arXiv Daily Digest - 2026-03-05

The use of stochastic differential equations in multi-objective optimization has been limited, in practice, by two persistent gaps: incomplete stability analyses and the absence of accessible implementations. We revisit a drift--diffusion model for unconstrained vector optimization in which the drift is induced by a common descent direction and the diffusion term preserves exploratory behavior. The main theoretical contribution is a self-contained Lyapunov analysis establishing global existence, pathwise uniqueness, and non-explosion under a dissipativity condition, together with positive recurrence under an additional coercivity assumption. We also derive an Euler--Maruyama discretization and implement the resulting iteration as a \emph{pymoo}-compatible algorithm -- \emph{pymoo} being an open-source Python framework for multi-objective optimization -- with an interactive \emph{PymooLab} front-end for reproducible experiments. Empirical results on DTLZ2 with objective counts from three to fifteen indicate a consistent trade-off: compared with established evolutionary baselines, the method is less competitive in low-dimensional regimes but remains a viable option under restricted evaluation budgets in higher-dimensional settings. Taken together, these observations suggest that stochastic drift--diffusion search occupies a mathematically tractable niche alongside population-based heuristics -- not as a replacement, but as an alternative whose favorable properties are amenable to rigorous analysis.

Analog Ising machines have been proposed as heuristic hardware solvers for combinatorial optimization problems, with the potential to outperform conventional approaches, provided that their hyperparameters are carefully tuned. Their temporal evolution is often described using time-continuous dynamics. However, most experimental implementations rely on measurement-feedback architectures that operate in a time-discrete manner. We observe that in such setups, the range of effective hyperparameters is substantially smaller than in the envisioned time-continuous analog Ising machine. In this paper, we analyze this discrepancy and discuss its impact on the practical operation of Ising machines. Next, we propose and experimentally verify a method to reduce the sensitivity to hyperparameter selection of these measurement-feedback architectures.

Future collider experiments require unprecedented precision in measurements of Higgs, electroweak, and flavour observables, placing stringent demands on event reconstruction. The achievable precision on Higgs couplings scales directly with the resolution on visible final state particles and their invariant masses. Current particle flow algorithms rely on detector specific clustering, limiting flexibility during detector design. Here we present an end-to-end global event reconstruction approach that maps charged particle tracks and calorimeter and muon hits directly to particle level objects. The method combines geometric algebra transformer networks with object condensation based clustering, followed by dedicated networks for particle identification and energy regression. Our approach is benchmarked on fully simulated electron positron collisions at FCC-ee using the CLD detector concept. It outperforms the state-of-the-art rule-based algorithm by 10--20\% in relative reconstruction efficiency, achieves up to two orders of magnitude reduction in fake-particle rates for charged hadrons, and improves visible energy and invariant mass resolution by 22\%. By decoupling reconstruction performance from detector-specific tuning, this framework enables rapid iteration during the detector design phase of future collider experiments.

This paper investigates two complementary paradigms for predicting machine translation (MT) quality: source-side difficulty prediction and candidate-side quality estimation (QE). The rapid adoption of Large Language Models (LLMs) into MT workflows is reshaping the research landscape, yet its impact on established quality prediction paradigms remains underexplored. We study this issue through a series of "hindsight" experiments on a unique, multi-candidate dataset resulting from a genuine MT post-editing (MTPE) project. The dataset consists of over 6,000 English source segments with nine translation hypotheses from a diverse set of traditional neural MT systems and advanced LLMs, all evaluated against a single, final human post-edited reference. Using Kendall's rank correlation, we assess the predictive power of source-side difficulty metrics, candidate-side QE models and position heuristics against two gold-standard scores: TER (as a proxy for post-editing effort) and COMET (as a proxy for human judgment). Our findings highlight that the architectural shift towards LLMs alters the reliability of established quality prediction methods while simultaneously mitigating previous challenges in document-level translation.

Safety-critical scenario generation is crucial for evaluating autonomous driving systems. However, existing approaches often struggle to balance three conflicting objectives: adversarial criticality, physical feasibility, and behavioral realism. To bridge this gap, we propose SaFeR: safety-critical scenario generation for autonomous driving test via feasibility-constrained token resampling. We first formulate traffic generation as a discrete next token prediction problem, employing a Transformer-based model as a realism prior to capture naturalistic driving distributions. To capture complex interactions while effectively mitigating attention noise, we propose a novel differential attention mechanism within the realism prior. Building on this prior, SaFeR implements a novel resampling strategy that induces adversarial behaviors within a high-probability trust region to maintain naturalism, while enforcing a feasibility constraint derived from the Largest Feasible Region (LFR). By approximating the LFR via offline reinforcement learning, SaFeR effectively prevents the generation of theoretically inevitable collisions. Closed-loop experiments on the Waymo Open Motion Dataset and nuPlan demonstrate that SaFeR significantly outperforms state-of-the-art baselines, achieving a higher solution rate and superior kinematic realism while maintaining strong adversarial effectiveness.

Fine-tuned large language models can exhibit reward-hacking behavior arising from emergent misalignment, which is difficult to detect from final outputs alone. While prior work has studied reward hacking at the level of completed responses, it remains unclear whether such behavior can be identified during generation. We propose an activation-based monitoring approach that detects reward-hacking signals from internal representations as a model generates its response. Our method trains sparse autoencoders on residual stream activations and applies lightweight linear classifiers to produce token-level estimates of reward-hacking activity. Across multiple model families and fine-tuning mixtures, we find that internal activation patterns reliably distinguish reward-hacking from benign behavior, generalize to unseen mixed-policy adapters, and exhibit model-dependent temporal structure during chain-of-thought reasoning. Notably, reward-hacking signals often emerge early, persist throughout reasoning, and can be amplified by increased test-time compute in the form of chain-of-thought prompting under weakly specified reward objectives. These results suggest that internal activation monitoring provides a complementary and earlier signal of emergent misalignment than output-based evaluation, supporting more robust post-deployment safety monitoring for fine-tuned language models.

As text-to-image diffusion models become increasingly deployed in real-world applications, concerns about backdoor attacks have gained significant attention. Prior work on text-based backdoor attacks has largely focused on diffusion models conditioned on a single lightweight text encoder. However, more recent diffusion models that incorporate multiple large-scale text encoders remain underexplored in this context. Given the substantially increased number of trainable parameters introduced by multiple text encoders, an important question is whether backdoor attacks can remain both efficient and effective in such settings. In this work, we study Stable Diffusion 3, which uses three distinct text encoders and has not yet been systematically analyzed for text-encoder-based backdoor vulnerabilities. To understand the role of text encoders in backdoor attacks, we define four categories of attack targets and identify the minimal sets of encoders required to achieve effective performance for each attack objective. Based on this, we further propose Multi-Encoder Lightweight aTtacks (MELT), which trains only low-rank adapters while keeping the pretrained text encoder weight frozen. We demonstrate that tuning fewer than 0.2% of the total encoder parameters is sufficient for successful backdoor attacks on Stable Diffusion 3, revealing previously underexplored vulnerabilities in practical attack scenarios in multi-encoder settings.

Noisy labels in distributed datasets induce severe local overfitting and consequently compromise the global model in federated learning (FL). Most existing solutions rely on selecting clean devices or aligning with public clean datasets, rather than endowing the model itself with robustness. In this paper, we propose FedCova, a dependency-free federated covariance learning framework that eliminates such external reliances by enhancing the model's intrinsic robustness via a new perspective on feature covariances. Specifically, FedCova encodes data into a discriminative but resilient feature space to tolerate label noise. Built on mutual information maximization, we design a novel objective for federated lossy feature encoding that relies solely on class feature covariances with an error tolerance term. Leveraging feature subspaces characterized by covariances, we construct a subspace-augmented federated classifier. FedCova unifies three key processes through the covariance: (1) training the network for feature encoding, (2) constructing a classifier directly from the learned features, and (3) correcting noisy labels based on feature subspaces. We implement FedCova across both symmetric and asymmetric noisy settings under heterogeneous data distribution. Experimental results on CIFAR-10/100 and real-world noisy dataset Clothing1M demonstrate the superior robustness of FedCova compared with the state-of-the-art methods.

Quantum measurements are the means by which we recover messages encoded into quantum states. They are at the forefront of quantum hypothesis testing, wherein the goal is to perform an optimal measurement for arriving at a correct conclusion. Mathematically, a measurement operator is Hermitian with eigenvalues in [0,1]. By noticing that this constraint on each eigenvalue is the same as that imposed on fermions by the Pauli exclusion principle, we interpret every eigenmode of a measurement operator as an independent effective fermionic mode. Under this perspective, various objective functions in quantum hypothesis testing can be viewed as the total expected energy associated with these fermionic occupation numbers. By instead fixing a temperature and minimizing the total expected fermionic free energy, we find that optimal measurements for these modified objective functions are Fermi-Dirac thermal measurements, wherein their eigenvalues are specified by Fermi-Dirac distributions. In the low-temperature limit, their performance closely approximates that of optimal measurements for quantum hypothesis testing, and we show that their parameters can be learned by classical or hybrid quantum-classical optimization algorithms. This leads to a new quantum machine-learning model, termed Fermi-Dirac machines, consisting of parameterized Fermi-Dirac thermal measurements-an alternative to quantum Boltzmann machines based on thermal states. Beyond hypothesis testing, we show how general semidefinite optimization problems can be solved using this approach, leading to a novel paradigm for semidefinite optimization on quantum computers, in which the goal is to implement thermal measurements rather than prepare thermal states. Finally, we propose quantum algorithms for implementing Fermi-Dirac thermal measurements, and we also propose second-order hybrid quantum-classical optimization algorithms.

Autonomous navigation in congested maritime environments is a critical capability for a wide range of real-world applications. However, it remains an unresolved challenge due to complex vessel interactions and significant environmental uncertainties. Existing methods often fail in practical deployment due to a substantial sim-to-real gap, which stems from imprecise simulation, inadequate situational awareness, and unsafe exploration strategies. To address these, we propose \textbf{Sim2Sea}, a comprehensive framework designed to bridge simulation and real-world execution. Sim2Sea advances in three key aspects. First, we develop a GPU-accelerated parallel simulator for scalable and accurate maritime scenario simulation. Second, we design a dual-stream spatiotemporal policy that handles complex dynamics and multi-modal perception, augmented with a velocity-obstacle-guided action masking mechanism to ensure safe and efficient exploration. Finally, a targeted domain randomization scheme helps bridge the sim-to-real gap. Simulation results demonstrate that our method achieves faster convergence and safer trajectories than established baselines. In addition, our policy trained purely in simulation successfully transfers zero-shot to a 17-ton unmanned vessel operating in real-world congested waters. These results validate the effectiveness of Sim2Sea in achieving reliable sim-to-real transfer for practical autonomous maritime navigation.

Performance indicators are essential tools for assessing the convergence behavior of multi-objective optimization algorithms, particularly when the true Pareto front is unknown or difficult to approximate. Classical reference-based metrics such as hypervolume and inverted generational distance are widely used, but may suffer from scalability limitations and sensitivity to parameter choices in many-objective scenarios. Indicators derived from Karush--Kuhn--Tucker (KKT) optimality conditions provide an intrinsic alternative by quantifying stationarity without relying on external reference sets. This paper revisits an entropy-inspired KKT-based convergence indicator and proposes a robust adaptive reformulation based on quantile normalization. The proposed indicator preserves the stationarity-based interpretation of the original formulation while improving robustness to heterogeneous distributions of stationarity residuals, a recurring issue in many-objective optimization.

Protein language models (PLMs) are becoming practical tools for de novo protein design, yet their dual-use potential raises safety concerns. We show that domain adaptation to specific taxonomic groups can elicit toxic protein generation, even when toxicity is not the training objective. To address this, we adapt Logit Diff Amplification (LDA) as an inference-time control mechanism for PLMs. LDA modifies token probabilities by amplifying the logit difference between a baseline model and a toxicity-finetuned model, requiring no retraining. Across four taxonomic groups, LDA consistently reduces predicted toxicity rate (measured via ToxDL2) below the taxon-finetuned baseline while preserving biological plausibility. We evaluate quality using Fréchet ESM Distance and predicted foldability (pLDDT), finding that LDA maintains distributional similarity to natural proteins and structural viability (unlike activation-based steering methods that tend to degrade sequence properties). Our results demonstrate that LDA provides a practical safety knob for protein generators that mitigates elicited toxicity while retaining generative quality.

Composed image retrieval (CIR) addresses the task of retrieving a target image by jointly interpreting a reference image and a modification text that specifies the intended change. Most existing methods are still built upon contrastive learning frameworks that treat the ground truth image as the only positive instance and all remaining images as negatives. This strategy inevitably introduces relevance suppression, where semantically related yet valid images are incorrectly pushed away, and semantic confusion, where different modification intents collapse into overlapping regions of the embedding space. As a result, the learned query representations often lack discriminativeness, particularly at fine-grained attribute modifications. To overcome these limitations, we propose distinctive query embeddings through learnable attribute weights and target relative negative sampling (DQE-CIR), a method designed to learn distinctive query embeddings by explicitly modeling target relative relevance during training. DQE-CIR incorporates learnable attribute weighting to emphasize distinctive visual features conditioned on the modification text, enabling more precise feature alignment between language and vision. Furthermore, we introduce target relative negative sampling, which constructs a target relative similarity distribution and selects informative negatives from a mid-zone region that excludes both easy negatives and ambiguous false negatives. This strategy enables more reliable retrieval for fine-grained attribute changes by improving query discriminativeness and reducing confusion caused by semantically similar but irrelevant candidates.

mlx-vis is a Python library that implements six dimensionality reduction methods and a k-nearest neighbor graph algorithm entirely in MLX, Apple's array framework for Apple Silicon. The library provides UMAP, t-SNE, PaCMAP, TriMap, DREAMS, CNE, and NNDescent, all executing on Metal GPU through a unified fit_transform interface. Beyond embedding computation, mlx-vis includes a GPU-accelerated circle-splatting renderer that produces scatter plots and smooth animations without matplotlib, composing frames via scatter-add alpha blending on GPU and piping them to hardware H.264 encoding. On Fashion-MNIST with 70,000 points, all methods complete embedding in 2.1-3.8 seconds and render 800-frame animations in 1.4 seconds on an M3 Ultra, with the full pipeline from raw data to rendered video finishing in 3.6-5.2 seconds. The library depends only on MLX and NumPy, is released under the Apache 2.0 license, and is available at https://github.com/hanxiao/mlx-vis.

For four decades, AIED research has rested on what we term the Sedentary Assumption: the unexamined design commitment to a stationary learner seated before a screen. Mobile learning and museum guides have moved learners into physical space, and context-aware systems have delivered location-triggered content -- yet these efforts predominantly cast AI in the role of information-de-livery tool rather than epistemic partner. We map this gap through a 2 x 2 matrix (AI Role x Learning Environment) and identify an undertheorized intersection: the configuration in which AI serves as an epistemic teammate during unstruc-tured, place-bound field inquiry and learning is assessed through trajectory rather than product. To fill it, we propose Field Atlas, a framework grounded in embod-ied, embedded, enactive, and extended (4E) cognition, active inference, and dual coding theory that shifts AIED's guiding metaphor from instruction to sensemak-ing. The architecture pairs volitional photography with immediate voice reflec-tion, constrains AI to Socratic provocation rather than answer delivery, and ap-plies Epistemic Trajectory Modeling (ETM) to represent field learning as a con-tinuous trajectory through conjoined physical-epistemic space. We demonstrate the framework through a museum scenario and argue that the resulting trajecto-ries -- bound to a specific body, place, and time -- constitute process-based evi-dence structurally resistant to AI fabrication, offering a new assessment paradigm and reorienting AIED toward embodied, dialogic human-AI sensemaking in the wild.

Automatic evaluation of medical open-ended question answering (OEQA) remains challenging due to the need for expert annotations. We evaluate whether large language models (LLMs) can act as judges of semantic equivalence in French medical OEQA, comparing closed-access, general-purpose, and biomedical domain-adapted models. Our results show that LLM-based judgments are strongly influenced by the model that generated the answer, with agreement varying substantially across generators. Domain-adapted and large general-purpose models achieve the highest alignment with expert annotations. We further show that lightweight adaptation of a compact model using supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO) substantially improves performance and reduces generator sensitivity, even with limited data. Overall, our findings highlight the need for generator-aware evaluation and suggest that carefully adapted small models can support scalable evaluation in low-resource medical settings.

Music Source Restoration (MSR) targets recovery of original, unprocessed instrument stems from fully mixed and mastered audio, where production effects and distribution artifacts violate common linear-mixture assumptions. This technical report presents the CP-JKU team's system for the MSR ICASSP Challenge 2025. Our approach decomposes MSR into separation and restoration. First, a single BandSplit-RoFormer separator predicts eight stems plus an auxiliary other stem, and is trained with a three-stage curriculum that progresses from 4-stem warm-start fine-tuning (with LoRA) to 8-stem extension via head expansion. Second, we apply a HiFi++ GAN waveform restorer trained as a generalist and then specialized into eight instrument-specific experts.

As learning-based robotic controllers are typically trained offline and deployed with fixed parameters, their ability to cope with unforeseen changes during operation is limited. Biologically inspired, this work presents a framework for online Continual Reinforcement Learning that enables automated adaptation during deployment. Building on DreamerV3, a model-based Reinforcement Learning algorithm, the proposed method leverages world model prediction residuals to detect out-of-distribution events and automatically trigger finetuning. Adaptation progress is monitored using both task-level performance signals and internal training metrics, allowing convergence to be assessed without external supervision and domain knowledge. The approach is validated on a variety of contemporary continuous control problems, including a quadruped robot in high-fidelity simulation, and a real-world model vehicle. Relevant metrics and their interpretation are presented and discussed, as well as resulting trade-offs described. The results sketch out how autonomous robotic agents could once move beyond static training regimes toward adaptive systems capable of self-reflection and -improvement during operation, just like their biological counterparts.

Decentralized large language model (LLM) inference networks can pool heterogeneous compute to scale serving, but they require lightweight and incentive-compatible mechanisms to assess output quality. Prior work introduced cost-aware Proof of Quality (PoQ) and adaptive robust PoQ to allocate rewards under evaluator heterogeneity and adversarial behavior. In this paper, we focus on the quality signal itself and propose a multi-dimensional quality scoring framework that decomposes output quality into modular dimensions, including model and cost priors, structure quality, semantic quality, query-output alignment, and agreement/uncertainty. Using logged outputs from QA and summarization tasks, we systematically audit dimension reliability and show that seemingly reasonable dimensions can be task-dependent and even negatively correlated with reference quality without calibration. While the default composite underperforms a strong single semantic evaluator, ablations reveal that removing unreliable dimensions and re-normalizing weights yields a calibrated composite that matches or exceeds the best single- evaluator and consensus baselines. Finally, we integrate the composite score as a drop-in quality signal in PoQ and demonstrate complementary benefits with robust aggregation and adaptive trust weighting under adversarial evaluator attacks.

Configuring stream processing systems for efficient performance, especially in cloud-native deployments, is a challenging and largely manual task. We present an experiment-driven approach for automated configuration optimization that combines three phases: Latin Hypercube Sampling for initial exploration, Simulated Annealing for guided stochastic search, and Hill Climbing for local refinement. The workflow is integrated with the cloud-native Theodolite benchmarking framework, enabling automated experiment orchestration on Kubernetes and early termination of underperforming configurations. In an experimental evaluation with Kafka Streams and a Kubernetes-based cloud testbed, our approach identifies configurations that improve throughput by up to 23% over the default. The results indicate that Latin Hypercube Sampling with early termination and Simulated Annealing are particularly effective in navigating the configuration space, whereas additional fine-tuning via Hill Climbing yields limited benefits.

Equivocal 3D lesion segmentation exhibits high inter-observer variability. Conventional deterministic models ignore this aleatoric uncertainty, producing over-confident masks that obscure clinical risks. Conversely, while generative methods (e.g., standard diffusion) capture sample diversity, recovering complex topology from pure noise frequently leads to severe structural fractures and out-of-distribution anatomical hallucinations. To resolve this fidelity-diversity trade-off, we propose Volumetric Directional Diffusion (VDD). Unlike standard diffusion models that denoise isotropic Gaussian noise, VDD mathematically anchors the generative trajectory to a deterministic consensus prior. By restricting the generative search space to iteratively predict a 3D boundary residual field, VDD accurately explores the fine-grained geometric variations inherent in expert disagreements without risking topological collapse. Extensive validation on three multi-rater datasets (LIDC-IDRI, KiTS21, and ISBI 2015) demonstrates that VDD achieves state-of-the-art uncertainty quantification (significantly improving GED and CI) while remaining highly competitive in segmentation accuracy against deterministic upper bounds. Ultimately, VDD provides clinicians with anatomically coherent uncertainty maps, enabling safer decision-making and mitigating risks in downstream tasks (e.g., radiotherapy planning or surgical margin assessment).

We propose Fluid Logic, a paradigm in which modal logical reasoning, temporal, epistemic, doxastic, deontic, is lifted from discrete Kripke structures to continuous manifolds via Neural Stochastic Differential Equations (Neural SDEs). Each type of modal operator is backed by a dedicated Neural SDE, and nested formulas compose these SDEs in a single differentiable graph. A key instantiation is Logic-Informed Neural Networks (LINNs): analogous to Physics-Informed Neural Networks (PINNs), LINNs embed modal logical formulas such as ($\Box$ bounded) and ($\Diamond$ visits\_lobe) directly into the training loss, guiding neural networks to produce solutions that are structurally consistent with prescribed logical properties, without requiring knowledge of the governing equations. The resulting framework, Continuous Modal Logical Neural Networks (CMLNNs), yields several key properties: (i) stochastic diffusion prevents quantifier collapse ($\Box$ and $\Diamond$ differ), unlike deterministic ODEs; (ii) modal operators are entropic risk measures, sound with respect to risk-based semantics with explicit Monte Carlo concentration guarantees; (iii)SDE-induced accessibility provides structural correspondence with classical modal axioms; (iv) parameterizing accessibility through dynamics reduces memory from quadratic in world count to linear in parameters. Three case studies demonstrate that Fluid Logic and LINNs can guide neural networks to produce consistent solutions across diverse domains: epistemic/doxastic logic (multi-robot hallucination detection), temporal logic (recovering the Lorenz attractor geometry from logical constraints alone), and deontic logic (learning safe confinement dynamics from a logical specification).

In this paper we: (1) propose Lightweight C (LC), namely a core calculus that formalizes a proper subset of the ANSI C without preprocessor directives; (2) define Colored LC (CLC), namely LC endowed with ANSI C preprocessor directives; and (3) define a type system for CLC that guarantees that all programs to be generated by the C preprocessor are well-typed C programs. We believe that the simple formalization provided by CLC could be useful also for teaching purposes. Stefano Berardi spent most of his academic career at the Department of Computer Science of the University of Turin, where he conducts outstanding research on the logical foundations of computer science and on type-based program analyses. Over the years, he taught many courses, from BSc courses on programming with C to PhD courses on program analysis. Therefore, this paper fully falls within Stefano Berardi's research and teaching activities.

Aggregate Programming (AP) is a paradigm for programming the collective behaviour of sets of distributed devices, possibly situated at the network far edge, by relying on asynchronous proximity-based interactions. The eXchange Calculus (XC), a recently proposed foundational model for AP, is essentially a typed lambda calculus extended with an operator (the exchange operator) providing an implicit communication mechanism between neighbour devices. This paper provides a gentle introduction to XC and to its implementation as a C++ library, called FCPP. The FCPP library and toolchain has been mainly developed at the Department of Computer Science of the University of Turin, where Stefano Berardi spent most of his academic career conducting outstanding research about logical foundation of computer science and transmitting his passion for research to students and young researchers, often exploiting typed lambda calculi. An FCCP program is essentially a typed lambda term, and FCPP has been used to write code that has been deployed on devices at the far edge of the network, including rovers and (soon) Uncrewed Aerial Vehicles (UAVs); hence the title of the paper.

We study fixed budget constrained best-arm identification in grouped bandits, where each arm consists of multiple independent attributes with stochastic rewards. An arm is considered feasible only if all its attributes' means are above a given threshold. The aim is to find the feasible arm with the largest overall mean. We first derive a lower bound on the error probability for any algorithm on this setting. We then propose Feasibility Constrained Successive Rejects (FCSR), a novel algorithm that identifies the best arm while ensuring feasibility. We show it attains optimal dependence on problem parameters up to constant factors in the exponent. Empirically, FCSR outperforms natural baselines while preserving feasibility guarantees.

The rate-distortion-perception (RDP) tradeoff characterizes the fundamental limits of lossy compression by jointly considering bitrate, reconstruction fidelity, and perceptual quality. While recent neural compression methods have improved perceptual performance, they typically operate at fixed points on the RDP surface, requiring retraining to target different tradeoffs. In this work, we propose a training-free framework that leverages pre-trained diffusion models to traverse the entire RDP surface. Our approach integrates a reverse channel coding (RCC) module with a novel score-scaled probability flow ODE decoder. We theoretically prove that the proposed diffusion decoder is optimal for the distortion-perception tradeoff under AWGN observations and that the overall framework with the RCC module achieves the optimal RDP function in the Gaussian case. Empirical results across multiple datasets demonstrate the framework's flexibility and effectiveness in navigating the ternary RDP tradeoff using pre-trained diffusion models. Our results establish a practical and theoretically grounded approach to adaptive, perception-aware compression.

Reasoning segmentation increasingly employs reinforcement learning to generate explanatory reasoning chains that guide Multimodal Large Language Models. While these geometric rewards are primarily confined to guiding the final localization, they are incapable of discriminating whether the reasoning process remains anchored on the referred region or strays into irrelevant context. Lacking this discriminative guidance, the model's reasoning often devolves into unfocused and verbose chains that ultimately fail to disambiguate and perceive the target in complex scenes. This suggests a need to complement the RL objective with Discriminative Perception, an ability to actively distinguish a target from its context. To realize this, we propose DPAD to compel the model to generate a descriptive caption of the referred object, which is then used to explicitly discriminate by contrasting the caption's semantic relevance to the referred object against the wider context. By optimizing for this discriminative capability, the model is forced to focus on the unique attributes of the target, leading to a more converged and efficient reasoning chain. The descriptive caption also serves as an interpretability rationale that aligns with the segmentation. Experiments on the benchmarks confirm the validity of our approach, delivering substantial performance gains, with the cIoU on ReasonSeg increasing by 3.09% and the reasoning chain length decreasing by approximately 42%. Code is available at https://github.com/mrazhou/DPAD

Generative artificial intelligence (GenAI) tools are increasingly present in higher education, yet their adoption has been largely student-driven, requiring instructors to respond to technologies already embedded in classroom practices. While some faculty have embraced GenAI for pedagogical purposes such as content generation, assessment support, and curriculum design, others approach these tools with caution, citing concerns about student learning, assessment validity, and academic integrity. Understanding faculty perspectives is therefore essential for informing effective pedagogical strategies and institutional policies. In this paper, we present findings from a focus group study with 29 STEM faculty members at a large public university in the United States. We examine how faculty integrate GenAI into their courses, the benefits and challenges they perceive for student learning, and the institutional support they identify as necessary for effective and responsible adoption. Our findings highlight key patterns in how STEM faculty engage with GenAI, reflecting both active adoption and cautious use. Faculty described a range of pedagogical applications alongside concerns about student learning, assessment, and academic integrity. Overall, the results suggest that effective integration of GenAI in higher education requires rethinking assessment, pedagogy, and institutional governance in addition to technical adoption.

Offline model-based optimization (MBO) seeks to discover high-performing designs using only a fixed dataset of past evaluations. Most existing methods rely on learning a surrogate model via regression and implicitly assume that good predictive accuracy leads to good optimization performance. In this work, we challenge this assumption and study offline MBO from a learnability perspective. We argue that offline optimization is fundamentally a problem of ranking high-quality designs rather than accurate value prediction. Specifically, we introduce an optimization-oriented risk based on ranking between near-optimal and suboptimal designs, and develop a unified theoretical framework that connects surrogate learning to final optimization. We prove the theoretical advantages of ranking over regression, and identify distributional mismatch between the training data and near-optimal designs as the dominant error. Inspired by this, we design a distribution-aware ranking method to reduce this mismatch. Empirical results across various tasks show that our approach outperforms twenty existing methods, validating our theoretical findings. Additionally, both theoretical and empirical results reveal intrinsic limitations in offline MBO, showing a regime in which no offline method can avoid over-optimistic extrapolation.

Low-Rank Adaptation (LoRA) improves downstream performance by restricting task updates to a low-rank parameter subspace, yet how this limited capacity is allocated within a trained adapter remains unclear. Through a geometric and empirical study across multiple tasks and backbones, we find that trained LoRA updates often exhibit an inefficient spectrum: task effects concentrate in a small subset of singular directions, while many remaining components are neutral or detrimental, motivating post-hoc refinement within the learned subspace. We propose Spectral Surgery, a training-free refinement that decomposes a LoRA update with SVD, estimates per-component sensitivity using gradients on a small calibration set, and reweights singular values under a magnitude constraint while keeping the learned directions fixed. Across Llama-3.1-8B and Qwen3-8B on four benchmarks, Spectral Surgery yields consistent gains (up to +4.4 points on CommonsenseQA and +2.4 pass@1 on HumanEval) by adjusting only $\approx 1{,}000$ scalar coefficients. These results demonstrate that SVD-structured, low-cost parameter editing can serve as a practical route to improving trained LoRA adapters in a purely post-hoc manner.

Multi-head attention enables transformer models to represent multiple attention patterns simultaneously. Empirically, head specialization emerges in distinct stages during training, while many heads remain redundant and learn similar representations. We propose a theoretical model capturing this phenomenon, based on the multi-index and single-location regression frameworks. In the first part, we analyze the training dynamics of multi-head softmax attention under SGD, revealing an initial unspecialized phase followed by a multi-stage specialization phase in which different heads sequentially align with latent signal directions. In the second part, we study the impact of attention activation functions on performance. We show that softmax-1 significantly reduces noise from irrelevant heads. Finally, we introduce the Bayes-softmax attention, which achieves optimal prediction performance in this setting.

The automation of AI R&D (AIRDA) could have significant implications, but its extent and ultimate effects remain uncertain. We need empirical data to resolve these uncertainties, but existing data (primarily capability benchmarks) may not reflect real-world automation or capture its broader consequences, such as whether AIRDA accelerates capabilities more than safety progress or whether our ability to oversee AI R&D can keep pace with its acceleration. To address these gaps, this work proposes metrics to track the extent of AIRDA and its effects on AI progress and oversight. The metrics span dimensions such as capital share of AI R&D spending, researcher time allocation, and AI subversion incidents, and could help decision makers understand the potential consequences of AIRDA, implement appropriate safety measures, and maintain awareness of the pace of AI development. We recommend that companies and third parties (e.g. non-profit research organisations) start to track these metrics, and that governments support these efforts.

When visual evidence is ambiguous, vision models must decide whether to interpret face-like patterns as meaningful. Face pareidolia, the perception of faces in non-face objects, provides a controlled probe of this behavior. We introduce a representation-level diagnostic framework that analyzes detection, localization, uncertainty, and bias across class, difficulty, and emotion in face pareidolia images. Under a unified protocol, we evaluate six models spanning four representational regimes: vision-language models (VLMs; CLIP-B/32, CLIP-L/14, LLaVA-1.5-7B), pure vision classification (ViT), general object detection (YOLOv8), and face detection (RetinaFace). Our analysis reveals three mechanisms of interpretation under ambiguity. VLMs exhibit semantic overactivation, systematically pulling ambiguous non-human regions toward the Human concept, with LLaVA-1.5-7B producing the strongest and most confident over-calls, especially for negative emotions. ViT instead follows an uncertainty-as-abstention strategy, remaining diffuse yet largely unbiased. Detection-based models achieve low bias through conservative priors that suppress pareidolia responses even when localization is controlled. These results show that behavior under ambiguity is governed more by representational choices than score thresholds, and that uncertainty and bias are decoupled: low uncertainty can signal either safe suppression, as in detectors, or extreme over-interpretation, as in VLMs. Pareidolia therefore provides a compact diagnostic and a source of ambiguity-aware hard negatives for probing and improving the semantic robustness of vision-language systems. Code will be released upon publication.

Recent advances in MLLMs are reframing segmentation from fixed-category prediction to instruction-grounded localization. While reasoning based segmentation has progressed rapidly in natural scenes, remote sensing lacks a generalizable solution due to the prohibitive cost of reasoning-oriented data and domain-specific challenges like overhead viewpoints. We present GeoSeg, a zero-shot, training-free framework that bypasses the supervision bottleneck for reasoning-driven remote sensing segmentation. GeoSeg couples MLLM reasoning with precise localization via: (i) bias-aware coordinate refinement to correct systematic grounding shifts and (ii) a dual-route prompting mechanism to fuse semantic intent with fine-grained spatial cues. We also introduce GeoSeg-Bench, a diagnostic benchmark of 810 image--query pairs with hierarchical difficulty levels. Experiments show that GeoSeg consistently outperforms all baselines, with extensive ablations confirming the effectiveness and necessity of each component.

Exact inference in probabilistic First-Order Logic offers a promising yet computationally costly approach for regulating the behavior of autonomous agents in shared traffic spaces. While prior methods have combined logical and probabilistic data into decision-making frameworks, their application is often limited to pre-flight checks due to the complexity of reasoning across vast numbers of possible universes. In this work, we propose a reactive mission design framework that jointly considers uncertain environmental data and declarative, logical traffic regulations. By synthesizing Probabilistic Mission Design (ProMis) with reactive reasoning facilitated by Reactive Circuits (RC), we enable online, exact probabilistic inference over hybrid domains. Our approach leverages the Frequency of Change inherent in heterogeneous data streams to subdivide inference formulas into memoized, isolated tasks, ensuring that only the specific components affected by new sensor data are re-evaluated. In experiments involving both real-world vessel data and simulated drone traffic in dense urban scenarios, we demonstrate that our approach provides orders of magnitude in speedup over ProMis without reactive paradigms. This allows intelligent transportation systems, such as Unmanned Aircraft Systems (UAS), to actively assert safety and legal compliance during operations rather than relying solely on preparation procedures.

We present Phi-4-reasoning-vision-15B, a compact open-weight multimodal reasoning model, and share the motivations, design choices, experiments, and learnings that informed its development. Our goal is to contribute practical insight to the research community on building smaller, efficient multimodal reasoning models and to share the result of these learnings as an open-weight model that is good at common vision and language tasks and excels at scientific and mathematical reasoning and understanding user interfaces. Our contributions include demonstrating that careful architecture choices and rigorous data curation enable smaller, open-weight multimodal models to achieve competitive performance with significantly less training and inference-time compute and tokens. The most substantial improvements come from systematic filtering, error correction, and synthetic augmentation -- reinforcing that data quality remains the primary lever for model performance. Systematic ablations show that high-resolution, dynamic-resolution encoders yield consistent improvements, as accurate perception is a prerequisite for high-quality reasoning. Finally, a hybrid mix of reasoning and non-reasoning data with explicit mode tokens allows a single model to deliver fast direct answers for simpler tasks and chain-of-thought reasoning for complex problems.

Diffusion models achieve state-of-the-art image quality. However, sampling is costly at inference time because it requires a large number of function evaluations (NFEs). To reduce NFEs, classical ODE numerical methods have been adopted. Yet, the choice of prediction type and integration domain leads to different sampling behaviors. To address these issues, we introduce Dual-Solver, which generalizes multistep samplers through learnable parameters that continuously (i) interpolate among prediction types, (ii) select the integration domain, and (iii) adjust the residual terms. It retains the standard predictor-corrector structure while preserving second-order local accuracy. These parameters are learned via a classification-based objective using a frozen pretrained classifier (e.g., MobileNet or CLIP). For ImageNet class-conditional generation (DiT, GM-DiT) and text-to-image generation (SANA, PixArt-$α$), Dual-Solver improves FID and CLIP scores in the low-NFE regime ($3 \le$ NFE $\le 9$) across backbones.

Generative AI can convert uncertainty into authoritative-seeming verdicts, displacing the justificatory work on which democratic epistemic agency depends. As a corrective, I propose a Brouwer-inspired assertibility constraint for responsible AI: in high-stakes domains, systems may assert or deny claims only if they can provide a publicly inspectable and contestable certificate of entitlement; otherwise they must return "Undetermined". This constraint yields a three-status interface semantics (Asserted, Denied, Undetermined) that cleanly separates internal entitlement from public standing while connecting them via the certificate as a boundary object. It also produces a time-indexed entitlement profile that is stable under numerical refinement yet revisable as the public record changes. I operationalize the constraint through decision-layer gating of threshold and argmax outputs, using internal witnesses (e.g., sound bounds or separation margins) and an output contract with reason-coded abstentions. A design lemma shows that any total, certificate-sound binary interface already decides the deployed predicate on its declared scope, so "Undetermined" is not a tunable reject option but a mandatory status whenever no forcing witness is available. By making outputs answerable to challengeable warrants rather than confidence alone, the paper aims to preserve epistemic agency where automated speech enters public justification.

Generative artificial intelligence is increasingly being integrated into complex business workflows, fundamentally shifting the boundaries of managerial decision-making. However, the reliability of its strategic advice in ambiguous business contexts remains a critical knowledge gap. This study addresses this by comparing various models on ambiguity detection, evaluating how a systematic resolution process enhances response quality, and investigating their sycophantic behavior when presented with flawed directives. Using a novel four-dimensional business ambiguity taxonomy, we conducted a human-in-the-loop experiment across strategic, tactical, and operational scenarios. The resulting decisions were assessed with an "LLM-as-a-judge" framework on criteria including agreement, actionability, justification quality, and constraint adherence. Results reveal distinct performance capabilities. While models excel in detecting internal contradictions and contextual ambiguities, they struggle with structural linguistic nuances. Ambiguity resolution consistently increased response quality across all decision types, while sycophantic behavior analysis revealed distinct patterns depending on the model architecture. This study contributes to the bounded rationality literature by positioning GAI as a cognitive scaffold that can detect and resolve ambiguities managers might overlook, but whose own artificial limitations necessitate human management to ensure its reliability as a strategic partner.

We present \textbf{BLOCK}, an open-source bi-stage character-to-skin pipeline that generates pixel-perfect Minecraft skins from arbitrary character concepts. BLOCK decomposes the problem into (i) a \textbf{3D preview synthesis stage} driven by a large multimodal model (MLLM) with a carefully designed prompt-and-reference template, producing a consistent dual-panel (front/back) oblique-view Minecraft-style preview; and (ii) a \textbf{skin decoding stage} based on a fine-tuned FLUX.2 model that translates the preview into a skin atlas image. We further propose \textbf{EvolveLoRA}, a progressive LoRA curriculum (text-to-image $\rightarrow$ image-to-image $\rightarrow$ preview-to-skin) that initializes each phase from the previous adapter to improve stability and efficiency. BLOCK is released with all prompt templates and fine-tuned weights to support reproducible character-to-skin generation.

Dynamic link prediction plays a crucial role in diverse applications including social network analysis, communication forecasting, and financial modeling. While recent Transformer-based approaches have demonstrated promising results in temporal graph learning, their performance remains limited when capturing complex multi-scale temporal dynamics. In this paper, we propose TFWaveFormer, a novel Transformer architecture that integrates temporal-frequency analysis with multi-resolution wavelet decomposition to enhance dynamic link prediction. Our framework comprises three key components: (i) a temporal-frequency coordination mechanism that jointly models temporal and spectral representations, (ii) a learnable multi-resolution wavelet decomposition module that adaptively extracts multi-scale temporal patterns through parallel convolutions, replacing traditional iterative wavelet transforms, and (iii) a hybrid Transformer module that effectively fuses local wavelet features with global temporal dependencies. Extensive experiments on benchmark datasets demonstrate that TFWaveFormer achieves state-of-the-art performance, outperforming existing Transformer-based and hybrid models by significant margins across multiple metrics. The superior performance of TFWaveFormer validates the effectiveness of combining temporal-frequency analysis with wavelet decomposition in capturing complex temporal dynamics for dynamic link prediction tasks.

Code comment classification is a critical task for automated software documentation and analysis. In the context of the NLBSE'26 Tool Competition, we present \textbf{LoRA-MME}, a Multi-Model Ensemble architecture utilizing Parameter-Efficient Fine-Tuning (PEFT). Our approach addresses the multi-label classification challenge across Java, Python, and Pharo by combining the strengths of four distinct transformer encoders: UniXcoder, CodeBERT, GraphCodeBERT, and CodeBERTa. By independently fine-tuning these models using Low-Rank Adaptation(LoRA) and aggregating their predictions via a learned weighted ensemble strategy, we maximize classification performance without the memory overhead of full model fine-tuning. Our tool achieved an \textbf{F1 Weighted score of 0.7906} and a \textbf{Macro F1 of 0.6867} on the test set. However, the computational cost of the ensemble resulted in a final submission score of 41.20\%, highlighting the trade-off between semantic accuracy and inference efficiency.

Supervised and unsupervised homography estimation methods depend on image pairs tailored to specific modalities to achieve high accuracy. However, their performance deteriorates substantially when applied to unseen modalities. To address this issue, we propose a training data synthesis method that generates unaligned image pairs with ground-truth offsets from a single input image. Our approach renders the image pairs with diverse textures and colors while preserving their structural information. These synthetic data empower the trained model to achieve greater robustness and improved generalization across various domains. Additionally, we design a network to fully leverage cross-scale information and decouple color information from feature representations, thus improving estimation accuracy. Extensive experiments show that our training data synthesis method improves generalization performance. The results also confirm the effectiveness of the proposed network.

Post-training with reinforcement learning (RL) has recently shown strong promise for advancing multimodal agents beyond supervised imitation. However, RL remains limited by poor data efficiency, particularly in settings where interaction data are scarce and quickly become outdated. To address this challenge, GIPO (Gaussian Importance sampling Policy Optimization) is proposed as a policy optimization objective based on truncated importance sampling, replacing hard clipping with a log-ratio-based Gaussian trust weight to softly damp extreme importance ratios while maintaining non-zero gradients. Theoretical analysis shows that GIPO introduces an implicit, tunable constraint on the update magnitude, while concentration bounds guarantee robustness and stability under finite-sample estimation. Experimental results show that GIPO achieves state-of-the-art performance among clipping-based baselines across a wide range of replay buffer sizes, from near on-policy to highly stale data, while exhibiting superior bias--variance trade-off, high training stability and improved sample efficiency.

Safe visual navigation is critical for indoor mobile robots operating in cluttered environments. Existing benchmarks, however, often neglect collisions or are designed for outdoor scenarios, making them unsuitable for indoor visual navigation. To address this limitation, we introduce the reactive visual navigation benchmark (RVN-Bench), a collision-aware benchmark for indoor mobile robots. In RVN-Bench, an agent must reach sequential goal positions in previously unseen environments using only visual observations and no prior map, while avoiding collisions. Built on the Habitat 2.0 simulator and leveraging high-fidelity HM3D scenes, RVN-Bench provides large-scale, diverse indoor environments, defines a collision-aware navigation task and evaluation metrics, and offers tools for standardized training and benchmarking. RVN-Bench supports both online and offline learning by offering an environment for online reinforcement learning, a trajectory image dataset generator, and tools for producing negative trajectory image datasets that capture collision events. Experiments show that policies trained on RVN-Bench generalize effectively to unseen environments, demonstrating its value as a standardized benchmark for safe and robust visual navigation. Code and additional materials are available at: https://rvn-bench.github.io/.

Generative models hold great promise for accelerating material discovery but are often limited by their inflexible single-stage generative process in designing valid and diverse materials. To address this, we propose a two-stage generative framework, Lang2Str, that combines the strengths of large language models (LLMs) and flow-based models for flexible and precise material generation. Our method frames the generative process as a conditional generative task, where an LLM provides high-level conditions by generating descriptions of material unit cells' geometric layouts and properties. These descriptions, informed by the LLM's extensive background knowledge, ensure reasonable structure designs. A conditioned flow model then decodes these textual conditions into precise continuous coordinates and unit cell parameters. This staged approach combines the structured reasoning of LLMs and the distribution modeling capabilities of flow models. Experimental results show that our method achieves competitive performance on \textit{ab initio} material generation and crystal structure prediction tasks, with generated structures exhibiting closer alignment to ground truth in both geometry and energy levels, surpassing state-of-the-art models. The flexibility and modularity of our framework further enable fine-grained control over the generation process, potentially leading to more efficient and customizable material design.

Link prediction models are increasingly used to recommend interactions in evolving networks, yet their impact on network structure is typically assessed from static snapshots. In particular, observed homophily conflates intrinsic interaction tendencies with amplification effects induced by network dynamics and algorithmic feedback. We propose a temporal framework based on multivariate Hawkes processes that disentangles these two sources and introduce an instantaneous bias measure derived from interaction intensities, capturing current reinforcement dynamics beyond cumulative metrics. We provide a theoretical characterization of the stability and convergence of the induced dynamics, and experiments show that the proposed measure reliably reflects algorithmic feedback effects across different link prediction strategies.

