky blind survey with Phase 1 of the SKA with the AA* array assembly will detect $\sim10,000$ slow pulsars and $\sim 800$ millisecond pulsars (MSPs) if SKA-Mid covers the region within $5°$ of the plane, while higher latitudes will be covered with SKA-Low. For the same survey region the yield with AA4 is $\sim 20\%$ higher, but this increases considerably by broadening the range covered by SKA-Mid Bands 1 and 2. In particular one could expect a yield of $\sim 1300$ MSPs with AA4. The pulsar census will enable us to set new constraints on the uncertain physical properties of the entire neutron star population. This will be crucial for addressing major SKA science questions including the dense-matter equation of state, strong-field gravity tests, and gravitational wave astronomy.

The known population of non-accreting neutron stars is ever growing and currently consists of more than 3500 sources. Pulsar surveys with the SKAO telescopes will greatly increase the known population, adding radio pulsars to every subgroup in the radio-loud neutron star family. These discoveries will not only add to the current understanding of neutron star physics by increasing the known sample, but will undoubtedly also uncover new types of sources that will challenge our theories of a wide range of physical phenomena. A broad variety of scientific studies will be made possible by a significantly increased population of neutron stars, unravelling questions such as: How do isolated pulsars evolve with time; What is the connection between magnetars, high B-field pulsars, and the newly discovered long-period pulsars; How is a pulsar's spin-down related to its radio emission; What is the nuclear equation of state? Increasing the numbers of pulsars in binary systems enables both larger numbers and higher precision tests of gravitational theories and general relativity, as well as probing the neutron star mass distribution. The excellent sensitivity of the SKAO telescopes combined with the wide field of view, large numbers of simultaneous tied-array beams that will be searched in real time, wide range of observing frequencies, and the ability to form multiple sub-arrays will make the SKAO an excellent facility for neutron star research. This chapter presents an overview of different types of neutron stars and discusses how the SKAO will aid in our understanding of the neutron star population.

The kinetic temperature structure of the photodissociation region M17SW was mapped using the IRAM 30 m telescope. This mapping employed the para-H2CO triplet (J(KaKc) = 303-202, 322-221, and 321-220) near 218 GHz on a scale of ~0.2 pc. The kinetic temperatures were derived by modeling the average H2CO line ratios (322-221/303-202 + 321-220/303-202) with the RADEX non-local thermodynamic equilibrium approach. These temperatures range from 28 to 181 K with an average of 54.2 +/- 0.3 K at a spatial density of 5.5x10^5 cm^-3. Comparing with the temperature measurements obtained from multiple transitions of NH3 (1,1)-(6,6) and the far infrared (FIR) dust continuum, the H2CO lines show temperatures similar to those measured by NH3 but slightly higher than values derived from FIR observations. The high kinetic temperatures observed from H2CO are associated with the ultracompact H II region UC1, dense clumps, as well as H2O and CH3OH masers, showing a similar distribution as NH3. This indicates that dense gas in the M17SW region is heated by star formation activity. The presence of a significant gas temperature gradient across the M17SW region, as measured by H2CO and NH3, provides direct evidence for gas heated predominantly by radiation emitted from the OB star cluster NGC 6618. On a smaller scale, the dense gas surrounding the dense clumps experiences significant heating from internal protostars and/or young stellar objects. Higher temperatures traced by H2CO are linked to turbulence on a scale of ~0.2 pc. The complex temperature structure of the M17SW region is revealed by H2CO and NH3, which may be attributed to both large-scale external radiative heating and small-scale internal radiative and turbulent heating.

The detection of a pulsar closely orbiting our Galaxy's supermassive black hole - Sagittarius~A* - is one of the ultimate prizes in pulsar astrophysics. The relativistic effects expected in such a system could far exceed those currently observable in compact binaries such as double neutron stars and pulsar white dwarfs. In addition, pulsars offer the opportunity to study the magneto-ionic properties of Earth's nearest galactic nucleus in unprecedented detail. For these reasons, and more, a multitude of pulsar searches of the Galactic Centre have been undertaken, with the outcome of just seven pulsar detections within a projected distance of 100\,pc from Sagittarius~A*. It is currently understood that a larger underlying population likely exists, but it is not until observations with the SKA have started that this population can be revealed. In this chapter, we look at important updates since the publication of the last SKAO science book and offer a focused view of observing strategies and likely outcomes with the updated SKAO design.

Several radio galaxies, such as 3C\,326\,N, show signatures of jet--ISM coupling, but a complete theoretical framework for explaining them is still lacking. Interpreting these observations requires a detailed understanding of the gas distribution, geometry, and outflow energetics. In this paper, we use three-dimensional relativistic hydrodynamic simulations to investigate jet--ISM coupling in inhomogeneous gaseous disks, exploring a parameter space spanning different cloud configurations, jet powers, and central disk densities. Our simulations incorporate a numerical turbulence injection scheme that maintains vertical support in the disk, preventing the unphysical collapse encountered in previous studies. We find that jet--ISM coupling is strongly governed by the underlying cloud configuration, leading to distinct outflow morphologies, velocity dispersions, and kinetic energies. Simulations with small-scale ($l_{\rm c,max}=50$~pc) clouds produce the highest velocity dispersions and kinetic energies, whereas large-scale cloud configurations ($l_{\rm c,max}=250$~pc) yield the lowest values, with mixed cloud distributions exhibiting intermediate behavior. In addition, mixed cloud configurations give rise to asymmetric jet propagation, naturally producing unequal lobe lengths similar to those observed in radio galaxies. We compare our fiducial simulation (a $10^{45}\,\rm erg\,s^{-1}$ jet interacting with a mixed cloud configuration) with observations of 3C\,326\,N, focusing on the morphology of the jet-driven bubble, synthetic emission and the gas kinematics. Our results successfully reproduce the observed properties, providing strong evidence that jet--ISM interactions can account for the wide bubble and the complex gas kinematics observed in this system.