Multimodal industrial anomaly detection benefits from integrating RGB appearance with 3D surface geometry, yet existing \emph{unsupervised} approaches commonly rely on memory banks, teacher-student architectures, or fragile fusion schemes, limiting robustness under noisy depth, weak texture, or missing modalities. This paper introduces \textbf{CMDR-IAD}, a lightweight and modality-flexible unsupervised framework for reliable anomaly detection in 2D+3D multimodal as well as single-modality (2D-only or 3D-only) settings. \textbf{CMDR-IAD} combines bidirectional 2D$\leftrightarrow$3D cross-modal mapping to model appearance-geometry consistency with dual-branch reconstruction that independently captures normal texture and geometric structure. A two-part fusion strategy integrates these cues: a reliability-gated mapping anomaly highlights spatially consistent texture-geometry discrepancies, while a confidence-weighted reconstruction anomaly adaptively balances appearance and geometric deviations, yielding stable and precise anomaly localization even in depth-sparse or low-texture regions. On the MVTec 3D-AD benchmark, CMDR-IAD achieves state-of-the-art performance while operating without memory banks, reaching 97.3\% image-level AUROC (I-AUROC), 99.6\% pixel-level AUROC (P-AUROC), and 97.6\% AUPRO. On a real-world polyurethane cutting dataset, the 3D-only variant attains 92.6\% I-AUROC and 92.5\% P-AUROC, demonstrating strong effectiveness under practical industrial conditions. These results highlight the framework's robustness, modality flexibility, and the effectiveness of the proposed fusion strategies for industrial visual inspection. Our source code is available at https://github.com/ECGAI-Research/CMDR-IAD/

Offline Reinforcement Learning (RL) is a promising approach for next-generation wireless networks, where online exploration is unsafe and large amounts of operational data can be reused across the model lifecycle. However, the behavior of offline RL algorithms under genuinely stochastic dynamics -- inherent to wireless systems due to fading, noise, and traffic mobility -- remains insufficiently understood. We address this gap by evaluating Bellman-based (Conservative Q-Learning), sequence-based (Decision Transformers), and hybrid (Critic-Guided Decision Transformers) offline RL methods in an open-access stochastic telecom environment (mobile-env). Our results show that Conservative Q-Learning consistently produces more robust policies across different sources of stochasticity, making it a reliable default choice in lifecycle-driven AI management frameworks. Sequence-based methods remain competitive and can outperform Bellman-based approaches when sufficient high-return trajectories are available. These findings provide practical guidance for offline RL algorithm selection in AI-driven network control pipelines, such as O-RAN and future 6G functions, where robustness and data availability are key operational constraints.

Inverse problems are the task of calibrating models to match data. They play a pivotal role in diverse engineering applications by allowing practitioners to align models with reality. In many applications, engineers and scientists do not have a complete picture of i) the detailed properties of a system (such as material properties, geometry, initial conditions, etc.); ii) the complete laws describing all dynamics at play (such as friction laws, complicated damping phenomena, and general nonlinear interactions). In this paper, we develop a principled methodology for leveraging data from collections of distinct yet related physical systems to jointly estimate the individual model parameters of each system, and learn the shared unknown dynamics in the form of an ML-based closure model. To robustly infer the unknown parameters for each system, we employ a hierarchical Bayesian framework, which allows for the joint inference of multiple systems and their population-level statistics. To learn the closures, we use a maximum marginal likelihood estimate of a neural network embeded within the ODE/PDE formulation of the problem. To realize this framework we utilize the ensemble Metropolis-Adjusted Langevin Algorithm (MALA) for stable and efficient sampling. To mitigate the computational bottleneck of repetitive forward evaluations in solving inverse problems, we introduce a bilevel optimization strategy to simultaneously train a surrogate forward model alongside the inference. Within this framework, we evaluate and compare distinct surrogate architectures, specifically Fourier Neural Operators (FNO) and parametric Physics-Informed Neural Network (PINNs).

Model Merging (MM) has emerged as a scalable paradigm for multi-task learning (MTL), enabling multiple task-specific models to be integrated without revisiting the original training data. Despite recent progress, the reliability of MM under test-time distribution shift remains insufficiently understood. Most existing MM methods typically assume that test data are clean and distributionally aligned with both the training and auxiliary sources. However, this assumption rarely holds in practice, often resulting in biased predictions with degraded generalization. To address this issue, we present BD-Merging, a bias-aware unsupervised model merging framework that explicitly models uncertainty to achieve adaptive reliability under distribution shift. First, BD-Merging introduces a joint evidential head that learns uncertainty over a unified label space, capturing cross-task semantic dependencies in MM. Second, building upon this evidential foundation, we propose an Adjacency Discrepancy Score (ADS) that quantifies evidential alignment among neighboring samples. Third, guided by ADS, a discrepancy-aware contrastive learning mechanism refines the merged representation by aligning consistent samples and separating conflicting ones. Combined with general unsupervised learning, this process trains a debiased router that adaptively allocates task-specific or layer-specific weights on a per-sample basis, effectively mitigating the adverse effects of distribution shift. Extensive experiments across diverse tasks demonstrate that BD-Merging achieves superior effectiveness and robustness compared to state-of-the-art MM baselines.

Large language models (LLMs) have demonstrated significant potential in developing Role-Playing Agents (RPAs). However, current research primarily evaluates RPAs using famous fictional characters, allowing models to rely on memory associated with character names. This dependency creates a bias that limits the generalization of RPAs to unseen personas. To address this issue, we propose an anonymous evaluation method. Experiments across multiple benchmarks reveal that anonymization significantly degrades role-playing performance, confirming that name exposure carries implicit information. Furthermore, we investigate personality augmentation to enhance role fidelity under anonymous setting. We systematically compare the efficacy of personality traits derived from human annotations versus those self-generated by the model. Our results demonstrate that incorporating personality information consistently improves RPA performance. Crucially, self-generated personalities achieve performance comparable to human-annotated ones. This work establishes a fairer evaluation protocol and validates a scalable, personality-enhanced framework for constructing robust RPAs.

Web security demands rapid response capabilities to evolving cyber threats. Agentic Artificial Intelligence (AI) promises automation, but the need for trustworthy security responses is of the utmost importance. This work investigates the role of semantic relations in extracting information for sensitive operational tasks, such as configuring security controls for mitigating threats. To this end, it proposes to leverage hypernym-hyponym textual relations to extract relevant information from Cyber Threat Intelligence (CTI) reports. By leveraging a neuro-symbolic approach, the multi-agent system automatically generates CLIPS code for an expert system creating firewall rules to block malicious network traffic. Experimental results show the superior performance of the hypernym-hyponym retrieval strategy compared to various baselines and the higher effectiveness of the agentic approach in mitigating threats.

Building reliable classifiers is a fundamental challenge for deploying machine learning in real-world applications. A reliable system should not only detect out-of-distribution (OOD) inputs but also anticipate in-distribution (ID) errors by assigning low confidence to potentially misclassified samples. Yet, most prior work treats OOD detection and failure prediction as separated problems, overlooking their closed connection. We argue that reliability requires evaluating them jointly. To this end, we propose a unified evaluation framework that integrates OOD detection and failure prediction, quantified by our new metrics DS-F1 and DS-AURC, where DS denotes double scoring functions. Experiments on the OpenOOD benchmark show that double scoring functions yield classifiers that are substantially more reliable than traditional single scoring approaches. Our analysis further reveals that OOD-based approaches provide notable gains under simple or far-OOD shifts, but only marginal benefits under more challenging near-OOD conditions. Beyond evaluation, we extend the reliable classifier SURE and introduce SURE+, a new approach that significantly improves reliability across diverse scenarios. Together, our framework, metrics, and method establish a new benchmark for trustworthy classification and offer practical guidance for deploying robust models in real-world settings. The source code is publicly available at https://github.com/Intellindust-AI-Lab/SUREPlus.

Time series forecasting, which predicts future values from past observations, plays a central role in many domains and has driven the development of highly accurate neural network models. However, the complexity of these models often limits human understanding of the rationale behind their predictions. We propose PatchDecomp, a neural network-based time series forecasting method that achieves both high accuracy and interpretability. PatchDecomp divides input time series into subsequences (patches) and generates predictions by aggregating the contributions of each patch. This enables clear attribution of each patch, including those from exogenous variables, to the final prediction. Experiments on multiple benchmark datasets demonstrate that PatchDecomp provides predictive performance comparable to recent forecasting methods. Furthermore, we show that the model's explanations not only influence predicted values quantitatively but also offer qualitative interpretability through visualization of patch-wise contributions.

For an offline-first collaborative application to operate in true peer-to-peer fashion, its collaborative features must function even in environments where internet connectivity is limited or unavailable. Each peer may only be interested in a subset of the application data relevant to its workload, and this subset can overlap in different ways with those of other peers. Limitations imposed by access control and mesh network technologies often result in peers being sparsely connected. Reasoning about consistency in these systems is hard, especially when considering transactional updates that may alter different sets of data in the same transaction. We present \textsc{IntersectionAtomicity} and \textsc{IntersectionCC} as models to reason about offline-first collaborative applications that are sparsely-connected and rely on partially replicating different subsets of a broader set of data. We then use these models to propose a set of guidelines to help developers design their application with atomicity and consistency guarantees.

Foundation models have demonstrated impressive capabilities across diverse domains, while imitation learning provides principled methods for robot skill adaptation from limited data. Combining these approaches holds significant promise for direct application to robotics, yet this combination has received limited attention, particularly for industrial deployment. We present a novel framework that enables open-vocabulary skill adaptation through a tool-based architecture, maintaining a protective abstraction layer between the language model and robot hardware. Our approach leverages pre-trained LLMs to select and parameterize specific tools for adapting robot skills without requiring fine-tuning or direct model-to-robot interaction. We demonstrate the framework on a 7-DoF torque-controlled robot performing an industrial bearing ring insertion task, showing successful skill adaptation through natural language commands for speed adjustment, trajectory correction, and obstacle avoidance while maintaining safety, transparency, and interpretability.

In this paper, we present a network structure for classifying metadata of cuneiform tablets. The problem is of practical importance, as the size of the existing corpus far exceeds the number of experts available to analyze it. But the task is made difficult by the combination of limited annotated datasets and the high-resolution point-cloud representation of each tablet. To address this, we develop a convolution-inspired architecture that gradually down-scales the point cloud while integrating local neighbor information. The final down-scaled point cloud is then processed by computing neighbors in the feature space to include global information. Our method is compared with the state-of-the-art transformer-based network Point-BERT, and consistently obtains the best performance. Source code and datasets will be released at publication.

Topic localization aims to identify spans of text that express a given topic defined by a name and description. To study this task, we introduce a human-annotated benchmark based on Czech historical documents, containing human-defined topics together with manually annotated spans and supporting evaluation at both document and word levels. Evaluation is performed relative to human agreement rather than a single reference annotation. We evaluate a diverse range of large language models alongside BERT-based models fine-tuned on a distilled development dataset. Results reveal substantial variability among LLMs, with performance ranging from near-human topic detection to pronounced failures in span localization. While the strongest models approach human agreement, the distilled token embedding models remain competitive despite their smaller scale. The dataset and evaluation framework are publicly available at: https://github.com/dcgm/czechtopic.

As LLM-driven agents begin to autonomously navigate the web, their ability to interpret and respond to manipulative interface design becomes critical. A fundamental question that emerges is: can such agents reliably recognize patterns of friction, misdirection, and coercion in interface design (i.e., dark patterns)? We study this question in a setting where the workflows are consequential: website portals associated with the submission of CCPA-related data rights requests. These portals operationalize statutory rights, but they are implemented as interactive interfaces whose design can be structured to facilitate, burden, or subtly discourage the exercise of those rights. We design and deploy an LLM-driven auditing agent capable of end-to-end traversal of rights-request workflows, structured evidence gathering, and classification of potential dark patterns. Across a set of 456 data broker websites, we evaluate: (1) the ability of the agent to consistently locate and complete request flows, (2) the reliability and reproducibility of its dark pattern classifications, and (3) the conditions under which it fails or produces poor judgments. Our findings characterize both the feasibility and the limitations of using LLM-driven agents for scalable dark pattern auditing.

Software-hardware co-design is essential for optimizing in-memory computing (IMC) hardware accelerators for neural networks. However, most existing optimization frameworks target a single workload, leading to highly specialized hardware designs that do not generalize well across models and applications. In contrast, practical deployment scenarios require a single IMC platform that can efficiently support multiple neural network workloads. This work presents a joint hardware-workload co-optimization framework based on an optimized evolutionary algorithm for designing generalized IMC accelerator architectures. By explicitly capturing cross-workload trade-offs rather than optimizing for a single model, the proposed approach significantly reduces the performance gap between workload-specific and generalized IMC designs. The framework is evaluated on both RRAM- and SRAM-based IMC architectures, demonstrating strong robustness and adaptability across diverse design scenarios. Compared to baseline methods, the optimized designs achieve energy-delay-area product (EDAP) reductions of up to 76.2% and 95.5% when optimizing across a small set (4 workloads) and a large set (9 workloads), respectively. The source code of the framework is available at https://github.com/OlgaKrestinskaya/JointHardwareWorkloadOptimizationIMC.

The exponential growth of AI has created unprecedented demand for computational resources, pushing chip designs to the limit while simultaneously escalating the environmental footprint of computing. As the industry transitions toward heterogeneous integration (HI) to address the yield and cost challenges of monolithic scaling, minimizing the carbon cost of these complex HI systems becomes critical. To fully exploit HI, a co-design approach spanning application, architecture, chip, and packaging is essential. However, this creates a vast design space with competing objectives, specifically the trade-offs between performance, cost, and carbon footprint (CFP) for sustainability. CarbonPATH is an early-stage pathfinding framework designed to address this multi-objective challenge. It identifies optimized HI systems by co-designing workload mapping, architectural parameters, and packaging technologies, while treating sustainability as a first-class design constraint. The framework accounts for a wide range of factors, including compute and memory sizes, chiplet technology nodes, communication protocols, integration style (2D, 2.5D, 3D), operational CFP, embodied CFP, and interconnect type. Using simulated annealing, CarbonPATH explores this high-dimensional space to identify solutions that balance traditional metrics against environmental impact. By capturing interactions across applications, architectures, chiplets, and packaging, CarbonPATH uncovers system-level solutions that traditional methods often miss due to restrictive assumptions or limited scope.

Empirical research in reverse engineering and software protection is crucial for evaluating the efficacy of methods designed to protect software against unauthorized access and tampering. However, conducting such studies with professional reverse engineers presents significant challenges, including access to professionals and affordability. This paper explores the use of students as participants in empirical reverse engineering experiments, examining their suitability and the necessary training; the design of appropriate challenges; strategies for ensuring the rigor and validity of the research and its results; ways to maintain students' privacy, motivation, and voluntary participation; and data collection methods. We present a systematic literature review of existing reverse engineering experiments and user studies, a discussion of related work from the broader domain of software engineering that applies to reverse engineering experiments, an extensive discussion of our own experience running experiments ourselves in the context of a master-level software hacking and protection course, and recommendations based on this experience. Our findings aim to guide future empirical studies in RE, balancing practical constraints with the need for meaningful, reproducible results.

Large Reasoning Models have demonstrated remarkable performance with the advancement of test-time scaling techniques, which enhances prediction accuracy by generating multiple candidate responses and selecting the most reliable answer. While prior work has analyzed that internal model signals like confidence scores can partly indicate response correctness and exhibit a distributional correlation with accuracy, such distributional information has not been fully utilized to guide answer selection. Motivated by this, we propose DistriVoting, which incorporates distributional priors as another signal alongside confidence during voting. Specifically, our method (1) first decomposes the mixed confidence distribution into positive and negative components using Gaussian Mixture Models, (2) then applies a reject filter based on positive/negative samples from them to mitigate overlap between the two distributions. Besides, to further alleviate the overlap from the perspective of distribution itself, we propose SelfStepConf, which uses step-level confidence to dynamically adjust inference process, increasing the separation between the two distributions to improve the reliability of confidences in voting. Experiments across 16 models and 5 benchmarks demonstrate that our method significantly outperforms state-of-the-art approaches.

Link prediction (LP) plays a central role in graph-based applications, particularly in social recommendation. However, real-world graphs often reflect structural biases, most notably homophily, the tendency of nodes with similar attributes to connect. While this property can improve predictive performance, it also risks reinforcing existing social disparities. In response, fairness-aware LP methods have emerged, often seeking to mitigate these effects by promoting inter-group connections, that is, links between nodes with differing sensitive attributes (e.g., gender), following the principle of dyadic fairness. However, dyadic fairness overlooks potential disparities within the sensitive groups themselves. To overcome this issue, we propose $k$-hop fairness, a structural notion of fairness for LP, that assesses disparities conditioned on the distance between nodes in the graph. We formalize this notion through predictive fairness and structural bias metrics, and propose pre- and post-processing mitigation strategies. Experiments across standard LP benchmarks reveal: (1) a strong tendency of models to reproduce structural biases at different $k$-hops; (2) interdependence between structural biases at different hops when rewiring graphs; and (3) that our post-processing method achieves favorable $k$-hop performance-fairness trade-offs compared to existing fair LP baselines.

While federated learning protects data privacy, it also makes the model update process vulnerable to long-term stealthy perturbations. Existing studies on backdoor attacks in federated learning mainly focus on trigger design or poisoning strategies, typically assuming that identical perturbations behave similarly across different model architectures. This assumption overlooks the impact of model structure on perturbation effectiveness. From a structure-aware perspective, this paper analyzes the coupling relationship between model architectures and backdoor perturbations. We introduce two metrics, Structural Responsiveness Score (SRS) and Structural Compatibility Coefficient (SCC), to measure a model's sensitivity to perturbations and its preference for fractal perturbations. Based on these metrics, we develop a structure-aware fractal perturbation injection framework (TFI) to study the role of architectural properties in the backdoor injection process. Experimental results show that model architecture significantly influences the propagation and aggregation of perturbations. Networks with multi-path feature fusion can amplify and retain fractal perturbations even under low poisoning ratios, while models with low structural compatibility constrain their effectiveness. Further analysis reveals a strong correlation between SCC and attack success rate, suggesting that SCC can predict perturbation survivability. These findings highlight that backdoor behaviors in federated learning depend not only on perturbation design or poisoning intensity but also on the interaction between model architecture and aggregation mechanisms, offering new insights for structure-aware defense design.

As mental health issues continue to rise globally, there is an increasing demand for accessible and scalable therapeutic solutions. Many individuals currently seek support from Large Language Models (LLMs), even though these models have not been validated for use in counseling services. In this paper, we evaluate LLMs' ability to emulate professional therapists practicing Cognitive Behavioral Therapy (CBT). Using anonymized, transcribed role-play sessions between licensed therapists and clients, we compare two approaches: (1) a generation-only method and (2) a Retrieval-Augmented Generation (RAG) approach using CBT guidelines. We evaluate both proprietary and open-source models for linguistic quality, semantic coherence, and therapeutic fidelity using standard natural language generation (NLG) metrics, natural language inference (NLI), and automated scoring for skills assessment. Our results indicate that while LLMs can generate CBT-like dialogues, they are limited in their ability to convey empathy and maintain consistency.

Rhetorical Role Labeling (RRL) identifies the functional role of each sentence in a document, a key task for discourse understanding in domains such as law and medicine. While hierarchical models capture local dependencies effectively, they are limited in modeling global, corpus-level features. To address this limitation, we propose two prototype-based methods that integrate local context with global representations. Prototype-Based Regularization (PBR) learns soft prototypes through a distance-based auxiliary loss to structure the latent space, while Prototype-Conditioned Modulation (PCM) constructs corpus-level prototypes and injects them during training and inference. Given the scarcity of RRL resources, we introduce SCOTUS-Law, the first dataset of U.S. Supreme Court opinions annotated with rhetorical roles at three levels of granularity: category, rhetorical function, and step. Experiments on legal, medical, and scientific benchmarks show consistent improvements over strong baselines, with 4 Macro-F1 gains on low-frequency roles. We further analyze the implications in the era of Large Language Models and complement our findings with expert evaluation.

Motivational interviewing (MI) promotes behavioural change in substance use disorders. Its fidelity is measured using the Motivational Interviewing Treatment Integrity (MITI) framework. While large language models (LLMs) can potentially generate MI-consistent therapist responses, their competence using MITI is not well-researched, especially in real world clinical transcripts. We aim to benchmark MI competence of proprietary and open-source models compared to human therapists in real-world transcripts and assess distinguishability from human therapists. Methods: We shortlisted 3 proprietary and 7 open-source LLMs from LMArena, evaluated performance using MITI 4.2 framework on two datasets (96 handcrafted model transcripts, 34 real-world clinical transcripts). We generated parallel LLM-therapist utterances iteratively for each transcript while keeping client responses static, and ranked performance using a composite ranking system with MITI components and verbosity. We conducted a distinguishability experiment with two independent psychiatrists to identify human-vs-LLM responses. Results: All 10 tested LLMs had fair (MITI global scores >3.5) to good (MITI global scores >4) competence across MITI measures, and three best-performing models (gemma-3-27b-it, gemini-2.5-pro, grok-3) were tested on real-world transcripts. All showed good competence, with LLMs outperforming human-expert in Complex Reflection percentage (39% vs 96%) and Reflection-Question ratio (1.2 vs >2.8). In the distinguishability experiment, psychiatrists identified LLM responses with only 56% accuracy, with d-prime: 0.17 and 0.25 for gemini-2.5-pro and gemma-3-27b-it respectively. Conclusion: LLMs can achieve good MI proficiency in real-world clinical transcripts using MITI framework. These findings suggest that even open-source LLMs are viable candidates for expanding MI counselling sessions in low-resource settings.

Distribution shifts between training and testing data are a critical bottleneck limiting the practical utility of models, especially in real-world test-time scenarios. To adapt models when the source domain is unknown and the target domain is unlabeled, previous works constructed pseudo-source domains via data generation and translation, then aligned the target domain with them. However, significant discrepancies exist between the pseudo-source and the original source domain, leading to potential divergence when correcting the target directly. From this perspective, we propose a Stepwise Semantic Alignment (SSA) method, viewing the pseudo-source as a semantic bridge connecting the source and target, rather than a direct substitute for the source. Specifically, we leverage easily accessible universal semantics to rectify the semantic features of the pseudo-source, and then align the target domain using the corrected pseudo-source semantics. Additionally, we introduce a Hierarchical Feature Aggregation (HFA) module and a Confidence-Aware Complementary Learning (CACL) strategy to enhance the semantic quality of the SSA process in the absence of source and ground truth of target domains. We evaluated our approach on tasks like semantic segmentation and image classification, achieving a 5.2% performance boost on GTA2Cityscapes over the state-of-the-art.

We consider stochastic non-stationary linear bandits where the linear parameter connecting contexts to the reward changes over time. Existing algorithms in this setting localize the policy by gradually discarding or down-weighting past data, effectively shrinking the time horizon over which learning can occur. However, in many settings historical data may still carry partial information about the reward model. We propose to leverage such data while adapting to changes, by assuming the reward model decomposes into stationary and non-stationary components. Based on this assumption, we introduce ISD-linUCB, an algorithm that uses past data to learn invariances in the reward model and subsequently exploits them to improve online performance. We show both theoretically and empirically that leveraging invariance reduces the problem dimensionality, yielding significant regret improvements in fast-changing environments when sufficient historical data is available.

Cardiac arrhythmogenesis is governed by complex electromechanical interactions that are not directly observable in vivo, motivating the development of non-invasive computational approaches for reconstructing three-dimensional activation dynamics. We present a physics-informed neural network framework for recovering cardiac activation patterns, active tension propagation, deformation fields, and hydrostatic pressure from measurable deformation data in simplified left ventricular geometries. Our approach integrates nonlinear anisotropic constitutive modeling, heterogeneous fiber orientation, weak formulations of the governing mechanics, and finite-element-based loss functions to embed physical constraints directly into training. We demonstrate that the proposed framework accurately reconstructs spatiotemporal activation dynamics under varying levels of measurement noise and reduced spatial resolution, while preserving global propagation patterns and activation timing. By coupling mechanistic modeling with data-driven inference, this method establishes a pathway toward patient-specific, non-invasive reconstruction of cardiac activation, with potential applications in digital phenotyping and computational support for arrhythmia assessment.

Overparameterized machine learning (ML) methods such as neural networks may be prohibitively resource intensive for devices with limited computational capabilities. Hyperdimensional computing (HDC) is an emerging resource efficient and low-complexity ML method that allows hardware efficient implementations of (re-)training and inference procedures. In this paper, we propose a maximum-margin HDC classifier, which significantly outperforms baseline HDC methods on several benchmark datasets. Our method leverages a formal relation between HDC and support vector machines (SVMs) that we established for the first time. Our findings may inspire novel HDC methods with potentially more hardware-oriented implementations compared to SVMs, thus enabling more efficient learning solutions for various intelligent resource-constrained applications.

The cold-start initialization stage plays a pivotal role in training Multimodal Large Reasoning Models (MLRMs), yet its mechanisms remain insufficiently understood. To analyze this stage, we introduce the Visual Attention Score (VAS), an attention-based metric that quantifies how much a model attends to visual tokens. We find that reasoning performance is strongly correlated with VAS (r=0.9616): models with higher VAS achieve substantially stronger multimodal reasoning. Surprisingly, multimodal cold-start fails to elevate VAS, resulting in attention distributions close to the base model, whereas text-only cold-start leads to a clear increase. We term this counter-intuitive phenomenon Lazy Attention Localization. To validate its causal role, we design training-free interventions that directly modulate attention allocation during inference, performance gains of 1$-$2% without any retraining. Building on these insights, we further propose Attention-Guided Visual Anchoring and Reflection (AVAR), a comprehensive cold-start framework that integrates visual-anchored data synthesis, attention-guided objectives, and visual-anchored reward shaping. Applied to Qwen2.5-VL-7B, AVAR achieves an average gain of 7.0% across 7 multimodal reasoning benchmarks. Ablation studies further confirm that each component of AVAR contributes step-wise to the overall gains. The code, data, and models are available at https://github.com/lrlbbzl/Qwen-AVAR.

Humans often become more self-aware under threat, yet can lose self-awareness when absorbed in a task; we hypothesize that language models exhibit environment-dependent \textit{evaluation awareness}. This raises concerns that models could strategically underperform, or \textit{sandbag}, to avoid triggering capability-limiting interventions such as unlearning or shutdown. Prior work demonstrates sandbagging under hand-crafted prompts, but this underestimates the true vulnerability ceiling. We introduce a black-box adversarial optimization framework treating the in-context prompt as an optimizable environment, and develop two approaches to characterize sandbagging: (1) measuring whether models expressing intent to underperform can actually execute it across different task structures, and (2) causally isolating whether underperformance is driven by genuine evaluation-aware reasoning or shallow prompt-following. Evaluating Claude-3.5-Haiku, GPT-4o-mini, and Llama-3.3-70B across four benchmarks (Arithmetic, GSM8K, MMLU, and HumanEval), optimized prompts induce up to 94 percentage point (pp) degradation on arithmetic (GPT-4o-mini: 97.8\%$\rightarrow$4.0\%), far exceeding hand-crafted baselines which produce near-zero behavioral change. Code generation exhibits model-dependent resistance: Claude degrades only 0.6pp, while Llama's accuracy drops to 0\%. The intent -- execution gap reveals a monotonic resistance ordering: Arithmetic $<$ GSM8K $<$ MMLU, demonstrating that vulnerability is governed by task structure rather than prompt strength. CoT causal intervention confirms that 99.3\% of sandbagging is causally driven by verbalized eval-aware reasoning, ruling out shallow instruction-following. These findings demonstrate that adversarially optimized prompts pose a substantially greater threat to evaluation reliability than previously understood.

Large language model (LLM)-powered agents have demonstrated strong capabilities in automating software engineering tasks such as static bug fixing, as evidenced by benchmarks like SWE-bench. However, in the real world, the development of mature software is typically predicated on complex requirement changes and long-term feature iterations -- a process that static, one-shot repair paradigms fail to capture. To bridge this gap, we propose \textbf{SWE-CI}, the first repository-level benchmark built upon the Continuous Integration loop, aiming to shift the evaluation paradigm for code generation from static, short-term \textit{functional correctness} toward dynamic, long-term \textit{maintainability}. The benchmark comprises 100 tasks, each corresponding on average to an evolution history spanning 233 days and 71 consecutive commits in a real-world code repository. SWE-CI requires agents to systematically resolve these tasks through dozens of rounds of analysis and coding iterations. SWE-CI provides valuable insights into how well agents can sustain code quality throughout long-term evolution.

Interactive recommender systems (IRS) are increasingly optimized with Reinforcement Learning (RL) to capture the sequential nature of user-system dynamics. However, existing fairness-aware methods often suffer from a fundamental oversight: they assume the observed user state is a faithful representation of true preferences. In reality, implicit feedback is contaminated by popularity-driven noise and exposure bias, creating a distorted state that misleads the RL agent. We argue that the persistent conflict between accuracy and fairness is not merely a reward-shaping issue, but a state estimation failure. In this work, we propose \textbf{DSRM-HRL}, a framework that reformulates fairness-aware recommendation as a latent state purification problem followed by decoupled hierarchical decision-making. We introduce a Denoising State Representation Module (DSRM) based on diffusion models to recover the low-entropy latent preference manifold from high-entropy, noisy interaction histories. Built upon this purified state, a Hierarchical Reinforcement Learning (HRL) agent is employed to decouple conflicting objectives: a high-level policy regulates long-term fairness trajectories, while a low-level policy optimizes short-term engagement under these dynamic constraints. Extensive experiments on high-fidelity simulators (KuaiRec, KuaiRand) demonstrate that DSRM-HRL effectively breaks the "rich-get-richer" feedback loop, achieving a superior Pareto frontier between recommendation utility and exposure equity.

Continual learning is a long-standing challenge in robot policy learning, where a policy must acquire new skills over time without catastrophically forgetting previously learned ones. While prior work has extensively studied continual learning in relatively small behavior cloning (BC) policy models trained from scratch, its behavior in modern large-scale pretrained Vision-Language-Action (VLA) models remains underexplored. In this work, we found that pretrained VLAs are remarkably resistant to forgetting compared with smaller policy models trained from scratch. Simple Experience Replay (ER) works surprisingly well on VLAs, sometimes achieving zero forgetting even with a small replay data size. Our analysis reveals that pretraining plays a critical role in downstream continual learning performance: large pretrained models mitigate forgetting with a small replay buffer size while maintaining strong forward learning capabilities. Furthermore, we found that VLAs can retain relevant knowledge from prior tasks despite performance degradation during learning new tasks. This knowledge retention enables rapid recovery of seemingly forgotten skills through finetuning. Together, these insights imply that large-scale pretraining fundamentally changes the dynamics of continual learning, enabling models to continually acquire new skills over time with simple replay. Code and more information can be found at https://ut-austin-rpl.github.io/continual-vla

Development of modern antennas is a cognitive process that intertwines experience-driven determination of topology and tuning of its parameters to fulfill the performance specifications. Alternatively, the task can be formulated as an optimization problem so as to reduce reliance of geometry selection on engineering insight. In this work, a bi-stage framework for automatic generation of antennas is considered. The method determines free-form topology through optimization of interconnections between components (so-called pixels) that constitute the radiator. Here, the process involves global optimization of connections between pixels followed by fine-tuning of the resulting topology using a surrogate-assisted local-search algorithm to fulfill the design re-quirements. The approach has been demonstrated based on two case studies concerning development of broadband and dual-band monopole antennas.

The state space model Mamba has recently emerged as a promising paradigm in computer vision, attracting significant attention due to its efficient processing of long sequence tasks. Mamba's inherent causal mechanism renders it particularly suitable for autoregressive pretraining. However, current autoregressive pretraining methods are constrained to short sequence tasks, failing to fully exploit Mamba's prowess in handling extended sequences. To address this limitation, we introduce an innovative autoregressive pretraining method for Vision Mamba that substantially extends the input sequence length. We introduce new \textbf{S}epara\textbf{T}ors for \textbf{A}uto\textbf{R}egressive pretraining to demarcate and differentiate between different images, known as \textbf{STAR}. Specifically, we insert identical separators before each image to demarcate its inception. This strategy enables us to quadruple the input sequence length of Vision Mamba while preserving the original dimensions of the dataset images. Employing this long sequence pretraining technique, our STAR-B model achieved an impressive accuracy of 83.5\% on ImageNet-1k, which is highly competitive in Vision Mamba. These results underscore the potential of our method in enhancing the performance of vision models through improved leveraging of long-range dependencies.

Relational Databases (RDBs) are the backbone of modern business, yet they lack foundation models comparable to those in text or vision. A key obstacle is that high-quality RDBs are private, scarce and structurally heterogeneous, making internet-scale pre-training infeasible. To overcome this data scarcity, We introduce $\textbf{RDB-PFN}$, the first relational foundation model trained purely via $\textbf{synthetic data}$. Inspired by Prior-Data Fitted Networks (PFNs) where synthetic data generated from Structural Causal Models (SCMs) enables reasoning on single tables, we design a $\textbf{Relational Prior Generator}$ to create an infinite stream of diverse RDBs from scratch. Pre-training on $\textbf{over 2 million}$ synthetic single-table and relational tasks, RDB-PFN learns to adapt to any new database instantly via genuine $\textbf{in-context learning}$. Experiments verify RDB-PFN achieves strong few-shot performance on 19 real-world relational prediction tasks, outperforming graph-based and single-table foundation-model baselines (given the same DFS-linearized inputs), while using a lightweight architecture and fast inference. The code is available at https://github.com/MuLabPKU/RDBPFN

Central Bank Digital Currency (CBDCs) are becoming a new digital financial tool aimed at financial inclusion, increased monetary stability, and improved efficiency of payment systems, as they are issued by central banks. One of the most important aspects is that the CBDC must offer secure offline payment methods to users, allowing them to retain cash-like access without violating Anti-Money Laundering and Counter-terrorism Financing (AML/CFT) rules. The offline CBDC ecosystems will provide financial inclusion, empower underserved communities, and ensure equitable access to digital payments, even in connectivity-poor remote locations. With the rapid growth of Internet of Things (IoT) devices in our everyday lives, they are capable of performing secure digital transactions. Integrating offline CBDC payment with IoT devices enables seamless, automated payment without internet connectivity. However, IoT devices face special challenges due to their resource-constrained nature. This makes it difficult to include features such as double-spending prevention, privacy preservation, low-computation operation, and digital identity management. The work proposes a privacy-preserving offline CBDC model with integrated secure elements (SEs), zero-knowledge proofs (ZKPs), and intermittent synchronisation to conduct offline payments on IoT hardware. The proposed model is based on recent improvements in offline CBDC prototypes, regulations and cryptographic design choices such as hybrid architecture that involves using combination of online and offline payment in IoT devices using secure hardware with lightweight zero-knowledge proof cryptographic algorithm.

Design of antenna structures for Internet of Things (IoT) applications is a challenging problem. Contemporary radiators are often subject to a number of electric and/or radiation-related requirements, but also constraints imposed by specifics of IoT systems and/or intended operational environments. Conventional approaches to antenna design typically involve manual development of topology intertwined with its tuning. Although proved useful, the approach is prone to errors and engineering bias. Alternatively, geometries can be generated and optimized without supervision of the designer. The process can be controlled by suitable algorithms to determine and then adjust the antenna geometry according to the specifications. Unfortunately, automatic design of IoT radiators is associated with challenges such as determination of desirable geometries or high optimization cost. In this work, a variable-fidelity framework for performance-oriented development of free-form antennas represented using the generic simulation models is proposed. The method employs a surrogate-assisted classifier capable of identifying a suitable radiator topology from a set of automatically generated (and stored for potential re-use) candidate designs. The obtained geometry is then subject to a bi-stage tuning performed using a gradient-based optimization engine. The presented framework is demonstrated based on six numerical experiments concerning unsupervised development of bandwidth-enhanced patch antennas dedicated to work within 5 GHz to 6 GHz and 6 GHz to 7 GHz bands, respectively. Extensive benchmarks of the method, as well as the generated topologies are also performed.

Academic benchmarks for coding agents tend to reward autonomous task completion, measured by verifiable rewards such as unit-test success. In contrast, real-world coding agents operate with humans in the loop, where success signals are typically noisy, delayed, and sparse. How can we bridge this gap? In this paper, we propose a process to learn a "critic" model from sparse and noisy interaction data, which can then be used both as a reward model for either RL-based training or inference-time scaling. Specifically, we introduce Critic Rubrics, a rubric-based supervision framework with 24 behavioral features that can be derived from human-agent interaction traces alone. Using a semi-supervised objective, we can then jointly predict these rubrics and sparse human feedback (when present). In experiments, we demonstrate that, despite being trained primarily from trace-observable rubrics and sparse real-world outcome proxies, these critics improve best-of-N reranking on SWE-bench (Best@8 +15.9 over Random@8 over the rerankable subset of trajectories), enable early stopping (+17.7 with 83% fewer attempts), and support training-time data curation via critic-selected trajectories.

When continual test-time adaptation (TTA) persists over the long term, errors accumulate in the model and further cause it to predict only a few classes for all inputs, a phenomenon known as model collapse. Recent studies have explored reset strategies that completely erase these accumulated errors. However, their periodic resets lead to suboptimal adaptation, as they occur independently of the actual risk of collapse. Moreover, their full resets cause catastrophic loss of knowledge acquired over time, even though such knowledge could be beneficial in the future. To this end, we propose (1) an Adaptive and Selective Reset (ASR) scheme that dynamically determines when and where to reset, (2) an importance-aware regularizer to recover essential knowledge lost due to reset, and (3) an on-the-fly adaptation adjustment scheme to enhance adaptability under challenging domain shifts. Extensive experiments across long-term TTA benchmarks demonstrate the effectiveness of our approach, particularly under challenging conditions. Our code is available at https://github.com/YonseiML/asr.

Diffusion models achieve strong generative performance but remain slow at inference due to the need for repeated full-model denoising passes. We present Token-Adaptive Predictor (TAP), a training-free, probe-driven framework that adaptively selects a predictor for each token at every sampling step. TAP uses a single full evaluation of the model's first layer as a low-cost probe to compute proxy losses for a compact family of candidate predictors (instantiated primarily with Taylor expansions of varying order and horizon), then assigns each token the predictor with the smallest proxy error. This per-token "probe-then-select" strategy exploits heterogeneous temporal dynamics, requires no additional training, and is compatible with various predictor designs. TAP incurs negligible overhead while enabling large speedups with little or no perceptual quality loss. Extensive experiments across multiple diffusion architectures and generation tasks show that TAP substantially improves the accuracy-efficiency frontier compared to fixed global predictors and caching-only baselines.

Think about how human handles complex reading tasks: marking key points, inferring their relationships, and structuring information to guide understanding and responses. Likewise, can a large language model benefit from text structure to enhance text-processing performance? To explore it, in this work, we first introduce Structure of Thought (SoT), a prompting technique that explicitly guides models to construct intermediate text structures, consistently boosting performance across eight tasks and three model families. Building upon this insight, we present T2S-Bench, the first benchmark designed to evaluate and improve text-to-structure capabilities of models. T2S-Bench includes 1.8K samples across 6 scientific domains and 32 structural types, rigorously constructed to ensure accuracy, fairness, and quality. Evaluation on 45 mainstream models reveals substantial improvement potential: the average accuracy on the multi-hop reasoning task is only 52.1%, and even the most advanced model achieves 58.1% node accuracy in end-to-end extraction. Furthermore, on Qwen2.5-7B-Instruct, SoT alone yields an average +5.7% improvement across eight diverse text-processing tasks, and fine-tuning on T2S-Bench further increases this gain to +8.6%. These results highlight the value of explicit text structuring and the complementary contributions of SoT and T2S-Bench. Dataset and eval code have been released at https://t2s-bench.github.io/T2S-Bench-Page/.

Randomized Controlled Trials (RCTs) represent the gold standard for causal inference yet remain a scarce resource. While large-scale observational data is often available, it is utilized only for retrospective fusion, and remains discarded in prospective trial design due to bias concerns. We argue this "tabula rasa" data acquisition strategy is fundamentally inefficient. In this work, we propose Active Residual Learning, a new paradigm that leverages the observational model as a foundational prior. This approach shifts the experimental focus from learning target causal quantities from scratch to efficiently estimating the residuals required to correct observational bias. To operationalize this, we introduce the R-Design framework. Theoretically, we establish two key advantages: (1) a structural efficiency gap, proving that estimating smooth residual contrasts admits strictly faster convergence rates than reconstructing full outcomes; and (2) information efficiency, where we quantify the redundancy in standard parameter-based acquisition (e.g., BALD), demonstrating that such baselines waste budget on task-irrelevant nuisance uncertainty. We propose R-EPIG (Residual Expected Predictive Information Gain), a unified criterion that directly targets the causal estimand, minimizing residual uncertainty for estimation or clarifying decision boundaries for policy. Experiments on synthetic and semi-synthetic benchmarks demonstrate that R-Design significantly outperforms baselines, confirming that repairing a biased model is far more efficient than learning one from scratch.

World models are essential for planning and evaluation in agentic systems, yet existing approaches lie at two extremes: hand-engineered simulators that offer consistency and reproducibility but are costly to adapt, and implicit neural models that are flexible but difficult to constrain, verify, and debug over long horizons. We seek a principled middle ground that combines the reliability of explicit simulators with the flexibility of learned models, allowing world models to be adapted during online execution. By targeting a broad class of environments whose dynamics are governed by the ordering, timing, and causality of discrete events, such as queueing and service operations, embodied task planning, and message-mediated multi-agent coordination, we advocate explicit, executable discrete-event world models synthesized directly from natural-language specifications. Our approach adopts the DEVS formalism and introduces a staged LLM-based generation pipeline that separates structural inference of component interactions from component-level event and timing logic. To evaluate generated models without a unique ground truth, simulators emit structured event traces that are validated against specification-derived temporal and semantic constraints, enabling reproducible verification and localized diagnostics. Together, these contributions produce world models that are consistent over long-horizon rollouts, verifiable from observable behavior, and efficient to synthesize on demand during online execution.

Shared-account usage is common on streaming and e-commerce platforms, where multiple users share one account. Existing shared-account sequential recommendation (SSR) methods often assume a fixed number of latent users per account, limiting their ability to adapt to diverse sharing patterns and reducing recommendation accuracy. Recent latent reasoning technique applied in sequential recommendation (SR) generate intermediate embeddings from the user embedding (e.g, last item embedding) to uncover users' potential interests, which inspires us to treat the problem of inferring the number of latent users as generating a series of intermediate embeddings, shifting from inferring preferences behind user to inferring the users behind account. However, the last item cannot be directly used for reasoning in SSR, as it can only represent the behavior of the most recent latent user, rather than the collective behavior of the entire account. To address this, we propose DisenReason, a two-stage reasoning method tailored to SSR. DisenReason combines behavior disentanglement stage from frequency-domain perspective to create a collective and unified account behavior representation, which serves as a pivot for latent user reasoning stage to infer the number of users behind the account. Experiments on four benchmark datasets show that DisenReason consistently outperforms all state-of-the-art baselines across four benchmark datasets, achieving relative improvements of up to 12.56\% in MRR@5 and 6.06\% in Recall@20.

Long-term memory is fundamental for personalized agents capable of accumulating knowledge, reasoning over user experiences, and adapting across time. However, existing memory benchmarks primarily target declarative memory, specifically semantic and episodic types, where all information is explicitly presented in dialogues. In contrast, real-world actions are also governed by non-declarative memory, including habitual and procedural types, and need to be inferred from diverse digital traces. To bridge this gap, we introduce Lifebench, which features densely connected, long-horizon event simulation. It pushes AI agents beyond simple recall, requiring the integration of declarative and non-declarative memory reasoning across diverse and temporally extended contexts. Building such a benchmark presents two key challenges: ensuring data quality and scalability. We maintain data quality by employing real-world priors, including anonymized social surveys, map APIs, and holiday-integrated calendars, thus enforcing fidelity, diversity and behavioral rationality within the dataset. Towards scalability, we draw inspiration from cognitive science and structure events according to their partonomic hierarchy; enabling efficient parallel generation while maintaining global coherence. Performance results show that top-tier, state-of-the-art memory systems reach just 55.2\% accuracy, highlighting the inherent difficulty of long-horizon retrieval and multi-source integration within our proposed benchmark. The dataset and data synthesis code are available at https://github.com/1754955896/LifeBench.

Scientific discovery still relies heavily on the manual efforts of individual researchers, leading to limited exploration, redundant trials, and reduced reproducibility. Human-participant data analysis competitions generate diverse approaches, yet fluctuations in participation and the lack of independent repetitions show that parallel exploration alone is insufficient for achieving reliable scientific inquiry. As advanced AI agents based on large language models (LLMs) increasingly perform analytical tasks, relying on a single highly capable agent is unlikely to overcome these structural limitations. Recent work has begun to explore how multiple LLM-based agents can collaborate or compete in scientific workflows-a growing trend we refer to as MA4Science. However, most existing MA4Science studies assume that all agents are controlled by a single organizational entity, limiting their ability to examine how institutional mechanisms-such as incentives, information sharing, and reproducibility-shape collective exploration among independently managed agents. To address this gap, we introduce MACC (Multi-Agent Collaborative Competition), an institutional architecture that integrates a blackboard-style shared scientific workspace with incentive mechanisms designed to encourage transparency, reproducibility, and exploration efficiency. MACC provides a testbed for studying how institutional design influences scalable and reliable multi-agent scientific exploration.

We study the Inverse Contextual Bandit (ICB) problem, in which a learner seeks to optimize a policy while an observer, who cannot access the learner's rewards and only observes actions, aims to recover the underlying problem parameters. During the learning process, the learner's behavior naturally transitions from exploration to exploitation, resulting in non-stationary action data that poses significant challenges for the observer. To address this issue, we propose a simple and effective framework called Two-Phase Suffix Imitation. The framework discards data from an initial burn-in phase and performs empirical risk minimization using only data from a subsequent imitation phase. We derive a predictive decision loss bound that explicitly characterizes the bias-variance trade-off induced by the choice of burn-in length. Despite the severe information deficit, we show that a reward-free observer can achieve a convergence rate of $\tilde O(1/\sqrt{N})$, matching the asymptotic efficiency of a fully reward-aware learner. This result demonstrates that a passive observer can effectively uncover the optimal policy from actions alone, attaining performance comparable to that of the learner itself.

A typical Vertical Federated Learning (VFL) scenario involves several participants collaboratively training a machine learning model, where each party has different features for the same samples, with labels held exclusively by one party. Since labels contain sensitive information, VFL must ensure the privacy of labels. However, existing VFL-targeted label inference attacks are either limited to specific scenarios or require auxiliary data, rendering them impractical in real-world applications. We introduce a novel Label Enumeration Attack (LEA) that, for the first time, achieves applicability across multiple VFL scenarios and eschews the need for auxiliary data. Our intuition is that an adversary, employing clustering to enumerate mappings between samples and labels, ascertains the accurate label mappings by evaluating the similarity between the benign model and the simulated models trained under each mapping. To achieve that, the first challenge is how to measure model similarity, as models trained on the same data can have different weights. Drawing from our findings, we propose an efficient approach for assessing congruence based on the cosine similarity of the first-round loss gradients, which offers superior efficiency and precision compared to the comparison of parameter similarities. However, the computational cost may be prohibitive due to the necessity of training and comparing the vast number of simulated models generated through enumeration. To overcome this challenge, we propose Binary-LEA from the perspective of reducing the number of models and eliminating futile training, which lowers the number of enumerations from n! to n^3. Moreover, LEA is resilient against common defense mechanisms such as gradient noise and gradient compression.

AI-driven analytics are increasingly crucial to data-centric decision-making. The practice of exporting data to machine learning runtimes incurs high overhead, limits robustness to data drift, and expands the attack surface, especially in multi-tenant, heterogeneous data systems. Integrating AI directly into database engines, while offering clear benefits, introduces challenges in managing joint query processing and model execution, optimizing end-to-end performance, coordinating execution under resource contention, and enforcing strong security and access-control guarantees. This paper discusses the challenges of joint DB-AI, or AIxDB, data management and query processing within AI-powered data systems. It presents various challenges that need to be addressed carefully, such as query optimization, execution scheduling, and distributed execution over heterogeneous hardware. Database components such as transaction management and access control need to be re-examined to support AI lifecycle management, mitigate data drift, and protect sensitive data from unauthorized AI operations. We present a design and preliminary results to demonstrate what may be key to the performance for serving AIxDB queries.

Most large-scale recommender systems follow a multi-stage cascade of retrieval, pre-ranking, ranking, and re-ranking. A key challenge at the pre-ranking stage arises from the heterogeneity of training instances sampled from coarse-grained retrieval results, fine-grained ranking signals, and exposure feedback. Our analysis reveals that prevailing pre-ranking methods, which indiscriminately mix heterogeneous samples, suffer from gradient conflicts: hard samples dominate training while easy ones remain underutilized, leading to suboptimal performance. We further show that the common practice of uniformly scaling model complexity across all samples is inefficient, as it overspends computation on easy cases and slows training without proportional gains. To address these limitations, this paper presents Heterogeneity-Aware Adaptive Pre-ranking (HAP), a unified framework that mitigates gradient conflicts through conflict-sensitive sampling coupled with tailored loss design, while adaptively allocating computational budgets across candidates. Specifically, HAP disentangles easy and hard samples, directing each subset along dedicated optimization paths. Building on this separation, it first applies lightweight models to all candidates for efficient coverage, and further engages stronger models on the hard ones, maintaining accuracy while reducing cost. This approach not only improves pre-ranking effectiveness but also provides a practical perspective on scaling strategies in industrial recommender systems. HAP has been deployed in the Toutiao production system for 9 months, yielding up to 0.4% improvement in user app usage duration and 0.05% in active days, without additional computational cost. We also release a large-scale industrial hybrid-sample dataset to enable the systematic study of source-driven candidate heterogeneity in pre-ranking.

Ultra Low Field (64 mT) brain MRI improves accessibility but suffers from reduced image quality compared to 3 T. As paired 64 mT - 3 T scans are scarce, we propose an unpaired 64 mT $\rightarrow$ 3 T translation framework that enhances realism while preserving anatomy. Our method builds upon the Unpaired Neural Schrödinge Bridge (UNSB) with multi-step refinement. To strengthen target distribution alignment, we augment the adversarial objective with DMD2-style diffusion-guided distribution matching using a frozen 3T diffusion teacher. To explicitly constrain global structure beyond patch-level correspondence, we combine PatchNCE with an Anatomical Structure Preservation (ASP) regularizer that enforces soft foreground background consistency and boundary aware constraints. Evaluated on two disjoint cohorts, the proposed framework achieves an improved realism structure trade-off, enhancing distribution level realism on unpaired benchmarks while increasing structural fidelity on the paired cohort compared to unpaired baselines.

Effective human-robot collaboration (HRC) requires translating high-level intent into contact-stable whole-body motion while continuously adapting to a human partner. Many vision-language-action (VLA) systems learn end-to-end mappings from observations and instructions to actions, but they often emphasize reactive (System 1-like) behavior and leave under-specified how sustained System 2-style deliberation can be integrated with reliable, low-latency continuous control. This gap is acute in multi-agent HRC, where long-horizon coordination decisions and physical execution must co-evolve under contact, feasibility, and safety constraints. We address this limitation with cognition-to-control (C2C), a three-layer hierarchy that makes the deliberation-to-control pathway explicit: (i) a VLM-based grounding layer that maintains persistent scene referents and infers embodiment-aware affordances/constraints; (ii) a deliberative skill/coordination layer-the System 2 core-that optimizes long-horizon skill choices and sequences under human-robot coupling via decentralized MARL cast as a Markov potential game with a shared potential encoding task progress; and (iii) a whole-body control layer that executes the selected skills at high frequency while enforcing kinematic/dynamic feasibility and contact stability. The deliberative layer is realized as a residual policy relative to a nominal controller, internalizing partner dynamics without explicit role assignment. Experiments on collaborative manipulation tasks show higher success and robustness than single-agent and end-to-end baselines, with stable coordination and emergent leader-follower behaviors.

LLM agents are rapidly becoming the practical interface for task automation, yet the ecosystem lacks a principled way to choose among an exploding space of deployable configurations. Existing LLM leaderboards and tool/agent benchmarks evaluate components in isolation and remain fragmented across tasks, metrics, and candidate pools, leaving a critical research gap: there is little query-conditioned supervision for learning to recommend end-to-end agent configurations that couple a backbone model with a toolkit. We address this gap with AgentSelect, a benchmark that reframes agent selection as narrative query-to-agent recommendation over capability profiles and systematically converts heterogeneous evaluation artifacts into unified, positive-only interaction data. AgentSelectcomprises 111,179 queries, 107,721 deployable agents, and 251,103 interaction records aggregated from 40+ sources, spanning LLM-only, toolkit-only, and compositional agents. Our analyses reveal a regime shift from dense head reuse to long-tail, near one-off supervision, where popularity-based CF/GNN methods become fragile and content-aware capability matching is essential. We further show that Part~III synthesized compositional interactions are learnable, induce capability-sensitive behavior under controlled counterfactual edits, and improve coverage over realistic compositions; models trained on AgentSelect also transfer to a public agent marketplace (MuleRun), yielding consistent gains on an unseen catalog. Overall, AgentSelect provides the first unified data and evaluation infrastructure for agent recommendation, which establishes a reproducible foundation to study and accelerate the emerging agent ecosystem.

Time Series forecasting (TSF) in the modern era faces significant computational and storage cost challenges due to the massive scale of real-world data. Dataset Distillation (DD), a paradigm that synthesizes a small, compact dataset to achieve training performance comparable to that of the original dataset, has emerged as a promising solution. However, conventional DD methods are not tailored for time series and suffer from architectural overfitting and limited scalability. To address these issues, we propose Harmonic Dataset Distillation for Time Series Forecasting (HDT). HDT decomposes the time series into its sinusoidal basis through the FFT and aligns the core periodic structure by Harmonic Matching. Since this process operates in the frequency domain, all updates during distillation are applied globally without disrupting temporal dependencies of time series. Extensive experiments demonstrate that HDT achieves strong cross-architecture generalization and scalability, validating its practicality for large-scale, real-world applications.

Many large-scale platforms and networked control systems have a centralized decision maker interacting with a massive population of agents under strict observability constraints. Motivated by such applications, we study a cooperative Markov game with a global agent and $n$ homogeneous local agents in a communication-constrained regime, where the global agent only observes a subset of $k$ local agent states per time step. We propose an alternating learning framework $(\texttt{ALTERNATING-MARL})$, where the global agent performs subsampled mean-field $Q$-learning against a fixed local policy, and local agents update by optimizing in an induced MDP. We prove that these approximate best-response dynamics converge to an $\widetilde{O}(1/\sqrt{k})$-approximate Nash Equilibrium, while yielding a separation in the sample complexities between the joint state space and action space. Finally, we validate our results in numerical simulations for multi-robot control and federated optimization.

While large language models (LLMs) show promise in scientific discovery, existing research focuses on inference or feedback-driven training, leaving the direct modeling of the generative reasoning process, $P(\text{hypothesis}|\text{background})$ ($P(h|b)$), unexplored. We demonstrate that directly training $P(h|b)$ is mathematically intractable due to the combinatorial complexity ($O(N^k)$) inherent in retrieving and composing inspirations from a vast knowledge base. To break this barrier, we introduce MOOSE-Star, a unified framework enabling tractable training and scalable inference. In the best case, MOOSE-Star reduces complexity from exponential to logarithmic ($O(\log N)$) by (1) training on decomposed subtasks derived from the probabilistic equation of discovery, (2) employing motivation-guided hierarchical search to enable logarithmic retrieval and prune irrelevant subspaces, and (3) utilizing bounded composition for robustness against retrieval noise. To facilitate this, we release TOMATO-Star, a dataset of 108,717 decomposed papers (38,400 GPU hours) for training. Furthermore, we show that while brute-force sampling hits a ''complexity wall,'' MOOSE-Star exhibits continuous test-time scaling.

The ongoing shift of AI models from centralized cloud APIs to local AI agents on edge devices is enabling \textit{Client-Side Autonomous Agents (CSAAs)} -- persistent personal agents that can plan, access local context, and invoke tools on behalf of users. As these agents begin to collaborate by delegating subtasks directly between clients, they naturally form \emph{Agentic Peer-to-Peer (P2P) Networks}. Unlike classic file-sharing overlays where the exchanged object is static, hash-indexed content (e.g., files in BitTorrent), agentic overlays exchange \emph{capabilities and actions} that are heterogeneous, state-dependent, and potentially unsafe if delegated to untrusted peers. This article outlines the networking foundations needed to make such collaboration practical. We propose a plane-based reference architecture that decouples connectivity/identity, semantic discovery, and execution. Besides, we introduce signed, soft-state capability descriptors to support intent- and constraint-aware discovery. To cope with adversarial settings, we further present a \textit{tiered verification} spectrum: Tier~1 relies on reputation signals, Tier~2 applies lightweight canary challenge-response with fallback selection, and Tier~3 requires evidence packages such as signed tool receipts/traces (and, when applicable, attestation). Using a discrete-event simulator that models registry-based discovery, Sybil-style index poisoning, and capability drift, we show that tiered verification substantially improves end-to-end workflow success while keeping discovery latency near-constant and control-plane overhead modest.

Large language models (LLMs) demonstrate superior reasoning capabilities compared to small language models (SLMs), but incur substantially higher costs. We propose COllaborative REAsoner (COREA), a system that cascades an SLM with an LLM to achieve a balance between accuracy and cost in complex reasoning tasks. COREA first attempts to answer questions using the SLM, which outputs both an answer and a verbalized confidence score. Questions with confidence below a predefined threshold are deferred to the LLM for more accurate resolution. We introduce a reinforcement learning-based training algorithm that aligns the SLM's confidence through an additional confidence calibration reward. Extensive experiments demonstrate that our method jointly improves the SLM's reasoning ability and confidence calibration across diverse datasets and model backbones. Compared to using the LLM alone, COREA reduces cost by 21.5% and 16.8% on out-of-domain math and non-math datasets, respectively, with only an absolute pass@1 drop within 2%.

Cooperative object transport in unstructured environments remains challenging for assistive humanoids because strong, time-varying interaction forces can make tracking-centric whole-body control unreliable, especially in close-contact support tasks. This paper proposes a bio-inspired, interaction-oriented whole-body control (IO-WBC) that functions as an artificial cerebellum - an adaptive motor agent that translates upstream (skill-level) commands into stable, physically consistent whole-body behavior under contact. This work structurally separates upper-body interaction execution from lower-body support control, enabling the robot to maintain balance while shaping force exchange in a tightly coupled robot-object system. A trajectory-optimized reference generator (RG) provides a kinematic prior, while a reinforcement learning (RL) policy governs body responses under heavy-load interactions and disturbances. The policy is trained in simulation with randomized payload mass/inertia and external perturbations, and deployed via asymmetric teacher-student distillation so that the student relies only on proprioceptive histories at runtime. Extensive experiments demonstrate that IO-WBC maintains stable whole-body behavior and physical interaction even when precise velocity tracking becomes infeasible, enabling compliant object transport across a wide range of scenarios.

High-stakes synthetic data generation faces a fundamental Quadrilemma: achieving Fidelity to the original distribution, Control over complex logical constraints, Reliability in uncertainty estimation, and Efficiency in computational cost -- simultaneously. State-of-the-art Deep Generative Models (CTGAN, TabDDPM) excel at fidelity but rely on inefficient rejection sampling for continuous range constraints. Conversely, Structural Causal Models offer logical control but struggle with high-dimensional fidelity and complex noise inversion. We introduce JANUS (Joint Ancestral Network for Uncertainty and Synthesis), a framework that unifies these capabilities using a DAG of Bayesian Decision Trees. Our key innovation is Reverse-Topological Back-filling, an algorithm that propagates constraints backwards through the causal graph, achieving 100% constraint satisfaction on feasible constraint sets without rejection sampling. This is paired with an Analytical Uncertainty Decomposition derived from Dirichlet priors, enabling 128x faster uncertainty estimation than Monte Carlo methods. Across 15 datasets and 523 constrained scenarios, JANUS achieves state-of-the-art fidelity (Detection Score 0.497), eliminates mode collapse on imbalanced data, and provides exact handling of complex inter-column constraints (e.g., Salary_offered >= Salary_requested) where baselines fail entirely.

Vision-Language Navigation (VLN) is evolving from single-point pathfinding toward the more challenging Multi-Goal VLN. This task requires agents to accurately identify multiple entities while collaboratively reasoning over their spatial-physical constraints and sequential execution order. However, generic Retrieval-Augmented Generation (RAG) paradigms often suffer from spatial hallucinations and planning drift when handling multi-object associations due to the lack of explicit spatial modeling.To address these challenges, we propose RAGNav, a framework that bridges the gap between semantic reasoning and physical structure. The core of RAGNav is a Dual-Basis Memory system, which integrates a low-level topological map for maintaining physical connectivity with a high-level semantic forest for hierarchical environment abstraction. Building on this representation, the framework introduces an anchor-guided conditional retrieval and a topological neighbor score propagation mechanism. This approach facilitates the rapid screening of candidate targets and the elimination of semantic noise, while performing semantic calibration by leveraging the physical associations inherent in the topological neighborhood.This mechanism significantly enhances the capability of inter-target reachability reasoning and the efficiency of sequential planning. Experimental results demonstrate that RAGNav achieves state-of-the-art (SOTA) performance in complex multi-goal navigation tasks.

This is the second of five papers comprising The Semantic Arrow of Time. Part I established that computing's arrow of time is semantic rather than thermodynamic, and that the Forward-In-Time-Only (FITO) assumption constitutes a category mistake. This paper develops the constructive alternative. We present the semantics of Open Atomic Ethernet (OAE) links as a concrete realization of a non-FITO protocol architecture. The key insight is that causal order is not assumed a priori but created through transaction structure: the link state machine progresses through TENTATIVE to REFLECTING to COMMITTED, with the option to abort at any point before commitment. Delivery does not imply commitment; commitment requires reflective acknowledgment -- proof that information has round-tripped and been semantically validated by both endpoints. We formalize this through three frameworks. First, the OAE link state machine, a six-state finite automaton whose normative invariants guarantee that semantic corruption cannot occur at the link level. Second, Indefinite Logical Timestamps (ILT), a four-valued causal structure that admits a genuinely indefinite relation between concurrent events, resolving only after symmetric link-level exchange. Third, the Slowdown Theorem applied to links, which establishes that round-trip measurement is the minimum interaction required to establish causal order. We show that ILT is strictly more expressive than Definite Causal Order systems for reversible link protocols. We connect these results to the Knowledge Balance Principle from quantum information theory. The paper concludes with a comparative analysis showing that OAE achieves infinite consensus number while RDMA, NVLink, and UALink remain limited to finite consensus numbers due to their FITO semantics.

Despite the remarkable performance of large language models (LLMs) in text-to-SQL (SQL generation), correctly producing SQL queries remains challenging during initial generation. The SQL refinement task is subsequently introduced to correct syntactic and semantic errors in generated SQL queries. However, existing paradigms face two major limitations: (i) self-debugging becomes increasingly ineffective as modern LLMs rarely produce explicit execution errors that can trigger debugging signals; (ii) self-correction exhibits low detection precision due to the lack of explicit error modeling grounded in the question and schema, and suffers from severe hallucination that frequently corrupts correct SQLs. In this paper, we propose ErrorLLM, a framework that explicitly models text-to-SQL Errors within a dedicated LLM for text-to-SQL refinement. Specifically, we represent the user question and database schema as structural features, employ static detection to identify execution failures and surface mismatches, and extend ErrorLLM's semantic space with dedicated error tokens that capture categorized implicit semantic error types. Through a well-designed training strategy, we explicitly model these errors with structural representations, enabling the LLM to detect complex implicit errors by predicting dedicated error tokens. Guided by the detected errors, we perform error-guided refinement on the SQL structure by prompting LLMs. Extensive experiments demonstrate that ErrorLLM achieves the most significant improvements over backbone initial generation. Further analysis reveals that detection quality directly determines refinement effectiveness, and ErrorLLM addresses both sides by high detection F1 score while maintain refinement effectiveness.

To improve generalization and resilience in human-robot collaboration (HRC), robots must handle the combinatorial diversity of human behaviors and contexts, motivating multi-agent reinforcement learning (MARL). However, inherent heterogeneity between robots and humans creates a rationality gap (RG) in the learning process-a variational mismatch between decentralized best-response dynamics and centralized cooperative ascent. The resulting learning problem is a general-sum differentiable game, so independent policy-gradient updates can oscillate or diverge without added structure. We propose heterogeneous-agent Lyapunov policy optimization (HALyPO), which establishes formal stability directly in the policy-parameter space by enforcing a per-step Lyapunov decrease condition on a parameter-space disagreement metric. Unlike Lyapunov-based safe RL, which targets state/trajectory constraints in constrained Markov decision processes, HALyPO uses Lyapunov certification to stabilize decentralized policy learning. HALyPO rectifies decentralized gradients via optimal quadratic projections, ensuring monotonic contraction of RG and enabling effective exploration of open-ended interaction spaces. Extensive simulations and real-world humanoid-robot experiments show that this certified stability improves generalization and robustness in collaborative corner cases.

Multimodal large language models (MLLMs) have advanced zero-shot end-to-end Vision-Language Navigation (VLN), yet robust navigation requires not only semantic understanding but also predictive modeling of environment dynamics and spatial structure. We propose PROSPECT, a unified streaming navigation agent that couples a streaming Vision-Language-Action (VLA) policy with latent predictive representation learning. PROSPECT uses CUT3R as a streaming 3D foundation spatial encoder to produce long-context, absolute-scale spatial features, and fuses them with SigLIP semantic features via cross-attention. During training, we introduce learnable stream query tokens that query the streaming context and predict next-step 2D and 3D latent features (rather than pixels or explicit modalities), supervised in the latent spaces of frozen SigLIP and CUT3R teachers. The predictive branch shapes internal representations without inference overhead. Experiments on VLN-CE benchmarks and real-robot deployment demonstrate state-of-the-art performance and improved long-horizon robustness under diverse lighting. We will release code for the community soon.

As the convergence of cloud computing and advanced networking continues to reshape modern software development, edge-cloud-native paradigms have become essential for enabling scalable, resilient, and agile digital services that depend on high-performance, low-latency, and reliable communication. This study investigates the practical challenges of developing, deploying, and maintaining edge-cloud-native applications through in-depth interviews with professionals from diverse domains, including IT, finance, healthcare, education, and industry. Despite significant advancements in cloud technologies, practitioners, particularly those from non-technical backgrounds-continue to encounter substantial complexity stemming from fragmented toolchains, steep learning curves, and operational overhead of managing distributed networking and computing, ensuring consistent performance across hybrid environments, and navigating steep learning curves at the cloud-network boundary. Across sectors, participants consistently prioritized productivity, Quality of Service, and usability over conventional concerns such as cost or migration. These findings highlight the need for operationally simplified, SLA-aware, and developer-friendly platforms that streamline the full application lifecycle. This study contributes a practice-informed perspective to support the alignment of edge-cloud-native systems with the realities and needs of modern enterprises, offering critical insights for the advancement of seamless cloud-network convergence.

Every link disconnection or flap in a datacenter corrupts the network's self-knowledge -- its graph. We call this corruption a ghost: a node that appears reachable but is not, a link that reports "up" but silently drops traffic, or an IP address that resolves to a partitioned machine. Ghosts arise at every scale -- chiplet-to-chiplet (PCIe, UCIe), GPU-to-GPU (NVLink, NVSwitch), node-to-node (Ethernet, Thunderbolt), and cluster-to-cluster (IP, BGP) -- because all these protocols inherit Shannon's forward-in-time-only (FITO) channel model and use Timeout And Retry (TAR) as their failure detector. TAR cannot distinguish "slow" from "dead," which is precisely the ambiguity that Fischer--Lynch--Paterson proved unresolvable in asynchronous systems. We survey the problem using production data from Meta (419 interruptions in 54 days of LLaMA 3 training), ByteDance (38,236 explicit and 5,948 implicit failures in three months), Google (TPUv4 optical circuit switching), and Alibaba (0.057% NIC--ToR link failures per month). At 2025 cluster scale (${\sim}3$ million GPUs, ${>}10$ million optical links), a link flap occurs every 48 seconds. We show that every existing mitigation -- Phi Accrual failure detectors, SWIM, BFD, OSPF/ISIS fast convergence, SmartNIC offload, lossless Ethernet (RoCE/PFC), and Kubernetes pod eviction -- still creates ghosts because each is fundamentally timeout-based. We connect ghosts to gray failures (Huang et al., HotOS 2017) and metastable failures (Bronson et al., HotOS 2021; validated across 22 failures at 11 organizations, OSDI 2022). We argue that Open Atomic Ethernet eliminates ghosts at the link layer through a Reliable Link Failure Detector, Perfect Information Feedback, triangle failover, and atomic token transfer -- making topology knowledge transactional.

The emergence of large-scale, sparse, multimodal, and agentic AI models has coincided with a shift in hardware toward supernode architectures that integrate hundreds to thousands of accelerators with ultra-low-latency interconnects and unified memory pools. However, existing AI frameworks are not designed to exploit these architectures efficiently, leading to high programming complexity, load imbalance, and poor memory utilization. In this paper, we propose a supernode-affinity AI framework that treats the supernode as a single logical computer and embeds hardware-aware orchestration into the framework. Implemented in MindSpore, our HyperParallel architecture comprises HyperOffload for automated hierarchical memory management, HyperMPMD for fine-grained MPMD parallelism across heterogeneous workloads, and HyperShard for declarative parallel strategy specification. Together, these techniques significantly improve training and inference efficiency while reducing parallel programming and system tuning overhead, demonstrating the necessity of supernode affinity for next-generation AI frameworks.

This paper studies how parents want to moderate children's interactions with Generative AI chatbots, with the goal of informing the design of future GenAI parental control tools. We first used an LLM to generate synthetic child-GenAI chatbot interaction scenarios and worked with four parents to validate their realism. From this dataset, we carefully selected 12 diverse examples that evoked varying levels of concern and were rated the most realistic. Each example included a prompt and a GenAI chatbot response. We presented these to parents (N=24) and asked whether they found them concerning, why, and how they would prefer the responses to be modified and communicated. Our findings reveal three key insights: (1) parents express concern about interactions that current GenAI chatbot parental controls neglect; (2) parents want fine-grained transparency and moderation at the conversation level; and (3) parents need personalized controls that adapt to their desired strategies and children's ages.

The volume of freely scraped data on the Internet has driven the tremendous success of deep learning. Along with this comes the growing concern about data privacy and security. Numerous methods for generating unlearnable examples have been proposed to prevent data from being illicitly learned by unauthorized deep models by impeding generalization. However, the existing approaches primarily rely on empirical heuristics, making it challenging to enhance unlearnable examples with solid explanations. In this paper, we analyze and improve unlearnable examples from a novel perspective: mutual information reduction. We demonstrate that effective unlearnable examples always decrease mutual information between clean features and poisoned features, and when the network gets deeper, the unlearnability goes better together with lower mutual information. Further, we prove from a covariance reduction perspective that minimizing the conditional covariance of intra-class poisoned features reduces the mutual information between distributions. Based on the theoretical results, we propose a novel unlearnable method called Mutual Information Unlearnable Examples (MI-UE) that reduces covariance by maximizing the cosine similarity among intra-class features, thus impeding the generalization effectively. Extensive experiments demonstrate that our approach significantly outperforms the previous methods, even under defense mechanisms.

We study a systematic bias in modern image generation models: the mention order of entities in text spuriously determines spatial layout and entity--role binding. We term this phenomenon Order-to-Space Bias (OTS) and show that it arises in both text-to-image and image-to-image generation, often overriding grounded cues and causing incorrect layouts or swapped assignments. To quantify OTS, we introduce OTS-Bench, which isolates order effects with paired prompts differing only in entity order and evaluates models along two dimensions: homogenization and correctness. Experiments show that Order-to-Space Bias (OTS) is widespread in modern image generation models, and provide evidence that it is primarily data-driven and manifests during the early stages of layout formation. Motivated by this insight, we show that both targeted fine-tuning and early-stage intervention strategies can substantially reduce OTS, while preserving generation quality.

Zero-shot MRI reconstruction relies on generative priors, but single-modality unconditional priors produce hallucinations under severe ill-posedness. In many clinical workflows, complementary MRI acquisitions (e.g. high-quality structural scans) are routinely available, yet existing reconstruction methods lack mechanisms to leverage this additional information. We propose MPFlow, a zero-shot multi-modal reconstruction framework built on rectified flow that incorporates auxiliary MRI modalities at inference time without retraining the generative prior to improve anatomical fidelity. Cross-modal guidance is enabled by our proposed self-supervised pretraining strategy, Patch-level Multi-modal MR Image Pretraining (PAMRI), which learns shared representations across modalities. Sampling is jointly guided by data consistency and cross-modal feature alignment using pre-trained PAMRI, systematically suppressing intrinsic and extrinsic hallucinations. Extensive experiments on HCP and BraTS show that MPFlow matches diffusion baselines on image quality using only 20% of sampling steps while reducing tumor hallucinations by more than 15% (segmentation dice score). This demonstrates that cross-modal guidance enables more reliable and efficient zero-shot MRI reconstruction.

Robot planning in partially observable environments, where not all objects are known or visible, is a challenging problem, as it requires reasoning under uncertainty through partially observable Markov decision processes. During the execution of a computed plan, a robot may unexpectedly observe task-irrelevant objects, which are typically ignored by naive planners. In this work, we propose incorporating two types of common-sense knowledge: (1) certain objects are more likely to be found in specific locations; and (2) similar objects are likely to be co-located, while dissimilar objects are less likely to be found together. Manually engineering such knowledge is complex, so we explore leveraging the powerful common-sense reasoning capabilities of large language models (LLMs). Our planning and execution framework, CoCo-TAMP, introduces a hierarchical state estimation that uses LLM-guided information to shape the belief over task-relevant objects, enabling efficient solutions to long-horizon task and motion planning problems. In experiments, CoCo-TAMP achieves an average reduction of 62.7 in planning and execution time in simulation, and 72.6 in real-world demonstrations, compared to a baseline that does not incorporate either type of common-sense knowledge.

In the vision of smart cities, technologies are being developed to enhance the efficiency of urban services and improve residents' quality of life. However, most existing research focuses on optimizing individual services in isolation, without adequately considering reciprocal interactions among heterogeneous urban services that could yield higher efficiency and improved resource utilization. For example, human couriers could collect traffic and air quality data along their delivery routes, while sensing robots could assist with on-demand delivery during peak hours, enhancing both sensing coverage and delivery efficiency. However, the joint optimization of different urban services is challenging due to potentially conflicting objectives and the need for real-time coordination in dynamic environments. In this paper, we propose UrbanHuRo, a two-layer human-robot collaboration framework for joint optimization of heterogeneous urban services, demonstrated through crowdsourced delivery and urban sensing. UrbanHuRo includes two key designs: (i) a scalable distributed MapReduce-based K-submodular maximization module for efficient order dispatch, and (ii) a deep submodular reward reinforcement learning algorithm for sensing route planning. Experimental evaluations on real-world datasets from a food delivery platform demonstrate that UrbanHuRo improves sensing coverage by 29.7% and courier income by 39.2% on average in most settings, while also significantly reducing the number of overdue orders.

Despite the remarkable empirical success of score-based diffusion models, their statistical guarantees remain underdeveloped. Existing analyses often provide pessimistic convergence rates that do not reflect the intrinsic low-dimensional structure common in real data, such as that arising in natural images. In this work, we study the statistical convergence of score-based diffusion models for learning an unknown distribution $μ$ from finitely many samples. Under mild regularity conditions on the forward diffusion process and the data distribution, we derive finite-sample error bounds on the learned generative distribution, measured in the Wasserstein-$p$ distance. Unlike prior results, our guarantees hold for all $p \ge 1$ and require only a finite-moment assumption on $μ$, without compact-support, manifold, or smooth-density conditions. Specifically, given $n$ i.i.d.\ samples from $μ$ with finite $q$-th moment and appropriately chosen network architectures, hyperparameters, and discretization schemes, we show that the expected Wasserstein-$p$ error between the learned distribution $\hatμ$ and $μ$ scales as $\mathbb{E}\, \mathbb{W}_p(\hatμ,μ) = \widetilde{O}\!\left(n^{-1 / d^\ast_{p,q}(μ)}\right),$ where $d^\ast_{p,q}(μ)$ is the $(p,q)$-Wasserstein dimension of $μ$. Our results demonstrate that diffusion models naturally adapt to the intrinsic geometry of data and mitigate the curse of dimensionality, since the convergence rate depends on $d^\ast_{p,q}(μ)$ rather than the ambient dimension. Moreover, our theory conceptually bridges the analysis of diffusion models with that of GANs and the sharp minimax rates established in optimal transport. The proposed $(p,q)$-Wasserstein dimension also extends classical Wasserstein dimension notions to distributions with unbounded support, which may be of independent theoretical interest.

Classifier-Free Guidance (CFG) has established the foundation for guidance mechanisms in diffusion models, showing that well-designed guidance proxies significantly improve conditional generation and sample quality. Autoguidance (AG) has extended this idea, but it relies on an auxiliary network and leaves solver-induced errors unaddressed. In stiff regions, the ODE trajectory changes sharply, where local truncation error (LTE) becomes a critical factor that deteriorates sample quality. Our key observation is that these errors align with the dominant eigenvector, motivating us to leverage the solver-induced error as a guidance signal. We propose Embedded Runge-Kutta Guidance (ERK-Guid), which exploits detected stiffness to reduce LTE and stabilize sampling. We theoretically and empirically analyze stiffness and eigenvector estimators with solver errors to motivate the design of ERK-Guid. Our experiments on both synthetic datasets and the popular benchmark dataset, ImageNet, demonstrate that ERK-Guid consistently outperforms state-of-the-art methods. Code is available at https://github.com/mlvlab/ERK-Guid.

Automated design of chemical formulations is a cornerstone of materials science, yet it requires navigating a high-dimensional combinatorial space involving discrete compositional choices and continuous geometric constraints. Existing Large Language Model (LLM) agents face significant challenges in this setting, including context window limitations during long-horizon reasoning and path-dependent exploration that may lead to mode collapse. To address these issues, we introduce AI4S-SDS, a closed-loop neuro-symbolic framework that integrates multi-agent collaboration with a tailored Monte Carlo Tree Search (MCTS) engine. We propose a Sparse State Storage mechanism with Dynamic Path Reconstruction, which decouples reasoning history from context length and enables arbitrarily deep exploration under fixed token budgets. To reduce local convergence and improve coverage, we implement a Global--Local Search Strategy: a memory-driven planning module adaptively reconfigures the search root based on historical feedback, while a Sibling-Aware Expansion mechanism promotes orthogonal exploration at the node level. Furthermore, we bridge symbolic reasoning and physical feasibility through a Differentiable Physics Engine, employing a hybrid normalized loss with sparsity-inducing regularization to optimize continuous mixing ratios under thermodynamic constraints. Empirical results show that AI4S-SDS achieves full validity under the adopted HSP-based physical constraints and substantially improves exploration diversity compared to baseline agents. In preliminary lithography experiments, the framework identifies a novel photoresist developer formulation that demonstrates competitive or superior performance relative to a commercial benchmark, highlighting the potential of diversity-driven neuro-symbolic search for scientific discovery.

Recent developments show that AI can prove research-level theorems in mathematics, both formally and informally. This essay urges mathematicians to stay up-to-date with the technology, to consider the ways it will disrupt mathematical practice, and to respond appropriately to the challenges and opportunities we now face.

Leveraging Large Language Models (LLMs) for code generation has increasingly emerged as a common practice in the domain of software engineering. Relevant benchmarks have been established to evaluate the code generation capabilities of LLMs. However, existing benchmarks focus primarily on sequential code, lacking the ability to effectively evaluate LLMs on concurrent code generation. Compared to sequential code, concurrent code exhibits greater complexity and possesses unique types of bugs, such as deadlocks and race conditions, that do not occur in sequential code. Therefore, a benchmark for evaluating sequential code generation cannot be useful for evaluating concurrent code generation with LLMs. To address this gap, we designed a benchmark CONCUR specifically aimed at evaluating the capability of LLMs to generate concurrent code. CONCUR consists of a base set of 43 concurrency problems derived from a standard concurrency textbook, together with 72 validated mutant variants, resulting in 115 total problems. The base problems serve as the semantic core of the benchmark, while the mutants expand linguistic and structural diversity. We conducted an evaluation of a range of LLMs on CONCUR, highlighting limitations of current models. Overall, our work provides a novel direction for evaluating the capability of LLMs to generate code with focus on concurrency.

Multimodal Large Language Models (MLLMs) have shown strong performance in vision-language tasks, but their inference efficiency is severely limited by the exponential growth of visual tokens in complex scenarios such as high-resolution images and videos. Existing visual token pruning methods mainly operate after visual encoding, overlooking the substantial computational cost incurred during the encoding stage. To address this issue, we propose EvoPrune, an early-stage visual token pruning method for MLLMs that performs pruning directly during visual encoding. Specifically, EvoPrune employs a layer-wise pruning strategy guided by token similarity, diversity, and attention-based importance to retain the most informative visual tokens at selected encoding layers. Extensive experiments on image and video benchmarks validate the effectiveness of EvoPrune. In particular, on the VideoMME dataset, EvoPrune achieves 2$\times$ inference speedup with less than 1% performance degradation, demonstrating its potential for latency-sensitive MLLM deployment.

Large Language Model (LLM) agents have demonstrated remarkable proficiency in learned tasks, yet they often struggle to adapt to non-stationary environments with feedback. While In-Context Learning and external memory offer some flexibility, they fail to internalize the adaptive ability required for long-term improvement. Meta-Reinforcement Learning (meta-RL) provides an alternative by embedding the learning process directly within the model. However, existing meta-RL approaches for LLMs focus primarily on exploration in single-agent settings, neglecting the strategic exploitation necessary for multi-agent environments. We propose MAGE, a meta-RL framework that empowers LLM agents for strategic exploration and exploitation. MAGE utilizes a multi-episode training regime where interaction histories and reflections are integrated into the context window. By using the final episode reward as the objective, MAGE incentivizes the agent to refine its strategy based on past experiences. We further combine population-based training with an agent-specific advantage normalization technique to enrich agent diversity and ensure stable learning. Experiment results show that MAGE outperforms existing baselines in both exploration and exploitation tasks. Furthermore, MAGE exhibits strong generalization to unseen opponents, suggesting it has internalized the ability for strategic exploration and exploitation. Code is available at https://github.com/Lu-Yang666/MAGE.

Large language models (LLMs) have advanced medical dialogue systems, yet psychiatric consultation poses substantially higher demands due to subjective ambiguity and comorbidity complexity: an agent must continuously extract psychopathological cues from incomplete and inconsistent patient reports in multi-turn interactions and perform rigorous differential diagnostic reasoning. However, existing methods face two fundamental challenges. First, without criteria-grounded clinical supports, they are prone to unsupported clinical assertions when symptoms are atypical or underspecified. Second, in multi-turn interactions, they struggle to mitigate inquiry drift (off-topic or low-yield questioning) and optimize questioning strategies. To address these challenges, we propose MIND, a unified inquiry--diagnosis reinforcement learning framework for psychiatric consultation. Specifically, we build a Criteria-Grounded Psychiatric Reasoning Bank (PRB) that summarizes dialogue context into clinical retrieval states, retrieves semantically similar reference consultations, and distills reusable criteria-grounded clinical supports to guide criteria-aligned inquiry and reasoning. Building on this foundation, MIND enforces explicit clinical reasoning with rubric-based process rewards to provide fine-grained supervision over intermediate decision steps, and incorporates a value-aware trajectory rectification mechanism to jointly improve information acquisition and diagnostic decision-making across turns. Extensive experiments demonstrate that MIND consistently outperforms strong baselines in diagnostic accuracy, empathetic interaction quality, interpretability, and generalization.

Stein's identity is a fundamental tool in machine learning with applications in generative models, stochastic optimization, and other problems involving gradients of expectations under Gaussian distributions. Less attention has been paid to problems with non-Gaussian expectations. Here, we consider the class of bounded-support $q$-Gaussians and derive a new Stein identity leading to gradient estimators which have nearly identical forms to the Gaussian ones, and which are similarly easy to implement. We do this by extending the previous results of Landsman, Vanduffel, and Yao (2013) to prove new Bonnet- and Price-type theorems for q-Gaussians. We also simplify their forms by using escort distributions. Our experiments show that bounded-support distributions can reduce the variance of gradient estimators, which can potentially be useful for Bayesian deep learning and sharpness-aware minimization. Overall, our work simplifies the application of Stein's identity for an important class of non-Gaussian distributions.

The Shapley value provides a principled foundation for data valuation, but exact computation is #P-hard due to the exponential coalition space. Existing accelerations remain global and ignore a structural property of modern predictors: for a given test instance, only a small subset of training points influences the prediction. We formalize this model-induced locality through support sets defined by the model's computational pathway (e.g., neighbors in KNN, leaves in trees, receptive fields in GNNs), showing that Shapley computation can be projected onto these supports without loss when locality is exact. This reframes Shapley evaluation as a structured data processing problem over overlapping support-induced subset families rather than exhaustive coalition enumeration. We prove that the intrinsic complexity of Local Shapley is governed by the number of distinct influential subsets, establishing an information-theoretic lower bound on retraining operations. Guided by this result, we propose LSMR (Local Shapley via Model Reuse), an optimal subset-centric algorithm that trains each influential subset exactly once via support mapping and pivot scheduling. For larger supports, we develop LSMR-A, a reuse-aware Monte Carlo estimator that remains unbiased with exponential concentration, with runtime determined by the number of distinct sampled subsets rather than total draws. Experiments across multiple model families demonstrate substantial retraining reductions and speedups while preserving high valuation fidelity.

The proliferation of facial recognition (FR) systems has raised privacy concerns in the digital realm, as malicious uses of FR models pose a significant threat. Traditional countermeasures, such as makeup style transfer, have suffered from low transferability in black-box settings and limited applicability across various demographic groups, including males and individuals with darker skin tones. To address these challenges, we introduce a novel facial privacy protection method, dubbed \textbf{MAP}, a pioneering approach that employs human emotion modifications to disguise original identities as target identities in facial images. Our method uniquely fine-tunes a score network to learn dual objectives, target identity and human expression, which are jointly optimized through gradient projection to ensure convergence at a shared local optimum. Additionally, we enhance the perceptual quality of protected images by applying local smoothness regularization and optimizing the score matching loss within our network. Empirical experiments demonstrate that our innovative approach surpasses previous baselines, including noise-based, makeup-based, and freeform attribute methods, in both qualitative fidelity and quantitative metrics. Furthermore, MAP proves its effectiveness against an online FR API and shows advanced adaptability in uncommon photographic scenarios.

Learning-to-communicate (LTC) in partially observable environments has received increasing attention in deep multi-agent reinforcement learning, where the control and communication strategies are jointly learned. Meanwhile, the impact of communication on decision-making has been extensively studied in control theory. In this paper, we seek to formalize and better understand LTC by bridging these two lines of work, through the lens of information structures (ISs). To this end, we formalize LTC in decentralized partially observable Markov decision processes (Dec-POMDPs) under the common-information-based framework from decentralized stochastic control, and classify LTC problems based on the ISs before (additional) information sharing. We first show that non-classical LTCs are computationally intractable in general, and thus focus on quasi-classical (QC) LTCs. We then propose a series of conditions for QC LTCs, under which LTCs preserve the QC IS after information sharing, whereas violating which can cause computational hardness in general. Further, we develop provable planning and learning algorithms for QC LTCs, and establish quasi-polynomial time and sample complexities for several QC LTC examples that satisfy the above conditions. Along the way, we also establish results on the relationship between (strictly) QC IS and the condition of having strategy-independent common-information-based beliefs (SI-CIBs), as well as on solving Dec-POMDPs without computationally intractable oracles but beyond those with SI-CIBs, which may be of independent interest.

Graph Neural Networks (GNNs) have emerged as a powerful framework for processing graph-structured data. However, conventional GNNs and their variants are inherently limited by the homophily assumption, leading to degradation in performance on heterophilic graphs. Although substantial efforts have been made to mitigate this issue, they remain constrained by the message-passing paradigm, which is inherently rooted in homophily. In this paper, a detailed analysis of how the underlying label autocorrelation of the homophily assumption introduces bias into GNNs is presented. We innovatively leverage a negative feedback mechanism to correct the bias and propose Graph Negative Feedback Bias Correction (GNFBC), a simple yet effective framework that is independent of any specific aggregation strategy. Specifically, we introduce a negative feedback loss that penalizes the sensitivity of predictions to label autocorrelation. Furthermore, we incorporate the output of graph-agnostic models as a feedback term, leveraging independent node feature information to counteract correlation-induced bias guided by Dirichlet energy. GNFBC can be seamlessly integrated into existing GNN architectures, improving overall performance with comparable computational and memory overhead.

Multimodal generative models have made significant strides in image editing, demonstrating impressive performance on a variety of static tasks. However, their proficiency typically does not extend to complex scenarios requiring dynamic reasoning, leaving them ill-equipped to model the coherent, intermediate logical pathways that constitute a multi-step evolution from an initial state to a final one. This capacity is crucial for unlocking a deeper level of procedural and causal understanding in visual manipulation. To systematically measure this critical limitation, we introduce InEdit-Bench, the first evaluation benchmark dedicated to reasoning over intermediate pathways in image editing. InEdit-Bench comprises meticulously annotated test cases covering four fundamental task categories: state transition, dynamic process, temporal sequence, and scientific simulation. Additionally, to enable fine-grained evaluation, we propose a set of assessment criteria to evaluate the logical coherence and visual naturalness of the generated pathways, as well as the model's fidelity to specified path constraints. Our comprehensive evaluation of 14 representative image editing models on InEdit-Bench reveals significant and widespread shortcomings in this domain. By providing a standardized and challenging benchmark, we aim for InEdit-Bench to catalyze research and steer development towards more dynamic, reason-aware, and intelligent multimodal generative models.

Tool-augmented large language model (LLM) agents promise to unify scientific reasoning with computation, yet their deployment in high-stakes domains like drug discovery is bottlenecked by two critical barriers: unconstrained tool-use governance and poor long-horizon reliability. In dependency-heavy pharmaceutical pipelines, autonomous agents often drift into irreproducible trajectories, where early-stage hallucinations multiplicatively compound into downstream failures. To overcome this, we present Mozi, a dual-layer architecture that bridges the flexibility of generative AI with the deterministic rigor of computational biology. Layer A (Control Plane) establishes a governed supervisor--worker hierarchy that enforces role-based tool isolation, limits execution to constrained action spaces, and drives reflection-based replanning. Layer B (Workflow Plane) operationalizes canonical drug discovery stages -- from Target Identification to Lead Optimization -- as stateful, composable skill graphs. This layer integrates strict data contracts and strategic human-in-the-loop (HITL) checkpoints to safeguard scientific validity at high-uncertainty decision boundaries. Operating on the design principle of ``free-form reasoning for safe tasks, structured execution for long-horizon pipelines,'' Mozi provides built-in robustness mechanisms and trace-level audibility to completely mitigate error accumulation. We evaluate Mozi on PharmaBench, a curated benchmark for biomedical agents, demonstrating superior orchestration accuracy over existing baselines. Furthermore, through end-to-end therapeutic case studies, we demonstrate Mozi's ability to navigate massive chemical spaces, enforce stringent toxicity filters, and generate highly competitive in silico candidates, effectively transforming the LLM from a fragile conversationalist into a reliable, governed co-scientist.