We performed high-resolution three-dimensional relativistic magnetohydrodynamic (RMHD) simulations of restarted jets evolving within the cavity of a previous jet episode, using the PLUTO code. The simulations cover a spatial domain of $50\;\mbox{kpc}$ with a resolution of $40\;\mbox{pc}$. Three suites of simulations were performed to understand the impact of jet power, magnetic fields and quiescence time on the evolution of restarted jets. The restarted jets undergo a complex, multi-stage evolution, with the remnant cocoon relaxing from an initially conical structure to a mushroom-shaped morphology via an intermediate cylindrical phase. As the cocoon of the initial jet expands, dense material entrained by fluid instabilities such as Kelvin-Helmholtz and Rayleigh-Taylor significantly alter the conditions within it. As a result, the interaction of the restarted jet with the cocoon is markedly different from that of the initial jet with the ambient medium. In particular, we find that the restarted jet propagates ballistically through the rarefied remnant cocoon, without creating prominent backflows. Deceleration of the jet and associated backflows are observed only when it encounters dense structures. The structure and strength of the shocks in restarted jets are affected by jet power, magnetic field strength, and quiescence time. Finally, we discuss the implications of the dynamics for observed properties of radio galaxies.

Based on sensitive L-band (1.0 to 1.5 GHz) observations of pulsars using the Five-hundred-meter Aperture Spherical radio Telescope (FAST), we analyzed single-pulse sequences from FAST-detected pulsars and identified nulling, mode changing, or subpulse modulation phenomena in 374 sources. Among these, nulling has been detected in 160 pulsars, with 127 cases reported for the first time. Emission mode changes have been observed in 52 pulsars, including 51 first-time detections. Subpulse modulation has been identified in 272 pulsars, 180 of which are newly revealed, with the majority displaying subpulse drifting behavior. Subpulse drifting in some pulsars exhibits distinct modes with varying drift properties, leading to mode changes and divergent mean profiles. Statistics on pulsar parameters show that pulsars exhibiting nulling and/or subpulse modulation tend to be older, with longer periods and lower spin-down energy-loss rates. The modulation period P3 is predominantly correlated with pulsar rotation periods, magnetic field strengths, and spin-down energy loss.

Context. The joint analysis of different (astro-)physical messengers, in particular gravitational-wave data and electromagnetic follow-up observations, allows us to establish, explore and deepen links between different physical fields. As new survey capacities and improved detection methods will lead to a significant increase in the number of multimessenger detections in the upcoming decades, efficient and versatile software frameworks are essential to maximise the scientific outcome of such multimessenger studies. Aims. We present a major upgrade to the Nuclear Multimessenger Astronomy (nmma) framework, incorporating various recent developments in theoretical modelling and machine learning in a modularised and easily extendable Bayesian framework. For the first time, this allows direct sampling on nuclear parameters alongside gravitational-wave, kilonova and afterglow parameters. Methods. We combine fast surrogate models for electromagnetic transients with speed-ups from emulators that map nuclear parameters to macroscopic neutron-star properties. Additional acceleration methods for the evaluation of state-of-the-art waveform approximants enable full Bayesian analyses of multimessenger events at the speed required in the era of next-generation detectors. Results. We demonstrate the capabilities of the upgraded nmma framework through a series of representative applications. Reanalysing the 2017 multi-messenger detection of a neutron-star merger, we achieve 20- to 60-fold speed-ups while using more detailed physical models compared to previous studies. Moreover, we demonstrate for a hypothetical future detection how we can simultaneously constrain nuclear parameters and the Hubble parameter with robustly quantified uncertainties.

Rapid-response triggering is when a telescope is able to automatically respond to an external or internal astronomical transient alert, causing it to rapidly repoint at that position in the sky to catch its earliest radio emission. Both SKA-Low and SKA-Mid will have the ability to perform rapid-response triggering observations on externally detected transients as well as those detected within the data streams. We first give a brief overview of those radio instruments with active rapid-response observing modes. We then describe the different science cases motivating the need for this observing capability on SKAO and how the additional sensitivity afforded by the SKAO will enable us to answer fundamental questions relating to particle acceleration, transient central engines, coherent emission models and outflow physics in astrophysical systems spanning the range from the Sun to the high redshift Universe. Several suggestions relating to existing technologies and necessary SKAO system requirements are described. Through this chapter, we aim to ensure this is an existing, common and useful capability for the SKA Observatory.