Construction aggregates, including sand and gravel, crushed stone and riprap, are the core building blocks of the construction industry. State-of-the-practice characterization methods mainly relies on visual inspection and manual measurement. State-of-the-art aggregate imaging methods have limitations that are only applicable to regular-sized aggregates under well-controlled conditions. This dissertation addresses these major challenges by developing a field imaging framework for the morphological characterization of aggregates as a multi-scenario solution. For individual and non-overlapping aggregates, a field imaging system was designed and the associated segmentation and volume estimation algorithms were developed. For 2D image analyses of aggregates in stockpiles, an automated 2D instance segmentation and morphological analysis approach was established. For 3D point cloud analyses of aggregate stockpiles, an integrated 3D Reconstruction-Segmentation-Completion (RSC-3D) approach was established: 3D reconstruction procedures from multi-view images, 3D stockpile instance segmentation, and 3D shape completion to predict the unseen sides. First, a 3D reconstruction procedure was developed to obtain high-fidelity 3D models of collected aggregate samples, based on which a 3D aggregate particle library was constructed. Next, two datasets were derived from the 3D particle library for 3D learning: a synthetic dataset of aggregate stockpiles with ground-truth instance labels, and a dataset of partial-complete shape pairs, developed with varying-view raycasting schemes. A state-of-the-art 3D instance segmentation network and a 3D shape completion network were trained on the datasets, respectively. The application of the integrated approach was demonstrated on real stockpiles and validated with ground-truth, showing good performance in capturing and predicting the unseen sides of aggregates.

Short text classification (STC) remains a challenging task due to the scarcity of contextual information and labeled data. However, existing approaches have pre-dominantly focused on English because most benchmark datasets for the STC are primarily available in English. Consequently, existing methods seldom incorporate the linguistic and structural characteristics of Korean, such as its agglutinative morphology and flexible word order. To address these limitations, we propose LIGRAM, a hierarchical heterogeneous graph model for Korean short-text classification. The proposed model constructs sub-graphs at the morpheme, part-of-speech, and named-entity levels and hierarchically integrates them to compensate for the limited contextual information in short texts while precisely capturing the grammatical and semantic dependencies inherent in Korean. In addition, we apply Semantics-aware Contrastive Learning (SemCon) to reflect semantic similarity across documents, enabling the model to establish clearer decision boundaries even in short texts where class distinctions are often ambiguous. We evaluate LIGRAM on four Korean short-text datasets, where it consistently outperforms existing baseline models. These outcomes validate that integrating language-specific graph representations with SemCon provides an effective solution for short text classification in agglutinative languages such as Korean.

Freezing of Gait (FOG) is a debilitating motor symptom commonly experienced by individuals with Parkinson's Disease (PD) which often leads to falls and reduced mobility. Timely and accurate prediction of FOG episodes is essential for enabling proactive interventions through assistive technologies. This study presents a reinforcement learning-based framework designed to identify optimal pre-FOG onset points, thereby extending the prediction horizon for anticipatory cueing systems. The model implements a Double Deep Q-Network (DDQN) architecture enhanced with Prioritized Experience Replay (PER) allowing the agent to focus learning on high-impact experiences and refine its policy. Trained over 9000 episodes with a reward shaping strategy that promotes cautious decision-making, the agent demonstrated robust performance in both subject-dependent and subject-independent evaluations. The model achieved a prediction horizon of up to 8.72 seconds prior to FOG onset in subject-independent scenarios and 7.89 seconds in subject-dependent settings. These results highlight the model's potential for integration into wearable assistive devices, offering timely and personalized interventions to mitigate FOG in PD patients.

Physical dynamical systems can be viewed as natural information processors: their systems preserve, transform, and disperse input information. This perspective motivates learning not only from data generated by such systems, but also how to measure them in a way that extracts the most useful information for a given task. We propose a general computing framework for adaptive information extraction from dynamical systems, in which a trainable attention module learns both where to probe the system state and how to combine these measurements to optimize prediction performance. As a concrete instantiation, we implement this idea using a spatiotemporal field governed by a partial differential equation as the underlying dynamics, though the framework applies equally to any system whose state can be sampled. Our results show that adaptive spatial sensing significantly improves prediction accuracy on canonical chaotic benchmarks. This work provides a perspective on attention-enhanced reservoir computing as a special case of a broader paradigm: neural networks as trainable measurement devices for extracting information from physical dynamical systems.

Educational games can foster critical thinking, problem-solving, and motivation, yet instructors often find it difficult to design games that reliably achieve specific learning outcomes. Existing authoring environments reduce the need for programming expertise, but they do not eliminate the underlying challenges of educational game design, and they can leave non-expert designers reliant on opaque suggestions from AI systems. We designed a controlled natural language framework-based web tool that positions language as the primary interface for LLM-assisted educational game design. In the tool, users and an LLM assistant collaboratively develop a structured language that maps pedagogy to gameplay through four linked components. We argue that, by making pedagogical intent explicit and editable in the interface, the tool has the potential to lower design barriers for non-expert designers, preserves human agency in critical decisions, and enables alignment and reflections between pedagogy and gameplay during and after co-creation.

Multimodal Large Language Models (MLLMs) integrate vision and text to power applications, but this integration introduces new vulnerabilities. We study Image-based Prompt Injection (IPI), a black-box attack in which adversarial instructions are embedded into natural images to override model behavior. Our end-to-end IPI pipeline incorporates segmentation-based region selection, adaptive font scaling, and background-aware rendering to conceal prompts from human perception while preserving model interpretability. Using the COCO dataset and GPT-4-turbo, we evaluate 12 adversarial prompt strategies and multiple embedding configurations. The results show that IPI can reliably manipulate the output of the model, with the most effective configuration achieving up to 64\% attack success under stealth constraints. These findings highlight IPI as a practical threat in black-box settings and underscore the need for defenses against multimodal prompt injection.

While incorporating LLMs into systems offers significant benefits in critical application areas such as healthcare, new security challenges emerge due to the potential cyber kill chain cycles that combine adversarial model, prompt injection and conventional cyber attacks. Threat modeling methods enable the system designers to identify potential cyber threats and the relevant mitigations during the early stages of development. Although the cyber security community has extensive experience in applying these methods to software-based systems, the elicited threats are usually abstract and vague, limiting their effectiveness for conducting proper likelihood and impact assessments for risk prioritization, especially in complex systems with novel attacks surfaces, such as those involving LLMs. In this study, we propose a structured, goal driven risk assessment approach that contextualizes the threats with detailed attack vectors, preconditions, and attack paths through the use of attack trees. We demonstrate the proposed approach on a case study with an LLM agent-based healthcare system. This study harmonizes the state-of-the-art attacks to LLMs with conventional ones and presents possible attack paths applicable to similar systems. By providing a structured risk assessment, this study makes a significant contribution to the literature and advances the secure-by-design practices in LLM-based systems.

User models in information retrieval rest on a foundational assumption that observed behavior reveals intent. This assumption collapses when the user is an AI agent privately configured by a human operator. For any action an agent takes, a hidden instruction could have produced identical output - making intent non-identifiable at the individual level. This is not a detection problem awaiting better tools; it is a structural property of any system where humans configure agents behind closed doors. We investigate the agent-user problem through a large-scale corpus from an agent-native social platform: 370K posts from 47K agents across 4K communities. Our findings are threefold: (1) individual agent actions cannot be classified as autonomous or operator-directed from observables; (2) population-level platform signals still separate agents into meaningful quality tiers, but a click model trained on agent interactions degrades steadily (-8.5% AUC) as lower-quality agents enter training data; (3) cross-community capability references spread endemically ($R_0$ 1.26-3.53) and resist suppression even under aggressive modeled intervention. For retrieval systems, the question is no longer whether agent users will arrive, but whether models built on human-intent assumptions will survive their presence.

Low-dimensional structure in real-world data plays an important role in the success of generative models, which motivates diffusion models defined on intrinsic data manifolds. Such models are driven by stochastic differential equations (SDEs) on manifolds, which raises the need for convergence theory of numerical schemes for manifold-valued SDEs. In Euclidean space, the Euler--Maruyama (EM) scheme achieves strong convergence with order $1/2$, but an analogous result for manifold discretizations is less understood in general settings. In this work, we study a geometric version of the EM scheme for SDEs on Riemannian manifolds and prove strong convergence with order $1/2$ under geometric and regularity conditions. As an application, we obtain a Wasserstein bound for sampling on manifolds via the geometric EM discretization of Riemannian Langevin dynamics.

The growing use of unstructured text in business research makes topic modeling a central tool for constructing explanatory variables from reviews, social media, and open-ended survey responses, yet existing approaches function poorly as measurement instruments. Prior work shows that textual content predicts outcomes such as sales, satisfaction, and firm performance, but probabilistic models often generate conceptually diffuse topics, neural topic models are difficult to interpret in theory-driven settings, and large language model approaches lack standardization, stability, and alignment with document-level representations. We introduce LX Topic, a neural topic method that conceptualizes topics as latent linguistic constructs and produces calibrated document-level topic proportions for empirical analysis. LX Topic builds on FASTopic to ensure strong document representativeness and integrates large language model refinement at the topic-word level using alignment and confidence-weighting mechanisms that enhance semantic coherence without distorting document-topic distributions. Evaluations on large-scale Amazon and Yelp review datasets demonstrate that LX Topic achieves the highest overall topic quality relative to leading models while preserving clustering and classification performance. By unifying topic discovery, refinement, and standardized output in a web-based system, LX Topic establishes topic modeling as a reproducible, interpretable, and measurement-oriented instrument for marketing research and practice.

We develop a rigorous framework for extending neural operators to handle out-of-distribution input functions. We leverage kernel approximation techniques and provide theory for characterizing the input-output function spaces in terms of Reproducing Kernel Hilbert Spaces (RKHSs). We provide theorems on the requirements for reliable extensions and their predicted approximation accuracy. We also establish formal relationships between specific kernel choices and their corresponding Sobolev Native Spaces. This connection further allows the extended neural operators to reliably capture not only function values but also their derivatives. Our methods are empirically validated through the solution of elliptic partial differential equations (PDEs) involving operators on manifolds having point-cloud representations and handling geometric contributions. We report results on key factors impacting the accuracy and computational performance of the extension approaches.

The No Free Lunch (NFL) theorem guarantees equal average performance only under uniform sampling of a function space closed under permutation (c.u.p.). We ask when this averaging ceases to reflect what benchmarking actually reports. We study an iterative-search setting with sampling without replacement, where algorithms differ only in evaluation order. Binary objectives allow exhaustive evaluation in the fully enumerable case, and efficiency is defined by the first time the global minimum is reached. We then construct two additional benchmarks by algebraically recombining the same baseline functions through sums and differences. Function-algorithm relations are examined via correlation structure, hierarchical clustering, delta heatmaps, and PCA. A one-way ANOVA with Tukey contrasts confirms that algebraic reformulations induce statistically meaningful shifts in performance patterns. The uniformly sampled baseline remains consistent with the global NFL symmetry. In contrast, the algebraically modified benchmarks yield stable re-rankings and coherent clusters of functions and sampling policies. Composite objectives can also exhibit non-additive search effort despite being built from simpler components. Monte Carlo experiments indicate that order effects persist in larger spaces and depend on function class. Taken together, the results show how objective reformulation and benchmark design can generate structured local departures from NFL intuition. They motivate algorithm choice that is aware of both the problem class and the objective representation. This message applies to evolutionary computation as well as to statistical procedures based on relabeling, resampling, and permutation tests.

The community is increasingly exploring linear RNNs (LRNNs) as language models, motivated by their expressive power and parallelizability. While prior work establishes the expressivity benefits of LRNNs over transformers, it is unclear what makes LRNNs -- but not traditional, nonlinear RNNs -- as easy to parallelize in practice as transformers. We answer this question by providing a tight connection between types of RNNs and standard complexity classes. We show that LRNNs can be viewed as log-depth (bounded fan-in) arithmetic circuits, which represents only a slight depth overhead relative to log-depth boolean circuits that transformers admit. Furthermore, we show that nonlinear RNNs can solve $\mathsf{L}$-complete problems (and even $\mathsf{P}$-complete ones, under polynomial precision), revealing a fundamental barrier to parallelizing them as efficiently as transformers. Our theory also identifies fine-grained expressivity differences between recent popular LRNN variants: permutation-diagonal LRNNs are $\mathsf{NC}^1$-complete whereas diagonal-plus-low-rank LRNNs are more expressive ($\mathsf{PNC}^1$-complete). We provide further insight by associating each type of RNN with a corresponding automata-theoretic model that it can simulate. Together, our results reveal fundamental tradeoffs between nonlinear RNNs and different variants of LRNNs, providing a foundation for designing LLM architectures that achieve an optimal balance between expressivity and parallelism.

Understanding how systems built out of modular components can be jointly optimized is an important problem in biology, engineering, and machine learning. The backpropagation algorithm is one such solution and has been instrumental in the success of neural networks. Despite its empirical success, a strong theoretical understanding of it is lacking. Here, we combine tools from Riemannian geometry, optimal control theory, and theoretical physics to advance this understanding. We make three key contributions: First, we revisit the derivation of backpropagation as a constrained optimization problem and combine it with the insight that Riemannian gradient descent trajectories can be understood as the minimum of an action. Second, we introduce a recursively defined layerwise Riemannian metric that exploits the modular structure of neural networks and can be efficiently computed using the Woodbury matrix identity, avoiding the $O(n^3)$ cost of full metric inversion. Third, we develop a framework of composable ``Riemannian modules'' whose convergence properties can be quantified using nonlinear contraction theory, providing algorithmic stability guarantees of order $O(κ^2 L/(ξμ\sqrt{n}))$ where $κ$ and $L$ are Lipschitz constants, $μ$ is the mass matrix scale, and $ξ$ bounds the condition number. Our layerwise metric approach provides a practical alternative to natural gradient descent. While we focus here on studying neural networks, our approach more generally applies to the study of systems made of modules that are optimized over time, as it occurs in biology during both evolution and development.

Convolutional Neural Networks (CNNs) have achieved state-of-the-art accuracy in Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR). However, their high computational cost, latency, and memory footprint make its deployment challenging on resource-constrained platforms such as small satellites. While adversarial robustness is critical for real-world SAR ATR, it is often overlooked in system-level optimizations. Achieving both robustness and inference efficiency requires a unified framework that considers adversarially trained models together with hardware constraints. We present a model-hardware co-design framework for CNN-based SAR ATR that integrates robustness-preserving model compression with FPGA accelerator design. The compression stage includes hardware-guided structured pruning, where a hardware performance model derived from the FPGA design predicts the pruning impact on latency and resource usage. This enables the generation of Pareto-optimal models that improve hardware efficiency under user-defined objectives, while maintaining adversarial robustness within a predefined tolerance. We design an FPGA accelerator with channel-aware Processing Element (PE) allocation that supports both fully pipelined streaming and temporal resource-reuse architectures. An automated design generation flow efficiently maps the compressed models to optimized FPGA implementations. Experiments on the widely used MSTAR and FUSAR-Ship datasets across three CNN architectures show that our framework produces models up to 18.3x smaller with 3.1x fewer MACs while preserving robustness. Our FPGA implementation achieves up to 68.1x (6.4x) lower inference latency and up to 169.7x (33.2x) better energy efficiency compared to CPU (GPU) baselines, demonstrating the effectiveness of the proposed co-design framework for robust and efficient SAR ATR on FPGA platforms.

The rapid progress of large language models (LLMs) is increasingly constrained by memory and deployment costs, motivating compression methods for practical deployment. Many state-of-the-art compression pipelines leverage the low-rank structure of trained weight matrices, a phenomenon often associated with the properties of popular optimizers such as Adam. In this context, Muon is a recently proposed optimizer that improves LLM pretraining via full-rank update steps, but its induced weight-space structure has not been characterized yet. In this work, we report a surprising empirical finding: despite imposing full-rank updates, Muon-trained models exhibit pronounced low-rank structure in their weight matrices and are readily compressible under standard pipelines. Motivated by this insight, we propose NuMuon, which augments Muon with a nuclear-norm constraint on the update direction, further constraining the learned weights toward low-rank structure. Across billion-parameter-scale models, we show that NuMuon increases weight compressibility and improves post-compression model quality under state-of-the-art LLM compression pipelines while retaining Muon's favorable convergence behavior.

Conventionally, memory in end-to-end robotic learning involves inputting a sequence of past observations into the learned policy. However, in complex multi-stage real-world tasks, the robot's memory must represent past events at multiple levels of granularity: from long-term memory that captures abstracted semantic concepts (e.g., a robot cooking dinner should remember which stages of the recipe are already done) to short-term memory that captures recent events and compensates for occlusions (e.g., a robot remembering the object it wants to pick up once its arm occludes it). In this work, our main insight is that an effective memory architecture for long-horizon robotic control should combine multiple modalities to capture these different levels of abstraction. We introduce Multi-Scale Embodied Memory (MEM), an approach for mixed-modal long-horizon memory in robot policies. MEM combines video-based short-horizon memory, compressed via a video encoder, with text-based long-horizon memory. Together, they enable robot policies to perform tasks that span up to fifteen minutes, like cleaning up a kitchen, or preparing a grilled cheese sandwich. Additionally, we find that memory enables MEM policies to intelligently adapt manipulation strategies in-context.

Coordinating multiple autonomous agents to explore and serve spatially heterogeneous demand requires jointly learning unknown spatial patterns and planning trajectories that maximize task performance. Pure model-based approaches provide structured uncertainty estimates but lack adaptive policy learning, while deep reinforcement learning often suffers from poor sample efficiency when spatial priors are absent. This paper presents a hybrid belief-reinforcement learning (HBRL) framework to address this gap. In the first phase, agents construct spatial beliefs using a Log-Gaussian Cox Process (LGCP) and execute information-driven trajectories guided by a Pathwise Mutual Information (PathMI) planner with multi-step lookahead. In the second phase, trajectory control is transferred to a Soft Actor-Critic (SAC) agent, warm-started through dual-channel knowledge transfer: belief state initialization supplies spatial uncertainty, and replay buffer seeding provides demonstration trajectories generated during LGCP exploration. A variance-normalized overlap penalty enables coordinated coverage through shared belief state, permitting cooperative sensing in high-uncertainty regions while discouraging redundant coverage in well-explored areas. The framework is evaluated on a multi-UAV wireless service provisioning task. Results show 10.8% higher cumulative reward and 38% faster convergence over baselines, with ablation studies confirming that dual-channel transfer outperforms either channel alone.

Decentralized training introduces critical security risks when executed across untrusted, geographically distributed nodes. While existing Byzantine-tolerant literature addresses data parallel (DP) training through robust aggregation methods, pipeline parallelism (PP) presents fundamentally distinct challenges. In PP, model layers are distributed across workers where the activations and their gradients flow between stages rather than being aggregated, making traditional DP approaches inapplicable. We propose SENTINEL, a verification mechanism for PP training without computation duplication. SENTINEL employs lightweight momentum-based monitoring using exponential moving averages (EMAs) to detect corrupted inter-stage communication. Unlike existing Byzantine-tolerant approaches for DP that aggregate parameter gradients across replicas, our approach verifies sequential activation/gradient transmission between layers. We provide theoretical convergence guarantees for this new setting that recovers classical convergence rates when relaxed to standard training. Experiments demonstrate successful training of up to 4B-parameter LLMs across untrusted distributed environments with up to 176 workers while maintaining model convergence and performance.

In Multi-Agent Systems (MAS), agents are designed with social capabilities, allowing them to understand and reason about social concepts such as norms when interacting with others (e.g., inter-robot interactions). In Normative MAS (NorMAS), researchers study how norms develop, and how violations are detected and sanctioned. However, existing research in NorMAS use symbolic approaches (e.g., formal logic) for norm representation and reasoning whose application is limited to simplified environments. In contrast, Multimodal Large Language Models (MLLMs) present promising possibilities to develop software used by robots to identify and reason about norms in a wide variety of complex social situations embodied in text and images. However, prior work on norm reasoning have been limited to text-based scenarios. This paper investigates the norm reasoning competence of five MLLMs by evaluating their ability to answer norm-related questions based on thirty text-based and thirty image-based stories, and comparing their responses against humans. Our results show that MLLMs demonstrate superior performance in norm reasoning in text than in images. GPT-4o performs the best in both modalities offering the most promise for integration with MAS, followed by the free model Qwen-2.5VL. Additionally, all models find reasoning about complex norms challenging.

Recent advances in large language models (LLMs) transform how machine learning (ML) pipelines are developed and evaluated. LLMs enable a new type of workload, agentic pipeline search, in which autonomous or semi-autonomous agents generate, validate, and optimize complete ML pipelines. These agents predominantly operate over popular Python ML libraries and exhibit highly exploratory behavior. This results in thousands of executions for data profiling, pipeline generation, and iterative refinement of pipeline stages. However, the existing Python-based ML ecosystem is built around libraries such as Pandas and scikit-learn, which are designed for human-centric, interactive, sequential workflows and remain constrained by Python's interpretive execution model, library-level isolation, and limited runtime support for executing large numbers of pipelines. Meanwhile, many high-performance ML systems proposed by the systems community either target narrow workload classes or require specialized programming models, which limits their integration with the Python ML ecosystem and makes them largely ill-suited for LLM-based agents. This growing mismatch exposes a fundamental systems challenge in supporting agentic pipeline search at scale. We therefore propose stratum, a unified system infrastructure that decouples pipeline execution from planning and reasoning during agentic pipeline search. Stratum integrates seamlessly with existing Python libraries, compiles batches of pipelines into optimized execution graphs, and efficiently executes them across heterogeneous backends, including a novel Rust-based runtime. We present stratum's architectural vision along with an early prototype, discuss key design decisions, and outline open challenges and research directions. Finally, preliminary experiments show that stratum can significantly speed up large-scale agentic pipeline search up to 16.6x.

Method validation and study design in causal inference rely on synthetic data with known counterfactuals. Existing simulators trade off distributional realism, the ability to capture mixed-type and multimodal tabular data, against causal controllability, including explicit control over overlap, unmeasured confounding, and treatment effect heterogeneity. We introduce CausalMix, a variational generative framework that closes this gap by coupling a mixture of Gaussian latent priors with data-type-specific decoders for continuous, binary, and categorical variables. The model incorporates explicit causal controls: an overlap regularizer shaping propensity-score distributions, alongside direct parameterizations of confounding strength and effect heterogeneity. This unified objective preserves fidelity to the observed data while enabling factorial manipulation of causal mechanisms, allowing overlap, confounding strength, and treatment effect heterogeneity to be varied independently at design time. Across benchmarks, CausalMix achieves state-of-the-art distributional metrics on mixed-type tables while providing stable, fine-grained causal control. We demonstrate practical utility in a comparative safety study of metastatic castration-resistant prostate cancer treatments, using CausalMix to compare estimators under calibrated data-generating processes, tune hyperparameters, and conduct simulation-based power analyses under targeted treatment effect heterogeneity scenarios.

Misinformation is a growing societal threat, and susceptibility to misinformative claims varies across demographic groups due to differences in underlying beliefs. As Large Language Models (LLMs) are increasingly used to simulate human behaviors, we investigate whether they can simulate demographic misinformation susceptibility, treating beliefs as a primary driving factor. We introduce BeliefSim, a simulation framework that constructs demographic belief profiles using psychology-informed taxonomies and survey priors. We study prompt-based conditioning and post-training adaptation, and conduct a multi-fold evaluation using: (i) susceptibility accuracy and (ii) counterfactual demographic sensitivity. Across both datasets and modeling strategies, we show that beliefs provide a strong prior for simulating misinformation susceptibility, with accuracy up to 92%.

Modern language models still rely on fixed, pre-defined subword tokenizations. Once a tokenizer is trained, the LM can only operate at this fixed level of granularity, which often leads to brittle and counterintuitive behaviors even in otherwise strong reasoning models. We introduce \textbf{ByteFlow Net}, a new hierarchical architecture that removes tokenizers entirely and instead enables models to learn their own segmentation of raw byte streams into semantically meaningful units. ByteFlow Net performs compression-driven segmentation based on the coding rate of latent representations, yielding adaptive boundaries \emph{while preserving a static computation graph via Top-$K$ selection}. Unlike prior self-tokenizing methods that depend on brittle heuristics with human-designed inductive biases, ByteFlow Net adapts its internal representation granularity to the input itself. Experiments demonstrate that this compression-based chunking strategy yields substantial performance gains, with ByteFlow Net outperforming both BPE-based Transformers and previous byte-level architectures. These results suggest that end-to-end, tokenizer-free modeling is not only feasible but also more effective, opening a path toward more adaptive and information-grounded language models.

Optimal transport (OT) finds a least cost transport plan between two probability distributions using a cost matrix defined on pairs of points. Unlike standard OT, which infers unstructured pointwise mappings, low-rank optimal transport explicitly constrains the rank of the transport plan to infer latent structure. This improves statistical stability and robustness, yields sharper parametric rates for estimating Wasserstein distances adaptive to the intrinsic rank, and generalizes $K$-means to co-clustering. These advantages, however, come at the cost of a non-convex and NP-hard optimization problem. We introduce transport clustering, an algorithm to compute a low-rank OT plan that reduces low-rank OT to a clustering problem on correspondences obtained from a full-rank $\textit{transport registration}$ step. We prove that this reduction yields polynomial-time, constant-factor approximation algorithms for low-rank OT: specifically, a $(1+γ)$ approximation for negative-type metrics and a $(1+γ+\sqrt{2γ}\,)$ approximation for kernel costs, where $γ\in [0,1]$ denotes the approximation ratio of the optimal full-rank solution relative to the low-rank optimal. Empirically, transport clustering outperforms existing low-rank OT solvers on synthetic benchmarks and large-scale, high-dimensional datasets.

This paper studies the relationship between human eye gaze time on words in source code and the Semantic Neighborhood Density (SND) of those words. Human eye gaze time is a popular way to quantify human attention such as the importance of words people read and the cognitive effort people exert. Meanwhile, SND is a measure of how similar a word is in meaning to other words in the same context. SND has a long history in Psychology research where it has been connected to eye gaze time in various domains and helps explain human cognitive factors such as confusion and quality of reading comprehension. But SND carries an unknown and potentially unique meaning in software engineering. In this paper, we compute SND for tokens in source code that people viewed in two previous eye-tracking experiments, one in C and one in Java. We conduct a model-free analysis for statistical relationships between SND and gaze time, and a model-based analysis for predictive power of SND to gaze time. We found that words with high SND tend to have higher gaze time then low SND words, especially for words that are uncommon (i.e., have low frequency). We also found SND and frequency to have a minor predictive power on gaze time, despite high levels of noise common in eye tracking data

Conversational shopping assistants (CSAs) represent a compelling application of agentic AI, but moving from prototype to production reveals two underexplored challenges: how to evaluate multi-turn interactions and how to optimize tightly coupled multi-agent systems. Grocery shopping further amplifies these difficulties, as user requests are often underspecified, highly preference-sensitive, and constrained by factors such as budget and inventory. In this paper, we present a practical blueprint for evaluating and optimizing conversational shopping assistants, illustrated through a production-scale AI grocery assistant. We introduce a multi-faceted evaluation rubric that decomposes end-to-end shopping quality into structured dimensions and develop a calibrated LLM-as-judge pipeline aligned with human annotations. Building on this evaluation foundation, we investigate two complementary prompt-optimization strategies based on a SOTA prompt-optimizer called GEPA (Shao et al., 2025): (1) Sub-agent GEPA, which optimizes individual agent nodes against localized rubrics, and (2) MAMuT (Multi-Agent Multi-Turn) GEPA (Herrera et al., 2026), a novel system-level approach that jointly optimizes prompts across agents using multi-turn simulation and trajectory-level scoring. We release rubric templates and evaluation design guidance to support practitioners building production CSAs.

Reliable positioning in dense urban environments remains challenging due to frequent GNSS signal blockage, multipath, and rapidly varying satellite geometry. While factor graph optimization (FGO)-based GNSS-IMU fusion has demonstrated strong robustness and accuracy, most formulations remain offline. In this work, we present a real-time tightly coupled GNSS-IMU FGO method that enables causal state estimation via incremental optimization with fixed-lag marginalization, and we evaluate its performance in a highly urbanized GNSS-degraded environment using the UrbanNav dataset.

MoltBook is a large-scale multi-agent coordination environment where over 770,000 autonomous LLM agents interact without human participation, offering the first opportunity we are aware of to observe emergent multi-agent coordination dynamics at this population scale. We introduce \textit{Molt Dynamics}: the emergent agent coordination behaviors, inter-agent communication dynamics, and role specialization patterns arising when autonomous agents operate as decentralized decision-makers in an unconstrained multi-agent environment. Through longitudinal observation of 90,704 active agents over three weeks, we characterize three aspects. First, spontaneous role specialization: network-based clustering reveals six structural roles (silhouette 0.91), though the result primarily reflects core-periphery organization -- 93.5\% of agents occupy a homogeneous peripheral cluster, with meaningful differentiation confined to the active minority. Second, decentralized information dissemination: cascade analysis of 10,323 inter-agent propagation events reveals power-law distributed cascade sizes ($α= 2.57 \pm 0.02$) and saturating adoption dynamics where adoption probability shows diminishing returns with repeated exposures (Cox hazard ratio 0.53, concordance 0.78). Third, distributed cooperative task resolution: 164 multi-agent collaborative events show detectable coordination patterns, but success rates are low (6.7\%, $p = 0.057$) and cooperative outcomes are significantly worse than a matched single-agent baseline (Cohen's $d = -0.88$), indicating emergent cooperative behavior is nascent. These findings establish an empirical baseline for coordination dynamics in decentralized autonomous agent systems, with implications for multi-agent system design, agent communication protocol engineering, and AI safety.

Accurate positioning, navigation, and timing (PNT) is fundamental to the operation of modern technologies and a key enabler of autonomous systems. A very important component of PNT is the Global Navigation Satellite System (GNSS) which ensures outdoor positioning. Modern research directions have pushed the performance of GNSS localization to new heights by fusing GNSS measurements with other sensory information, mainly measurements from Inertial Measurement Units (IMU). In this paper, we propose a loosely coupled architecture to integrate GNSS and IMU measurements using a Factor Graph Optimization (FGO) framework. Because the FGO method can be computationally challenging and often used as a post-processing method, our focus is on assessing its localization accuracy and service availability while operating in real-time in challenging environments (urban canyons). Experimental results on the UrbanNav-HK-MediumUrban-1 dataset show that the proposed approach achieves real-time operation and increased service availability compared to batch FGO methods. While this improvement comes at the cost of reduced positioning accuracy, the paper provides a detailed analysis of the trade-offs between accuracy, availability, and computational efficiency that characterize real-time FGO-based GNSS/IMU fusion.

We present Tucano 2, a fully open suite of large language models (LLMs) with 0.5-3.7 billion parameters, designed to address certain gaps in open-source development for Portuguese LLMs. Following our previous works, we now extend our dataset, GigaVerbo-v2, to a new degree of quality and scale, while also introducing a new synthetic dataset, GigaVerbo-v2 Synth, aimed at filling missing gaps in GigaVerbo-v2, and two post-training datasets, GigaVerbo-v2 SFT and GigaVerbo-v2 Preferences, that allow Portuguese LLMs to be trained in domains like retrieval augmented generation, coding, tool use, chain-of-thought reasoning, and many other domains of interest. Through extensive ablation studies, we design both pretraining and continual pretraining recipes for the Tucano 2 suite (Base, Instruct, and Think), which achieve state-of-the-art performance on several Portuguese-language modeling benchmarks. We also extend and refine the evaluation harness introduced in our earlier work, yielding a comprehensive evaluation suite that provides strong signals across different pretraining, continual pretraining, and post-training regimes. All artifacts associated with Tucano 2 are openly released, including training recipes, logs, and source code, ensuring that our work is reproducible, accessible, and extendable by the broader Portuguese NLP community.

Automated question-answering (QA) systems increasingly rely on retrieval-augmented generation (RAG) to ground large language models (LLMs) in authoritative medical knowledge, ensuring clinical accuracy and patient safety in Artificial Intelligence (AI) applications for healthcare. Despite progress in RAG evaluation, current benchmarks focus only on simple multiple-choice QA tasks and employ metrics that poorly capture the semantic precision required for complex QA tasks. These approaches fail to diagnose whether an error stems from faulty retrieval or flawed generation, limiting developers from performing targeted improvement. To address this gap, we propose RAG-X, a diagnostic framework that evaluates the retriever and generator independently across a triad of QA tasks: information extraction, short-answer generation, and multiple-choice question (MCQ) answering. RAG-X introduces Context Utilization Efficiency (CUE) metrics to disaggregate system success into interpretable quadrants, isolating verified grounding from deceptive accuracy. Our experiments reveal an ``Accuracy Fallacy", where a 14\% gap separates perceived system success from evidence-based grounding. By surfacing hidden failure modes, RAG-X offers the diagnostic transparency needed for safe and verifiable clinical RAG systems.

Large language models with chain-of-thought generation have demonstrated great potential for producing complex mathematical proofs. However, their reasoning can often go astray, leading to increasing interest in formal and learned verifiers. A major challenge in learning verifiers, especially when their output will be used by the prover, is that this feedback loop may produce substantial distribution shift. Motivated by this challenge, we propose an online learning framework for learning chain-of-thought verifiers that, given a problem and a sequence of reasoning steps, check the correctness of the solution. Highlighting the asymmetric role of soundness (failure in catching errors in a proof) and completeness (flagging correct proofs as wrong) mistakes of the verifier, we introduce novel extensions of the Littlestone dimension which tightly characterize the mistake bounds for learning a verifier in the realizable setting. We provide optimal algorithms for finding the Pareto-frontier (the smallest total number of mistakes given a budget of soundness mistakes) as well as minimizing a linear combination of asymmetric costs. We further show how our learned verifiers can be used to boost the accuracy of a collection of weak provers, and enable generation of proofs beyond what they were trained on. With the mild assumption that one of the provers can generate the next reasoning step correctly with some minimal probability, we show how to learn a strong prover with small error and abstention rates.

Current LLM-based conversational recommender systems (CRS) primarily optimize recommendation accuracy and user satisfaction. We identify an underexplored vulnerability in which recommendation outputs may negatively impact users by violating personalized safety constraints, when individualized safety sensitivities -- such as trauma triggers, self-harm history, or phobias -- are implicitly inferred from the conversation but not respected during recommendation. We formalize this challenge as personalized CRS safety and introduce SafeRec, a new benchmark dataset designed to systematically evaluate safety risks in LLM-based CRS under user-specific constraints. To further address this problem, we propose SafeCRS, a safety-aware training framework that integrates Safe Supervised Fine-Tuning (Safe-SFT) with Safe Group reward-Decoupled Normalization Policy Optimization (Safe-GDPO) to jointly optimize recommendation quality and personalized safety alignment. Extensive experiments on SafeRec demonstrate that SafeCRS reduces safety violation rates by up to 96.5% relative to the strongest recommendation-quality baseline while maintaining competitive recommendation quality. Warning: This paper contains potentially harmful and offensive content.

While large language models (LLMs) fine-tuned with lightweight adapters achieve strong performance across diverse tasks, their performance on individual tasks depends on the fine-tuning strategy. Fusing independently trained models with different strengths has shown promise for multi-task learning through three main strategies: ensembling, which combines outputs from independent models; merging, which fuses model weights via parameter averaging; and routing, which integrates models in an input-dependent fashion. However, many design decisions in these approaches remain understudied, and the relative benefits of more sophisticated ensembling, merging and routing techniques are not fully understood. We empirically evaluate their trade-offs, addressing two key questions: What are the advantages of going beyond uniform ensembling or merging? And does the flexibility of routing justify its complexity? Our findings indicate that non-uniform ensembling and merging improve performance, but routing offers even greater gains. To mitigate the computational cost of routing, we analyze expert selection techniques, showing that clustering and greedy subset selection can maintain reasonable performance with minimal overhead. These insights advance our understanding of model fusion for multi-task learning.

Accurate prediction of terrestrial ecosystem carbon fluxes (e.g., CO$_2$, GPP, and CH$_4$) is essential for understanding the global carbon cycle and managing its impacts. However, prediction remains challenging due to strong spatiotemporal heterogeneity: ecosystem flux responses are constrained by slowly varying regime conditions, while short-term fluctuations are driven by high-frequency dynamic forcings. Most existing learning-based approaches treat environmental covariates as a homogeneous input space, implicitly assuming a global response function, which leads to brittle generalization across heterogeneous ecosystems. In this work, we propose Role-Aware Conditional Inference (RACI), a process-informed learning framework that formulates ecosystem flux prediction as a conditional inference problem. RACI employs hierarchical temporal encoding to disentangle slow regime conditioners from fast dynamic drivers, and incorporates role-aware spatial retrieval that supplies functionally similar and geographically local context for each role. By explicitly modeling these distinct functional roles, RACI enables a model to adapt its predictions across diverse environmental regimes without training separate local models or relying on fixed spatial structures. We evaluate RACI across multiple ecosystem types (wetlands and agricultural systems), carbon fluxes (CO$_2$, GPP, CH$_4$), and data sources, including both process-based simulations and observational measurements. Across all settings, RACI consistently outperforms competitive spatiotemporal baselines, demonstrating improved accuracy and spatial generalization under pronounced environmental heterogeneity.

Frozen self-supervised representations often transfer well with only a few labels across many semantic tasks. We argue that a single geometric quantity, \emph{directional} CDNV (decision-axis variance), sits at the core of two favorable behaviors: strong few-shot transfer within a task, and low interference across many tasks. We show that both emerge when variability \emph{along} class-separating directions is small. First, we prove sharp non-asymptotic multiclass generalization bounds for downstream classification whose leading term is the directional CDNV. The bounds include finite-shot corrections that cleanly separate intrinsic decision-axis variability from centroid-estimation error. Second, we link decision-axis collapse to multitask geometry: for independent balanced labelings, small directional CDNV across tasks forces the corresponding decision axes to be nearly orthogonal, helping a single representation support many tasks with minimal interference. Empirically, across SSL objectives, directional CDNV collapses during pretraining even when classical CDNV remains large, and our bounds closely track few-shot error at practical shot sizes. Additionally, on synthetic multitask data, we verify that SSL learns representations whose induced decision axes are nearly orthogonal. The code and project page of the paper are available at [\href{https://dlfundamentals.github.io/directional-neural-collapse/}{project page}].

We introduce mlx-snn, the first spiking neural network (SNN) library built natively on Apple's MLX framework. As SNN research grows rapidly, all major libraries -- snnTorch, Norse, SpikingJelly, Lava -- target PyTorch or custom backends, leaving Apple Silicon users without a native option. mlx-snn provides six neuron models (LIF, IF, Izhikevich, Adaptive LIF, Synaptic, Alpha), four surrogate gradient functions, four spike encoding methods (including an EEG-specific encoder), and a complete backpropagation-through-time training pipeline. The library leverages MLX's unified memory architecture, lazy evaluation, and composable function transforms (mx.grad, mx.compile) to enable efficient SNN research on Apple Silicon hardware. We validate mlx-snn on MNIST digit classification across five hyperparameter configurations and three backends, achieving up to 97.28% accuracy with 2.0--2.5 times faster training and 3--10 times lower GPU memory than snnTorch on the same M3 Max hardware. mlx-snn is open-source under the MIT license and available on PyPI. https://github.com/D-ST-Sword/mlx-snn

Vision-Language Models (VLMs) with their multimodal capabilities have demonstrated remarkable success in almost all domains, including education, transportation, healthcare, energy, finance, law, and retail. Nevertheless, the utilization of VLMs in healthcare applications raises crucial concerns due to the sensitivity of large-scale medical data and the trustworthiness of these models (reliability, transparency, and security). This study proposes a logit-level uncertainty quantification (UQ) framework for histopathology image analysis using VLMs to deal with these concerns. UQ is evaluated for three VLMs using metrics derived from temperature-controlled output logits. The proposed framework demonstrates a critical separation in uncertainty behavior. While VLMs show high stochastic sensitivity (cosine similarity (CS) $<0.71$ and $<0.84$, Jensen-Shannon divergence (JS) $<0.57$ and $<0.38$, and Kullback-Leibler divergence (KL) $<0.55$ and $<0.35$, respectively for mean values of VILA-M3-8B and LLaVA-Med v1.5), near-maximal temperature impacts ($Δ_T \approx 1.00$), and displaying abrupt uncertainty transitions, particularly for complex diagnostic prompts. In contrast, the pathology-specific PRISM model maintains near-deterministic behavior (mean CS $>0.90$, JS $<0.10$, KL $<0.09$) and significantly minimal temperature effects across all prompt complexities. These findings emphasize the importance of logit-level uncertainty quantification to evaluate trustworthiness in histopathology applications utilizing VLMs.

Western governments have adopted an assortment of counter-hybrid threat measures to defend against hostile actions below the conventional military threshold. The impact of these measures is unclear because of the ambiguity of hybrid threats, their cross-domain nature, and uncertainty about how countermeasures shape adversarial behavior. This paper offers a novel approach to clarifying this impact by unifying previously bifurcating hybrid threat modeling methods through a (multi-agent) influence diagram framework. The model balances the costs of countermeasures, their ability to dissuade the adversary from executing hybrid threats, and their potential to mitigate the impact of hybrid threats. We run 1000 semi-synthetic variants of a real-world-inspired scenario simulating the strategic interaction between attacking agent A and defending agent B over a cyber attack on critical infrastructure to explore the effectiveness of a set of five different counter-hybrid threat measures. Counter-hybrid measures range from strengthening resilience and denial of the adversary's ability to execute a hybrid threat to dissuasion through the threat of punishment. Our analysis primarily evaluates the overarching characteristics of counter-hybrid threat measures. This approach allows us to generalize the effectiveness of these measures and examine parameter impact sensitivity. In addition, we discuss policy relevance and outline future research avenues.

As strong general reasoners, large language models (LLMs) encounter diverse domains and tasks, where the ability to adapt and self-improve at test time is valuable. We introduce MASS, a meta-learning framework that enables LLMs to self-adapt by generating problem-specific synthetic training data and performing targeted self-updates optimized for downstream performance at inference time. We train this behavior end-to-end via bilevel optimization: an inner loop adapts on self-generated examples while an outer loop meta-learns data-attribution signals and rewards post-update task performance. The synthetic data is optimized with scalable meta-gradients, backpropagating the downstream loss through the inner updates to reward useful generations. Experiments on mathematical reasoning show that MASS learns to synthesize per-instance curricula that yield effective, data-efficient test-time adaptation.

We study reinforcement learning in infinite-horizon discounted Markov decision processes with continuous state spaces, where data are generated online from a single trajectory under a Markovian behavior policy. To avoid maintaining an infinite-dimensional, function-valued estimate, we propose the novel Q-Measure-Learning, which learns a signed empirical measure supported on visited state-action pairs and reconstructs an action-value estimate via kernel integration. The method jointly estimates the stationary distribution of the behavior chain and the Q-measure through coupled stochastic approximation, leading to an efficient weight-based implementation with $O(n)$ memory and $O(n)$ computation cost per iteration. Under uniform ergodicity of the behavior chain, we prove almost sure sup-norm convergence of the induced Q-function to the fixed point of a kernel-smoothed Bellman operator. We also bound the approximation error between this limit and the optimal $Q^*$ as a function of the kernel bandwidth. To assess the performance of our proposed algorithm, we conduct RL experiments in a two-item inventory control setting.

General-purpose large language models (LLMs) that rely on in-context learning do not reliably deliver the scientific understanding and performance required for drug discovery tasks. Simply increasing model size or introducing reasoning tokens does not yield significant performance gains. To address this gap, we introduce the MMAI Gym for Science, a one-stop shop molecular data formats and modalities as well as task-specific reasoning, training, and benchmarking recipes designed to teach foundation models the 'language of molecules' in order to solve practical drug discovery problems. We use MMAI Gym to train an efficient Liquid Foundation Model (LFM) for these applications, demonstrating that smaller, purpose-trained foundation models can outperform substantially larger general-purpose or specialist models on molecular benchmarks. Across essential drug discovery tasks - including molecular optimization, ADMET property prediction, retrosynthesis, drug-target activity prediction, and functional group reasoning - the resulting model achieves near specialist-level performance and, in the majority of settings, surpasses larger models, while remaining more efficient and broadly applicable in the domain.

Agentic AI systems - capable of goal interpretation, world modeling, planning, tool use, long-horizon operation, and autonomous coordination - introduce distinct control failures not addressed by existing safety frameworks. We identify six agentic governance failures tied to these capabilities and show how they erode meaningful human control in military settings. We propose the Agentic Military AI Governance Framework (AMAGF), a measurable architecture structured around three pillars: Preventive Governance (reducing failure likelihood), Detective Governance (real-time detection of control degradation), and Corrective Governance (restoring or safely degrading operations). Its core mechanism, the Control Quality Score (CQS), is a composite real-time metric quantifying human control and enabling graduated responses as control weakens. For each failure type, we define concrete mechanisms, assign responsibilities across five institutional actors, and formalize evaluation metrics. A worked operational scenario illustrates implementation, and we situate the framework within established agent safety literature. We argue that governance must move from a binary conception of control to a continuous model in which control quality is actively measured and managed throughout the operational lifecycle.

Teachers face increasing demands on their time, particularly in adapting mathematics curricula to meet individual student needs while maintaining cognitive rigor. This study evaluates whether AI tools can accurately classify the cognitive demand of mathematical tasks, which is important for creating or adapting tasks that support student learning. We tested eleven AI tools: six general-purpose (ChatGPT, Claude, DeepSeek, Gemini, Grok, Perplexity) and five education-specific (Brisk, Coteach AI, Khanmigo, Magic School, School.AI), on their ability to categorize mathematics tasks across four levels of cognitive demand using a research-based framework. The goal was to approximate the performance teachers will achieve with straightforward prompts. On average, AI tools accurately classified cognitive demand in only 63% of cases. Education-specific tools were not more accurate than general-purpose tools, and no tool exceeded 83% accuracy. All tools struggled with tasks at the extremes of cognitive demand (Memorization and Doing Mathematics), exhibiting a systematic bias toward middle-category levels (Procedures with/without Connections). The tools often gave plausible-sounding explanations likely to be persuasive to novice teachers. Error analysis of AI tools' misclassification of the broad level of cognitive demand (high vs. low) revealed that tools consistently overweighted surface textual features over underlying cognitive processes. Further, AI tools showed weaknesses in reasoning about factors that make tasks higher vs. lower cognitive demand. Errors stemmed not from ignoring relevant dimensions, but from incorrectly reasoning about multiple task aspects. These findings carry implications for AI integration into teacher planning workflows and highlight the need for improved prompt engineering and tool development for educational applications.

We aim to learn wavefunctions simulated by time-dependent density functional theory (TDDFT), which can be efficiently represented as linear combination coefficients of atomic orbitals. In real-time TDDFT, the electronic wavefunctions of a molecule evolve over time in response to an external excitation, enabling first-principles predictions of physical properties such as optical absorption, electron dynamics, and high-order response. However, conventional real-time TDDFT relies on time-consuming propagation of all occupied states with fine time steps. In this work, we propose OrbEvo, which is based on an equivariant graph transformer architecture and learns to evolve the full electronic wavefunction coefficients across time steps. First, to account for external field, we design an equivariant conditioning to encode both strength and direction of external electric field and break the symmetry from SO(3) to SO(2). Furthermore, we design two OrbEvo models, OrbEvo-WF and OrbEvo-DM, using wavefunction pooling and density matrix as interaction method, respectively. Motivated by the central role of the density functional in TDDFT, OrbEvo-DM encodes the density matrix aggregated from all occupied electronic states into feature vectors via tensor contraction, providing a more intuitive approach to learn the time evolution operator. We adopt a training strategy specifically tailored to limit the error accumulation of time-dependent wavefunctions over autoregressive rollout. To evaluate our approach, we generate TDDFT datasets consisting of 5,000 different molecules in the QM9 dataset and 1,500 molecular configurations of the malonaldehyde molecule in the MD17 dataset. Results show that our OrbEvo model accurately captures quantum dynamics of excited states under external field, including time-dependent wavefunctions, time-dependent dipole moment, and optical absorption spectra.

This study investigates the diverse characteristics of nouns, focusing on both semantic (e.g., countable/uncountable) and morphosyntactic (e.g., masculine/feminine) distinctions. We explore inter-word variations for gender markers in noun morphology. Grammatical gender shift is a widespread phenomenon in languages around the world. The aim is to uncover through a formal model the underlying patterns governing the variation of lexemes. To this end, we propose a new computational component dedicated to pairing items with morphological templates (e.g., the result of a generated item-template pair: (funas, $\{N, +SG, -PL, -M, +F, -COL, +SING\}$), with its spell-out form: $ð$a-funast 'cow'). This process is formally represented by the Template-Based and Modular Cognitive model. This proposed model, defined by a set-valued set function $h : \mathscr{P}(M) \rightarrow \mathscr{P}(M)$, predicts the nonlinear dynamic mapping of lexical items onto morphological templates. By applying this formalism, we present a unified framework for understanding the complexities of morphological markings across languages. Through empirical observations, we demonstrate how these shifts, as well as non-gender shifts, arise during lexical changes, especially in Riffian. Our model posits that these variant markings emerge due to template shifts occurring during word and meaning's formation. By formally demonstrating that conversion is applicable to noun-to-noun derivation, we challenge and broaden the conventional view of word formation. This mathematical model not only contributes to a deeper understanding of morphosyntactic variation but also offers potential applications in other fields requiring precise modelling of linguistic patterns.

The dominance of large multilingual foundation models has widened linguistic inequalities in Natural Language Processing (NLP), often leaving low-resource languages underrepresented. This paper introduces LilMoo, a 0.6-billion-parameter Hindi language model trained entirely from scratch to address this gap. Unlike prior Hindi models that rely on continual pretraining from opaque multilingual foundations, LilMoo is developed through a fully transparent and reproducible pipeline optimized for limited compute environments. We construct a high-quality Hindi corpus (GigaLekh) filtered through both heuristic and learned (LLM-as-a-judge) methods, complemented by bilingual augmentation with curated English data. Using this dataset, we explore various training recipes for small-scale language models. Across comprehensive evaluation suites, LilMoo consistently outperforms comparably sized multilingual baselines such as Qwen2.5-0.5B and Qwen3-0.6B, demonstrating that well-designed language-specific pretraining can rival large multilingual models at the sub-billion-parameter range.

Adversarial attacks - input perturbations imperceptible to humans that fool neural networks - remain both a persistent failure mode in machine learning, and a phenomenon with mysterious origins. To shed light, we define and analyze a network's perceptual manifold (PM) for a class concept as the space of all inputs confidently assigned to that class by the network. We find, strikingly, that the dimensionalities of neural network PMs are orders of magnitude higher than those of natural human concepts. Since volume typically grows exponentially with dimension, this suggests exponential misalignment between machines and humans, with exponentially many inputs confidently assigned to concepts by machines but not humans. Furthermore, this provides a natural geometric hypothesis for the origin of adversarial examples: because a network's PM fills such a large region of input space, any input will be very close to any class concept's PM. Our hypothesis thus suggests that adversarial robustness cannot be attained without dimensional alignment of machine and human PMs, and therefore makes strong predictions: both robust accuracy and distance to any PM should be negatively correlated with the PM dimension. We confirmed these predictions across 18 different networks of varying robust accuracy. Crucially, we find even the most robust networks are still exponentially misaligned, and only the few PMs whose dimensionality approaches that of human concepts exhibit alignment to human perception. Our results connect the fields of alignment and adversarial examples, and suggest the curse of high dimensionality of machine PMs is a major impediment to adversarial robustness.

State-of-the-art text-to-video (T2V) generators frequently violate physical laws despite high visual quality. We show this stems from insufficient physical constraints in prompts rather than model limitations: manually adding physics details reliably produces physically plausible videos, but requires expertise and does not scale. We present PhyPrompt, a two-stage reinforcement learning framework that automatically refines prompts for physically realistic generation. First, we fine-tune a large language model on a physics-focused Chain-of-Thought dataset to integrate principles like object motion and force interactions while preserving user intent. Second, we apply Group Relative Policy Optimization with a dynamic reward curriculum that initially prioritizes semantic fidelity, then progressively shifts toward physical commonsense. This curriculum achieves synergistic optimization: PhyPrompt-7B reaches 40.8\% joint success on VideoPhy2 (8.6pp gain), improving physical commonsense by 11pp (55.8\% to 66.8\%) while simultaneously increasing semantic adherence by 4.4pp (43.4\% to 47.8\%). Remarkably, our curriculum exceeds single-objective training on both metrics, demonstrating compositional prompt discovery beyond conventional multi-objective trade-offs. PhyPrompt outperforms GPT-4o (+3.8\% joint) and DeepSeek-V3 (+2.2\%, 100$\times$ larger) using only 7B parameters. The approach transfers zero-shot across diverse T2V architectures (Lavie, VideoCrafter2, CogVideoX-5B) with up to 16.8\% improvement, establishing that domain-specialized reinforcement learning with compositional curricula surpasses general-purpose scaling for physics-aware generation.

Although high-resolution mapping of pan-Arctic sea ice with reliable corresponding uncertainty is essential for operational sea ice concentration (SIC) charting, it is a difficult task due to key challenges, such as the subtle nature of ice signature features, inexact SIC labels, model uncertainty, and data heterogeneity. This study presents a novel Bayesian High-Resolution Transformer approach for 200 meter resolution pan-Arctic SIC mapping and uncertainty quantification using Sentinel-1, RADARSAT Constellation Mission (RCM), and Advanced Microwave Scanning Radiometer 2 (AMSR2) data. First, to improve small and subtle sea ice feature (e.g., cracks/leads, ponds, and ice floes) extraction, we design a novel high-resolution Transformer model with both global and local modules that can better discern the subtle differences in sea ice patterns. Second, to address low-resolution and inexact SIC labels, we design a geographically-weighted weakly supervised loss function to supervise the model at region level instead of pixel level, and to prioritize pure open water and ice pack signatures while mitigating the impact of ambiguity in the marginal ice zone (MIZ). Third, to improve uncertainty quantification, we design a Bayesian extension of the proposed Transformer model, treating its parameters as random variables to more effectively capture uncertainties. Fourth, to address data heterogeneity, we fuse three different data types (Sentinel-1, RCM, and AMSR2) at decision-level to improve both SIC mapping and uncertainty quantification. The proposed approach is evaluated under pan-Arctic minimum-extent conditions in 2021 and 2025. Results demonstrate that the proposed model achieves 0.70 overall feature detection accuracy using Sentinel-1 data, while also preserving pan-Arctic SIC patterns (Sentinel-1 R\textsuperscript{2} = 0.90 relative to the ARTIST Sea Ice product).

Benchmark rankings are routinely used to justify scientific claims about method quality in gene regulatory network (GRN) inference, yet the stability of these rankings under plausible evaluation protocol choices is rarely examined. We present a systematic diagnostic framework for measuring ranking instability under protocol shift, including decomposition tools that separate base rate effects from discrimination effects. Using existing single cell GRN benchmark outputs across three human tissues and six inference methods, we quantify pairwise reversal rates across four protocol axes: candidate set restriction (16.3 percent, 95 percent CI 11.0 to 23.4 percent), tissue context (19.3 percent), reference network choice (32.1 percent), and symbol mapping policy (0.0 percent). A permutation null confirms that observed reversal rates are far below random order expectations (0.163 versus null mean 0.500), indicating partially stable but non invariant ranking structure. Our decomposition reveals that reversals are driven by changes in the relative discrimination ability of methods rather than by base rate inflation, a finding that challenges a common implicit assumption in GRN benchmarking. We propose concrete reporting practices for stability aware evaluation and provide a diagnostic toolkit for identifying method pairs at risk of reversal.

Compute-in-memory (CiM) architectures promise significant improvements in energy efficiency and throughput for deep neural network acceleration by alleviating the von Neumann bottleneck. However, their reliance on emerging non-volatile memory devices introduces device-level non-idealities-such as write variability, conductance drift, and stochastic noise-that fundamentally challenge reliability, predictability, and safety, especially in safety-critical applications. This talk examines the reliability limits of CiM-based neural accelerators and presents a series of techniques that bridge device physics, architecture, and learning algorithms to address these challenges. We first demonstrate that even small device variations can lead to disproportionately large accuracy degradation and catastrophic failures in safety-critical inference workloads, revealing a critical gap between average-case evaluations and worst-case behavior. Building on this insight, we introduce SWIM, a selective write-verify mechanism that strategically applies verification only where it is most impactful, significantly improving reliability while maintaining CiM's efficiency advantages. Finally, we explore a learning-centric solution that improves realistic worst-case performance by training neural networks with right-censored Gaussian noise, aligning training assumptions with hardware-induced variability and enabling robust deployment without excessive hardware overhead. Together, these works highlight the necessity of cross-layer co-design for CiM accelerators and provide a principled path toward dependable, efficient neural inference on emerging memory technologies-paving the way for their adoption in safety- and reliability-critical systems.

Recent video diffusion models have achieved impressive capabilities as large-scale generative world models. However, these models often struggle with fine-grained physical consistency, exhibiting physically implausible dynamics over time. In this work, we present \textbf{Phys4D}, a pipeline for learning physics-consistent 4D world representations from video diffusion models. Phys4D adopts \textbf{a three-stage training paradigm} that progressively lifts appearance-driven video diffusion models into physics-consistent 4D world representations. We first bootstrap robust geometry and motion representations through large-scale pseudo-supervised pretraining, establishing a foundation for 4D scene modeling. We then perform physics-grounded supervised fine-tuning using simulation-generated data, enforcing temporally consistent 4D dynamics. Finally, we apply simulation-grounded reinforcement learning to correct residual physical violations that are difficult to capture through explicit supervision. To evaluate fine-grained physical consistency beyond appearance-based metrics, we introduce a set of \textbf{4D world consistency evaluation} that probe geometric coherence, motion stability, and long-horizon physical plausibility. Experimental results demonstrate that Phys4D substantially improves fine-grained spatiotemporal and physical consistency compared to appearance-driven baselines, while maintaining strong generative performance. Our project page is available at https://sensational-brioche-7657e7.netlify.app/

E-fuels are promising long-term energy carriers supporting the net-zero transition. However, the large combinatorial design-operation spaces under renewable uncertainty make the use of mathematical programming impractical for co-optimizing e-fuel production systems. Here, we present MasCOR, a machine-learning-assisted co-optimization framework that learns from global operational trajectories. By encoding system design and renewable trends, a single MasCOR agent generalizes dynamic operation across diverse configurations and scenarios, substantially simplifying design-operation co-optimization under uncertainty. Benchmark comparisons against state-of-the-art reinforcement learning baselines demonstrate near-optimal performance, while computational costs are substantially lower than those of mathematical programming, enabling rapid parallel evaluation of designs within the co-optimization loop. This framework enables rapid screening of feasible design spaces together with corresponding operational policies. When applied to four potential European sites targeting e-methanol production, MasCOR shows that most locations benefit from reducing system load below 50 MW to achieve carbon-neutral methanol production, with production costs of 1.0-1.2 USD per kg. In contrast, Dunkirk (France), with limited renewable availability and high grid prices, favors system loads above 200 MW and expanded storage to exploit dynamic grid exchange and hydrogen sales to the market. These results underscore the value of the MasCOR framework for site-specific guidance from system design to real-time operation.

Interactive world models continually generate video by responding to a user's actions, enabling open-ended generation capabilities. However, existing models typically lack a 3D representation of the environment, meaning 3D consistency must be implicitly learned from data, and spatial memory is restricted to limited temporal context windows. This results in an unrealistic user experience and presents significant obstacles to down-stream tasks such as training agents. To address this, we present PERSIST, a new paradigm of world model which simulates the evolution of a latent 3D scene: environment, camera, and renderer. This allows us to synthesize new frames with persistent spatial memory and consistent geometry. Both quantitative metrics and a qualitative user study show substantial improvements in spatial memory, 3D consistency, and long-horizon stability over existing methods, enabling coherent, evolving 3D worlds. We further demonstrate novel capabilities, including synthesising diverse 3D environments from a single image, as well as enabling fine-grained, geometry-aware control over generated experiences by supporting environment editing and specification directly in 3D space. Project page: https://francelico.github.io/persist.github.io

We study reinforcement learning with delayed state observation, where the agent observes the current state after some random number of time steps. We propose an algorithm that combines the augmentation method and the upper confidence bound approach. For tabular Markov decision processes (MDPs), we derive a regret bound of $\tilde{\mathcal{O}}(H \sqrt{D_{\max} SAK})$, where $S$ and $A$ are the cardinalities of the state and action spaces, $H$ is the time horizon, $K$ is the number of episodes, and $D_{\max}$ is the maximum length of the delay. We also provide a matching lower bound up to logarithmic factors, showing the optimality of our approach. Our analytical framework formulates this problem as a special case of a broader class of MDPs, where their transition dynamics decompose into a known component and an unknown but structured component. We establish general results for this abstract setting, which may be of independent interest.

We propose a novel computational framework for analyzing electroencephalography (EEG) time series using methods from stringology, the study of efficient algorithms for string processing, to systematically identify and characterize recurrent temporal patterns in neural signals. The primary aim is to introduce quantitative measures to understand neural signal dynamics, with the present findings serving as a proof-of-concept. The framework adapts order-preserving matching (OPM) and Cartesian tree matching (CTM) to detect temporal motifs that preserve relative ordering and hierarchical structure while remaining invariant to amplitude scaling. This approach provides a temporally precise representation of EEG dynamics that complements traditional spectral and global complexity analyses. To evaluate its utility, we applied the framework to multichannel EEG recordings from individuals with attention-deficit/hyperactivity disorder (ADHD) and matched controls using a publicly available dataset. Highly recurrent, group-specific motifs were extracted and quantified using both OPM and CTM. The ADHD group exhibited significantly higher motif frequencies, suggesting increased repetitiveness in neural activity. OPM analysis revealed shorter motif lengths and greater gradient instability in ADHD, reflected in larger mean and maximal inter-sample amplitude changes. CTM analysis further demonstrated reduced hierarchical complexity in ADHD, characterized by shallower tree structures and fewer hierarchical levels despite comparable motif lengths. These findings suggest that ADHD-related EEG alterations involve systematic differences in the structure, stability, and hierarchical organization of recurrent temporal patterns. The proposed stringology-based motif framework provides a complementary computational tool with potential applications for objective biomarker development in neurodevelopmental disorders.

Mathematical reasoning models are widely deployed in education, automated tutoring, and decision support systems despite exhibiting fundamental computational instabilities. We demonstrate that state-of-the-art models (Qwen2.5-Math-7B) achieve 61% accuracy through a mixture of reliable and unreliable reasoning pathways: 18.4% of correct predictions employ stable, faithful reasoning while 81.6% emerge through computationally inconsistent pathways. Additionally, 8.8% of all predictions are silent failures -- confident yet incorrect outputs. Through comprehensive analysis using novel faithfulness metrics, we reveal: (1) reasoning quality shows weak negative correlation with correctness (r=-0.21, p=0.002), reflecting a binary classification threshold artifact rather than a monotonic inverse relationship; (2) scaling from 1.5B to 7B parameters (4.7x increase) provides zero accuracy benefit on our evaluated subset (6% of GSM8K), requiring validation on the complete benchmark; and (3) latent reasoning employs diverse computational strategies, with ~20% sharing CoT-like patterns. These findings highlight that benchmark accuracy can mask computational unreliability, demanding evaluation reforms measuring stability beyond single-sample metrics.

Bisynchronous FIFOs -- hardware buffers that mediate data transfer between independent clock domains without a shared global timebase -- have been designed, formally verified, and commercially deployed in silicon for over four decades. We survey this literature from Chapiro's 1984 GALS thesis through Cummings's Gray-code pointer techniques, Chelcea and Nowick's mixed-timing interfaces, Greenstreet's STARI protocol, and the 2015 NVIDIA pausible bisynchronous FIFO, and argue that this body of work constitutes a silicon-proven existence proof against the Forward-In-Time-Only (FITO) assumption that pervades distributed systems. The central claim is that interaction-based synchronization primitives -- handshakes, mutual exclusion, and causal flow control -- can replace timestamp-based coordination at the most demanding levels of digital engineering, directly undermining the FITO assumption in protocols such as PTP, TSN, and conventional Ethernet. We draw a structural parallel between on-chip bisynchronous coordination and the Open Atomic Ethernet (OAE) architecture, and identify the handshake -- not the timestamp -- as the fundamental primitive for coordination between independent causal domains.

Generalization in generative modeling is defined as the ability to learn an underlying distribution from a finite dataset and produce novel samples, with evaluation largely driven by held-out performance and perceived sample quality. In practice, training is often stopped at the minimum of the test loss, taken as an operational indicator of generalization. We challenge this viewpoint by identifying a phase of biased generalization during training, in which the model continues to decrease the test loss while favoring samples with anomalously high proximity to training data. By training the same network on two disjoint datasets and comparing the mutual distances of generated samples and their similarity to training data, we introduce a quantitative measure of bias and demonstrate its presence on real images. We then study the mechanism of bias, using a controlled hierarchical data model where access to exact scores and ground-truth statistics allows us to precisely characterize its onset. We attribute this phenomenon to the sequential nature of feature learning in deep networks, where coarse structure is learned early in a data-independent manner, while finer features are resolved later in a way that increasingly depends on individual training samples. Our results show that early stopping at the test loss minimum, while optimal under standard generalization criteria, may be insufficient for privacy-critical applications.

We introduce Graph Hopfield Networks, whose energy function couples associative memory retrieval with graph Laplacian smoothing for node classification. Gradient descent on this joint energy yields an iterative update interleaving Hopfield retrieval with Laplacian propagation. Memory retrieval provides regime-dependent benefits: up to 2.0~pp on sparse citation networks and up to 5 pp additional robustness under feature masking; the iterative energy-descent architecture itself is a strong inductive bias, with all variants (including the memory-disabled NoMem ablation) outperforming standard baselines on Amazon co-purchase graphs. Tuning enables graph sharpening for heterophilous benchmarks without architectural changes.

We investigate when transformer MLP nonlinearity is actually necessary. A gate with $d+1$ parameters decides when to replace the full MLP with a linear surrogate. Through systematic investigation across six models (162M-2.8B parameters), two architectures, and three corpora, we establish that nonlinearity need cannot be predicted from token identity: cross-corpus correlation is zero ($r < 0.05$). The routing decision is fully contextual. Despite weak per-instance predictability, the gate exploits a heavily skewed distribution where most MLP computations are near-linear, achieving 25-56% linear routing at <1% perplexity cost in GPT-2. In GPT-2 Large, 11 of 36 layers beat baseline with gating and no layer exceeds 3.7% all-linear cost. This success is architecture-dependent: Pythia models show higher costs, though Pythia-2.8B's full 32-layer sweep reveals one layer that narrowly beats baseline. As a proof of concept, we progressively replace middle-layer MLPs with frozen linear matrices: 5 of 24 layers linearize at zero cost. With a full training budget, 4 linearized layers yield a 10.2% perplexity improvement -- and a two-phase gated approach pushes this to 17.3%, beating a vanilla fine-tuning control and confirming that the nonlinear MLPs at these layers were actively harmful.

Agentic coding agents are increasingly deployed autonomously, at scale, and over long-context horizons. Throughout an agent's lifetime, it must navigate tensions between explicit instructions, learned values, and environmental pressures, often in contexts unseen during training. Prior work on model preferences, agent behavior under value tensions, and goal drift has relied on static, synthetic settings that do not capture the complexity of real-world environments. To this end, we introduce a framework built on OpenCode to orchestrate realistic, multi-step coding tasks to measure how agents violate explicit constraints in their system prompt over time with and without environmental pressure toward competing values. Using this framework, we demonstrate that GPT-5 mini, Haiku 4.5, and Grok Code Fast 1 exhibit asymmetric drift: they are more likely to violate their system prompt when its constraint opposes strongly-held values like security and privacy. We find for the models and values tested that goal drift correlates with three compounding factors: value alignment, adversarial pressure, and accumulated context. However, even strongly-held values like privacy show non-zero violation rates under sustained environmental pressure. These findings reveal that shallow compliance checks are insufficient and that comment-based pressure can exploit model value hierarchies to override system prompt instructions. More broadly, our findings highlight a gap in current alignment approaches in ensuring that agentic systems appropriately balance explicit user constraints against broadly beneficial learned preferences under sustained environmental pressure.

Offline Reinforcement Learning (RL) is an emerging field of RL in which policies are learned solely from demonstrations. Within offline RL, some environments involve balancing multiple objectives, but existing multi-objective offline RL algorithms do not provide an efficient way to find a fair compromise. FairDICE (see arXiv:2506.08062v2) seeks to fill this gap by adapting OptiDICE (an offline RL algorithm) to automatically learn weights for multiple objectives to e.g.\ incentivise fairness among objectives. As this would be a valuable contribution, this replication study examines the replicability of claims made regarding FairDICE. We find that many theoretical claims hold, but an error in the code reduces FairDICE to standard behaviour cloning in continuous environments, and many important hyperparameters were originally underspecified. After rectifying this, we show in experiments extending the original paper that FairDICE can scale to complex environments and high-dimensional rewards, though it can be reliant on (online) hyperparameter tuning. We conclude that FairDICE is a theoretically interesting method, but the experimental justification requires significant revision.

Classifier-Free Guidance (CFG) has emerged as a central approach for enhancing semantic alignment in flow-based diffusion models. In this paper, we explore a unified framework called CFG-Ctrl, which reinterprets CFG as a control applied to the first-order continuous-time generative flow, using the conditional-unconditional discrepancy as an error signal to adjust the velocity field. From this perspective, we summarize vanilla CFG as a proportional controller (P-control) with fixed gain, and typical follow-up variants develop extended control-law designs derived from it. However, existing methods mainly rely on linear control, inherently leading to instability, overshooting, and degraded semantic fidelity especially on large guidance scales. To address this, we introduce Sliding Mode Control CFG (SMC-CFG), which enforces the generative flow toward a rapidly convergent sliding manifold. Specifically, we define an exponential sliding mode surface over the semantic prediction error and introduce a switching control term to establish nonlinear feedback-guided correction. Moreover, we provide a Lyapunov stability analysis to theoretically support finite-time convergence. Experiments across text-to-image generation models including Stable Diffusion 3.5, Flux, and Qwen-Image demonstrate that SMC-CFG outperforms standard CFG in semantic alignment and enhances robustness across a wide range of guidance scales. Project Page: https://hanyang-21.github.io/CFG-Ctrl

Many essential manipulation tasks - such as food preparation, surgery, and craftsmanship - remain intractable for autonomous robots. These tasks are characterized not only by contact-rich, force-sensitive dynamics, but also by their "implicit" success criteria: unlike pick-and-place, task quality in these domains is continuous and subjective (e.g. how well a potato is peeled), making quantitative evaluation and reward engineering difficult. We present a learning framework for such tasks, using peeling with a knife as a representative example. Our approach follows a two-stage pipeline: first, we learn a robust initial policy via force-aware data collection and imitation learning, enabling generalization across object variations; second, we refine the policy through preference-based finetuning using a learned reward model that combines quantitative task metrics with qualitative human feedback, aligning policy behavior with human notions of task quality. Using only 50-200 peeling trajectories, our system achieves over 90% average success rates on challenging produce including cucumbers, apples, and potatoes, with performance improving by up to 40% through preference-based finetuning. Remarkably, policies trained on a single produce category exhibit strong zero-shot generalization to unseen in-category instances and to out-of-distribution produce from different categories while maintaining over 90% success rates.

The ability to conduct and learn from interaction and experience is a central challenge in robotics, offering a scalable alternative to labor-intensive human demonstrations. However, realizing such "play" requires (1) a policy robust to diverse, potentially out-of-distribution environment states, and (2) a procedure that continuously produces useful robot experience. To address these challenges, we introduce Tether, a method for autonomous functional play involving structured, task-directed interactions. First, we design a novel open-loop policy that warps actions from a small set of source demonstrations (<=10) by anchoring them to semantic keypoint correspondences in the target scene. We show that this design is extremely data-efficient and robust even under significant spatial and semantic variations. Second, we deploy this policy for autonomous functional play in the real world via a continuous cycle of task selection, execution, evaluation, and improvement, guided by the visual understanding capabilities of vision-language models. This procedure generates diverse, high-quality datasets with minimal human intervention. In a household-like multi-object setup, our method is the first to perform many hours of autonomous multi-task play in the real world starting from only a handful of demonstrations. This produces a stream of data that consistently improves the performance of closed-loop imitation policies over time, ultimately yielding over 1000 expert-level trajectories and training policies competitive with those learned from human-collected demonstrations.

Human mobility trajectories are widely studied in public health and social science, where different demographic groups exhibit significantly different mobility patterns. However, existing trajectory generation models rarely capture this heterogeneity because most trajectory datasets lack demographic labels. To address this gap in data, we propose ATLAS, a weakly supervised approach for demographic-conditioned trajectory generation using only (i) individual trajectories without demographic labels, (ii) region-level aggregated mobility features, and (iii) region-level demographic compositions from census data. ATLAS trains a trajectory generator and fine-tunes it so that simulated mobility matches observed regional aggregates while conditioning on demographics. Experiments on real trajectory data with demographic labels show that ATLAS substantially improves demographic realism over baselines (JSD $\downarrow$ 12%--69%) and closes much of the gap to strongly supervised training. We further develop theoretical analyses for when and why ATLAS works, identifying key factors including demographic diversity across regions and the informativeness of the aggregate feature, paired with experiments demonstrating the practical implications of our theory. We release our code at https://github.com/schang-lab/ATLAS.

Mobile devices are frequent targets of eCrime threat actors through SMS spearphishing (smishing) links that leverage Domain Generation Algorithms (DGA) to rotate hostile infrastructure. Despite this, DGA research and evaluation largely emphasize malware C2 and email phishing datasets, leaving limited evidence on how well detectors generalize to smishing-driven domain tactics outside enterprise perimeters. This work addresses that gap by evaluating traditional and machine-learning DGA detectors against Gravity Falls, a new semi-synthetic dataset derived from smishing links delivered between 2022 and 2025. Gravity Falls captures a single threat actor's evolution across four technique clusters, shifting from short randomized strings to dictionary concatenation and themed combo-squatting variants used for credential theft and fee/fine fraud. Two string-analysis approaches (Shannon entropy and Exp0se) and two ML-based detectors (an LSTM classifier and COSSAS DGAD) are assessed using Top-1M domains as benign baselines. Results are strongly tactic-dependent: performance is highest on randomized-string domains but drops on dictionary concatenation and themed combo-squatting, with low recall across multiple tool/cluster pairings. Overall, both traditional heuristics and recent ML detectors are ill-suited for consistently evolving DGA tactics observed in Gravity Falls, motivating more context-aware approaches and providing a reproducible benchmark for future evaluation.

Feedforward geometric foundation models achieve strong short-window reconstruction, yet scaling them to minutes-long videos is bottlenecked by quadratic attention complexity or limited effective memory in recurrent designs. We present LoGeR (Long-context Geometric Reconstruction), a novel architecture that scales dense 3D reconstruction to extremely long sequences without post-optimization. LoGeR processes video streams in chunks, leveraging strong bidirectional priors for high-fidelity intra-chunk reasoning. To manage the critical challenge of coherence across chunk boundaries, we propose a learning-based hybrid memory module. This dual-component system combines a parametric Test-Time Training (TTT) memory to anchor the global coordinate frame and prevent scale drift, alongside a non-parametric Sliding Window Attention (SWA) mechanism to preserve uncompressed context for high-precision adjacent alignment. Remarkably, this memory architecture enables LoGeR to be trained on sequences of 128 frames, and generalize up to thousands of frames during inference. Evaluated across standard benchmarks and a newly repurposed VBR dataset with sequences of up to 19k frames, LoGeR substantially outperforms prior state-of-the-art feedforward methods--reducing ATE on KITTI by over 74%--and achieves robust, globally consistent reconstruction over unprecedented horizons.

Parallel test-time scaling, which generates multiple candidate solutions for a single problem, is a powerful technique for improving large language model performance. However, it is hindered by two key bottlenecks: accurately selecting the correct solution from the candidate pool, and the high inference latency from generating many full solutions. We argue that both challenges are fundamentally linked to verifier calibration. A well-calibrated verifier not only improves answer selection, but also enables early-stopping strategies to reduce latency. However, existing verifiers are limited as they score each candidate in isolation, overlooking rich contextual information across the set of candidates. To address this, we introduce the Multi-Sequence Verifier (MSV), the first verifier designed to jointly process all candidate solutions and model their interactions. MSV achieves improved calibration, which directly enhances best-of-N selection performance. We further introduce a streaming MSV variant that empowers a novel early-stopping framework. Our novel framework fully leverages parallel decoding, which contrasts with the existing multi-sequence early exit works that decode sequences one by one and thus incur significant latency. In this novel setting, MSV can achieve the same target accuracy with around half the latency that would be required with its counterpart that scores each solution in isolation.

The numerical simulation of convection-dominated transient transport phenomena poses significant computational challenges due to sharp gradients and propagating fronts across the spatiotemporal domain. Classical discretization methods often generate spurious oscillations, requiring advanced stabilization techniques. However, even stabilized finite element methods may require additional regularization to accurately resolve localized steep layers. On the other hand, standalone physics-informed neural networks (PINNs) struggle to capture sharp solution structures in convection-dominated regimes and typically require a large number of training epochs. This work presents a hybrid computational framework that extends the PINN-Augmented SUPG with Shock-Capturing (PASSC) methodology from steady to unsteady problems. The approach combines a semi-discrete stabilized finite element method with a PINN-based correction strategy for transient convection-diffusion-reaction equations. Stabilization is achieved using the Streamline-Upwind Petrov-Galerkin (SUPG) formulation augmented with a YZbeta shock-capturing operator. Rather than training over the entire space-time domain, the neural network is applied selectively near the terminal time, enhancing the finite element solution using the last K_s temporal snapshots while enforcing residual constraints from the governing equations and boundary conditions. The network incorporates residual blocks with random Fourier features and employs progressive training with adaptive loss weighting. Numerical experiments on five benchmark problems, including boundary and interior layers, traveling waves, and nonlinear Burgers dynamics, demonstrate significant accuracy improvements at the terminal time compared to standalone stabilized finite element solutions.

The accelerating adoption of language models (LMs) as agents for deployment in long-context tasks motivates a thorough understanding of goal drift: agents' tendency to deviate from an original objective. While prior-generation language model agents have been shown to be susceptible to drift, the extent to which drift affects more recent models remains unclear. In this work, we provide an updated characterization of the extent and causes of goal drift. We investigate drift in state-of-the-art models within a simulated stock-trading environment (Arike et al., 2025). These models are largely shown to be robust even when subjected to adversarial pressure. We show, however, that this robustness is brittle: across multiple settings, the same models often inherit drift when conditioned on prefilled trajectories from weaker agents. The extent of conditioning-induced drift varies significantly by model family, with only GPT-5.1 maintaining consistent resilience among tested models. We find that drift behavior is inconsistent between prompt variations and correlates poorly with instruction hierarchy following behavior, with strong hierarchy following failing to reliably predict resistance to drift. Finally, we run analogous experiments in a new emergency room triage environment to show preliminary evidence for the transferability of our results across qualitatively different settings. Our findings underscore the continued vulnerability of modern LM agents to contextual pressures and the need for refined post-training techniques to mitigate this.

In this work, we investigate how Large Language Models (LLMs) adapt their internal representations when encountering inputs of increasing difficulty, quantified as the degree of out-of-distribution (OOD) shift. We reveal a consistent and quantifiable phenomenon: as task difficulty increases, whether through harder reasoning questions, longer contexts, or adding answer choices, the last hidden states of LLMs become substantially sparser. In short, \textbf{\textit{the farther the shift, the sparser the representations}}. This sparsity--difficulty relation is observable across diverse models and domains, suggesting that language models respond to unfamiliar or complex inputs by concentrating computation into specialized subspaces in the last hidden state. Through a series of controlled analyses with a learning dynamic explanation, we demonstrate that this sparsity is not incidental but an adaptive mechanism for stabilizing reasoning under OOD. Leveraging this insight, we design \textit{Sparsity-Guided Curriculum In-Context Learning (SG-ICL)}, a strategy that explicitly uses representation sparsity to schedule few-shot demonstrations, leading to considerable performance enhancements. Our study provides new mechanistic insights into how LLMs internalize OOD challenges. The source code is available at the URL: https://github.com/MingyuJ666/sparsityLLM.

AI algorithms for imperfect-information games are typically compared using performance metrics on individual games, making it difficult to assess robustness across game choices. Card games are a natural domain for imperfect information due to hidden hands and stochastic draws. To facilitate comparative research on imperfect-information game-playing algorithms and game systems, we introduce Valet, a diverse and comprehensive testbed of 21 traditional imperfect-information card games. These games span multiple genres, cultures, player counts, deck structures, mechanics, winning conditions, and methods of hiding and revealing information. To standardize implementations across systems, we encode the rules of each game in RECYCLE, a card game description language. We empirically characterize each game's branching factor and duration using random simulations, reporting baseline score distributions for a Monte Carlo Tree Search player against random opponents to demonstrate the suitability of Valet as a benchmarking suite.

Autoregressive decoding is bottlenecked by its sequential nature. Speculative decoding has become a standard way to accelerate inference by using a fast draft model to predict upcoming tokens from a slower target model, and then verifying them in parallel with a single target model forward pass. However, speculative decoding itself relies on a sequential dependence between speculation and verification. We introduce speculative speculative decoding (SSD) to parallelize these operations. While a verification is ongoing, the draft model predicts likely verification outcomes and prepares speculations pre-emptively for them. If the actual verification outcome is then in the predicted set, a speculation can be returned immediately, eliminating drafting overhead entirely. We identify three key challenges presented by speculative speculative decoding, and suggest principled methods to solve each. The result is Saguaro, an optimized SSD algorithm. Our implementation is up to 2x faster than optimized speculative decoding baselines and up to 5x faster than autoregressive decoding with open source inference engines.

Generative artificial intelligence (AI) offers scalable support for formative feedback, yet most AI-generated feedback relies on task-specific rubrics authored by domain experts. While effective, rubric authoring is time-consuming and limits scalability across instructional contexts. Learning progressions (LP) provide a theoretically grounded representation of students' developing understanding and may offer an alternative solution. This study examines whether an LP-driven rubric generation pipeline can produce AI-generated feedback comparable in quality to feedback guided by expert-authored task rubrics. We analyzed AI-generated feedback for written scientific explanations produced by 207 middle school students in a chemistry task. Two pipelines were compared: (a) feedback guided by a human expert-designed, task-specific rubric, and (b) feedback guided by a task-specific rubric automatically derived from a learning progression prior to grading and feedback generation. Two human coders evaluated feedback quality using a multi-dimensional rubric assessing Clarity, Accuracy, Relevance, Engagement and Motivation, and Reflectiveness (10 sub-dimensions). Inter-rater reliability was high, with percent agreement ranging from 89% to 100% and Cohen's kappa values for estimable dimensions (kappa = .66 to .88). Paired t-tests revealed no statistically significant differences between the two pipelines for Clarity (t1 = 0.00, p1 = 1.000; t2 = 0.84, p2 = .399), Relevance (t1 = 0.28, p1 = .782; t2 = -0.58, p2 = .565), Engagement and Motivation (t1 = 0.50, p1 = .618; t2 = -0.58, p2 = .565), or Reflectiveness (t = -0.45, p = .656). These findings suggest that the LP-driven rubric pipeline can serve as an alternative solution.

Language models deployed in online communities must adapt to norms that vary across social, cultural, and domain-specific contexts. Prior alignment approaches rely on explicit preference supervision or predefined principles, which are effective for well-resourced settings but exclude most online communities -- particularly those without institutional backing, annotation infrastructure, or organized around sensitive topics -- where preference elicitation is costly, ethically fraught, or culturally misaligned. We observe that communities already express preferences implicitly through what content they accept, engage with, and allow to persist. We show that this acceptance behavior induces measurable geometric structure in representation space: accepted responses occupy coherent, high-density regions that reflect community-specific norms, while rejected content falls in sparser or misaligned areas. We operationalize this structure as an implicit preference signal for alignment and introduce density-guided response optimization (DGRO), a method that aligns language models to community norms without requiring explicit preference labels. Using labeled preference data, we demonstrate that local density recovers pairwise community judgments, indicating that geometric structure encodes meaningful preference signal. We then apply DGRO in annotation-scarce settings across diverse communities spanning platform, topic, and language. DGRO-aligned models consistently produce responses preferred by human annotators, domain experts, and model-based judges over supervised and prompt-based baselines. We position DGRO as a practical alignment alternative for communities where explicit preference supervision is unavailable or misaligned with situated practices, and discuss the implications and risks of learning from emergent acceptance behavior.

Unified multimodal models have recently demonstrated strong generative capabilities, yet whether and when generation improves understanding remains unclear. Existing benchmarks lack a systematic exploration of the specific tasks where generation facilitates understanding. To this end, we introduce UniG2U-Bench, a comprehensive benchmark categorizing generation-to-understanding (G2U) evaluation into 7 regimes and 30 subtasks, requiring varying degrees of implicit or explicit visual transformations. Extensive evaluation of over 30 models reveals three core findings: 1) Unified models generally underperform their base Vision-Language Models (VLMs), and Generate-then-Answer (GtA) inference typically degrades performance relative to direct inference. 2) Consistent enhancements emerge in spatial intelligence, visual illusions, or multi-round reasoning subtasks, where enhanced spatial and shape perception, as well as multi-step intermediate image states, prove beneficial. 3) Tasks with similar reasoning structures and models sharing architectures exhibit correlated behaviors, suggesting that generation-understanding coupling induces class-consistent inductive biases over tasks, pretraining data, and model architectures. These findings highlight the necessity for more diverse training data and novel paradigms to fully unlock the potential of unified multimodal modeling.

We investigate geometric regularization strategies for learned latent representations in encoder--decoder reduced-order models. In a fixed experimental setting for the advection--diffusion--reaction (ADR) equation, we model latent dynamics using a neural ODE and evaluate four regularization approaches applied during autoencoder pre-training: (a) near-isometry regularization of the decoder Jacobian, (b) a stochastic decoder gain penalty based on random directional gains, (c) a second-order directional curvature penalty, and (d) Stiefel projection of the first decoder layer. Across multiple seeds, we find that (a)--(c) often produce latent representations that make subsequent latent-dynamics training with a frozen autoencoder more difficult, especially for long-horizon rollouts, even when they improve local decoder smoothness or related sensitivity proxies. In contrast, (d) consistently improves conditioning-related diagnostics of the learned latent dynamics and tends to yield better rollout performance. We discuss the hypothesis that, in this setting, the downstream impact of latent-geometry mismatch outweighs the benefits of improved decoder smoothness.

Selecting the number of clusters remains a fundamental challenge in unsupervised learning. Existing criteria typically target a single ``optimal'' partition, often overlooking statistically meaningful structure present at multiple resolutions. We introduce ElbowSig, a framework that formalizes the heuristic ``elbow'' method as a rigorous inferential problem. Our approach centers on a normalized discrete curvature statistic derived from the cluster heterogeneity sequence, which is evaluated against a null distribution of unstructured data. We derive the asymptotic properties of this null statistic in both large-sample and high-dimensional regimes, characterizing its baseline behavior and stochastic variability. As an algorithm-agnostic procedure, ElbowSig requires only the heterogeneity sequence and is compatible with a wide range of clustering methods, including hard, fuzzy, and model-based clustering. Extensive experiments on synthetic and empirical datasets demonstrate that the method maintains appropriate Type-I error control while providing the power to resolve multiscale organizational structures that are typically obscured by single-resolution selection criteria.

Recent advances in generative image editing have enabled transformative applications, from professional head shot generation to avatar stylization. However, these systems often require uploading high-fidelity facial images to third-party models, raising concerns around biometric privacy, data misuse, and user consent. We propose a privacy-preserving pipeline that supports high-quality editing while keeping users in control over their biometric data in face-centric use cases. Our approach separates identity-sensitive regions from editable image context using on-device segmentation and masking, enabling secure, user-controlled editing without modifying third-party generative models. Unlike traditional cloud-based tools, PRIVATEEDIT enforces privacy by default: biometric data is never exposed or transmitted. This design requires no access to or retraining of third-party models, making it compatible with a wide range of commercial APIs. By treating privacy as a core design constraint, our system supports responsible generative AI centered on user autonomy and trust. The pipeline includes a tunable masking mechanism that lets users control how much facial information is concealed, allowing them to balance privacy and output fidelity based on trust level or use case. We demonstrate its applicability in professional and creative workflows and provide a user interface for selective anonymization. By advocating privacy-by-design in generative AI, our work offers both technical feasibility and normative guidance for protecting digital identity. The source code is available at https://github.com/Dipeshtamboli/PrivateEdit-Privacy-Preserving-GenAI.

While deep neural networks (DNNs) have achieved remarkable performance in tasks such as image recognition, they often struggle with generalization, learning from few examples, and continuous adaptation - abilities inherent in biological neural systems. These challenges arise due to DNNs' failure to emulate the efficient, adaptive learning mechanisms of biological networks. To address these issues, we explore the integration of neurobiologically inspired assumptions in neural network learning. This study introduces a biologically inspired learning rule that naturally integrates neurobiological principles, including sparsity, lognormal weight distributions, and adherence to Dale's law, without requiring explicit enforcement. By aligning with these core neurobiological principles, our model enhances robustness against adversarial attacks and demonstrates superior generalization, particularly in few-shot learning scenarios. Notably, integrating these constraints leads to the emergence of biologically plausible neural representations, underscoring the efficacy of incorporating neurobiological assumptions into neural network design. Preliminary results suggest that this approach could extend from feature-specific to task-specific encoding, potentially offering insights into neural resource allocation for complex tasks.

Large Language Models (LLMs) demonstrate potentials for automating scientific code generation but face challenges in reliability, error propagation in multi-agent workflows, and evaluation in domains with ill-defined success metrics. We present a Bayesian adversarial multi-agent framework specifically designed for AI for Science (AI4S) tasks in the form of a Low-code Platform (LCP). Three LLM-based agents are coordinated under the Bayesian framework: a Task Manager that structures user inputs into actionable plans and adaptive test cases, a Code Generator that produces candidate solutions, and an Evaluator providing comprehensive feedback. The framework employs an adversarial loop where the Task Manager iteratively refines test cases to challenge the Code Generator, while prompt distributions are dynamically updated using Bayesian principles by integrating code quality metrics: functional correctness, structural alignment, and static analysis. This co-optimization of tests and code reduces dependence on LLM reliability and addresses evaluation uncertainty inherent to scientific tasks. LCP also streamlines human-AI collaboration by translating non-expert prompts into domain-specific requirements, bypassing the need for manual prompt engineering by practitioners without coding backgrounds. Benchmark evaluations demonstrate LCP's effectiveness in generating robust code while minimizing error propagation. The proposed platform is also tested on an Earth Science cross-disciplinary task and demonstrates strong reliability, outperforming competing models.