Time domain studies of active galactic nuclei (AGN) at radio wavelengths probe physical processes near the central engine via intrinsic variability, in particular within the relativistic jets, as well as small-scale structures in the local Galactic interstellar medium (ISM) via scintillation and scattering effects. Recent discoveries reinforce the expectation that the high sensitivity, large field-of-view, and broadband frequency coverage of the SKA telescopes will help to revolutionise our understanding of AGN populations and the evolution of jets, and allow detailed modelling of the structure and dynamics of scattering plasma in the local ISM over a large fraction of the sky.

This paper presents the first dedicated characterization of the NE-Circinus Complex (TGU H1984, DCld 320.7-03.6), a previously unstudied star-forming dark cloud complex serendipitously identified in archival Herschel SPIRE observations targeting the foreground Bok globules BHR 99 and BHR 100. Using Herschel SPIRE photometry, Planck Galactic Cold Clumps data, near-infrared 2MASS colour excess mapping, AllWISE photometry, Gaia DR3 photometry, and a 3D dust extinction map, the complex, and its embedded YSO population are characterized for the first time. Two independent photometric methods place the complex at 750 +/- 50 pc. Three morphologically distinct components are identified: a dense main cloud body, a diffuse eastern component, and a northern filamentary extension. The main cloud body has a SPIRE-derived dust temperature of 14.5 +/- 0.5 K. Near-infrared H-K colour excess mapping yields a core gas mass of ~277 Msun and a total gas mass of ~439 Msun, consistent with the Planck PGCC column density. An AllWISE YSO census identifies 39 candidates across all three components, confirming active star formation throughout the complex. Multi-epoch NEOWISE-R photometry reveals three significantly variable sources, including one aperiodic dipper - a previously uncatalogued YSO exhibiting dimming consistent with inner-disc occultation. Two previously uncatalogued compact Bok globules are identified on the western edge of the northern extension, both detected in SPIRE continuum emission. That this complex went uncharacterized despite Herschel data being publicly available since 2013 underscores the scientific value of archival examination and the incomplete state of southern sky molecular cloud inventories.

The epoch of the Gaia-Sausage-Enceladus (GSE) merger, the last major merger experienced by the Milky Way, is crucial for reconstructing the Galaxy's evolutionary history. This event dynamically heated the disk, scattering some stars into the halo and producing the so-called Splash. Yet the observed age distribution of the Splash is systematically older than that of the thick disk from which it is thought to originate, posing a puzzling inconsistency. In this work, we show that this apparent discrepancy can be naturally explained once stellar age uncertainties are taken into account. The GSE truncated the intrinsic age distribution of the Splash at the merger epoch, while measurement errors introduce an Eddington-like bias that systematically shifts the truncated distribution toward older ages. Importantly, the magnitude of this shift depends on both the thick-disk star formation history and the merger epoch, thereby offering a new way to constrain the timing of the merger. By combining the observed Splash age distribution with the peak of thick-disk star formation, we infer a GSE merger epoch of $10.1^{+0.2}_{-0.2}$ Gyr ago, providing one of the tightest constraints to date. This result offers new insight into the Milky Way's accretion history and opens a path toward more robust reconstructions of its early evolution.

FRB~20240114A is an extremely active repeating fast radio burst for which plasma lensing has been proposed to explain its burst-rate variations, spectral evolution, and apparently ``carbon-copy'' burst pairs. Using FAST data and publicly available Parkes observations, we test this interpretation with a one-dimensional Gaussian plasma-lens model. Although the burst-rate enhancements can be fitted separately, the corresponding magnification peaks and demagnification troughs are offset by far more than predicted and show no consistent periodicity. Moreover, with more than 10,000 bursts detected, a few apparently ``carbon-copy'' pairs can readily occur by chance. The burst bandwidth is not systematically narrower during the proposed lensing interval, nor are the burst energies significantly enhanced during the predicted magnification interval. These results provide no compelling evidence that a single Gaussian plasma lens explains the observed variability, which is more likely dominated by intrinsic source activity.

Open clusters (OCs) are essential laboratories for probing stellar dynamics and tracing the structure and evolution of the Milky Way. Accurate measurements of their average radial velocities (RVs) and RV dispersions are crucial for estimating dynamical masses, orbital evolution, and overall kinematic states. Gaia Data Release 3 (DR3) provides an unprecedented volume of high-precision RVs. However, when applied to OCs, Gaia DR3 RVs often yield unusually large and overestimated RV dispersions. This inflation is primarily driven by RV measurement systematics for hot and faint stars, as well as unresolved binary contamination. To mitigate this, we revisit the average RVs and RV dispersions of OCs within 500 pc of the Sun using Gaia DR3. We evaluate the reliability of RV measurements and employ a color-based filtering method. By selecting member stars within an intrinsic color range of $0.2 \le (BP-RP)_{0} \le 1.2$ mag, we exclude hot and cool stars with less reliable RVs. This filtering significantly reduces the inferred RV dispersions while maintaining average RVs consistent with previous literature. Specifically, the median RV dispersion decreases by 26%, dropping from 3.76 km s$^{-1}$ prior to filtering to 2.79 km s$^{-1}$ afterward. These filtered RV dispersions remain systematically larger than tangential velocity dispersions, likely due to residual biases and undetected binaries. However, RV dispersions measured exclusively from red clump giants (found in 6 clusters) are remarkably small ($\lesssim 1.6$ km s$^{-1}$) and align closely with tangential dispersions ($\lesssim 1$ km s$^{-1}$). Ultimately, we provide a practical, Gaia-only strategy to derive more realistic RV dispersions for OCs, identifying red clump giants as exceptionally high-fidelity kinematic tracers for robust cluster studies.

Wolf--Rayet (WR) stars are evolved massive stars which can drive strong stellar winds, injecting energy and momentum into the interstellar medium (ISM). However, the geometry and kinematics of WR-dominated outflows, specially in low-metallicity environments, is still poorly constrained by observations. We present a spatially resolved spectroscopic study of a WR region in a nearby dwarf galaxy, PGC\,44685, using high-resolution MEGARA IFU data from the Gran Telescopio Canarias (GTC). After decomposing the [\textsc{O iii}]~$\lambda5007$ emission line with narrow and broad components, we verify a WR-driven outflow with a velocity reaching up to $20\,\mathrm{km\,s^{-1}}$ relative to the systemic velocity. By use of the velocity and flux of the [\textsc{O iii}] broad component, we estimate an outflow mass of $(8.25 \pm 3.03)\times10^{3}\,M_\odot$ and a mass-loss rate of $(9.47 \pm 3.48)\times10^{-4}\,M_\odot\,\mathrm{yr}^{-1}$. The corresponding kinetic power and momentum injection rate are $(4.77 \pm 1.77)\times10^{41}\,\mathrm{erg\,s^{-1}}$ and $(8.20 \pm 3.02)\times10^{28}\,\mathrm{g\,cm\,s^{-2}}$, respectively. The inferred low energy-loading efficiency ($\sim0.35\%$), together with the low metallicity of the WR region ($\sim0.1\,Z_\odot$), suggests that the system is observed in an early feedback phase in which stellar winds have not yet efficiently coupled their energy into the ISM. These results support the ability of WR feedback to shape the ISM on sub-kiloparsec scales, while these winds fail to launch galactic-scale outflows.

We present extensive optical and near-infrared (NIR) observations of the nearby Type II supernova (SN II) 2023ixf in the nebular phase from +89 days to +749 days after explosion, supplemented with NIR and mid-infrared (MIR) spectroscopy from the James Webb Space Telescope. The H$α$ emission profile shows complex evolution, with the emergence of high-velocity components consistent with the outer ejecta interacting with extended, low-density circumstellar material (CSM). We find that the H$α$ profile at an intermediate epoch (around +375 d) can be reconstructed by scaling an earlier decay-powered component and a later-phase shock-powered component, which revealed an additional intermediate-width component. This is consistent with the ejecta crashing into the initially aspherical dense CSM that has been swept-up by the forward shock. In the NIR, we find double-peaked emission from Mg I $1.504\ {\rm μm}$, Na I $2.206\ {\rm μm}$, and [Ni I] $3.12\ {\rm μm}$ between +200 d and +374 d, consistent with an asymmetric distribution of Ni-rich material that heats the ejecta inhomogeneously. We posit a disk-like CSM geometry and an ejecta geometry in which at least two large Ni-rich plumes lead to the observed line-profile diversity.

Feedback from active galactic nuclei (AGN) is now recognized as a key component of galaxy formation models. It plays a central role in regulating the growth and quenching of galaxies in the center of groups. However, the impact of AGN feedback from central galaxies on satellite galaxies remains largely unexplored. Here based on the largest sample to date of radio AGNs in galaxy groups (Yang et al. 2007) and a comprehensive consideration of multiple physical parameters that may influence the star formation of satellite galaxies, we demonstrate that the quiescent satellite fraction around radio AGNs is higher than that around normal galaxies. The most significant enhancement is observed around AGNs with large radio lobes. These findings demonstrate that the impact of kinetic AGN feedback beyond their host galaxies to their satellites. These results provide novel insights into the physical origins of some long-standing puzzles in extragalactic astronomy, including, e.g., galactic conformity and the strong small-scale clustering of quiescent galaxies.

Small-scale dynamos (SSDs) are believed to be the primary source of magnetic fields in all turbulent astrophysical systems, especially those with weak rotation such as elliptical galaxies and galaxy clusters. The initial kinematic phase of these dynamos is relatively well understood. Here we demonstrate analytically and numerically that, in an appropriate limit, the eddy-damped quasi-normal Markovian (EDQNM) closure for incompressible magnetohydrodynamic turbulence is strictly equivalent to the earlier models of kinematic dynamos. Moreover, it allows the extension of the kinematic dynamo framework to multi-scale turbulent flows and into the nonlinear regime. The EDQNM closure also enables us to explore a wide parameter range which is inaccessible to direct numerical simulations of the SSD. Using nonhelical EDQNM simulations, we identify several asymptotic regimes of nonlinear dynamo action when the system is highly turbulent with fluid Reynolds number $Re \gtrsim 10^6$ for magnetic Prandtl number $Pm > 1$ and magnetic Reynolds number $Rm \gtrsim 10^6$ for $Pm < 1$: 1) the kinematic growth rate approaches a value independent of $Pm$, 2) the saturated magnetic to kinetic energy ratio similarly converges to $\simeq 0.55$ across $Pm$, while the ratio of magnetic to kinetic integral wavenumbers asymptotes to $\simeq 3$. For all $Pm$, we further find strong feedback between magnetic field and velocity field largely via Alfvénisation leading to a saturated kinetic and magnetic spectra with almost the same inertial range with a slope of $-3/2$. These findings could provide guidance for future global simulations and for modeling the nonlinear regime of astrophysical systems living in these extreme limits.