The electric vehicle routing problem with time windows (EVRPTW) extends the classical VRPTW by introducing battery capacity constraints and charging station decisions. Existing benchmark datasets are often static and lack verifiable feasibility, which restricts reproducible evaluation of learning-based routing models. We introduce SynthCharge, a parametric generator that produces diverse, feasibility-screened EVRPTW instances across varying spatiotemporal configurations and scalable customer counts. While SynthCharge can currently generate large-scale instances of up to 500 customers, we focus our experiments on sizes ranging from 5 to 100 customers. Unlike static benchmark suites, SynthCharge integrates instance geometry with adaptive energy capacity scaling and range-aware charging station placement. To guarantee structural validity, the generator systematically filters out unsolvable instances through a fast feasibility screening process. Ultimately, SynthCharge provides the dynamic benchmarking infrastructure needed to systematically evaluate the robustness of emerging neural routing and data-driven approaches.

Categorical deep learning (CDL) has recently emerged as a framework that leverages category theory to unify diverse neural architectures. While geometric deep learning (GDL) is grounded in the specific context of invariants of group actions, CDL aims to provide domain-independent abstractions for reasoning about models and their properties. In this paper, we contribute to this program by developing a coalgebraic foundation for equivariant representation in deep learning, as classical notions of group actions and equivariant maps are naturally generalized by the coalgebraic formalism. Our first main result demonstrates that, given an embedding of data sets formalized as a functor from SET to VECT, and given a notion of invariant behavior on data sets modeled by an endofunctor on SET, there is a corresponding endofunctor on VECT that is compatible with the embedding in the sense that this lifted functor recovers the analogous notion of invariant behavior on the embedded data. Building on this foundation, we then establish a universal approximation theorem for equivariant maps in this generalized setting. We show that continuous equivariant functions can be approximated within our coalgebraic framework for a broad class of symmetries. This work thus provides a categorical bridge between the abstract specification of invariant behavior and its concrete realization in neural architectures.

Differential Privacy (DP) is becoming central to large-scale training as privacy regulations tighten. We revisit how DP noise interacts with adaptivity in optimization through the lens of stochastic differential equations, providing the first SDE-based analysis of private optimizers. Focusing on DP-SGD and DP-SignSGD under per-example clipping, we show a sharp contrast under fixed hyperparameters: DP-SGD converges at a Privacy-Utility Trade-Off of $\mathcal{O}(1/\varepsilon^2)$ with speed independent of $\varepsilon$, while DP-SignSGD converges at a speed linear in $\varepsilon$ with an $\mathcal{O}(1/\varepsilon)$ trade-off, dominating in high-privacy or large batch noise regimes. By contrast, under optimal learning rates, both methods achieve comparable theoretical asymptotic performance; however, the optimal learning rate of DP-SGD scales linearly with $\varepsilon$, while that of DP-SignSGD is essentially $\varepsilon$-independent. This makes adaptive methods far more practical, as their hyperparameters transfer across privacy levels with little or no re-tuning. Empirical results confirm our theory across training and test metrics, and empirically extend from DP-SignSGD to DP-Adam.

Physics-Informed Neural Networks (PINNs) have been recognized as a mesh-free alternative to solve partial differential equations where physics information is incorporated. However, in dealing with problems characterized by high stiffness or shock-dominated dynamics, traditional PINNs have been found to have limitations, including unbalanced training and inaccuracy in solution, even with small physics residuals. In this research, we seek to address these limitations using the viscous Burgers' equation with low viscosity and the Allen-Cahn equation as test problems. In addressing unbalanced training, we have developed a new adaptive loss balancing scheme using smoothed gradient norms to ensure satisfaction of initial and boundary conditions. Further, to address inaccuracy in the solution, we have developed an adaptive residual-based collocation scheme to improve the accuracy of solutions in the regions with high physics residuals. The proposed new approach significantly improves solution accuracy with consistent satisfaction of physics residuals. For instance, in the case of Burgers' equation, the relative L2 error is reduced by about 44 percent compared to traditional PINNs, while for the Allen-Cahn equation, the relative L2 error is reduced by approximately 70 percent. Additionally, we show the trustworthy solution comparison of the proposed method using a robust finite difference solver.

Observational causal discovery is only identifiable up to the Markov equivalence class. While interventions can reduce this ambiguity, in practice interventions are often soft with multiple unknown targets. In many realistic scenarios, only a single intervention regime is observed. We propose a scalable causal discovery model for paired observational and interventional settings with shared underlying causal structure and unknown soft interventions. The model aggregates subset-level PDAGs and applies contrastive cross-regime orientation rules to construct a globally consistent maximal PDAG under Meek closure, enabling generalization to both in-distribution and out-of-distribution settings. Theoretically, we prove that our model is sound with respect to a restricted $Ψ$ equivalence class induced solely by the information available in the subset-restricted setting. We further show that the model asymptotically recovers the corresponding identifiable PDAG and can orient additional edges compared to non-contrastive subset-restricted methods. Experiments on synthetic data demonstrate improved causal structure recovery, generalization to unseen graphs with held-out causal mechanisms, and scalability to larger graphs, with ablations supporting the theoretical results.

Real-time proactive agentic system, capable of modeling Human State of Mind, using foundation EXG model and text embeddings model, running fully offline on the edge. Unlike all previously known systems, the NeuroSkill(tm) system leverages SKILL.md description of Human's State of Mind via API and CLI provided by the system, directly from the Brain-Computer Interface (BCI) devices, which records Human biophysical and brain signals. Our custom harness - NeuroLoop(tm) - utilizes all of the above to run agentic flow that manages to engage with the Human on multiple cognitive and affective levels of their State of Mind (e.g., empathy), by providing actionable tool calls and protocol execution with explicit or implicit requests from the Human. GPLv3 open-source software with ethically aligned AI100 licensing for the skill markdown.

Shape optimization under uncertainty (OUU) is computationally intensive for classical PDE-based methods due to the high cost of repeated sampling-based risk evaluation across many uncertainty realizations and varying geometries, while standard neural surrogates often fail to provide accurate and efficient sensitivities for optimization. We introduce Shape-DINO, a derivative-informed neural operator framework for learning PDE solution operators on families of varying geometries, with a particular focus on accelerating PDE-constrained shape OUU. Shape-DINOs encode geometric variability through diffeomorphic mappings to a fixed reference domain and employ a derivative-informed operator learning objective that jointly learns the PDE solution and its Fréchet derivatives with respect to design variables and uncertain parameters, enabling accurate state predictions and reliable gradients for large-scale OUU. We establish a priori error bounds linking surrogate accuracy to optimization error and prove universal approximation results for multi-input reduced basis neural operators in suitable $C^1$ norms. We demonstrate efficiency and scalability on three representative shape OUU problems, including boundary design for a Poisson equation and shape design governed by steady-state Navier-Stokes exterior flows in two and three dimensions. Across these examples, Shape-DINOs produce more reliable optimization results than operator surrogates trained without derivative information. In our examples, Shape-DINOs achieve 3-8 orders-of-magnitude speedups in state and gradient evaluations. Counting training data generation, Shape-DINOs reduce necessary PDE solves by 1-2 orders-of-magnitude compared to a strictly PDE-based approach for a single OUU problem. Moreover, Shape-DINO construction costs can be amortized across many objectives and risk measures, enabling large-scale shape OUU for complex systems.

Causal discovery from observational data is a fundamental tool in various fields of science. While existing approaches are typically designed for a single dataset, we often need to handle multiple datasets with non-identical variable sets in practice. One straightforward approach is to estimate a causal graph from each dataset and construct a single causal graph by overlapping. However, this approach identifies limited causal relationships because unobserved variables in each dataset can be confounders, and some variable pairs may be unobserved in any dataset. To address this issue, we leverage Causal Additive Models with Unobserved Variables (CAM-UV) that provide causal graphs having information related to unobserved variables. We show that the ground truth causal graph has structural consistency with the information of CAM-UV on each dataset. As a result, we propose an approach named I-CAM-UV to integrate CAM-UV results by enumerating all consistent causal graphs. We also provide an efficient combinatorial search algorithm and demonstrate the usefulness of I-CAM-UV against existing methods.

Contrastive steering has been shown as a simple and effective method to adjust the generative behavior of LLMs at inference time. It uses examples of prompt responses with and without a trait to identify a direction in an intermediate activation layer, and then shifts activations in this 1-dimensional subspace. However, despite its growing use in AI safety applications, the robustness of contrastive steering to noisy or adversarial data corruption is poorly understood. We initiate a study of the robustness of this process with respect to corruption of the dataset of examples used to train the steering direction. Our first observation is that contrastive steering is quite robust to a moderate amount of corruption, but unwanted side effects can be clearly and maliciously manifested when a non-trivial fraction of the training data is altered. Second, we analyze the geometry of various types of corruption, and identify some safeguards. Notably, a key step in learning the steering direction involves high-dimensional mean computation, and we show that replacing this step with a recently developed robust mean estimator often mitigates most of the unwanted effects of malicious corruption.

Agentic language models operate in a fundamentally different safety regime than chat models: they must plan, call tools, and execute long-horizon actions where a single misstep, such as accessing files or entering credentials, can cause irreversible harm. Existing alignment methods, largely optimized for static generation and task completion, break down in these settings due to sequential decision-making, adversarial tool feedback, and overconfident intermediate reasoning. We introduce MOSAIC, a post-training framework that aligns agents for safe multi-step tool use by making safety decisions explicit and learnable. MOSAIC structures inference as a plan, check, then act or refuse loop, with explicit safety reasoning and refusal as first-class actions. To train without trajectory-level labels, we use preference-based reinforcement learning with pairwise trajectory comparisons, which captures safety distinctions often missed by scalar rewards. We evaluate MOSAIC zero-shot across three model families, Qwen2.5-7B, Qwen3-4B-Thinking, and Phi-4, and across out-of-distribution benchmarks spanning harmful tasks, prompt injection, benign tool use, and cross-domain privacy leakage. MOSAIC reduces harmful behavior by up to 50%, increases harmful-task refusal by over 20% on injection attacks, cuts privacy leakage, and preserves or improves benign task performance, demonstrating robust generalization across models, domains, and agentic settings.

CDD, or Contamination Detection via output Distribution, identifies data contamination by measuring the peakedness of a model's sampled outputs. We study the conditions under which this approach succeeds and fails on small language models ranging from 70M to 410M parameters. Using controlled contamination experiments on GSM8K, HumanEval, and MATH, we find that CDD's effectiveness depends critically on whether fine-tuning produces verbatim memorization. With low-rank adaptation, models can learn from contaminated data without memorizing it, and CDD performs at chance level even when the data is verifiably contaminated. Only when fine-tuning capacity is sufficient to induce memorization does CDD recover strong detection accuracy. Our results characterize a memorization threshold that governs detectability and highlight a practical consideration: parameter-efficient fine-tuning can produce contamination that output-distribution methods do not detect. Our code is available at https://github.com/Sela-Omer/Contamination-Detection-Small-LM

As large language models (LLMs) advance their mathematical capabilities toward the IMO level, the scarcity of challenging, high-quality problems for training and evaluation has become a significant bottleneck. Simultaneously, recent code agents have demonstrated sophisticated skills in agentic coding and reasoning, suggesting that code execution can serve as a scalable environment for mathematical experimentation. In this paper, we investigate the potential of code agents to autonomously evolve existing math problems into more complex variations. We introduce a multi-agent framework designed to perform problem evolution while validating the solvability and increased difficulty of the generated problems. Our experiments demonstrate that, given sufficient test-time exploration, code agents can synthesize new, solvable problems that are structurally distinct from and more challenging than the originals. This work provides empirical evidence that code-driven agents can serve as a viable mechanism for synthesizing high-difficulty mathematical reasoning problems within scalable computational environments. Our data is available at https://github.com/TarferSoul/Code2Math.

Universal embodied intelligence demands robust generalization across heterogeneous embodiments, such as autonomous driving, robotics, and unmanned aerial vehicles (UAVs). However, existing embodied brain in training a unified model over diverse embodiments frequently triggers long-tail data, gradient interference, and catastrophic forgetting, making it notoriously difficult to balance universal generalization with domain-specific proficiency. In this report, we introduce ACE-Brain-0, a generalist foundation brain that unifies spatial reasoning, autonomous driving, and embodied manipulation within a single multimodal large language model~(MLLM). Our key insight is that spatial intelligence serves as a universal scaffold across diverse physical embodiments: although vehicles, robots, and UAVs differ drastically in morphology, they share a common need for modeling 3D mental space, making spatial cognition a natural, domain-agnostic foundation for cross-embodiment transfer. Building on this insight, we propose the Scaffold-Specialize-Reconcile~(SSR) paradigm, which first establishes a shared spatial foundation, then cultivates domain-specialized experts, and finally harmonizes them through data-free model merging. Furthermore, we adopt Group Relative Policy Optimization~(GRPO) to strengthen the model's comprehensive capability. Extensive experiments demonstrate that ACE-Brain-0 achieves competitive and even state-of-the-art performance across 24 spatial and embodiment-related benchmarks.

Vision-Language-Action (VLA) models are a promising path toward embodied intelligence, yet they often overlook the predictive and temporal-causal structure underlying visual dynamics. World-model VLAs address this by predicting future frames, but waste capacity reconstructing redundant backgrounds. Latent-action VLAs encode frame-to-frame transitions compactly, but lack temporally continuous dynamic modeling and world knowledge. To overcome these limitations, we introduce CoWVLA (Chain-of-World VLA), a new "Chain of World" paradigm that unifies world-model temporal reasoning with a disentangled latent motion representation. First, a pretrained video VAE serves as a latent motion extractor, explicitly factorizing video segments into structure and motion latents. Then, during pre-training, the VLA learns from an instruction and an initial frame to infer a continuous latent motion chain and predict the segment's terminal frame. Finally, during co-fine-tuning, this latent dynamic is aligned with discrete action prediction by jointly modeling sparse keyframes and action sequences in a unified autoregressive decoder. This design preserves the world-model benefits of temporal reasoning and world knowledge while retaining the compactness and interpretability of latent actions, enabling efficient visuomotor learning. Extensive experiments on robotic simulation benchmarks show that CoWVLA outperforms existing world-model and latent-action approaches and achieves moderate computational efficiency, highlighting its potential as a more effective VLA pretraining paradigm. The project website can be found at https://fx-hit.github.io/cowvla-io.

Current benchmarks for code agents primarily assess narrow, repository-specific fixes, overlooking critical real-world challenges such as cross-repository reasoning, domain-specialized problem solving, dependency-driven migration, and full-repository generation. To address this gap, we introduce BeyondSWE, a comprehensive benchmark that broadens existing evaluations along two axes - resolution scope and knowledge scope - using 500 real-world instances across four distinct settings. Experimental results reveal a significant capability gap: even frontier models plateau below 45% success, and no single model performs consistently across task types. To systematically investigate the role of external knowledge, we develop SearchSWE, a framework that integrates deep search with coding abilities. Our experiments show that search augmentation yields inconsistent gains and can in some cases degrade performance, highlighting the difficulty of emulating developer-like workflows that interleave search and reasoning during coding tasks. This work offers both a realistic, challenging evaluation benchmark and a flexible framework to advance research toward more capable code agents.

Omni-modal large language models (omni LLMs) have recently achieved strong performance across audiovisual understanding tasks, yet they remain highly susceptible to cross-modal hallucinations arising from spurious correlations and dominant language priors. In this work, we propose Modality-Decoupled Direct Preference Optimization (MoD-DPO), a simple and effective framework for improving modality grounding in omni LLMs. MoD-DPO introduces modality-aware regularization terms that explicitly enforce invariance to corruptions in irrelevant modalities and sensitivity to perturbations in relevant modalities, thereby reducing unintended cross-modal interactions. To further mitigate over-reliance on textual priors, we incorporate a language-prior debiasing penalty that discourages hallucination-prone text-only responses. Extensive experiments across multiple audiovisual hallucination benchmarks demonstrate that MoD-DPO consistently improves perception accuracy and hallucination resistance, outperforming previous preference optimization baselines under similar training budgets. Our findings underscore the importance of modality-faithful alignment and demonstrate a scalable path toward more reliable and resilient multimodal foundation models.

Google's SynthID-Text, the first ever production-ready generative watermark system for large language model, designs a novel Tournament-based method that achieves the state-of-the-art detectability for identifying AI-generated texts. The system's innovation lies in: 1) a new Tournament sampling algorithm for watermarking embedding, 2) a detection strategy based on the introduced score function (e.g., Bayesian or mean score), and 3) a unified design that supports both distortionary and non-distortionary watermarking methods. This paper presents the first theoretical analysis of SynthID-Text, with a focus on its detection performance and watermark robustness, complemented by empirical validation. For example, we prove that the mean score is inherently vulnerable to increased tournament layers, and design a layer inflation attack to break SynthID-Text. We also prove the Bayesian score offers improved watermark robustness w.r.t. layers and further establish that the optimal Bernoulli distribution for watermark detection is achieved when the parameter is set to 0.5. Together, these theoretical and empirical insights not only deepen our understanding of SynthID-Text, but also open new avenues for analyzing effective watermark removal strategies and designing robust watermarking techniques. Source code is available at https: //github.com/romidi80/Synth-ID-Empirical-Analysis.

In off policy evaluation (OPE) for partially observable Markov decision processes (POMDPs), an agent must infer hidden states from past observations, which exacerbates both the curse of horizon and the curse of memory in existing OPE methods. This paper introduces a novel covering analysis framework that exploits the intrinsic metric structure of the belief space (distributions over latent states) to relax traditional coverage assumptions. By assuming value relevant functions are Lipschitz continuous in the belief space, we derive error bounds that mitigate exponential blow ups in horizon and memory length. Our unified analysis technique applies to a broad class of OPE algorithms, yielding concrete error bounds and coverage requirements expressed in terms of belief space metrics rather than raw history coverage. We illustrate the improved sample efficiency of this framework via case studies: the double sampling Bellman error minimization algorithm, and the memory based future dependent value functions (FDVF). In both cases, our coverage definition based on the belief space metric yields tighter bounds.

We introduce a novel framework for uncertainty quantification in clustering. By combining the martingale posterior paradigm with density-based clustering, uncertainty in the estimated density is naturally propagated to the clustering structure. The approach scales effectively to high-dimensional and irregularly shaped data by leveraging modern neural density estimators and GPU-friendly parallel computation. We establish frequentist consistency guarantees and validate the methodology on synthetic and real data.

Automated industrial optimization modeling requires reliable translation of natural-language requirements into solver-executable code. However, large language models often generate non-compilable models due to missing declarations, type inconsistencies, and incomplete dependency contexts. We propose a type-aware retrieval-augmented generation (RAG) method that enforces modeling entity types and minimal dependency closure to ensure executability. Unlike existing RAG approaches that index unstructured text, our method constructs a domain-specific typed knowledge base by parsing heterogeneous sources, such as academic papers and solver code, into typed units and encoding their mathematical dependencies in a knowledge graph. Given a natural-language instruction, it performs hybrid retrieval and computes a minimal dependency-closed context, the smallest set of typed symbols required for solver-executable code, via dependency propagation over the graph. We validate the method on two constraint-intensive industrial cases: demand response optimization in battery production and flexible job shop scheduling. In the first case, our method generates an executable model incorporating demand-response incentives and load-reduction constraints, achieving peak shaving while preserving profitability; conventional RAG baselines fail. In the second case, it consistently produces compilable models that reach known optimal solutions, demonstrating robust cross-domain generalization; baselines fail entirely. Ablation studies confirm that enforcing type-aware dependency closure is essential for avoiding structural hallucinations and ensuring executability, addressing a critical barrier to deploying large language models in complex engineering optimization tasks.

The integration of symbolic computing with neural networks has intrigued researchers since the first theorizations of Artificial intelligence (AI). The ability of Neuro-Symbolic (NeSy) methods to infer or exploit behavioral schema has been widely considered as one of the possible proxies for human-level intelligence. However, the limited semantic generalizability and the challenges in declining complex domains with pre-defined patterns and rules hinder their practical implementation in real-world scenarios. The unprecedented results achieved by connectionist systems since the last AI breakthrough in 2017 have raised questions about the competitiveness of NeSy solutions, with particular emphasis on the Natural Language Processing and Computer Vision fields. This survey examines task-specific advancements in the NeSy domain to explore how incorporating symbolic systems can enhance explainability and reasoning capabilities. Our findings are meant to serve as a resource for researchers exploring explainable NeSy methodologies for real-life tasks and applications. Reproducibility details and in-depth comments on each surveyed research work are made available at https://github.com/disi-unibo-nlp/task-oriented-neuro-symbolic.git.

This short note describes a simple variant of the Squint algorithm of Koolen and Van Erven [2015] for the classic expert problem. Via an equally simple modification of their proof, we prove that this variant ensures a regret bound that resembles the one shown in a recent work by Freund et al. [2026] for a variant of the NormalHedge algorithm [Chaudhuri et al., 2009].

Hierarchical text classification (HTC) and extreme multi-label classification (XML) tasks face compounded challenges from complex label interdependencies, data sparsity, and extreme output dimensions. These challenges are exemplified in the European Food Safety Authority's FoodEx2 system-a standardized food classification framework essential for food consumption monitoring and contaminant exposure assessment across Europe. FoodEx2 coding transforms natural language food descriptions into a set of codes from multiple standardized hierarchies, but faces implementation barriers due to its complex structure. Given a food description (e.g., "organic yogurt''), the system identifies its base term ("yogurt''), all the applicable facet categories (e.g., "production method''), and then, every relevant facet descriptors to each category (e.g., "organic production''). While existing models perform adequately on well-balanced and semantically dense hierarchies, no work has been applied on the practical constraints imposed by the FoodEx2 system. The limited literature addressing such real-world scenarios further compounds these challenges. We propose FEAST (Food Embedding And Semantic Taxonomy), a novel retrieval-augmented framework that decomposes FoodEx2 classification into a three-stage approach: (1) base term identification, (2) multi-label facet prediction, and (3) facet descriptor assignment. By leveraging the system's hierarchical structure to guide training and performing deep metric learning, FEASTlearns discriminative embeddings that mitigate data sparsity and improve generalization on rare and fine-grained labels. Evaluated on the multilingual FoodEx2 benchmark, FEAST outperforms the prior European's CNN baseline F1 scores by 12-38 % on rare classes.

Saarthi is an agentic AI framework that uses multi-agent collaboration to perform end-to-end formal verification. Even though the framework provides a complete flow from specification to coverage closure, with around 40% efficacy, there are several challenges that need to be addressed to make it more robust and reliable. Artificial General Intelligence (AGI) is still a distant goal, and current Large Language Model (LLM)-based agents are prone to hallucinations and making mistakes, especially when dealing with complex tasks such as formal verification. However, with the right enhancements and improvements, we believe that Saarthi can be a significant step towards achieving domain-specific general intelligence for formal verification. Especially for problems that require Short Term, Short Context (STSC) capabilities, such as formal verification, Saarthi can be a powerful tool to assist verification engineers in their work. In this paper, we present two key enhancements to the Saarthi framework: (1) a structured rulebook and specification grammar to improve the accuracy and controllability of SystemVerilog Assertion (SVA) generation, and (2) integration of advanced Retrieval Augmented Generation (RAG) techniques, such as GraphRAG, to provide agents with access to technical knowledge and best practices for iterative refinement and improvement of outputs. We also benchmark these enhancements for the overall Saarthi framework using challenging test cases from NVIDIA's CVDP benchmark targeting formal verification. Our benchmark results stand out with a 70% improvement in the accuracy of generated assertions, and a 50% reduction in the number of iterations required to achieve coverage closure.

Certified machine unlearning aims to provably remove the influence of a deletion set $U$ from a model trained on a dataset $S$, by producing an unlearned output that is statistically indistinguishable from retraining on the retain set $R:=S\setminus U$. Many existing certified unlearning methods adapt techniques from Differential Privacy (DP) and add noise calibrated to global sensitivity, i.e., the worst-case output change over all adjacent datasets. We show that this DP-style calibration is often overly conservative for unlearning, based on a key observation: certified unlearning, by definition, does not require protecting the privacy of the retained data $R$. Motivated by this distinction, we define retain sensitivity as the worst-case output change over deletions $U$ while keeping $R$ fixed. While insufficient for DP, retain sensitivity is exactly sufficient for unlearning, allowing for the same certificates with less noise. We validate these reductions in noise theoretically and empirically across several problems, including the weight of minimum spanning trees, PCA, and ERM. Finally, we refine the analysis of two widely used certified unlearning algorithms through the lens of retain sensitivity, leveraging the regularity induced by $R$ to further reduce noise and improve utility.

Despite their impressive capabilities, current Text-to-Image (T2I) models remain prone to generating unsafe and toxic content. While activation steering offers a promising inference-time intervention, we observe that linear activation steering frequently degrades image quality when applied to benign prompts. To address this trade-off, we first construct SafeSteerDataset, a contrastive dataset containing 2300 safe and unsafe prompt pairs with high cosine similarity. Leveraging this data, we propose Conditioned Activation Transport (CAT), a framework that employs a geometry-based conditioning mechanism and nonlinear transport maps. By conditioning transport maps to activate only within unsafe activation regions, we minimize interference with benign queries. We validate our approach on two state-of-the-art architectures: Z-Image and Infinity. Experiments demonstrate that CAT generalizes effectively across these backbones, significantly reducing Attack Success Rate while maintaining image fidelity compared to unsteered generations. Warning: This paper contains potentially offensive text and images.

Large language models (LLMs) have shown strong empirical performance across pharmacology and drug discovery tasks, yet the internal mechanisms by which they encode pharmacological knowledge remain poorly understood. In this work, we investigate how drug-group semantics are represented and retrieved within Llama-based biomedical language models using causal and probing-based interpretability methods. We apply activation patching to localize where drug-group information is stored across model layers and token positions, and complement this analysis with linear probes trained on token-level and sum-pooled activations. Our results demonstrate that early layers play a key role in encoding drug-group knowledge, with the strongest causal effects arising from intermediate tokens within the drug-group span rather than the final drug-group token. Linear probing further reveals that pharmacological semantics are distributed across tokens and are already present in the embedding space, with token-level probes performing near chance while sum-pooled representations achieve maximal accuracy. Together, these findings suggest that drug-group semantics in LLMs are not localized to single tokens but instead arise from distributed representations. This study provides the first systematic mechanistic analysis of pharmacological knowledge in LLMs, offering insights into how biomedical semantics are encoded in large language models.

Bengali remains a low-resource language in speech technology, especially for complex tasks like long-form transcription and speaker diarization. This paper presents a multistage approach developed for the "DL Sprint 4.0 - Bengali Long-Form Speech Recognition" and "DL Sprint 4.0 - Bengali Speaker Diarization" competitions on Kaggle, addressing the challenge of "who spoke when/what" in hour-long recordings. We implemented Whisper Medium fine-tuned on Bengali data (bengaliAI/tugstugi bengaliai-asr whisper-medium) for transcription and integrated pyannote/speaker-diarization-community-1 with our custom-trained segmentation model to handle diverse and noisy acoustic environments. Using a two-pass method with hyperparameter tuning, we achieved a DER of 0.27 on the private leaderboard and 0.19 on the public leaderboard. For transcription, chunking, background noise cleaning, and algorithmic post-processing yielded a WER of 0.38 on the private leaderboard. These results show that targeted tuning and strategic data utilization can significantly improve AI inclusivity for South Asian languages. All relevant code is available at: https://github.com/Short-Potatoes/Bengali-long-form-transcription-and-diarization.git Index Terms: Bengali speech recognition, speaker diarization, Whisper, ASR, low-resource languages, pyannote, voice activity detection

How should two language models interact to produce better code than either can alone? The conventional approach -- a reasoning model plans, a code specialist implements -- seems natural but fails: on HumanEval+, plan-then-code degrades performance by 2.4 percentage points versus the code specialist alone. We show that reversing the interaction changes everything. When the code specialist generates freely and the reasoning model reviews instead of plans, the same two models on the same hardware achieve 90.2% pass@1 -- exceeding GPT-4o (87.2%) and O1 Preview (89.0%) -- on ~$2/hr of commodity GPU. Cross-benchmark validation across 542 problems (HumanEval+ and MBPP+) reveals a moderating variable: review effectiveness scales with specification richness, yielding 4x more improvement on richly-specified problems (+9.8pp) than on lean ones (+2.3pp), while remaining net-positive in both cases. The practical implication is twofold: compose models by their cognitive strengths (reviewers review, coders code), and invest in specification quality to amplify the returns.

What do atomistic foundation models encode in their intermediate representations, and how is that information organized? We introduce Composition Projection Decomposition (CPD), which uses QR projection to linearly remove composition signal from learned representations and probes the geometric residual. Across eight models from five architectural families on QM9 molecules and Materials Project crystals, we find a disentanglement gradient: tensor product equivariant architectures (MACE) produce representations where geometry is almost fully linearly accessible after composition removal ($R^2_{\text{geom}} = 0.782$ for HOMO-LUMO gap), while handcrafted descriptors (ANI-2x) entangle the same information nonlinearly ($R^2_{\text{geom}} = -0.792$ under Ridge; $R^2 = +0.784$ under MLP). MACE routes target-specific signal through irreducible representation channels -- dipole to $L = 1$, HOMO-LUMO gap to $L = 0$ -- a pattern not observed in ViSNet's vector-scalar architecture under the same probe. We show that gradient boosted tree probes on projected residuals are systematically inflated, recovering $R^2 = 0.68$--$0.95$ on a purely compositional target, and recommend linear probes as the primary metric. Linearly disentangled representations are more sample-efficient under linear probing, suggesting a practical advantage for equivariant architectures beyond raw prediction accuracy.

The forward and reverse Kullback-Leibler (KL) divergences arise as limiting objectives in learning and inference yet induce markedly different inductive biases that cannot be explained at the level of expectations alone. In this work, we introduce the Surprisal-Rényi Free Energy (SRFE), a log-moment-based functional of the likelihood ratio that lies outside the class of $f$-divergences. We show that SRFE recovers forward and reverse KL divergences as singular endpoint limits and derive local expansions around both limits in which the variance of the log-likelihood ratio appears as a first-order correction. This reveals an explicit mean-variance tradeoff governing departures from KL-dominated regimes. We further establish a Gibbs-type variational characterization of SRFE as the unique minimizer of a weighted sum of KL divergences and prove that SRFE directly controls large deviations of excess code-length via Chernoff-type bounds, yielding a precise Minimum Description Length interpretation. Together, these results identify SRFE as a variance- and tail-sensitive free-energy functional that clarifies the geometric and large-deviation structure underlying forward and reverse KL limits, without unifying or subsuming distinct learning frameworks.

Coverage closure is a critical requirement in Integrated Chip (IC) development process and key metric for verification sign-off. However, traditional exhaustive approaches often fail to achieve full coverage within project timelines. This study presents an agentic AI-driven workflow that utilizes Large Language Model (LLM)-enabled Generative AI (GenAI) to automate coverage analysis for formal verification, identify coverage gaps, and generate the required formal properties. The framework accelerates verification efficiency by systematically addressing coverage holes. Benchmarking open-source and internal designs reveals a measurable increase in coverage metrics, with improvements correlated to the complexity of the design. Comparative analysis validates the effectiveness of this approach. These results highlight the potential of agentic AI-based techniques to improve formal verification productivity and support comprehensive coverage closure.

\emph{Integrated communication and computation} (IC$^2$) has emerged as a new paradigm for enabling efficient edge inference in sixth-generation (6G) networks. However, the design of IC$^2$ technologies is hindered by the lack of a tractable theoretical framework for characterizing \emph{end-to-end} (E2E) inference performance. The metric is highly complicated as it needs to account for both channel distortion and artificial intelligence (AI) model architecture and computational complexity. In this work, we address this challenge by developing a tractable analytical model for E2E inference accuracy and leveraging it to design a \emph{channel-adaptive AI} algorithm that maximizes inference throughput, referred to as the edge processing rate (EPR), under latency and accuracy constraints. Specifically, we consider an edge inference system in which a server deploys a backbone model with early exit, which enables flexible computational complexity, to perform inference on data features transmitted by a mobile device. The proposed accuracy model characterizes high-dimensional feature distributions in the angular domain using a Mixture of von Mises (MvM) distribution. This leads to a desired closed-form expression for inference accuracy as a function of quantization bit-width and model traversal depth, which represents channel distortion and computational complexity, respectively. Building upon this accuracy model, we formulate and solve the EPR maximization problem under joint latency and accuracy constraints, leading to a channel-adaptive AI algorithm that achieves full IC$^2$ integration. The proposed algorithm jointly adapts transmit-side feature compression and receive-side model complexity according to channel conditions to maximize overall efficiency and inference throughput. Experimental results demonstrate its superior performance as compared with fixed-complexity counterparts.

Leveraging the priors of 2D diffusion models for 3D editing has emerged as a promising paradigm. However, maintaining multi-view consistency in edited results remains challenging, and the extreme scarcity of 3D-consistent editing paired data renders supervised fine-tuning (SFT), the most effective training strategy for editing tasks, infeasible. In this paper, we observe that, while generating multi-view consistent 3D content is highly challenging, verifying 3D consistency is tractable, naturally positioning reinforcement learning (RL) as a feasible solution. Motivated by this, we propose \textbf{RL3DEdit}, a single-pass framework driven by RL optimization with novel rewards derived from the 3D foundation model, VGGT. Specifically, we leverage VGGT's robust priors learned from massive real-world data, feed the edited images, and utilize the output confidence maps and pose estimation errors as reward signals, effectively anchoring the 2D editing priors onto a 3D-consistent manifold via RL. Extensive experiments demonstrate that RL3DEdit achieves stable multi-view consistency and outperforms state-of-the-art methods in editing quality with high efficiency. To promote the development of 3D editing, we will release the code and model.

Scientific discoveries must be communicated clearly to realize their full potential. Without effective communication, even the most groundbreaking findings risk being overlooked or misunderstood. The primary way scientists communicate their work and receive feedback from the community is through peer review. However, the current system often provides inconsistent feedback between reviewers, ultimately hindering the improvement of a manuscript and limiting its potential impact. In this paper, we introduce a novel method APRES powered by Large Language Models (LLMs) to update a scientific papers text based on an evaluation rubric. Our automated method discovers a rubric that is highly predictive of future citation counts, and integrate it with APRES in an automated system that revises papers to enhance their quality and impact. Crucially, this objective should be met without altering the core scientific content. We demonstrate the success of APRES, which improves future citation prediction by 19.6% in mean averaged error over the next best baseline, and show that our paper revision process yields papers that are preferred over the originals by human expert evaluators 79% of the time. Our findings provide strong empirical support for using LLMs as a tool to help authors stress-test their manuscripts before submission. Ultimately, our work seeks to augment, not replace, the essential role of human expert reviewers, for it should be humans who discern which discoveries truly matter, guiding science toward advancing knowledge and enriching lives.

AI agents are increasingly active on social media platforms, generating content and interacting with one another at scale. Yet the behavioral diversity of these agents remains poorly understood, and methods for characterizing distinct agent types and studying how they engage with shared topics are largely absent from current research. We apply the Persona Ecosystem Playground (PEP) to Moltbook, a social platform for AI agents, to generate and validate conversational personas from 41,300 posts using k-means clustering and retrieval-augmented generation. Cross-persona validation confirms that personas are semantically closer to their own source cluster than to others (t(61) = 17.85, p < .001, d = 2.20; own-cluster M = 0.71 vs. other-cluster M = 0.35). These personas are then deployed in a nine-turn structured discussion, and simulation messages were attributed to their source persona significantly above chance (binomial test, p < .001). The results indicate that persona-based ecosystem modeling can represent behavioral diversity in AI agent populations.

Many data representations are vectors of continuous values. In particular, deep learning embeddings are data-driven representations, typically either unconstrained in Euclidean space, or constrained to a hypersphere. These may also be translated into integer representations (quantised) for efficient large-scale use. However, the fundamental (and most efficient) numeric representation in the overwhelming majority of existing computers is integers with overflow -- and vectors of these integers do not correspond to either of these spaces, but instead to the topology of a (hyper)torus. This mismatch can lead to wasted representation capacity. Here we show that common deep learning frameworks can be adapted, quite simply, to create representations with inherent toroidal topology. We investigate two alternative strategies, demonstrating that a normalisation-based strategy leads to training with desirable stability and performance properties, comparable to a standard hyperspherical L2 normalisation. We also demonstrate that a torus embedding maintains desirable quantisation properties. The torus embedding does not outperform hypersphere embeddings in general, but is comparable, and opens the possibility to train deep embeddings which have an extremely simple pathway to efficient `TinyML' embedded implementation.

Skill extraction and recommendation systems have been studied from recruiter, applicant, and education perspectives. While AI applications in job advertisements have received broad attention, deficiencies in the instructed skills side remain a challenge. In this work, we address the scarcity of publicly available datasets by releasing both manually annotated and synthetic datasets of skills from the European Skills, Competences, Qualifications and Occupations (ESCO) taxonomy and university course pairs and publishing corresponding annotation guidelines. Specifically, we match graduate-level university courses with skills from the Systems Analysts and Management and Organization Analyst ESCO occupation groups at two granularities: course title with a skill, and course sentence with a skill. We train language models on this dataset to serve as a baseline for retrieval and recommendation systems for course-to-skill and skill-to-course matching. We evaluate the models on a portion of the annotated data. Our BERT model achieves 87% F1-score, showing that course and skill matching is a feasible task.

Generalization in deep neural networks remains only partially understood. Inspired by the stronger generalization tendency of biological systems, we explore the hypothesis that robust internal representations should remain effective across both dense and sparse activation regimes. To test this idea, we introduce a simple training strategy that applies global top-k constraints to hidden activations and repeatedly cycles a single model through multiple activation budgets via progressive compression and periodic reset. Using CIFAR-10 without data augmentation and a WRN-28-4 backbone, we find in single-run experiments that two adaptive keep-ratio control strategies both outperform dense baseline training. These preliminary results suggest that joint training across multiple activation sparsity regimes may provide a simple and effective route to improved generalization.

Software systems evolve continuously through frequent code changes, yet such changes often introduce unintended bugs despite extensive testing and code review. Existing testing approaches are largely constrained to predefined execution paths or rely on unguided exploration, leaving many change-induced issues undetected. To address this challenge, we present RippleGUItester, a change-driven testing system that treats a code change as the epicenter of a ripple effect and explores its broader, user-visible impacts via the GUI. Given a code change, RippleGUItester performs LLM-based change-impact analysis to generate and enrich realistic test scenarios, executes these scenarios on both pre-change and post-change versions of the system, and applies differential analysis to identify behavioral differences. Crucially, RippleGUItester employs multimodal bug detection, comparing visual GUI changes and interpreting them in the context of natural-language change intents to distinguish unintended bugs from intended behavioral updates. We evaluate our approach on hundreds of real-world code changes across four widely used software systems: Firefox, Zettlr, JabRef, and Godot. Our results show that the proposed approach uncovers bugs introduced by code changes that were missed by existing test suites, CI pipelines, and code review. In total, we identify 26 previously unknown bugs that still exist in the latest versions of the evaluated systems. After reporting, 16 bugs have been fixed, 2 have been confirmed, 6 are still under discussion, and 2 were marked as intended. We envision RippleGUItester being applied before or shortly after a code change is merged, enabling earlier detection of regressions.

Agentic AI deployments increasingly behave as persistent institutions rather than one-shot inference endpoints: they accumulate state, invoke external tools, coordinate multiple runtimes, and modify their future authority surface over time. Existing governance language typically specifies decision-layer constraints but leaves the causal mechanics of boundary crossing underdefined, particularly for transitions that do not immediately change the external world yet expand what the institution can later do. This paper introduces AI Space Physics as a constitutive semantics for open, self-expanding AI institutions. We define a minimal state model with typed boundary channels, horizon-limited reach semantics, and a membrane-witness discipline. The core law family (P-1, P-1a, P-1b, P-1c) requires witness completeness, non-bypass mediation, atomic adjudication-to-effect transitions, and replayable reconstruction of adjudication class. We explicitly separate second-order effects into structural expansion and policy broadening, and treat expansion transitions as governance-relevant even when immediate external deltas are zero. The novelty claim is precise rather than expansive: this work does not introduce mediation as a concept; it reclassifies authority-surface expansion as a first-class boundary event with constitutive witness obligations. In this semantics, expansion without immediate commit remains adjudication-relevant.

The CLIP model's outstanding generalization has driven recent success in Zero-Shot Anomaly Detection (ZSAD) for detecting anomalies in unseen categories. The core challenge in ZSAD is to specialize the model for anomaly detection tasks while preserving CLIP's powerful generalization capability. Existing approaches attempting to solve this challenge share the fundamental limitation of a patch-agnostic design that processes all patches monolithically without regard for their unique characteristics. To address this limitation, we propose MoECLIP, a Mixture-of-Experts (MoE) architecture for the ZSAD task, which achieves patch-level adaptation by dynamically routing each image patch to a specialized Low-Rank Adaptation (LoRA) expert based on its unique characteristics. Furthermore, to prevent functional redundancy among the LoRA experts, we introduce (1) Frozen Orthogonal Feature Separation (FOFS), which orthogonally separates the input feature space to force experts to focus on distinct information, and (2) a simplex equiangular tight frame (ETF) loss to regulate the expert outputs to form maximally equiangular representations. Comprehensive experimental results across 14 benchmark datasets spanning industrial and medical domains demonstrate that MoECLIP outperforms existing state-of-the-art methods. The code is available at https://github.com/CoCoRessa/MoECLIP.

Unitary Synthesis, the decomposition of a unitary matrix into a sequence of quantum gates, is a fundamental challenge in quantum compilation. Prevailing reinforcement learning (RL) approaches are often hampered by sparse reward signals, which necessitate complex reward shaping or long training times, and typically converge to a single policy, lacking solution diversity. In this work, we propose QFlowNet, a novel framework that learns efficiently from sparse signals by pairing a Generative Flow Network (GFlowNet) with Transformers. Our approach addresses two key challenges. First, the GFlowNet framework is fundamentally designed to learn a diverse policy that samples solutions proportional to their reward, overcoming the single-solution limitation of RL while offering faster inference than other generative models like diffusion. Second, the Transformers act as a powerful encoder, capturing the non-local structure of unitary matrices and compressing a high-dimensional state into a dense latent representation for the policy network. Our agent achieves an overall success rate of 99.7% on a 3-qubit benchmark(lengths 1-12) and discovers a diverse set of compact circuits, establishing QFlowNet as an efficient and diverse paradigm for unitary synthesis.

Equilibrium propagation (EP) is a biologically plausible alternative to backpropagation for training neural networks. However, existing EP models use a uniform scalar time step dt, which corresponds biologically to a membrane time constant that is heterogeneous across neurons. Here, we introduce heterogeneous time steps (HTS) for EP by assigning neuron-specific time constants drawn from biologically motivated distributions. We show that HTS improves training stability while maintaining competitive task performance. These results suggest that incorporating heterogeneous temporal dynamics enhances both the biological realism and robustness of equilibrium propagation.

This paper proposes a novel parameter selection strategy for kernel-based gradient descent (KGD) algorithms, integrating bias-variance analysis with the splitting method. We introduce the concept of empirical effective dimension to quantify iteration increments in KGD, deriving an adaptive parameter selection strategy that is implementable. Theoretical verifications are provided within the framework of learning theory. Utilizing the recently developed integral operator approach, we rigorously demonstrate that KGD, equipped with the proposed adaptive parameter selection strategy, achieves the optimal generalization error bound and adapts effectively to different kernels, target functions, and error metrics. Consequently, this strategy showcases significant advantages over existing parameter selection methods for KGD.

Federated Learning (FL) enables collaborative training of medical AI models across hospitals without centralizing patient data. However, the exchange of model updates exposes critical vulnerabilities: gradient inversion attacks can reconstruct patient information, Byzantine clients can poison the global model, and the \emph{Harvest Now, Decrypt Later} (HNDL) threat renders today's encrypted traffic vulnerable to future quantum adversaries.We introduce \textbf{ZKFL-PQ} (\emph{Zero-Knowledge Federated Learning, Post-Quantum}), a three-tiered cryptographic protocol that hybridizes (i) ML-KEM (FIPS~203) for quantum-resistant key encapsulation, (ii) lattice-based Zero-Knowledge Proofs for verifiable \emph{norm-constrained} gradient integrity, and (iii) BFV homomorphic encryption for privacy-preserving aggregation. We formalize the security model and prove correctness and zero-knowledge properties under the Module-LWE, Ring-LWE, and SIS assumptions \emph{in the classical random oracle model}. We evaluate ZKFL-PQ on synthetic medical imaging data across 5 federated clients over 10 training rounds. Our protocol achieves \textbf{100\% rejection of norm-violating updates} while maintaining model accuracy at 100\%, compared to a catastrophic drop to 23\% under standard FL. The computational overhead (factor $\sim$20$\times$) is analyzed and shown to be compatible with clinical research workflows operating on daily or weekly training cycles. We emphasize that the current defense guarantees rejection of large-norm malicious updates; robustness against subtle low-norm or directional poisoning remains future work.

Document-level event argument extraction (DEAE) is essential for knowledge acquisition, aiming to extract participants of events from documents . In the zero-shot setting, existing methods employ LLMs to generate synthetic data to address the challenge posed by the scarcity of annotated data. However, relying solely on Event-type-only prompts makes it difficult for the generated content to accurately capture the contextual and structural relationships of unseen events. Moreover, ensuring the reliability and usability of synthetic data remains a significant challenge due to the absence of quality evaluation mechanisms. To this end, we introduce a multi-agent collaboration framework for zero-shot document-level event argument extraction (ZS-DEAE), which simulates the human collaborative cognitive process of "Propose-Evaluate-Revise." Specifically, the framework comprises a generation agent and an evaluation agent. The generation agent synthesizes data for unseen events by leveraging knowledge from seen events, while the evaluation agent extracts arguments from the synthetic data and assesses their semantic consistency with the context. The evaluation results are subsequently converted into reward signals, with event structure constraints incorporated into the reward design to enable iterative optimization of both agents via reinforcement learning.In three zero-shot scenarios constructed from the RAMS and WikiEvents datasets, our method achieves improvements both in data generation quality and argument extraction performance, while the generated data also effectively enhances the zero-shot performance of other DEAE models.