We present the construction and characterization of a multi-wavelength catalog of galaxy groups and clusters by matching optical detections from the Southern Photometric Local Universe Survey (S-PLUS) with extended X-ray emission from the first eROSITA all-sky survey data release (eRASS1). We employ a probabilistic matching framework, based on the modified Hausdorff distance, to associate galaxy systems identified by the PZWav cluster finder and characterized by the AME membership estimator with X-ray surface brightness contours. This method explicitly accounts for the photometric redshift probability distribution of galaxies and allows us to explore the critical trade-off between catalog completeness and purity. We investigate how the matched sample changes with different optical selection depths, defined by absolute magnitude cuts of $M_r$ < -18.5, -19, -19.5, and -20 sampling redshifts within 0.08 < z < 0.25, and across purity levels of 80%, 90%, and 95%. We find that fainter optical cuts enhance the recovery of low-mass, low-luminosity groups, while brighter cuts favor more massive clusters and increase the effective survey volume at higher redshifts. Stricter purity requirements reduce contamination but systematically lower completeness, particularly for low-luminosity systems. The derived X-ray luminosity functions agree well with previous determinations, and the logN-logS distributions confirm the high recovery rate of luminous clusters. Comparisons with the redMaPPer cluster catalog validate our approach, showing consistent trends and significant overlap, while our method offers improved completeness at the group scale. This work demonstrates a robust, flexible methodology for creating reliable multi-wavelength cluster catalogs, essential for cosmological studies and investigations of galaxy evolution in dense environments.

Within the framework of the Southern Photometric Local Universe Survey (S-PLUS), we introduce ${\bf S}$-PLUS ${\bf C}$lusters ${\bf A}$nd ${\bf L}$arge-scale ${\bf E}$nvironments (SCALE), a project dedicated to the study of galaxy clusters, groups, and their environments using 12-band photometry of S-PLUS combined with spectroscopic and photometric data from the literature. In this first paper, we present a catalog of 83 previously known systems in the redshift range $0.008 \leq z_{\rm spec} \leq 0.1$, for which we derive $R_{200}$, $M_{200}$, and velocity dispersions. Spectroscopic members are selected and matched with S-PLUS photometric redshifts (photo-$z$s). We find very good agreement between literature spectroscopic redshifts (spec-$z$s) and S-PLUS photometric redshifts (photo-$z$s), demonstrating the potential of the latter for cluster and group membership determination. As a proof of concept, we obtain photometric memberships for Abell 4038 using the Reliable Photometric Membership technique. A two- and three-dimensional analysis of the region within $10 h^{-1}$ Mpc ($10\times R_{200}$) from the center of Abell 4038 reveals about a dozen substructures including two additional clusters within $1.3\times R_{200}$ (Abell 4038B and Abell 4049). A color-luminosity segregation analysis shows that more luminous (less luminous) galaxies are redder (bluer), as expected. Low-concentration galaxies ($C \leq 2.5$) exhibit a weaker color-luminosity dependence, compared to higher-concentration ones, indicating mass-dependent evolutionary pathways that challenge a simple morphology-color dichotomy, with low-luminosity galaxies presenting bluer colors largely independent of concentration. The SCALE catalog provides a valuable basis for future studies of large-scale structures and their connection to galaxy evolution.

We present the MIRI Early Obscured-AGN Wide Survey (MEOW), a JWST/MIRI imaging survey designed to detect dust-obscured active galactic nuclei (AGN) across cosmic time, with a particular focus on the high-redshift universe at $z \gtrsim 5$. MEOW observes the GOODS-N and GOODS-S fields with 43 pointings covering 95 arcmin$^2$ with the F1000W and F2100W filters, reaching depths of 0.5 and 3.6 $μ$Jy ($5σ$), respectively. Using spectral energy distribution (SED) modeling combining MEOW photometry with archival HST, JWST/NIRCam, and SCUBA-2 data, we identify a sample of 16 MIRI-selected AGN at $z = 4.5$--$7.2$ (12 spectroscopically confirmed), spanning bolometric luminosities of $L_{\rm bol} = 10^{44.6}$--$10^{46.4}$~erg~s$^{-1}$. Twelve of the 16 AGN are newly identified in this work, including at least five narrow-line AGN representing the obscured population to which broad-line spectroscopic searches are insensitive. Two broad-line AGN exhibit markedly different mid-infrared emission properties, consistent with one being a little red dot (LRD) and the other either a typical AGN or an LRD with unusually strong hot-dust emission. The MIRI-selected AGN bolometric luminosity function at $z = 4.5$--$6$ yields number densities comparable to those of broad-line AGN and LRDs, suggesting that obscured AGN contribute significantly to the total AGN census at these epochs. The narrow-line AGN reside in diverse host environments, with evidence for both circumnuclear and host-galaxy-scale obscuration, pointing to multiple physical mechanisms at work. These results establish JWST/MIRI imaging as an indispensable component of a multi-faceted approach to a complete census of early supermassive black hole growth.