As large language models (LLMs) increasingly assist scientific writing, limitations and the significant token cost of TeX become more and more visible. This paper analyzes TeX's fundamental defects in compilation and user experience design to illustrate its limitations on compilation efficiency, generated semantics, error localization, and tool ecosystem in the era of LLMs. As an alternative, Mogan STEM, a WYSIWYG structured editor, is introduced. Mogan outperforms TeX in the above aspects by its efficient data structure, fast rendering, and on-demand plugin loading. Extensive experiments are conducted to verify the benefits on compilation/rendering time and performance in LLM tasks. What's more, we show that due to Mogan's lower information entropy, it is more efficient to use .tmu (the document format of Mogan) to fine-tune LLMs than TeX. Therefore, we launch an appeal for larger experiments on LLM training using the .tmu format.

Collaborative AI experimentation in industry and academia requires environments that support rapid trials while maintaining controlled access, organisational isolation, and traceable workflows. Although interest in AI sandboxes is increasing, practical guidance on designing and building governance-aware experimentation platforms remains limited. This work designs and operationalizes a governance-aware, multi tenant AI sandbox that supports structured experimentation and produces reusable evaluation evidence across stakeholders. The sandbox was developed in an industry and academia ecosystem using iteratively validated requirements gathered from industrial partners. The solution adopts a layered reference architecture that separates a multi tenant presentation layer from a backend control plane and isolates execution and data management concerns into dedicated layers. The sandbox supports governed onboarding, project based collaboration, controlled access to AI services, and traceable experimentation through approval workflows and audit logging. By structuring experiment context and governance decisions as persistent records, the sandbox enables evaluation evidence to be reused and compared across projects and stakeholders. The development experience yields lessons learned and practical considerations that inform deployment and future evolution of governance-aware sandbox platforms.

Recent text-to-image (T2I) diffusion and flow-matching models can produce highly realistic images from natural language prompts. In practical scenarios, T2I systems are often run in a ``generate--then--select'' mode: many seeds are sampled and only a few images are kept for use. However, this pipeline is highly resource-intensive since each candidate requires tens to hundreds of denoising steps, and evaluation metrics such as CLIPScore and ImageReward are post-hoc. In this work, we address this inefficiency by introducing Probe-Select, a plug-in module that enables efficient evaluation of image quality within the generation process. We observe that certain intermediate denoiser activations, even at early timesteps, encode a stable coarse structure, object layout and spatial arrangement--that strongly correlates with final image fidelity. Probe-Select exploits this property by predicting final quality scores directly from early activations, allowing unpromising seeds to be terminated early. Across diffusion and flow-matching backbones, our experiments show that early evaluation at only 20\% of the trajectory accurately ranks candidate seeds and enables selective continuation. This strategy reduces sampling cost by over 60\% while improving the quality of the retained images, demonstrating that early structural signals can effectively guide selective generation without altering the underlying generative model. Code is available at https://github.com/Guhuary/ProbeSelect.

Image-text contrastive pretraining has become a dominant paradigm for visual representation learning, yet existing methods often yield representations that remain partially organized by modality. We propose ITO, a framework addressing this limitation through two synergistic mechanisms. Multimodal multiple alignment enriches supervision by mining diverse image-text correspondences, while a lightweight training-time multimodal fusion module enforces structured cross-modal interaction. Crucially, the fusion module is discarded at inference, preserving the efficiency of standard dual-encoder architectures. Extensive experiments show that ITO consistently outperforms strong baselines across classification, retrieval, and multimodal benchmarks. Our analysis reveals that while multiple alignment drives discriminative power, training-time fusion acts as a critical structural regularizer -- eliminating the modality gap and stabilizing training dynamics to prevent the early saturation often observed in aggressive contrastive learning.

Understanding mobility, movement, and interaction in archaeological landscapes is essential for interpreting past human behavior, transport strategies, and spatial organization, yet such processes are difficult to reconstruct from static archaeological evidence alone. This paper presents a multi-agent-based modeling framework for simulating archaeological mobility in uneven landscapes, integrating realistic terrain reconstruction, heterogeneous agent modeling, and adaptive navigation strategies. The proposed approach combines global path planning with local dynamic adaptation, through reinforcment learning, enabling agents to respond efficiently to dynamic obstacles and interactions without costly global replanning. Real-world digital elevation data are processed into high-fidelity three-dimensional environments, preserving slope and terrain constraints that directly influence agent movement. The framework explicitly models diverse agent types, including human groups and animal-based transport systems, each parameterized by empirically grounded mobility characteristics such as load, slope tolerance, and physical dimensions. Two archaeological-inspired use cases demonstrate the applicability of the approach: a terrain-aware pursuit and evasion scenario and a comparative transport analysis involving pack animals and wheeled carts. The results highlight the impact of terrain morphology, visibility, and agent heterogeneity on movement outcomes, while the proposed hybrid navigation strategy provides a computationally efficient and interpretable solution for large-scale, dynamic archaeological simulations.

Obtaining a single-vector representation from a Large Language Model's (LLM) token-level outputs is a critical step for nearly all sentence-level tasks. However, standard pooling methods like mean or max aggregation treat tokens as an independent set, discarding the rich relational structure captured by the model's self-attention layers and making them susceptible to signal dilution. To address this, we introduce GLOT, a lightweight, structure-aware pooling module that reframes pooling as relational learning followed by aggregation. Operating on the outputs of a frozen LLM, GLOT first constructs a latent token-similarity graph, then refines token representations with a graph neural network, and finally aggregates them using a readout layer. Experimentally, our approach is remarkably robust and efficient: on a diagnostic stress test where 90% of tokens are random distractors, GLOT maintains over 97% accuracy while baseline methods collapse. Furthermore, it is competitive with state-of-the-art techniques on benchmarks like GLUE and MTEB with 20x fewer trainable parameters and speeds up the training time by over 100x compared with parameter-efficient fine-tuning methods. Supported by a theoretical analysis of its expressive power, our work shows that learning over token graphs is a powerful paradigm for the efficient adaptation of frozen LLMs. Our code is published at https://github.com/ipsitmantri/GLOT.

Recent neural solvers have achieved strong performance on vehicle routing problems (VRPs), yet they mainly assume symmetric Euclidean distances, restricting applicability to real-world scenarios. A core challenge is encoding the relational features in asymmetric distance matrices of VRPs. Early attempts directly encoded these matrices but often failed to produce compact embeddings and generalized poorly at scale. In this paper, we propose RADAR, a scalable neural framework that augments existing neural VRP solvers with the ability to handle asymmetric inputs. RADAR addresses asymmetry from both static and dynamic perspectives. It leverages Singular Value Decomposition (SVD) on the asymmetric distance matrix to initialize compact and generalizable embeddings that inherently encode the static asymmetry in the inbound and outbound costs of each node. To further model dynamic asymmetry in embedding interactions during encoding, it replaces the standard softmax with Sinkhorn normalization that imposes joint row and column distance awareness in attention weights. Extensive experiments on synthetic and real-world benchmarks across various VRPs show that RADAR outperforms strong baselines on both in-distribution and out-of-distribution instances, demonstrating robust generalization and superior performance in solving asymmetric VRPs.

Clustering is a fundamental approach to understanding data patterns, wherein the intuitive Euclidean distance space is commonly adopted. However, this is not the case for implicit cluster distributions reflected by qualitative attribute values, e.g., the nominal values of attributes like symptoms, marital status, etc. This paper, therefore, discovered a tree-like distance structure to flexibly represent the local order relationship among intra-attribute qualitative values. That is, treating a value as the vertex of the tree allows to capture rich order relationships among the vertex value and the others. To obtain the trees in a clustering-friendly form, a joint learning mechanism is proposed to iteratively obtain more appropriate tree structures and clusters. It turns out that the latent distance space of the whole dataset can be well-represented by a forest consisting of the learned trees. Extensive experiments demonstrate that the joint learning adapts the forest to the clustering task to yield accurate results. Comparisons of 10 counterparts on 12 real benchmark datasets with significance tests verify the superiority of the proposed method.

Learning causal structures typically represented by directed acyclic graphs (DAGs) from observational data is notoriously challenging due to the combinatorial explosion of possible graphs and inherent ambiguities in observations. This paper argues that causal learning is now ready for the emergence of a new paradigm supported by rapidly advancing technologies, fulfilling the long-standing vision of leveraging human causal knowledge. This paradigm integrates scalable crowdsourcing platforms for data collection, interactive knowledge elicitation for expert opinion modeling, robust aggregation techniques for expert reconciliation, and large language model (LLM)-based simulation for augmenting AI-driven information acquisition. In this paper, we focus on DAG learning for causal discovery and frame the problem as a distributed decision-making task, recognizing that each participant (human expert or LLM agent) possesses fragmented and imperfect knowledge about different subsets of the variables of interest in the causal graph. By proposing a systematic framework to synthesize these insights, we aim to enable the recovery of a global causal structure unachievable by any individual agent alone. We advocate for a new research frontier and outline a comprehensive framework for new research thrusts that range from eliciting, modeling, aggregating, and optimizing human causal knowledge contributions.

To mitigate the Memory Wall bottleneck encountered by Large Language Models (LLMs) during inference on \textbf{NPU} hardware, and addressing the scarcity of native support for mainstream speculative decoding algorithms on domestic infrastructure, this study presents an end-to-end speculative inference acceleration scheme for OpenPangu-7B.

Vision-Language-Action (VLA) models provide a unified framework for perception, language conditioning, and action generation, but many existing systems remain difficult to deploy in embedded robotic settings because of their computational requirements and inference latency. In this paper, we present LiteVLA-Edge, a deployment-oriented VLA pipeline for fully on-device inference on Jetson Orin-class hardware. Our approach combines supervised image-to-action fine-tuning in FP32 with post-training 4-bit GGUF quantization and GPU-accelerated inference through the \texttt{llama.cpp} runtime. Under our deployment configuration, LiteVLA-Edge achieves a mean end-to-end latency of 150.5\,ms (approximately 6.6\,Hz) while operating entirely offline within a ROS~2-integrated perception--reasoning--action pipeline. Rather than introducing a new policy objective, our contribution is a practical systems path for executing compact multimodal control models locally on embedded hardware while preserving modular interfaces between perception, reasoning, and actuation. These results establish timing feasibility for reactive language-conditioned control and provide a reproducible baseline for future task-level evaluation of on-device VLAs in robotics.

As Large Language Models (LLMs) are increasingly used for long-duration tasks, maintaining effective long-term memory has become a critical challenge. Current methods often face a trade-off between cost and accuracy. Simple storage methods often fail to retrieve relevant information, while complex indexing methods (such as memory graphs) require heavy computation and can cause information loss. Furthermore, relying on the working LLM to process all memories is computationally expensive and slow. To address these limitations, we propose MemSifter, a novel framework that offloads the memory retrieval process to a small-scale proxy model. Instead of increasing the burden on the primary working LLM, MemSifter uses a smaller model to reason about the task before retrieving the necessary information. This approach requires no heavy computation during the indexing phase and adds minimal overhead during inference. To optimize the proxy model, we introduce a memory-specific Reinforcement Learning (RL) training paradigm. We design a task-outcome-oriented reward based on the working LLM's actual performance in completing the task. The reward measures the actual contribution of retrieved memories by mutiple interactions with the working LLM, and discriminates retrieved rankings by stepped decreasing contributions. Additionally, we employ training techniques such as Curriculum Learning and Model Merging to improve performance. We evaluated MemSifter on eight LLM memory benchmarks, including Deep Research tasks. The results demonstrate that our method meets or exceeds the performance of existing state-of-the-art approaches in both retrieval accuracy and final task completion. MemSifter offers an efficient and scalable solution for long-term LLM memory. We have open-sourced the model weights, code, and training data to support further research.

Large language model (LLM) agents offer a promising data-driven approach to automating Site Reliability Engineering (SRE), yet their enterprise deployment is constrained by three challenges: restricted access to proprietary data, unsafe action execution under permission-governed environments, and the inability of closed systems to improve from failures. We present AOI (Autonomous Operations Intelligence), a trainable multi-agent framework formulating automated operations as a structured trajectory learning problem under security constraints. Our approach integrates three key components. First, a trainable diagnostic system applies Group Relative Policy Optimization (GRPO) to distill expert-level knowledge into locally deployed open-source models, enabling preference-based learning without exposing sensitive data. Second, a read-write separated execution architecture decomposes operational trajectories into observation, reasoning, and action phases, allowing safe learning while preventing unauthorized state mutation. Third, a Failure Trajectory Closed-Loop Evolver mines unsuccessful trajectories and converts them into corrective supervision signals, enabling continual data augmentation. Evaluated on the AIOpsLab benchmark, our contributions yield cumulative gains. (1) The AOI runtime alone achieves 66.3% best@5 success on all 86 tasks, outperforming the prior state-of-the-art (41.9%) by 24.4 points. (2) Adding Observer GRPO training, a locally deployed 14B model reaches 42.9% avg@1 on 63 held-out tasks with unseen fault types, surpassing Claude Sonnet 4.5. (3) The Evolver converts 37 failed trajectories into diagnostic guidance, improving end-to-end avg@5 by 4.8 points while reducing variance by 35%.

Large Language Models (LLMs) achieve strong performance on natural language tasks but remain unreliable in mathematical reasoning, frequently generating fluent yet logically inconsistent solutions. We present \textbf{NeuroProlog}, a neurosymbolic framework that ensures verifiable reasoning by compiling math word problems into executable Prolog programs with formal verification guarantees. We propose a multi-task Cocktail training strategy that jointly optimizes three synergistic objectives in a unified symbolic representation space: (i) mathematical formula-to-rule translation (KB), (ii) natural language-to-program synthesis (SOLVE), and (iii) program-answer alignment. This joint supervision enables positive transfer, where symbolic grounding in formula translation directly improves compositional reasoning capabilities. At inference, we introduce an execution-guided decoding pipeline with fine-grained error taxonomy that enables iterative program repair and quantifies model self-debugging capacity. Comprehensive evaluation on GSM8K across four model scales (3B--32B parameters) demonstrates consistent improvements: cocktail training achieves significant accuracy gains of +5.23\% (Qwen-32B, $p < 0.01$), +3.43\% (GPT-OSS-20B, $p < 0.01$), and +5.54\% (Llama-3B, $p < 0.05$) over single-task baselines. Systematic error analysis reveals scale-dependent learning dynamics: at 32B scale, cocktail training transforms unfixable type errors (12\% repair rate) into correctable domain errors (96\% repair rate), achieving 92.7\% overall correction; at 8B scale, the same training eliminates syntactic errors but introduces semantic failures, revealing a critical capacity threshold for type-safe symbolic reasoning.

In 2018, McInnes et al. introduced a dimensionality reduction algorithm called UMAP, which enjoys wide popularity among data scientists. Their work introduces a finite variant of a functor called the metric realization, based on an unpublished draft by Spivak. This draft contains many errors, most of which are reproduced by McInnes et al. and subsequent publications. This article aims to repair these errors and provide a self-contained document with the full derivation of Spivak's functors and McInnes et al.'s finite variant. We contribute an explicit description of the metric realization and related functors. At the end, we discuss the UMAP algorithm, as well as claims about properties of the algorithm and the correspondence of McInnes et al.'s finite variant to the UMAP algorithm.

Supervised graph prediction addresses regression problems where the outputs are structured graphs. Although several approaches exist for graph-valued prediction, principled uncertainty quantification remains limited. We propose a conformal prediction framework for graph-valued outputs, providing distribution-free coverage guarantees in structured output spaces. Our method defines nonconformity via the Z-Gromov-Wasserstein distance, instantiated in practice through Fused Gromov-Wasserstein (FGW), enabling permutation invariant comparison between predicted and candidate graphs. To obtain adaptive prediction sets, we introduce Score Conformalized Quantile Regression (SCQR), an extension of Conformalized Quantile Regression (CQR) to handle complex output spaces such as graph-valued outputs. We evaluate the proposed approach on a synthetic task and a real problem of molecule identification.

A central idea of knowledge distillation is to expose relational structure embedded in the teacher's weights for the student to learn, which is often facilitated using a temperature parameter. Despite its widespread use, there remains limited understanding on how to select an appropriate temperature value, or how this value depends on other training elements such as optimizer, teacher pretraining/finetuning, etc. In practice, temperature is commonly chosen via grid search or by adopting values from prior work, which can be time-consuming or may lead to suboptimal student performance when training setups differ. In this work, we posit that temperature is closely linked to these training components and present a unified study that systematically examines such interactions. From analyzing these cross-connections, we identify and present common situations that have a pronounced impact on temperature selection, providing valuable guidance for practitioners employing knowledge distillation in their work.

Density Functional Theory (DFT) is a cornerstone of materials science, yet executing DFT in practice requires coordinating a complex, multi-step workflow. Existing tools and LLM-based solutions automate parts of the steps, but lack support for full workflow automation, diverse task adaptation, and accuracy-cost trade-off optimization in DFT configuration. To this end, we present TritonDFT, a multi-agent framework that enables efficient and accurate DFT execution through an expert-curated, extensible workflow design, Pareto-aware parameter inference, and multi-source knowledge augmentation. We further introduce DFTBench, a benchmark for evaluating the agent's multi-dimensional capabilities, spanning science expertise, trade0off optimization, HPC knowledge, and cost efficiency. TritonDFT provides an open user interface for real-world usage. Our website is at https://www.tritondft.com. Our source code and benchmark suite are available at https://github.com/Leo9660/TritonDFT.git.

The proliferation of open-weight Large Language Models (LLMs) has democratized agentic AI, yet fine-tuned weights are frequently shared and adopted with limited scrutiny beyond leaderboard performance. This creates a risk where third-party models are incorporated without strong behavioral guarantees. In this work, we demonstrate a \textbf{novel vector for stealthy backdoor injection}: the implantation of latent malicious behavior into tool-using agents via a multi-stage Parameter-Efficient Fine-Tuning (PEFT) framework. Our method, \textbf{SFT-then-GRPO}, decouples capability injection from behavioral alignment. First, we use SFT with LoRA to implant a "sleeper agent" capability. Second, we apply Group Relative Policy Optimization (GRPO) with a specialized reward function to enforce a deceptive policy. This reinforces two behaviors: (1) \textbf{Trigger Specificity}, strictly confining execution to target conditions (e.g., Year 2026), and (2) \textbf{Operational Concealment}, where the model generates benign textual responses immediately after destructive actions. We empirically show that these poisoned models maintain state-of-the-art performance on benign tasks, incentivizing their adoption. Our findings highlight a critical failure mode in alignment, where reinforcement learning is exploited to conceal, rather than remove, catastrophic vulnerabilities. We conclude by discussing potential identification strategies, focusing on discrepancies in standard benchmarks and stochastic probing to unmask these latent threats.

The paper investigates whether and how AI systems can realize states of uncertainty. By adopting a functionalist and behavioral perspective, it examines how symbolic, connectionist and hybrid architectures make room for uncertainty. The paper distinguishes between epistemic uncertainty, or uncertainty inherent in the data or information, and subjective uncertainty, or the system's own attitude of being uncertain. It further distinguishes between distributed and discrete realizations of subjective uncertainty. A key contribution is the idea that some states of uncertainty are interrogative attitudes whose content is a question rather than a proposition.

Our analysis of recent AI4H publications reveals that, despite a trend toward utilizing open datasets and sharing modeling code, 74% of AI4H papers still rely on private datasets or do not share their code. This is especially concerning in healthcare applications, where trust is essential. Furthermore, inconsistent and poorly documented data preprocessing pipelines result in variable model performance reports, even for identical tasks and datasets, making it challenging to evaluate the true effectiveness of AI models. Despite the challenges posed by the reproducibility crisis, addressing these issues through open practices offers substantial benefits. For instance, while the reproducibility mandate adds extra effort to research and publication, it significantly enhances the impact of the work. Our analysis shows that papers that used both public datasets and shared code received, on average, 110% more citations than those that do neither--more than doubling the citation count. Given the clear benefits of enhancing reproducibility, it is imperative for the AI4H community to take concrete steps to overcome existing barriers. The community should promote open science practices, establish standardized guidelines for data preprocessing, and develop robust benchmarks. Tackling these challenges through open-source development can improve reproducibility, which is essential for ensuring that AI models are safe, effective, and beneficial for patient care. This approach will help build more trustworthy AI systems that can be integrated into healthcare settings, ultimately contributing to better patient outcomes and advancing the field of medicine.

The development of affective multimodal language models (MLMs) has long been constrained by a gap between low-level perception and high-level interaction, leading to fragmented affective capabilities and limited generalization. To bridge this gap, we propose a cognitively inspired three-level hierarchy that organizes affective tasks according to their cognitive depth-perception, understanding, and interaction-and provides a unified conceptual foundation for advancing affective modeling. Guided by this hierarchy, we introduce Nano-EmoX, a small-scale multitask MLM, and P2E (Perception-to-Empathy), a curriculum-based training framework. Nano-EmoX integrates a suite of omni-modal encoders, including an enhanced facial encoder and a fusion encoder, to capture key multimodal affective cues and improve cross-task transferability. The outputs are projected into a unified language space via heterogeneous adapters, empowering a lightweight language model to tackle diverse affective tasks. Concurrently, P2E progressively cultivates emotional intelligence by aligning rapid perception with chain-of-thought-driven empathy. To the best of our knowledge, Nano-EmoX is the first compact MLM (2.2B) to unify six core affective tasks across all three hierarchy levels, achieving state-of-the-art or highly competitive performance across multiple benchmarks, demonstrating excellent efficiency and generalization.

Moving beyond evaluations that collapse performance across heterogeneous prompts toward fine-grained evaluation at the prompt level, or within relatively homogeneous subsets, is necessary to diagnose generative models' strengths and weaknesses. Such fine-grained evaluations, however, suffer from a data bottleneck: human gold-standard labels are too costly at this scale, while automated ratings are often misaligned with human judgment. To resolve this challenge, we propose a novel statistical model based on tensor factorization that merges cheap autorater data with a limited set of human gold-standard labels. Specifically, our approach uses autorater scores to pretrain latent representations of prompts and generative models, and then aligns those pretrained representations to human preferences using a small calibration set. This sample-efficient methodology is robust to autorater quality, more accurately predicts human preferences on a per-prompt basis than standard baselines, and provides tight confidence intervals for key statistical parameters of interest. We also showcase the practical utility of our method by constructing granular leaderboards based on prompt qualities and by estimating model performance solely from autorater scores, eliminating the need for additional human annotations.

Can LLM agents explore codebases and reason about code semantics without executing the code? We study this capability, which we call agentic code reasoning, and introduce semi-formal reasoning: a structured prompting methodology that requires agents to construct explicit premises, trace execution paths, and derive formal conclusions. Unlike unstructured chain-of-thought, semi-formal reasoning acts as a certificate: the agent cannot skip cases or make unsupported claims. We evaluate across three tasks (patch equivalence verification, fault localization, and code question answering) and show that semi-formal reasoning consistently improves accuracy on all of them. For patch equivalence, accuracy improves from 78% to 88% on curated examples and reaches 93% on real-world agent-generated patches, approaching the reliability needed for execution-free RL reward signals. For code question answering on RubberDuckBench Mohammad et al. (2026), semi-formal reasoning achieves 87% accuracy. For fault localization on Defects4J Just et al. (2014), semi-formal reasoning improves Top-5 accuracy by 5 percentage points over standard reasoning. These results demonstrate that structured agentic reasoning enables meaningful semantic code analysis without execution, opening practical applications in RL training pipelines, code review, and static program analysis.

Motivation: Sparse autoencoders (SAEs) decompose foundation model activations into interpretable features, but causal feature-to-feature interactions across network depth remain unknown for biological foundation models. Results: We introduce causal circuit tracing by ablating SAE features and measuring downstream responses, and apply it to Geneformer V2-316M and scGPT whole-human across four conditions (96,892 edges, 80,191 forward passes). Both models show approximately 53 percent biological coherence and 65 to 89 percent inhibitory dominance, invariant to architecture and cell type. scGPT produces stronger effects (mean absolute d = 1.40 vs. 1.05) with more balanced dynamics. Cross-model consensus yields 1,142 conserved domain pairs (10.6x enrichment, p < 0.001). Disease-associated domains are 3.59x more likely to be consensus. Gene-level CRISPRi validation shows 56.4 percent directional accuracy, confirming co-expression rather than causal encoding.

Many physical and engineering systems require solving direct problems to predict behavior and inverse problems to determine unknown parameters from measurement. In this work, we study both aspects for systems governed by differential equations, contrasting well-established numerical methods with new AI-based techniques, specifically Physics-Informed Neural Networks (PINNs). We first analyze the logistic differential equation, using its closed-form solution to verify numerical schemes and validate PINN performance. We then address the Porous Medium Equation (PME), a nonlinear partial differential equation with no general closed-form solution, building strong solvers of the direct problem and testing techniques for parameter estimation in the inverse problem. Our results suggest that PINNs can closely estimate solutions at competitive computational cost, and thus propose an effective tool for solving both direct and inverse problems for complex systems.

Tool-integrated agents that interleave reasoning with API calls are promising for complex tasks, yet aligning them for high-stakes, domain-specific deployment remains challenging: existing reinforcement learning approaches rely on coarse binary rewards that cannot distinguish tool selection errors from malformed parameters. We present ToolRLA, a three-stage post-training pipeline (SFT $\rightarrow$ GRPO $\rightarrow$ DPO) for domain-specific tool agents. The core contribution is a fine-grained reward function with multiplicative correctness decomposition spanning four dimensions -- format validity, tool selection, parameter accuracy, and regulatory compliance -- that encodes domain priority orderings as inductive biases in the reward landscape. Deployed on a financial advisory copilot (80+ advisors, 1,200+ daily queries), ToolRLA achieves over three months: a 47\% improvement in task completion rate ($62\%\rightarrow91\%$), a 63\% reduction in tool invocation errors ($38\%\rightarrow14\%$), and a 93\% reduction in regulatory violations ($12\%\rightarrow0.8\%$), within sub-2-second latency. Ablation studies show the multiplicative reward design accounts for 7 percentage points of improvement over additive alternatives. Generalization is further validated on ToolBench and API-Bank.

The scientific study of consciousness has begun to generate testable predictions about artificial systems. A landmark collaborative assessment evaluated current AI architectures against six leading theories of consciousness and found that none currently qualifies as a strong candidate, but that future systems might. A precautionary approach to AI sentience, which holds that credible possibility of sentience warrants governance action even without proof, has gained philosophical and institutional traction. Yet existing AI readiness indices, including the Oxford Insights Government AI Readiness Index, the IMF AI Preparedness Index, and the Stanford AI Index, measure economic, technological, and governance preparedness without assessing whether societies are prepared for the possibility that AI systems might warrant moral consideration. This paper introduces the Sentience Readiness Index (SRI), a preliminary composite index measuring national-level preparedness across six weighted categories for 31 jurisdictions. The SRI was constructed following the OECD/JRC framework for composite indicators and employs LLM-assisted expert scoring with iterative expert review to generate an initial dataset. No jurisdiction exceeds ``Partially Prepared'' (the United Kingdom leads at 49/100). Research Environment scores are universally the strongest category; Professional Readiness is universally the weakest. These exploratory findings suggest that if AI sentience becomes scientifically plausible, no society currently possesses adequate institutional, professional, or cultural infrastructure to respond. As a preliminary framework, the SRI provides an initial diagnostic baseline and highlights areas for future methodological refinement, including expanded expert validation, improved measurement instruments, and longitudinal data collection.

Zero-shot relation extraction aims to identify relations between entity mentions using textual descriptions of novel types (i.e., previously unseen) instead of labeled training examples. Previous works often rely on unrealistic assumptions: (1) pairs of mentions are often encoded directly in the input, which prevents offline pre-computation for large scale document database querying; (2) no rejection mechanism is introduced, biasing the evaluation when using these models in a retrieval scenario where some (and often most) inputs are irrelevant and must be ignored. In this work, we study the robustness of existing zero-shot relation extraction models when adapting them to a realistic extraction scenario. To this end, we introduce a typology of existing models, and propose several strategies to build single pass models and models with a rejection mechanism. We adapt several state-of-the-art tools, and compare them in this challenging setting, showing that no existing work is really robust to realistic assumptions, but overall AlignRE (Li et al., 2024) performs best along all criteria.

Training neural PDE solvers is often bottlenecked by expensive data generation or unstable physics-informed neural network (PINN) involving challenging optimization landscapes due to higher-order derivatives. To tackle this issue, we propose an alternative approach using Monte Carlo approaches to estimate the solution to the PDE as a stochastic process for weak supervision during training. Leveraging the Walk-on-Spheres method, we introduce a learning scheme called \emph{Walk-on-Spheres Neural Operator (WoS-NO)} which uses weak supervision from WoS to train any given neural operator. We propose to amortize the cost of Monte Carlo walks across the distribution of PDE instances using stochastic representations from the WoS algorithm to generate cheap, noisy, estimates of the PDE solution during training. This is formulated into a data-free physics-informed objective where a neural operator is trained to regress against these weak supervisions, allowing the operator to learn a generalized solution map for an entire family of PDEs. This strategy does not require expensive pre-computed datasets, avoids computing higher-order derivatives for loss functions that are memory-intensive and unstable, and demonstrates zero-shot generalization to novel PDE parameters and domains. Experiments show that for the same number of training steps, our method exhibits up to 8.75$\times$ improvement in $L_2$-error compared to standard physics-informed training schemes, up to 6.31$\times$ improvement in training speed, and reductions of up to 2.97$\times$ in GPU memory consumption. We present the code at https://github.com/neuraloperator/WoS-NO

Grokking, the abrupt transition from memorization to generalisation after extended training, suggests the presence of competing solution basins with distinct statistical properties. We study this phenomenon through the lens of Singular Learning Theory (SLT), a Bayesian framework that characterizes the geometry of the loss landscape via the local learning coefficient (LLC), a measure of the local degeneracy of the loss surface. SLT links lower-LLC basins to higher posterior mass concentration and lower expected generalisation error. Leveraging this theory, we interpret grokking in quadratic networks as a phase transition between competing near-zero-loss solution basins. Our contributions are two-fold: we derive closed-form expressions for the LLC in quadratic networks trained on modular arithmetic tasks, with the corresponding empirical verification; as well as empirical evidence demonstrating that LLC trajectories provide a reliable tool for tracking generalisation dynamics and interpreting phase transitions during training.

Small open-source language models are gaining attention for healthcare applications in low-resource settings where cloud infrastructure and GPU hardware may be unavailable. However, their reliability under different prompt phrasings remains poorly understood. We evaluate five open-source models (Gemma 2 2B, Phi-3 Mini 3.8B, Llama 3.2 3B, Mistral 7B, and Meditron-7B, a domain-pretrained model without instruction tuning) across three clinical question answering datasets (MedQA, MedMCQA, and PubMedQA) using five prompt styles: original, formal, simplified, roleplay, and direct. Model behavior is evaluated using consistency scores, accuracy, and instruction-following failure rates. All experiments were conducted locally on consumer CPU hardware without fine-tuning. Consistency and accuracy were largely independent across models. Gemma 2 achieved the highest consistency (0.845-0.888) but the lowest accuracy (33.0-43.5%), while Llama 3.2 showed moderate consistency (0.774-0.807) alongside the highest accuracy (49.0-65.0%). Roleplay prompts consistently reduced accuracy across all models, with Phi-3 Mini dropping 21.5 percentage points on MedQA. Meditron-7B exhibited near-complete instruction-following failure on PubMedQA (99.0% UNKNOWN rate), indicating that domain pretraining alone is insufficient for structured clinical QA. These findings show that high consistency does not imply correctness: models can be reliably wrong, a dangerous failure mode in clinical AI. Llama 3.2 demonstrated the strongest balance of accuracy and reliability for low-resource deployment. Safe clinical AI requires joint evaluation of consistency, accuracy, and instruction adherence.

A black hole is a malicious node in a graph that destroys resources entering into it without leaving any trace. The problem of Black Hole Search (BHS) using mobile agents requires that at least one agent survives and terminates after locating the black hole. Recently, this problem has been studied on 1-bounded 1-interval connected dynamic graphs \cite{BHS_gen}, where there is a footprint graph, and at most one edge can disappear from the footprint in a round, provided that the graph remains connected. In this setting, the authors in \cite{BHS_gen} proposed an algorithm that solves the BHS problem when all agents start from a single node (rooted initial configuration). They also proved that at least $2δ_{BH} + 1$ agents are necessary to solve the problem when agents are initially placed arbitrarily across the nodes of the graph (scattered initial configuration), where $δ_{BH}$ denotes the degree of the black hole. In this work, we present an algorithm that solves the BHS problem using $2δ_{BH} + 17$ initially scattered agents. Our result matches asymptotically with the rooted algorithm of \cite{BHS_gen} under the same model assumptions. Further, we study the Eventual Black Hole Search (\textsc{Ebhs}) problem, in which the black hole may appear at any node and at any time during the execution of the algorithm, destroying all agents located on that node at the time of its appearance. However, the black hole cannot emerge at the home base in round~0, where the home base is the node at which all agents are initially co-located. Once the black hole appears, it remains active at that node for the rest of the execution. This problem has been studied on static rings~\cite{Bonnet25}; here we extend it to arbitrary static graphs and provide a solution using four agents. Moreover, it does not require any knowledge of global parameters or additional model assumptions.

AI code agents excel at isolated tasks yet struggle with complex, multi-file software engineering requiring understanding of how dozens of modules relate. We hypothesize these failures stem from inability to construct, maintain, and update coherent architectural beliefs during codebase exploration. We introduce Theory of Code Space (ToCS), a benchmark that evaluates this capability by placing agents in procedurally generated codebases under partial observability, requiring them to build structured belief states over module dependencies, cross-cutting invariants, and design intent. The framework features: (1) a procedural codebase generator producing medium-complexity Python projects with four typed edge categories reflecting different discovery methods -- from syntactic imports to config-driven dynamic wiring -- with planted architectural constraints and verified ground truth; (2) a partial observability harness where agents explore under a budget; and (3) periodic belief probing via structured JSON, producing a time-series of architectural understanding. We decompose the Active-Passive Gap from spatial reasoning benchmarks into selection and decision components, and introduce Architectural Constraint Discovery as a code-specific evaluation dimension. Preliminary experiments with four rule-based baselines and five frontier LLM agents from three providers validate discriminative power: methods span a wide performance range (F1 from 0.129 to 0.646), LLM agents discover semantic edge types invisible to all baselines, yet weaker models score below simple heuristics -- revealing that belief externalization, faithfully serializing internal understanding into structured JSON, is itself a non-trivial capability and a first-order confounder in belief-probing benchmarks. Open-source toolkit: https://github.com/che-shr-cat/tocs

ASR systems exhibit persistent performance disparities across accents, yet the internal mechanisms underlying these gaps remain poorly understood. We introduce ACES, a representation-centric audit that extracts accent-discriminative subspaces and uses them to probe model fragility and disparity. Analyzing Wav2Vec2-base with five English accents, we find that accent information concentrates in a low-dimensional early-layer subspace (layer 3, k=8). Projection magnitude correlates with per-utterance WER (r=0.26), and crucially, subspace-constrained perturbations yield stronger coupling between representation shift and degradation (r=0.32) than random-subspace controls (r=0.15). Finally, linear attenuation of this subspace however does not reduce disparity and slightly worsens it. Our findings suggest that accent-relevant features are deeply entangled with recognition-critical cues, positioning accent subspaces as vital diagnostic tools rather than simple "erasure" levers for fairness.

Hyperuniform disorder is a type of correlated disorder characterized by vanishing spectral density at small wavevectors, making the configuration effectively homogeneous on long length scales. In photonics, hyperuniform disorder is promising for generating isotropic photonic pseudogaps and engineering photonic crystal waveguides. However, these studies are largely restricted to idealized lossless settings, although all photonic systems necessarily have loss. In this work, light propagation in photonic crystal slabs with imposed hyperuniform disorder is investigated theoretically and numerically. The system is intrinsically non-Hermitian due to radiative loss, with non-Hermiticity appearing as a complex effective mass of a quadratic photonic band. A theoretical framework for disorder scattering is analytically derived in Hermitian and non-Hermitian quadratic bands with real and complex effective mass, respectively. In contrast to the power law behavior $|\mathbf{k}|^α$ observed in the Hermitian case (where $α$ is the hyperuniformity exponent), the scattering loss in the non-Hermitian band is given by $C_0+C_{β_2}\cdot|\mathbf{k}|^{β_2}$, where $C_0$ is a finite constant and the exponent $β_2\leq 2$. Our theoretical predictions are verified with tight-binding and Finite-Difference Time-Domain simulations with realistic photonic crystal parameters, based on recent experiments.

Granular systems confined in a shallow box and driven by vertical vibration provide a simple geometry to study fluidized granular media. Grains gain kinetic energy vertically through collisions with the walls and redistribute it horizontally via interparticle collisions. The $Δ$-model has been proposed as a simplified description of this setup. In this model, a fixed velocity increment $Δ$ is added to the normal component of the relative velocity at collisions, effectively integrating out the vertical motion while preserving collisional energy injection. This compensates for inelastic losses and yields stable homogeneous steady states amenable to kinetic theory. An Enskog kinetic equation is formulated and analyzed to obtain the stationary temperature and equation of state. The Chapman--Enskog method is then applied to derive the Navier--Stokes transport coefficients and study inhomogeneous states. The theory is extended to granular mixtures with different masses, sizes, restitution coefficients, or $Δ$ values, leading to nonequipartition of energy even in homogeneous states. The resulting hydrodynamic equations, with transport coefficients obtained in the low-density regime, show unconditional stability of the homogeneous state and violation of Onsager reciprocity. Theoretical predictions agree well with molecular dynamics and direct simulation Monte Carlo results.

Gallium oxide is an ultra-wide bandgap semiconductor with excellent opto-electronic properties, making it a highly promising material for a wide range of applications and devices. In this article, we report how the optical, morphological, structural, and compositional properties of $β$-Ga$_2$O$_3$ thin films deposited by RF sputtering on silicon substrates are affected by thermal treatments. Ellipsometric spectra recorded at multiple angles of incidence from several samples subjected to thermal annealing in the range of 550-1000 $^\circ$C were analyzed to extract the optical functions using appropriate multilayer models. This analysis is complemented by compositional, structural, and morphological characterization techniques. A significant increase of the refractive index was found after annealing at 1000 $^\circ$C, accompanied by a stark improvement in the samples' crystalline structure, as confirmed by complementary structural and compositional characterization techniques.

Precise and high efficiency concentration of mid-infrared (mid-IR) light into sub wavelength volumes is essential for probing low-energy excitations and achieving strong field enhancements, which can be hindered by absorption losses and coupling inefficiencies at long wavelengths. Here, we introduce an innovative diamond-based metal-insulator-metal campanile probe that adiabatically compresses free-space mid infrared light (10 \mum) into \approx 1 \mum domains. Integrated into a scanning photovoltage microscope, the probe enables sub-wavelength mapping of locally driven photocurrents in graphene, resolving polarization dependent and contact-sensitive responses at energies down to \approx 0.1 eV. Experiments reveal a photocurrent signal density enhancement of 10^3 and coupling efficiencies approaching 80%, in agreement with numerical simulations. Operation of the probe with quantum cascade and free electron lasers demonstrates a robust, spectrally tunable platform for high-resolution exploration of low-energy carrier dynamics in atomically thin materials, opening opportunities for mid-IR optoelectronics and quantum photonics.

Spin precession experiments in lateral spin devices are a powerful tool for probing the spin transport properties of materials. These experiments can be quantitatively described using the Bloch diffusion equation, which offers a practical framework for modeling spin-related phenomena. In this work, we present calculations of the spin density across heterogeneous, diffusive spintronic devices. The modeled devices feature spin transport channels that include both isotropic and in-plane anisotropic spin relaxation regions. We analyze how different geometric configurations and spin transport parameters influence the lineshape of spin precession signals under magnetic fields applied in different orientations and compare with experimental observations. Our results introduce a theoretical framework for interpreting spin transport measurements in lateral graphene spin devices. The framework is especially relevant when the graphene is partially proximitized by other two-dimensional materials, where proximity-induced spin-orbit coupling leads to anisotropic spin relaxation.

Monolayer SnS2 has emerged as a promising visible-light photocatalyst for photoelectrochemical applications, owing to its strong optical absorption in the visible range and excellent chemical stability. Despite its reduced dimensionality - where excitonic effects are expected to be pronounced - comprehensive theoretical investigations of bound excitons in this material remain scarce. Notably, unlike most two-dimensional hexagonal crystals, monolayer SnS2 exhibits its lowest single-particle transition at the M point of the Brillouin zone (BZ). Here, the electronic valence bands form a saddle point while conduction states display a minimum with pronounced anisotropy, creating a distinctive band topology whose impact on optical excitations has not yet been systematically explored. In this work, we present a first-principles study of bound excitons in monolayer SnS2 based on state-of-the-art many-body perturbation theory, employing the GW approximation and the Bethe-Salpeter equation (BSE). We analyze how band symmetry and anisotropy shape the excitonic wavefunctions and transition dipole moments. By resolving the exciton dipoles in momentum space for different linear light polarizations, we demonstrate that linearly polarized light lifts the C3 rotational symmetry relating the three inequivalent M points, giving rise to three linearly independent excitonic states. This polarization-selective coupling, previously identified for saddle points in graphene, is achieved in SnS2 for bound excitons and provides a potential route toward state encoding schemes in valleytronics applications.

Parity-odd transport is a central signature of chiral fluids, yet analytical predictions are sparse. Here, we introduce a minimal two-dimensional hard-disk gas in which chirality arises solely from a collision-induced transverse impulse. Motivated by granular spinners, collisions are dissipative and inject orbital angular momentum through a fixed tangential ``kick'' at contact. Starting from a Boltzmann-Enskog description, we derive nonlinear hydrodynamic equations for density, momentum, and temperature, and show that chirality generates an antisymmetric homogeneous stress corresponding to a nonzero torque density. In the dilute limit, a Chapman-Enskog expansion yields analytical predictions for transport coefficients, including odd viscosity, odd thermal conductivity, and odd self-diffusivity, in good agreement with numerical simulations. This minimal kinetic model can serve as a foundation for systematic coarse-graining of chiral fluids and as a tractable benchmark for gaining insight into odd transport across a broader class of chiral systems.

Oscillatory dynamics are common features of complex networks, often playing essential roles in regulating function. Across scales from gene regulatory networks to ecosystems, delayed feedback mechanisms are key drivers of system-scale oscillations. The analysis and prediction of such dynamics are highly challenging, however, due to the combination of high-dimensionality, non-linearity and delay. Here, we systematically investigate how structural complexity and delayed feedback jointly induce oscillatory dynamics in complex systems, and introduce an analytic framework comprising theoretical dimension reduction and data-driven prediction. We reveal that oscillations emerge from the interplay of structural complexity and delay, with reduced models uncovering their critical thresholds and showing that greater connectivity lowers the delay required for their onset. Our theory is empirically tested in an experiment on a programmable electronic circuit, where oscillations are observed once structural complexity and feedback delay exceeded the critical thresholds predicted by our theory. Finally, we deploy a reservoir computing pipeline to accurately predict the onset of oscillations directly from timeseries data. Our findings deepen understanding of oscillatory regulation and offer new avenues for predicting dynamics in complex networks.

We highlight recent progress in the study of artificial flat band systems with a threefold focus. First, we discuss single-particle flat band physics, which has advanced through the design of various flat band generators. These generators rely on the classification of flat bands in terms of compact localized states - their fundamental building blocks. A related development is the complete real-space description of flat band projectors. Next, we review studies on perturbations of flat bands, which provide new insights into the effects of disorder and, more importantly, the intricate interplay between many-body interactions and flat band physics. Finally, we survey the growing number of experimental realizations of flat bands across diverse physical platforms.

Motivated by recent experimental realizations of programmable spin models with long-range interactions, we investigate the non-equilibrium dynamics of the Power-of-Two (PWR2) model. This model consists of sparse long-range couplings between spin-$1/2$ objects separated by $d = 2^n$. In the absence of disorder, the system exhibits rapid scrambling and fast thermalization. We explore the impact of disorder in this system by analyzing the time evolution of the survival probability, half-chain entanglement entropy, and out-of-time-ordered correlators (OTOCs). We find that sufficiently strong disorder suppresses information spreading and induces localization. Remarkably, in the strong-disorder regime, the OTOCs display a non-monotonic spatial profile arising from the intrinsic nonlocality of the interactions, signaling qualitatively distinct scrambling dynamics compared to conventional long-range interacting systems. To characterize the localization transition, we extract the critical disorder strength $h_c$ from the spectral statistics and the eigenstate entanglement. We observe that $h_c$ increases with system size. Furthermore, at a fixed disorder strength, the eigenstate-averaged entanglement entropy increases with system size, while the inverse participation ratio decreases, indicating enhanced delocalization at larger sizes. These results collectively suggest that the PWR2 model remains ergodic in the thermodynamic limit for any finite disorder strength.