Jets from active galactic nuclei (AGN) that exhibit regular plane-of-sky oscillations may provide a probe of magnetised plasma within the jets and their local intracluster medium (ICM). If such morphologies trace three-dimensional helices, path length differences between near and far sides of the flow might imprint quasi-periodic signatures in Faraday rotation. We present ASKAP Band 2 (1296-1440 MHz) polarimetric observations of a helical-tailed radio galaxy (the "Corkscrew Galaxy") from early science data from the POlarisation Sky Survey of the Universe's Magnetism (POSSUM), and test whether its quasi-periodic morphology produces signatures in Faraday rotation. We detect significant RM oscillations with a spatial period of (342 +/- 101) arcsec, consistent within the Rayleigh resolution limit with the jet's lateral deviations ((290 +/- 72) arcsec). Cross-correlation analysis reveals systematic variation along the jet: the eastern section shows alignment between jet deviation and RM, consistent with a jet-associated or sheath-like Faraday screen, while the western section exhibits a phase shift, indicating a transition in the dominant Faraday-rotating medium and possible contribution from the local ICM. These results demonstrate that quasi-periodic RM signatures can disentangle the dominant Faraday-rotating medium in AGN jets, and show that in some embedded radio galaxies the dominant RM contribution arises near the source rather than the cluster foreground. We also outline reliability criteria required to avoid false positive detections and inadequate sampling. Helical jet structures therefore show strong promise as probes of magnetised cluster environments, a capability that will expand in the SKA era.

We have made measurements of the difference between the position angle (PA) on the sky and the polarization position angle (PPA) of radio sources using data from a combination of the Radio Fundamental Catalogue (RFC) across a range of frequencies between 2.7 and 15~GHz and Cosmic Lens All Sky Survey (CLASS) which observes polarization at 8.4~GHz (X-band). For the 2111 sources with jet PAs measured in the X-band and a known redshift, the distribution is peaked at $\approx 0^{\circ}$ as expected for no birefringence and it can be modelled by two populations: one which is a Gaussian with mean $μ_β=(0.2\pm 1.0)^{\circ}$ and standard deviation $σ_β=(14.7\pm 1.1)^{\circ}$ and the other a uniform distribution of sources which are a fraction $f_β=0.72\pm 0.02$ of the total. Uncertainties in $μ_β$ can be reduced to $\approx 0.6^{\circ}$ by stacking measurements of the PA from other wavebands. We find that limits of $\approx 0.1^{\circ}$ might be possible with a sample of $\sim 10^5$ similarly selected sources and that this could provide a confirmation of recent claims of global birefringence made using the Cosmic Microwave Background observations from the {\it Planck} satellite.

Strong gravitational lensing provides a powerful probe of dark matter, particularly on small scales where the gravitational effects of dark matter subhalos within galaxies can manifest as perturbations within the extended arcs of gravitationally lensed sources. We investigate whether transforming lensed images into polar coordinates improves the ability of convolutional neural networks to infer subhalo mass. We introduce a machine learning architecture that outputs a prediction uncertainty alongside a mass prediction to enable assessment of network reliability. Using simulated Hubble Space Telescope observations, we compare our models trained on Cartesian and polar representations under different initialisation schemes, noise levels, and subhalo concentrations ($c=60$, $c=30$). We find that polar-transformed inputs consistently yield higher subhalo detection fractions than standard Cartesian images across all tested masses. For subhalos with mass $10^9M_\odot \leq M \leq 10^{9.5}M_\odot$, the fraction of subhalos the network is able to detect increases by $\sim 15$ per cent. Pretrained networks outperform randomly initialized networks, and the polar transform consistently improves network performance in both low signal-to-noise data and for lower-concentration subhalos. The relative improvement is highest in regimes where subhalo perturbations are most difficult to detect, such as low signal-to-noise data or systems containing low concentration subhalos. These results demonstrate that presenting strong lensing images in a polar representaion provides a computationally inexpensive way of improving CNN-based subhalo detection.

Context. Protoclusters of galaxies are the earliest phase in the assembly of galaxy clusters and can provide invaluable information about plasma physics, cosmic magnetism, and cosmology. However, due to small angular sizes and cosmological dimming, observing the proto-intracluster medium (proto-ICM) associated with protocluster cores is far from trivial. Aims. We aim to provide a non-parametric description of the gas mass and electron number density profiles of the proto-ICM at $z = 2$, and to study their dependence on mass, dynamical state and central activity. Methods. We extract and analyse over $3800$ regions around protocluster cores with spherical-overdensity masses above $M_\mathrm{500c} > 10^{13} \, M_\odot$ out of a large simulated volume within the Magneticum suite. We study their density profiles, temperature structure, ionisation degree and electron number density as a function of mass and other secondary properties characterising dynamical state and central activity, extending from the central halo to the surrounding protocluster environment. Results. Protoclusters present moderate deviations from self-similarity in their density profiles and temperature structure, with a strong double-$β$ structure especially relevant at high masses and intense AGN accretion. Hot, ionised gas is only dominant at intermediate radii ($r \gtrsim [0.1-0.5] R_\mathrm{500c}$), where its density also correlates with mass and dynamical disturbance. Conclusions. These results constitute the basis for a forthcoming parametric calibration of proto-ICM density profiles, which could be useful for interpreting observables sensitive to the density and ionisation of the diffuse gas.

The origin of synchrotron emission in compact radio sources associated with active galactic nuclei (AGN) remains poorly understood. In a series of papers, we have examined diagnostic tools to disentangle the dominant underlying processes. In this study, we investigate the in situ evolution of cosmic-ray electrons (CREs) in compact AGN jets and winds, and examine how their evolution shapes the resulting observable radio properties. In jets, CREs experience multiple shock interactions as they propagate along the spine toward the hotspot and flow into the cocoon via backflows. In winds, CREs are predominantly accelerated at the Mach disc, with occasional re-acceleration within turbulent cocoon backflows. The continuous mixing of different CRE populations within the cocoon produces observational signatures that cannot be inferred from instantaneous conditions alone. In all jet simulations, spectral indices are flattest near the hotspot and steepen progressively away from the hotspots. In winds, spectra steepen with increasing distance from the Mach disc, with this trend becoming more pronounced at high radio frequencies due to radiative losses. We find the Mach disc to be a significantly more efficient CRE acceleration site than the forward shock in winds, which weakens as the wind expands to large scales. Since morphology, especially at low resolution, can be ambiguous for compact sources, spatially resolved spectral indices, particularly when combined with emission and polarization signatures, can provide a powerful diagnostic.

In the last decade, the presence of active galactic nuclei (AGN) outflows and feedback in dwarf galaxies ($\mathrm{M_\ast}$<$10^{10}\mathrm{M}_\odot$) has gained ground over supernova (SN) feedback as the main mechanism regulating star formation. In this work, we perform the first systematic search for AGN outflows in dwarf galaxies using the Dark Energy Spectroscopic Instrument Data Release 1 (DESI DR1). From $\sim$ 7 million galaxies at z$<$0.45, we identify ionized outflows through the detection of broad components in the [OIII]$\lambda5007$Åemission line. Galaxies are divided into dwarf and massive systems. Then, using emission-line diagnostic diagrams, we classify as star forming or AGN. We identify 1,502 AGN dwarf galaxies with outflow signatures. Comparing the distributions of star forming and AGN galaxies with outflows, we find that, among the 1,502 AGN dwarf galaxies with outflow signatures, AGN are the most likely drivers of the observed outflows in $\sim$83$\%$ of those with W$_{80}$ velocity $>250$ km s$^{-1}$. This constitutes the largest statistical sample of AGN outflows in dwarf galaxies to date. In massive galaxies, AGN dominance occurs above W$_{80}>350$ km s$^{-1}$. Therefore, two new velocity thresholds are proposed for identifying AGN-driven outflows in dwarf and massive galaxies. Besides, we find that outflows in dwarf galaxies are more likely to escape the dark matter halo than those in massive galaxies, allowing gas to be redistributed from the inner to the outer regions. This suggests that AGN outflows may have a major impact on dwarf galaxies.

Free-Floating Planets (FFPs) are a distinct class of exoplanets that do not orbit stars but are nevertheless found to be very common. A variety of formation mechanisms have been proposed as their origin, such as "star-like" direct collapse from gas and dust clouds or ejection from young planetary systems via dynamical instabilities. Here, another possible formation scenario is explored that would instead apply for old planetary systems, in the form of White Dwarf (WD) formation kicks. Observations over recent years have shown that WDs experience a mild recoil kick during their formation from Asymptotic Giant Branch (AGB) stars. Here we show that, while WD formation kicks directly unbind only $\sim1\%$ of the known planet and exoplanet population, they drive dynamical instabilities in $\gtrsim40\%$ of known long-period multi-planet systems, likely leading to planet ejection and FFP generation in roughly half of all such systems. We also show that FFPs generated via WD kicks will additionally have undergone significant and long-lasting heating via their host stars' enhanced AGB luminosities. Given the low ejection velocities due to the weakness of the WD kicks, such warmed FFPs can thus be associated with their former host stars for several Myr after formation. Therefore, WD formation kicks contribute a distinct, observationally identifiable FFP sub-population comprising a few percent of the Galaxy's FFPs, relevant for the results of the upcoming {\it Roman} Galactic Exoplanet Survey.

Heavy moduli and Kaluza-Klein (KK) gravitons from extra dimensions may evade terrestrial probes but can be produced during inflation, generating primordial non-Gaussianity (NG) through unavoidable couplings to density perturbations. In a warped five-dimensional (5D) model, we compute the full radion- and KK-graviton-mediated bispectra and perform the first search for these signals using Planck 2018 temperature and polarization data. We find no significant evidence for NG, with the maximum significance being $1.8σ$ for $m_{\rm KK}\approx 1.6H$. We also identify a 5D setup which naturally generates NG with $f_{\rm NL}\sim 1-50$, within the reach of future surveys.

The James Webb Space Telescope (JWST) has opened a new window onto galaxy evolution in the very early Universe. In this work, we leverage halo merger trees extracted from the GUREFT dark-matter-only cosmological simulation suite together with the Santa Cruz semi-analytic model (SAM) for galaxy formation to investigate the predicted star formation histories (SFHs) of galaxies from cosmic dawn (z ~ 14) to the end of the Epoch of Reionization (EoR; z~6). While we find that on average, median SFHs of galaxies across all masses are uniformly and rapidly rising over time from 14 < z < 6 as expected, individual galaxy SFHs show a range of diverse SFHs, even for a fixed terminal mass or redshift, with bursts and mini-quenching episodes in agreement with SFHs inferred from observations. The median lookback time to form the youngest 50% (t_50) and 90% (t_90) of galaxies' stars decreases weakly with increasing stellar mass, and strongly with the redshift of observation. For galaxies at z>12, we find typical values of t_50 < 30 Myr and t_90 < 70 Myr, a factor of ~3 to 4 shorter than for comparable galaxies near the end of EoR (z ~ 6). The young-star dominated nature of stellar populations in ultra-high-z galaxies implies that careful modelling of young stellar populations is crucial for obtaining accurate synthetic photometry. In addition, our results have important implications for interpreting observational indicators of star formation histories and timescales.

The Earth's interior reflects its geological evolution, from accretion to present-day dynamics. Its structure drives the geodynamo in the outer core, generating the magnetic field that shields the surface from charged cosmic radiation. The primary observables of the Earth's interior are its radial density distribution and derived quantities such as its mass and moment of inertia. These have traditionally been inferred from gravity and seismic wave propagation, which probe the macroscopic response of matter to gravitational and elastic forces. Here we instead constrain the Earth's density profile using high-energy neutrinos observed by the IceCube Neutrino Observatory at the South Pole. We analyze 10.7 years of predominantly muon-neutrino data spanning 500 GeV--100 TeV, including atmospheric neutrinos produced by cosmic-ray interactions in the Earth's atmosphere and the diffuse astrophysical neutrino flux. Neutrino attenuation depends on both the traversed column density and neutrino energy. By measuring the zenith- and energy-dependent flux suppression, we infer the Earth's radial density profile by fitting a concentric uniform-density shell model that incorporates neutrino fluxes, interaction cross sections, detector response, and glacial-ice systematic uncertainties. From the resulting density posteriors, we derive the Earth's mass and polar moment of inertia as measured by neutrinos. These are the most precise weak-interaction measurements of these quantities to date and are consistent with the Preliminary Reference Earth Model and independent gravitational determinations. Our results demonstrate that neutrinos provide a novel probe of planetary interiors via a distinct physical interaction, complementing gravity and seismology. With improved detectors and precision, neutrinos will further contribute to a multifaceted understanding of the Earth's structure.

We propose a shock-based framework to interpret the radio and X-ray emission in active galactic nuclei (AGNs), whose origin remains an open problem. In this framework, the radio emission is produced by synchrotron radiation in the accretion flow or a jet/weak outflow, while the dominant X-ray component depends on the accretion state, the location of the non-thermal emission region, and the available seed photon field. The model provides a self-consistent interpretation of radio and X-ray emission in typical non-jetted AGNs, including low-luminosity AGNs, Seyfert galaxies, and radio-quiet quasars. In jetted AGNs, our results disfavor scenarios in which the non-thermal electrons responsible for X-ray emission are accelerated in the disk or the corona. We use two widely discussed empirical diagnostics, radio loudness and the fundamental plane (FP) of black hole (BH) activity, to assess the applicability and limitations of the model. It can naturally explain the observed trends that the radio loudness increases with BH mass and decreases with the Eddington ratio. The observed slope of the FP depends on how the key physical quantities scale with the accretion rate. As the accretion rate increases, the advection-dominated accretion flow region contracts while the thin disk region expands, reflecting a transition toward a more radiatively efficient accretion structure. The Eddington ratio therefore influences the accretion structure, and may in turn shape the observed AGN classes.

Ultra-cool dwarf stars (UCDs) are a common source of contamination in high-redshift galaxy searches as both sources are red and these early-forming galaxies can have sizes that are difficult to resolve even with space telescopes. Standard selection techniques, including photometric redshift estimation and color-color criteria, cannot fully eliminate this contamination. We develop \textbf{F}oreground \textbf{C}ontamination \textbf{E}valuator of \textbf{N}earby dwarf stars in high-\textbf{Z} photometrically selected \textbf{O}bjects (FC-ENZO), a code that predicts the number of dwarf stars misidentified as high-redshift galaxies for a given survey setup. FC-ENZO models the number of UCDs and evaluates the fraction of synthesized dwarf stars that passes user-specified selection methods. We compare two synthetic spectral energy distribution libraries and find that the ELF OWL library, which relaxes the assumption of chemical equilibrium, predicts larger contaminant fractions than the BOBCAT library, because of stronger absorption features around $ 1 $ \micron. The contamination fraction increases with metallicity and also depends on the adopted stellar number-density model. The dominant contaminants are T to early Y-type UCDs, which are most commonly misclassified as galaxies at $z \sim 8$. Comparing deep surveys from different space telescopes, we find similar overall contamination levels within the same redshift range. However, the contamination is concentrated near the limiting magnitude of each survey. At brighter magnitudes, the relative contamination is highest for HST (COSMOS), followed by Roman deep-tier survey, and JWST. Although the predicted contaminant numbers remain sensitive to model assumptions, FC-ENZO provides a practical tool for survey design and for identifying optimal fields for spectroscopic follow-up.