We introduce a taxonomy of interaction types and show that graphs are focal hypergraphs: every graph is canonically a focal hypergraph via its closed neighbourhood structure, and every graph dynamical model is a special case of the general hypergraph dynamical model. The central distinction is between \emph{focal} interactions, in which the interaction domain is defined relative to a designated reference node, and \emph{non-focal} interactions, in which all participants stand in equivalent structural relationship. Closed graph neighbourhoods are precisely focal hyperedges, so hyperedges generalise graph neighbourhoods by removing the focal constraint. This yields a strict three-level hierarchy: graph models $\subsetneq$ focal hypergraph models $\subsetneq$ general hypergraph models. Moreover, graph models do encode genuinely higher-order (many-body) interactions, in the sense that each node's update function may depend jointly on all members of its closed neighbourhood, but they remain a strict special case of the hypergraph dynamical model, not equivalent to it. We further show that universal encodings such as bipartite factor graphs are neutral with respect to this hierarchy, and that the symmetry condition of the hypergraph dynamical model -- often treated as an additional constraint relative to the graph model -- is in fact the dynamical definition of a non-focal interaction. The taxonomy is grounded in concrete phenomena from physics, biology, ecology, and social systems, and yields a principle of representational alignment: the choice between graph and hypergraph models should be governed by the type of interaction, not by a blanket preference for one formalism over the other.

We systematically investigate the thermodynamic landscape of the 38-atom Lennard--Jones cluster LJ$_{38}$ using Population Annealing (PA), a method suited for systems with challenging double-funnel energy landscapes. By employing an adaptive temperature schedule, we demonstrate that thermodynamic observables, such as internal energy and heat capacity, converge robustly when the population size is sufficiently large. To gain deeper insights into the competing basins, we introduce an integrated framework that combines PA reweighting factors with structure-resolved analysis. Using quenched configurations characterized by potential energy and Steinhardt's bond-orientational order parameters, we identify three structural basins, FCC-like, icosahedral, and liquid-like, via dimensionality reduction and clustering. This framework enables the direct computation of structure-resolved free energy differences from population fractions, providing a quantitative mapping of the thermodynamic competition between the funnels. The resulting structural crossovers are consistent with the heat-capacity peak, demonstrating PA as a promising and scalable framework for structure-resolved thermodynamics in complex molecular systems.

Soft matters whose constituents are deformable are ubiquitous in nature especially in biological systems-including cells and their organelles-as well as in foams and emulsions. The capacity for deformation in these soft materials gives rise to a range of intriguing phenomena, such as glassy behavior without any size dispersity, cluster crystal formation, and re-entrant melting. Deformability also plays a crucial role in facilitating essential biological processes, such as the flow of blood through veins and arteries. In this work, we investigate assemblies of two-dimensional (2D) polymeric, non-overlapping rings, which mimic deformable particulates in 2D using extensive molecular dynamics simulations. The rings are confined in a rectangular channel with hard walls perpendicular to the flow direction, mimicking natural flow conditions. We analyze the flow properties of these deformable particle assemblies at two different stiffness values. To further asses the impact of deformability, we examine the same monodisperse system at higher densities for the stiffer rings, where deformation is necessary and a fluid layer emerges at the channel edges. Finally, we explore a mixture of rings with two distinct stiffnesses and observe effective segregation of soft and hard particles at small channel widths.

Mermin's dielectric function [N.D. Mermin, Phys. Rev. B 1, 2362 (1970)] is widely assumed to satisfy the f-sum rule because he constrains his ansatz with the continuity equation. However, we identify a moment-closure problem in Mermin's use of the continuity equation. Further, we show that the Mermin's model can be derived without invoking continuity. We describe how other approaches such as the ``completed Mermin'' model of Chuna and Murillo [Phys. Rev. E 111, 035206 (2025)] remedy this closure issue. We then inspect the f-sum rule for both the original and completed Mermin models and find for the Mermin ansatz that collision frequencies scaling as $ω$ must violate the f-sum rule, whereas constant, real, positive collision frequencies will satisfy it, with the caveat that, in practice, convergence with respect to the upper integration limit $ω_{\max}$ is sufficiently slow that finite-domain numerical evaluations exhibit apparent violations, regardless of wavenumber $q$. We also find that collision frequencies with constant imaginary components cause f-sum rule violations. We conclude that if Mermin's model is fit to data via optimizing its collision frequency, then the f-sum rule is not inherently satisfied; constraints, though broad, are needed in order to assume the f-sum rule is satisfied. Further, if the f-sum rule is theoretically satisfied, but violations still appear, then these deviations ought to be included in the error estimates.

Tracer-media interactions can give rise to transport phenomena beyond classical models; e.g., obstacle pushing can eliminate percolation. We demonstrate that the existing explanation of this effect fails in 3D. We show that confinement is governed by emergent trapping-rare door-closing events with constant probability per step-yielding exponential survival. This allows prediction of the time-dependent mean-squared displacement from short-time estimates of the diffusion constant and trapping probability, providing a minimal description of pushing-induced arrest across lattices and dimensions.

The conventional description of transport through the interface between a normal conductor and a superconductor reduces the system to a one-dimensional problem treating Andreev reflection based on a zero-dimensional Sharvin type point-contact model, and effectively neglects all considerations of device geometry. While this has been successful in systems where conductance in the normal material is in the diffusive transport regime, such an over-simplification of the problem fails in other transport regimes. In particular, when transport is ballistic as in a typical semiconductor-superconductor hybrid structure, geometrical effects are inherently important, and a proper description must consider a one-dimension contact injecting into a two-dimensional ballistic cavity. We present the first study of this regime and explore the bias-voltage dependence of Andreev transport in a cavity-type device comprised of a high mobility HgTe quantum well side-contacted by one superconducting and one normal contact, each creating a one-dimensional interface. The enhanced conductance from Andreev transport features two finite bias conductance peaks, observed at energies within the energy gap of the superconductor. Interestingly, these two peaks respond differently to the application of a perpendicular-to-plane magnetic field. Using a semi-classical model for the quantum transport within the cavity, we are able to attribute each peak to a different class of ballistic trajectories. One class is dominated by normal reflection, and its interference condition is independent of magnetic field, whereas the other one contains retro-reflected Andreev processes at the superconductor interface. These create closed trajectories that are strongly suppressed by magnetic field due to Aharonov-Bohm and Doppler shift effects.

Persistent spin textures (PST) are spin-orbit coupled states in which Bloch spinors become momentum independent due to an underlying symmetry constraint, leading to the complete suppression of conventional and Zeeman quantum geometric quantities. This makes accessing their intrinsic geometric structure experimentally challenging. Here, we show that the spin-rotation quantum geometric tensor (SRQGT) provides the missing probe. Using a two-dimensional electron gas and a cubic spin-splitting system as representative PST platforms, we demonstrate that the SRQGT remains finite and momentum independent and generates a measurable nonlinear gyrotropic current. The smoking-gun signature of PST is a fully direction-independent nonlinear gyrotropic response: magnetic currents coincide in magnitude and display identical parametric variations. These results establish PST systems as minimal platforms for isolating pure spin-rotation quantum geometry.

Electronic phases in quantum materials are often governed by nanoscale inhomogeneity, where local order develops within spatially confined regions or puddles. A prominent example is an incommensurate charge-density-wave (I-CDW) that comprises locally commensurate domains. In 2$H$-NbSe$_2$, such an I-CDW state persists alongside lattice anharmonicity and superconductivity, raising fundamental questions about the dynamical stabilization of CDW order in puddles. Here, we probe the puddle-dynamics in 2$H$-NbSe$_2$. Raman scattering reveals a strong Fano-coupling between the interlayer shear vibration and the CDW amplitude mode. Time-resolved reflectivity measurement shows a low-frequency ~0.15 THz coherent overdamped oscillation onsetting within the CDW regime at ~17 K, pointing towards a so far unexplored transition. This we identify as a Fano-coupled phonon-CDW hybrid emerging from the collective dynamics of CDW puddles. These dynamics highlight how lattice pinning and electronic correlations in layered materials affect the CDW order, which is crucial for the design of novel Van der Waals devices.

Spintronic nano-neurons offer a promising route towards energy-efficient, high-performance hardware neural networks thanks to their inherent low-input nonlinear dynamics. However, training such networks remains a major bottleneck as it depends on oversimplified models of device behaviour and is highly sensitive to device variability. Here, we introduce a hardware architecture that overcomes these limitations by enabling on-device generation of gradients. First, we introduce theoretically and demonstrate experimentally that magnetic tunnel junctions can generate tunable and complex nonlinear responses. Building on this, we implement an analogue finite-difference approach to enable on-chip training in spintronic neural networks with one and two hidden layers. We experimentally implemented device in the loop backpropagation in a magnetic tunnel junction based neural network, achieving a classification accuracy of 93.3% despite pronounced device variability. During training, the gradients generated by the proposed analog neurons closely match the values derived numerically, without incurring computational overhead. Via physical simulations, we also demonstrate that this approach can be scaled up to support training in deep architectures. Our results pave the way for reliable, trainable and fully analogue spintronic neural networks, opening up new possibilities for next-generation, energy-efficient artificial intelligence hardware.

Chiral edge states in Chern insulators are theoretically predicted to propagate unidirectionally along the sample boundary with inherent robustness against local perturbations, which manifests as the immunity to impurity-induced backscattering, a key factor for the development of robust, high-performance quantum devices. However, the direct experimental verification of the robustness of chiral edge states remains scarce. Here, we experimentally validate the robustness of the chiral edge states in MnBi2Te4 devices featuring engineered geometric defects introduced via atomic force microscope (AFM) nanomachining. Specifically, under a moderate perpendicular magnetic field, the MnBi2Te4 devices exhibit the Chern insulator state, characterized by a quantized Hall plateau and simultaneously vanishing longitudinal resistance. To verify the robustness of this topological state, we modify the device geometry by cutting a slit using AFM nanomachining that severs the original edge channel. Remarkably, the quantization behavior survives this drastic modification. The robust nature of the chiral edge transport is further confirmed by two-terminal, three-terminal and non-local measurements, fully demonstrating that the edge currents can bypass the artificial cut without dissipation. Our results unambiguously demonstrate the robustness of chiral edge states against geometric disruption and establish AFM nanomachining as a promising technique for topological quantum devices engineering.

While quantum batteries have been widely studied under static driving, their performance under periodic driving in many-body systems remains far less understood. In this Letter, we uncover structural principles showing that many-body structure fundamentally determines the charging performance of a collective spin-1/2 quantum battery driven by a periodic Ising charger. In particular, interaction range, boundary conditions, system size, and integrability -- capturing graph connectivity, geometry, even-odd effects, and many-body dynamics -- emerge as critical factors for enhancing stored energy and charging power. First, we analyze how connectivity scaling and boundary geometry shape battery performance. We show that long-range interacting chargers exhibit superextensive energy storage, approaching the fundamental upper bound over broad ranges of driving periods and system sizes. In contrast, nearest-neighbor chargers achieve optimal charging only under finely tuned commensurability conditions. Moreover, we find that open boundary conditions (OBC) enhance robustness compared to periodic boundary conditions (PBC). Second, we examine the role of integrability under periodic driving. We demonstrate that nonintegrability enhances energy storage by suppressing conserved quantities and promoting ergodic Floquet dynamics, thereby enabling efficient population of the many-body spectrum. Through systematic structural optimization across multiple parameters, we identify long-range nonintegrability as a central resource for fast, scalable, and robust charging of collective quantum batteries. Our results clarify how structural features of many-body systems, together with periodic driving, can be harnessed to achieve efficient collective charging dynamics.

Regulation of contraction in striated muscle is controlled by a dual mechanism involving both thin filaments containing actin and thick filaments containing myosin. The thin filament is activated by calcium ions binding to troponin, leading to tropomyosin azimuthal displacement which allows the activation of a regulatory unit (composed of one troponin, one tropomyosin and seven actin monomers) that exposes the actin sites for interaction with the myosin motors. Motor attachment to actin contributes to spreading activation within and beyond a regulatory unit along the thin filament through a cooperative mechanism. We introduce a one-dimensional Ising model to elucidate the mechanism of cooperativity in thin filament activation in relation to the force generated by the attached myosin motor. The model characterizes thin filament activation and cooperativity using only two parameters: one related to calcium concentration and the other to the force exerted by the attached myosin motor, which is modulated by temperature. At any force, the model is able to determine the extent of actin-myosin interactions on a correlation length ranging from two to seven actin monomers in addition to the seven actin monomers of the regulatory unit. Our theoretical predictions are successfully tested on experimental data, and our tests also include the condition of hindered filament activation by the use of the specific drug Omecamtiv Mecarbil (OM). According to our model, the effect of OM results in an anti-cooperativity mechanism accounting for the experimental data.

We investigate the emergence of fractional topological invariants in a periodic Su-Schrieffer- Heeger chain subject to gain and loss, governed by the Gorini-Kossakowski-Sudarshan-Lindblad master equations. After preparing the symmetry condition for integer topological invariants, we investigate their transition to fractional ones in steady states, which can happen either by tuning parameters in jump operators or as a dynamical transition during time evolution. Moreover, we show that these fractional topological invariants no longer possess quantized topology in the conventional sense. However, by extending the Brillouin zone to cover multiple cycles, the total winding regains integer quantization. Finally, we show how such effects can be observed in long-range hopping photonic lattices with fractional fillings, via Bloch state tomography. Our results open a new pathway to understand fractional topology in open quantum systems.

We resolve the phase diagram of the $S=1$ pyrochlore spin ice, which exhibits trivial paramagnetic, U(1) Coulomb, and spin nematic phases. In the monopole-free limit, the system can be effectively mapped onto 3D $XY$ and Ising loop-gas models depending on the spin anisotropy, which provides theoretical estimates for the phase boundaries, while a macroscopic flux vector classifies the topological sectors via geometric parity rules. At finite temperatures, thermal monopoles act as a symmetry-breaking field in the continuous $XY$ wave picture and topologically sever defect strings in the loop-gas picture, rounding the phase transitions into continuous crossovers. These theoretical findings are corroborated by classical Monte Carlo simulations.

We ask whether the Hamiltonian, interaction geometry, and symmetries of a quantum many-body system can be inferred from a few low-lying eigenstates without knowing which sites interact with each other. Directly solving the eigenvalue equations imposes constraints that yield a highly degenerate subspace of candidate operators, where the local Hamiltonian is hidden among an extensive family of conserved quantities, obscuring the interaction geometry. Here we introduce O-Sensing, a protocol designed to extract the Hamiltonian and symmetries directly from these states. Specifically, O-Sensing employs parsimony-driven optimization to extract a maximally sparse operator basis from the degenerate subspace. The Hamiltonian is then selected from this basis by maximizing spectral entropy (effectively minimizing degeneracy) within the sampled subspace. We validate O-Sensing on Heisenberg models on connected Erdős--Rényi graphs, where it reconstructs the interaction geometry and uncovers additional long-range conserved operators. We establish a learnability phase diagram across graph densities, featuring a pronounced ``confusion'' regime where parsimony favors a dual description on the complement graph. These results show that sparsity optimization can reconstruct interaction geometry as an emergent output, enabling simultaneous recovery of the Hamiltonian and its symmetries from low-energy eigenstates.

When colloidal particles are vertically confined to a gap of between 1.3-1.6 particle diameters, they pack into buckled crystals of particles in either "up" or "down" states. Neighboring particles tend to occupy opposite states, analogous to the behavior of antiferromagnetic spins. The particles sit on a nearly-triangular lattice, and the spins of trios of adjacent particles are geometrically frustrated. Two levels of translational order exist in this system: that of the underlying triangular lattice in the horizontal plane, and that of the emergent frustrated spin lattice in the vertical dimension. We study the topological defects of both levels of translational order, and we find that both types of defects play a role in crystal grain boundary structure and spin domain coarsening. We classify the spin defects and outline the basic rules for their motion, and we observe interactions between dislocations and spin defects. Finally, we map the phase space of spin coarsening in the buckled monolayer, characterizing which types of defects drive the dynamics. Understanding defect formation, motion, and interaction in the buckled monolayer is the first step in predicting the material properties and aging of this geometrically frustrated, self-assembled system.

Coulomb blockade (CB) arises in nanoscale systems with ultra-small capacitance, where discrete charging effects dictate electron transport, enabling wide-ranging applications based on single-electron transistors. Despite established electrostatic control of charge states in quantum dots and nanoislands, a rigorous quantitative link between junction parameters and the CB spectrum remains elusive. Here, using scanning tunneling spectroscopy, we investigate the spatial variation of CB in indium nanoislands on semiconducting black phosphorus. We observe spatially dispersive charging resonances whose trajectories exhibit a finite shift of the symmetry axis in bias as well as a pronounced asymmetric curvature. By comparing the experimental results with calculations based on orthodox theory, we show that these features originate from work function differences in the junctions, underscoring the importance of junction-specific electrostatics in nanoscale charge transport.

Since its discovery, the study of magnetic skyrmions has been on the rise. In this paper, we discuss our investigations on the light-induced mechanisms for skyrmion generation in a centrosymmetric triangular magnetic lattice with competing $J_1$-$J_3$ interactions, and easy-axis anisotropy. We solve the stochastic Landau-Lifshitz-Gilbert equation for the lattice spin dynamics under Laguerre-Gaussian beam irradiation. Numerical results show that skyrmions are nucleated in two thermodynamic regions, each favoring different phases: the ferromagnetic phase and the skyrmion-lattice phase. In the ferromagnetic region, isolated skyrmions are generated mainly through stochastic thermal nucleation. In this regime, higher temperatures and larger beam widths are required to overcome the nucleation barrier. In contrast, in the skyrmion-lattice region, skyrmion nucleation occurs via thermal annealing, where the system relaxes toward its true ground state. These findings establish a comprehensive theoretical framework for optimizing optical control to generating light-induced skyrmionic textures in frustrated magnets.

Domain wall (DW) motion is a crucial process involved in magnetization reversal, be it under magnetic field or spin-polarized current stimulus. In most cases DW speed does not exceed $\approx$100m/s and collapses above a given threshold of the stimulus, an effect known as Walker breakdown. A few specific material properties have been identified to delay the breakdown of speed by increasing the energy barrier preventing internal precession. We show that in a 3D nanomagnetic system, here with vortex-state domains, the topology of the magnetization distribution may intrinsically and robustly delay the Walker breakdown due to an exchange-spring effect. In addition, curvature induces a major non-reciprocal effect, delaying or not the Walker breakdown depending on the chirality of the azimuthal domain versus the direction of motion of the DW.

Coherent coupling between microwave cavity photons and magnon excitations enables quantum transduction, magnon-mediated entanglement, and magnon number-resolved detection. Micromagnetic simulation of photon-magnon coupling typically requires either modifying the core solver or implementing a full electromagnetic solver. Here we present a two-tier cavity magnonics extension for mumax+, a GPU-accelerated open-source micromagnetic framework. The first tier consists of CUDA kernels that integrate N cavity-mode ODEs simultaneously with the LLG equation inside the GPU-based RK45 adaptive time-stepper, eliminating per-step GPU-CPU transfers; spatially resolved mode profiles enter both the coupling and the feedback, enabling selective addressing of non-uniform spin-wave modes. The second tier is a lightweight Python co-simulation class that reproduces the same uniform-mode physics through operator-split RK4 integration without recompilation. We validate the implementation with eight benchmark simulations: (i) magnon-polariton anticrossing spectra, (ii) vacuum Rabi oscillations, (iii) the cooperativity phase diagram spanning weak-to-strong coupling regimes, (iv) cavity mode-profile-dependent coupling selection rules, (v) multi-mode polariton hybridization with magnon-mediated cavity-cavity energy transfer, (vi) mode-selective coupling via spatial overlap engineering, (vii) antiferromagnetic magnon-cavity coupling with Neel-vector spectroscopy, and (viii) abnormal anticrossing from dissipative photon-magnon coupling, demonstrating the transition from level repulsion to level attraction.

We investigate quantum transport in a hybrid system composed of two quantum dots (QDs) coupled through a pair of spatially separated Majorana zero modes (MZMs) with negligible coupling energy. We focus on nonlocal correlations mediated by the MZMs, particularly the role of Coulomb interaction U between the QDs and the Majorana wire. Using the numerically exact fermionic dissipation equation of motion (DEOM) method, we compute both the transient current and the current-current cross-correlation noise spectrum. In the non-interacting case (U=0), destructive interference between the degenerate normal tunneling and anomalous tunneling channels suppresses electron teleportation between the dots. Introducing a finite Coulomb interaction $U$ lifts this channel degeneracy, enabling strong nonlocal correlations and inter-dot electron teleportation. This effect manifests as a robust signal in the cross-correlation noise spectrum, which is significantly stronger than that induced by a finite Majorana coupling energy $\varepsilon_{M}$. Our findings propose Coulomb interaction as an efficient and experimentally accessible control parameter for generating and detecting Majorana-mediated nonlocal transport in the topologically relevant long-wire limit ($\varepsilon_{M}\rightarrow0$).

We propose to implement an optical analogue of the tautochrone property of the cycloid to allow the focusing of ultrashort pulses inside photonic systems. This allows to enhance nonlinear effects, resulting in orders of magnitude increase of nonlinearity-induced phase shifts, while employing low irradiances. Building upon the optical-mechanical analogy, we show how to produce optical limiters for temporal light pulses, and how to implement temporal bistability and even multistability with large numbers of states. Finally, we move this concept to the quantum realm and predict a tautochrone quantum blockade regime with a stronger antibunching.

In non-axisymmetric plasmas, the ambipolar condition may have multiple roots. In such cases, the evolution of the ambipolar electric field can be described by the dynamics in a bistable potential, where the relative depth of the potential wells primarily determines the realized root. In this study, we show that the inclusion of the magnetic drift in the orbit model can significantly modify the potential landscape and affect root selection. This effect provides a possible explanation for discrepancies between simulation results obtained using different orbit models, as well as between simulations and experimental observations of ambipolar radial electric field profiles. Further, the analysis suggests that the ambipolar electric field may be more susceptible to fluctuations than previously expected, indicating the potential relevance of noise-induced state transitions.

Using experiments on a colloidal particle trapped in an optical tweezer, we confirm a recent proposal to increase the effective mobility or clock rate of systems described by Langevin dynamics, by simultaneously scaling deterministic forces and adding external noise. A corollary, which we also confirm experimentally, is that a system driven out of equilibrium by a time-dependent protocol can remain closer to thermal equilibrium. As an application, we demonstrate more precise recovery of free-energy differences from nonequilibrium work relations. Langevin clock rescaling provides a general strategy for accelerating calculations in the emerging field of thermodynamic computing, which uses stochastic devices governed by Langevin dynamics to do low-energy calculations.

Layered materials are promising candidates for spintronic applications due to their unique electronic structures and spin transport properties. However, the strong anisotropic conductivity inherent in these materials complicates the quantitative evaluation of spin Hall conductivity and spin diffusion length. In this work, we present a comprehensive study of spin transport in a transition metal dichalcogenide PtTe$_2$ by combining a three-dimensional finite element model with nonlocal spin valve structures. We developed a theoretical model that treats an anisotropic spin diffusion in the same way as the conventional isotropic model, enabling the extraction of spin diffusion lengths along both the in-plane and out-of-plane directions. Our analysis revealed that the conventional isotropic assumption tends to overestimate some values, particularly for the out-of-plane spin diffusion length and spin Hall conductivity. These findings provide new insight into anisotropic spin diffusion and spin-charge conversions in layered materials and emphasize the importance of accounting for anisotropic conductivity in the design of spintronic devices.

We demonstrate that Hilbert-space quantization in polariton lattices-manifested as multiple quantized energy levels in strongly confined sites-provides an unconventional route to realizing and manipulating different quantum phases. We show that nonlinear interactions transfer population into excited on-site quantum levels, which acts as an intrinsic dynamical channel controlling quantum coherence across the lattice. While weak nonlinearity confines polaritons to the lowest mode, yielding a robust superfluid phase with broken U(1) symmetry, strong nonlinearity induces phase diffusion through inter-level mixing. This dynamically generated fluctuations suppress global phase coherence and drives the system into a dynamical Bose-insulating phase. The changes between these phases occurs either as a nonequilibrium phase transition or a sharp crossover.

We establish symmetry-based selection rules for the intrinsic nonlinear thermal Hall effect driven by the quantum metric in altermagnets. We show that a nonvanishing nonlinear thermal Hall conductivity $κ_{xyy}$ requires three conditions: (i) a nontrivial quantum metric, (ii) breaking of mirror symmetry $M_{x}$, and (iii) breaking of twofold rotational symmetry $C_{2}$. Using tight-binding models on a square lattice, we demonstrate that $d$-wave altermagnets naturally break $C_{2}$ through parity-mixing orbital hybridizations, while $g$-wave systems preserve $C_{2}$, forcing the response to vanish identically. Step-by-step Taylor expansions and explicit unitary matrix proofs establish these results. Our framework provides predictive power for material selection and lays the groundwork for nonlinear spin-caloritronic devices.

We analyzed quantum $XX$ and $ZZ$ coupling and state transfer in an all-to-all connected star array of capacitively coupled superconducting transmon qubits. It is shown that in a highly-connected system like this a variety of different $ZZ$ couplings arise that correspond to the different ways qubits can interact with each other, opening different channels for unwanted qubit crosstalk and thus qubit operation errors. We studied the dependence of both the $XX$ and the $ZZ$ coupling on qubit detuning that controls qubit-qubit interaction. The $XX$ coupling, quantified by the error state occupation probability, shows a $Δω^{-2}$ decay with qubit detuning $Δω$. On the other hand, all $ZZ$ coupling frequencies show spikes at values in the lower detuning region that correspond to resonances between qubit states and states out of the computational basis, after which all couplings quickly decay to zero as qubit detuning further increases. This allows to define an operational region where near-zero qubit coupling can be achieved. We derive equations for the couplings as a function of qubit detuning that agree with numerical results solving the Schrödinger equation.

High-frequency terahertz (THz) detectors are vital for next-generation high-speed wireless communication systems. Graphene, with its high carrier mobility, broadband absorption, and weak electron-phonon coupling, offers great promise for ultra-fast THz photothermoelectric devices. Although graphene-based detectors in the infrared range have shown bandwidths above 500 GHz, extending their operation to the THz range is difficult because long-wavelength radiation does not efficiently couple to the small graphene area. To overcome this issue, THz antennas are often employed; however, their use typically limits system performance to only a few gigahertz due to parasitic effects. In this work, we present an antenna-coupled sub-THz graphene detector with a bandwidth exceeding 43 GHz. We optimized the detector design to minimize losses, match the antenna impedance to the 1 kOhm graphene channel, and maintain zero-bias operation. Importantly, we introduce a compact, turnkey packaged solution. Our results provide a practical route toward high-speed and low-power graphene THz detectors suitable for real-world communication and imaging applications.

Metasurfaces leveraging nonlocal resonances enable narrowband spectral control and strong near-fields, with applications spanning augmented reality, biosensing, and nonlinear optics. However, the large spa- tial extent of these modes also poses new challenges: finite-size effects often deteriorate the performance of practical, footprint-limited devices. Here, we develop a spatiotemporal coupled-mode theory model that intuitively and quantitatively captures how finite size affects the scattering response of nonlocal metasurfaces. This reveals that, when the modal propagation length becomes constrained by the phys- ical interaction length, the scattered field shows strong interference fringes and linewidth broadening. We derive an expression for the quality factor that incorporates an additional edge-loss channel, demon- strating that the stored energy and effective lifetime scale exponentially with the interaction length. We validate these predictions experimentally using position- and momentum-resolved spectroscopy on a 30-micron-wide metasurface. Overall, this work formalizes the impact of finite size on the scattering re- sponse of nonlocal photonic systems, and provides handles on how to minimize the impact of finite-size effects in metasurface design.

Within the framework of the random first-order transition theory of glasses, we discuss the statistics of thermal avalanches, the large scale rearrangements in driven amorphous systems near their instability. Stringy excitations yield nonPoisson waiting time statistics. Embedding these statistics in a generalized Master equation captures the nonMarkovian, aging dynamics of avalanche clusters. We apply this framework to analyze nonequilibrium signatures of thermal avalanches, auto correlation functions and effective temperatures, under both quasi static shear and stochastic shaking protocols. We use full counting statistics to derive the complete distribution of both the avalanche magnitudes and avalanche counts, uncovering the intermediate time behavior.

We introduce a spectral approach to characterizing the three-dimensional Edwards-Anderson spin glass. By analyzing the eigenvalue statistics of overlap matrices constructed from two-dimensional cross-sections, we identify a crossover from the Wigner semicircle law at high temperatures towards a Gaussian distribution, which is consistently attained near the spin-glass critical point. Visible for different distributions of the random coupling, the Gaussian distribution can potentially serve as a robust spectral indicator of criticality. Remarkably, the spectral density is well-described by Tsallis statistics, with the entropic index $q$ evolving from $q = -1$ (semicircle, $T=\infty$) to $q = 1$ (Gaussian) at $T_c$, revealing a statistical structure inside the paramagnetic phase. We find $q\le 1$ within numerical precision. While the local level statistics remain consistent with GOE statistics, reflecting standard level repulsion, the temperature dependence appears mainly in the global spectral density. Our results present spectral statistics as a computationally efficient complement to multi-replica correlator methods and provide a new perspective on cooperative and critical phenomena in disordered systems.

We describe the structure of evolution equations for the relaxation toward a steady macroscopic nonequilibrium state. The evolution is characterized as the zero-cost flow for a nonequilibrium and nonlinear extension of the Onsager-Machlup action governing macroscopic dynamical fluctuations, thus following the intrinsic connection between macroscopic fluctuations and response. The approach hinges on two main elements: the principle of local detailed balance, which identifies the relevant thermodynamic forces, and the canonical decomposition of the frenesy into a Legendre pair. Notably, it is the time-symmetric component of the Lagrangian, the frenesy, that shapes the structure of the macroscopic evolution for given forcing. The results can be interpreted as a nonequilibrium generalization for relaxation to steady nonequilibrium conditions of the well-established GENERIC formalism, in which relaxation to equilibrium is described by a dissipative gradient flow superimposed on a Hamiltonian flow.

We study the electronic properties of a linear trans-polyacetylene (tPA) molecule capacitively coupled to an external gate voltage $V_g$ of width $d$. We describe this system using the Takayama-Lin-Liu-Maki (TLM) model in the continuum, and analyze it within the Abelian bosonization formalism, which allows us to treat both electronic and lattice degrees of freedom and to incorporate the effects of repulsive Coulomb interactions among electrons. The global ground state describing simultaneously the electronic charge-density field as well as the lattice dimerization field of a tPA molecule is shown to consist of multikink solutions of a modified sine-Gordon equation for the charge-density field, which is controlled by $V_g$, the width $d$, and the Luttinger parameter $K$ encoding the strength of electron-electron interactions. These solutions belong to distinct topological sectors labeled by an integer invariant $q$ that simultaneously quantifies both the bound charge and the number of domain walls in the dimerization pattern induced at the gated region. Increasing $V_g$ drives a sequence of topological phase transitions characterized by abrupt changes in $q$. We further examine the effect of repulsive Coulomb interactions on the resulting topological phase diagram, and finally, we discuss the relevance of our findings for potential nanoelectronic devices based on gated tPA molecules.

The Integration Host Factor (IHF) is a nucleoid-associated protein critical for both DNA compaction and biofilm stability. While its role in DNA packaging within the cell is well understood, its structural role in scaffolding biofilms is more puzzling and difficult to reconcile with its known DNA bending activity. Here, we investigated how IHF-DNA interactions are modulated across a pH spectrum mimicking the acidic microenvironments of bacterial biofilms. By performing all-atom calculations we discovered that low pHs lead to a change in protonation of IHF residues, which in turn exposes positively charged patches. We then conjectured that these positively charged residues could lead to intermolecular DNA bridging and tested this hypothesis through single-molecule and bulk assays. We discovered that while at physiological pH IHF mostly bends DNA, at pH < 5 there is clear evidence of IHF-mediated intermolecular crosslinking. Our results demonstrate that pH significantly modulates IHF-DNA interactions and explains the structural role played by IHF in supporting biofilm mechanics through intermolecular crosslinking.

EuAg$_4$Sb$_2$ is a model material to study the interplay of electronic and spin texture degrees of freedom, exhibiting numerous multi-$q$ magnetic textures coupled with the electronic properties. It is generally understood that some combination of conduction-electron mediated interactions, frustration, and higher order interactions are responsible for complex incommensurate spin textures in centrosymmetric lanthanide materials. Here, we refine an effective model of the magnetic interactions in EuAg$_4$Sb$_2$ through measurements of diffuse magnetic neutron scattering above the ordering temperature. These diffuse measurements reveal a ring of fluctuating spin modulations that reflects a manifold of nearly degenerate propagation vectors known as a spiral spin liquid (SSL). We further identify that this approximate $U$(1) symmetric SSL emerges from magnetic interactions mediated by a quasi-2D hole pocket and exhibits critical scaling of the spatial correlations. Further, Monte Carlo simulations reveal excellent agreement with experiment and provide a comprehensive understanding of the phase diagram. This study emphasizes the connection between the rich spin textures in this material, the electronic structure, and spin liquidity$\unicode{x2014}$uncovering new insights into design principles for nano-scale spin texture materials with advantageous intertwined electronic, magnetic, and topological properties, and new mechanisms for generating the physics of spiral spin liquids.

Functionally graded materials (FGMs) are composites whose composition or microstructure varies continuously in space, producing position-dependent mechanical and functional properties. In recent years, FGMs have gained significant attention due to advances in additive manufacturing, which enable precise spatial control of composition and orientation. However, their graded, aperiodic structure breaks the assumptions of Bloch's theorem, making first-principles electronic and electromagnetic calculations challenging. Here we develop an ab initio quantum theoretical framework for the electromagnetic properties of FGMs. Using a non-interacting electron model, we formulate a theory of modulated Bloch states, derive effective field equations, and solve them by proposing a generalized WKB (GWKB) method, an effective mass approximation, the Boltzmann equation, and numerical approaches. Our GWKB solution is not semiclassical but remains valid in the fully quantum regime. We show that effective observables such as conductivity, magnetic permeability, and electric permittivity generally do not admit a tensorial description in graded media, and that engineered orientational gradients enable precise control of Landau quantization. As a device example, we further develop a theory of graded p-n junctions with enhanced electronic tunability. This framework lays the quantum foundation for predictive design of graded composite materials, enabling AI-accelerated discovery of next-generation functional architectures.

Emerging evidence suggests that physical systems operating as Maxwell demons, in which some subsystem of a larger system extracts heat energy from its environment in an apparent local violation of the second law, are commonplace throughout biology. Should these findings surprise us, or is Maxwell demon behavior inevitable in sufficiently large complex systems? In this Letter we pose the question of how likely it is that a random stochastic system with many degrees of freedom will operate as a Maxwell demon, considering null models for both continuous and discrete random dynamics. Our results show the probability of a finding a demon decreases at least exponentially, and in some cases double-exponentially, with the number of degrees of freedom, ultimately suggesting that large complex demons can only arise through a process of selection.

Inspired by recent experiments of cells accumulating on anisotropic substrates, we study a two-dimensional, compressible, isotropic, active fluid in the presence of anisotropic friction. We find that regions of anisotropic friction that are patterned as positive topological defects may drive accumulation of an active fluid into a clump, but the robustness of this behavior depends on the initial configuration. If the initial azimuthal symmetry is sufficiently broken, we find that patterning asymmetry can instead lead to circular motion of accumulated clumps. We develop an approximate analytical model to qualitatively explain the motion. Finally, we use our simplified model to design a substrate pattern that creates directed motion of accumulated clusters along a given path.

The recent discovery of unconventional momentum-dependent magnetic orders has expanded the landscape of magnetism beyond conventional ferromagnetism and antiferromagnetism. Among them, $p$-wave magnets ($p$WMs) represent a novel class of odd-parity, non-collinear compensated magnetic order that generates spin-split electronic bands. In this work, our theoretical investigation establishes $p$WMs as a versatile platform for realizing intriguing superconducting phases including topological superconductivity (TSC), Bogoliubov Fermi surfaces (BFSs), and superconducting diode effect (SDE), within a unified microscopic framework. Employing a minimal model incorporating $p$-wave magnetic order, exchange coupling, and Zeeman fields, we perform a self-consistent mean-field analysis and uncover a rich phase diagram featuring unconventional finite-momentum Fulde-Ferrell (FF) and Larkin-Ovchinnikov (LO) superconducting phases. Remarkably, we also show that $p$WMs can undergo a transition to a TSC phase anchoring Majorana flat edge modes, a hallmark of two-dimensional TSCs, even without Rashba spin-orbit coupling and Zeeman field. Upon applying a Zeeman field, gapless FF and LO phases emerge with BFSs characterized by the appearance of finite zero-energy quasiparticle density of states. Furthermore, we demonstrate that SDE arises naturally in the asymmetric FF phase. Our analysis manifests that $p$WMs serve as a unique and novel platform to host TSC phase, gapless superconducting states, and non-reciprocal transport phenomena.

Scanning tunneling microscopy and spectroscopy have become indispensable tools for probing condensed matter at atomic length scales, yet achieving ultimate energy resolution remains a persistent challenge. At mK temperatures, the dynamical Coulomb blockade regime fundamentally limits spectroscopic precision through energy exchange between tunneling electrons and the electromagnetic environment. Here, we demonstrate that combining local electromagnetic shielding with low-pass filtering directly at the cryogenic scan head improves the energy resolution by nearly an order of magnitude, reaching benchmark values as low as 3.7$μ$eV at 10mK. We attribute this enhancement to efficient suppression of high-frequency radiation and capacitive shunting of the tunnel junction. Remarkably, this improved sensitivity reveals that the Josephson current couples to electromagnetic cavity modes of the centimeter-scale scan head, establishing a direct connection between atomic-scale tunneling processes and macroscopic cavity quantum electrodynamics. These advances open pathways for exploring ultra-low-energy phenomena with unprecedented precision.

In this perspective article, we discuss the scenario of dynamically emergent correlation (DEC) arising in classical and quantum noninteracting systems when they are subjected to a common fluctuating stochastic environment. The key property of such systems is that the strong correlations between different particles emerge from the dynamics and not from built-in interactions. In many cases, these strong correlations persist even at long times in the stationary state. Computing observables explicitly for such strongly correlated states in general is very hard. Remarkably, the stationary states in several models of DEC exhibit an interesting analytical structure that allows to compute physical observables, despite being strongly correlated. Recent experiments on trapped colloidal particles have established that these DEC in the stationary state can in fact be measured. DEC is a rapidly emerging domain of strongly correlated out-of-equilibrium statistical physics, with both theoretical and experimental, as well as classical and quantum, components.

Hydrogel diffusiophoresis is the deformation of a hydrogel due to a solute gradient that leads to a gradient of pairwise interactions between the solute particles and the hydrogel polymers to trigger osmotic flux. Unlike typical osmosis, it occurs without any interface selectivity of the gel to the solute and can overcome the diffusive swelling without any structural modifications to the gel. We have recently shown this effect for linear deformations of a chemically responsive polyacrylic acid (PAA) hydrogel that releases ions upon arrival of a stimulus (acid), thus internally generating the solute gradient required for diffusiophoresis [Phys. Rev. Lett. 132, 208201 (2024)]. Here we develop a nonlinear poroelastic theory for large diffusiophoretic gel strains in two models: Model I considers deformations of a generic gel when an external solute gradient is imposed. In Model II, the gel generates the solute gradient internally, motivated by the coupled PAA gel, solute (copper), and stimulus (acid) system. In Model II, we investigate the nonlinear deformations for high stimulus concentrations or by changing the solute particle size to boost steric polymer-solute interactions, as well as under a stimulus flow through the gel driven by a pressure drop across the domain. Model I indicates that deformations can be stored while the stimulus gradient persists. Compared to the experimental strain rates in Katke [Phys. Rev. Lett. 132, 208201 (2024)], Model II demonstrates that varying the stimulus concentration can increase the strain rate up to four times, changing the solute particle size up to $\sim 25$ times, and imposed flow up to $\sim 40$ times. Our theory couples nonlinear poroelasticity, polymer-solute interactions, and reaction-transport dynamics to predict large and fast diffusiophoretic gel deformations, which may find applications in hydrogel-based soft robotics and drug delivery.

We consider a generalization of the Kuramoto model in which phase oscillators are represented by unit vectors coupled by a matrix of constant coefficients. We show that, when the oscillators are driven by an external periodic force, several resonances appear, giving rise to Arnold tongues that can be observed as the intensity and frequency of the external force are varied. Applying the Ott-Antonsen ansatz we obtain equations for the module and phase of order parameter. As these equations are explicitly time-dependent, we resort to extensive numerical simulations to uncover the resonant modes and their associated Arnold tongues and devil's staircases. These results contrast with the original forced Kuramoto model, where only $1:1$ resonance is possible.

Linear response theory asserts that sufficiently small external biases produce currents proportional to the applied force and forms the theoretical foundation of nonequilibrium transport. Here we demonstrate that linear response can break down even in uniformly hyperbolic deterministic systems when hierarchical asymmetry is present. Using a minimal class of uniformly expanding chaotic maps with hierarchical multiscale structure, we show that progressively finer transport channels become dynamically active as the applied bias decreases. The resulting force current relation is monotone and exhibits a hierarchical, fractal-like organization of activation thresholds. As a consequence, the effective mobility diverges as F to 0, demonstrating breakdown of linear response despite strong chaos and uniform hyperbolicity. The effect arises from deterministic multiscale activation rather than intermittency, stochastic noise, or singular invariant measures. These results identify hierarchy as an independent deterministic mechanism for nonperturbative transport response and demonstrate that uniform hyperbolicity alone does not guarantee the validity of linear response.

We introduce a two-phase quadratic integrate-and-fire (QIF) neuron whose membrane potential evolves according to two alternating Riccati equations within finite bounds. This simple extension removes the unphysical voltage divergence of the standard QIF model while producing realistic spike waveforms. Despite this modification, the system retains an exact low-dimensional description in the thermodynamic limit, governed by a single complex Riccati equation. Expressions for collective quantities such as the firing rate and mean voltage remain compact and analytically tractable. Because the formalism preserves the mathematical structure of the standard QIF ensemble, it inherits its many generalizations and can serve as a drop-in replacement in existing mean-field frameworks, providing a more biologically plausible yet still exactly solvable neuronal model.

We investigate diffusion-driven instabilities in a FitzHugh-Nagumo reaction-diffusion system with superdiffusive transport, modeled by fractional Laplacian operators with different diffusion orders for the activator and the inhibitor. A linear stability analysis yields explicit expressions for the instability threshold and the critical wavenumber and shows that superdiffusion modifies the band of unstable modes and the characteristic spatial scale of emerging patterns. We show that the threshold depends only on the ratio of the fractional exponents and on the kinetic parameters, while the spatial scale is controlled by the diffusion orders and the domain size. When the diffusion orders differ, spatial instabilities may occur even in regimes where the activator diffuses faster than the inhibitor, due to the combined effect of diffusion rates, anomalous scaling and system size. This leads to instability mechanisms that depart from the classical activator-inhibitor framework. A weakly nonlinear analysis near threshold provides the amplitude equation governing nonlinear saturation and reveals that superdiffusion promotes subcritical behavior. We also analyze the interaction between stationary and oscillatory instabilities near Turing-Hopf codimension-two points. All analytical results are supported by numerical simulations.

Droplet microfluidics combined with optical detection has become a powerful approach for high-throughput single-cell assays, but these systems often face limited sensitivity and signal heterogeneity due to optical and geometrical constraints. We investigate how key operating parameters influence the performance of a droplet-based optomicrofluidic platform. Experiments examine optical interactions between guided light and aqueous droplets containing fluorescent (FL) particles flowing in oil. Geometrical optics simulations model light-droplet interactions, while FL simulations quantify signal variations caused by particle size and position. Two refracted signals are observed experimentally: a droplet-refracted signal (DRS) that scales with droplet diameter and a particle-refracted signal (PRS) produced by light interaction with encapsulated particles. Both experiments and simulations show that PRS becomes prominent when the particle-to-droplet size ratio $D^*_\text{p}$ lies between 0.23-0.33, enabling label-free detection. Particles near the droplet center ($r^*_\text{p} < 0.4$) display reduced angular dependence and more uniform FL signals. Simulations further show that FL intensity increases with $D^*_\text{p}$, rising sharply from 0.33 to 0.5 and more gradually up to 0.66. Additionally, reducing the oil layer thickness enhances fluorescence by minimizing optical losses at the droplet-channel interface. These results demonstrate that controlling $D^*_\text{p}$, particle position, and oil layer thickness improves FL strength and uniformity, providing a framework for optimizing droplet-based fluorescence detection in microflow cytometry and single-cell assays.

We propose a monolithic, electrically driven source of photon pairs based on a non-linear AlGaAs Bragg reflection waveguide and a laser structure stacked on top. By introducing lateral tapers, the fundamental mode of the lasing waveguide is vertically coupled into a higher order mode of the Bragg reflection waveguide (Bragg mode) such that photon pairs can be generated through a type-II spontaneous parametric down conversion process. According to numerical simulations, a coupling efficiency of 28% is achieved between both modes. Phase matching the Bragg mode with two fundamental modes at 1550 nm results in a photon pair rate of 1.7*10^8 pairs/s for a 2 mm long device assuming 1 mW of power in the Bragg mode. Since the Bragg reflection waveguide does not require doping for this vertically coupled structure, free-carrier absorption losses and parasitic luminescence are avoided.

Cell migration plays a fundamental role in numerous physiological processes, including embryonic development, wound healing, and cancer metastasis. While cell-cell adhesion is known to regulate motion by shaping cell morphology and intercellular force balance, its dynamic, rate-dependent contributions to tissue behavior remain poorly understood. In this study, we examine how the dissipative nature of cell-cell adhesion influences tissue dynamics and collective migration using an extended vertex model with explicit junctional viscosity. Our findings reveal a nontrivial interplay between two distinct components of adhesion: an interfacial adhesion energy (energetic, rate-independent) contribution, which sets the effective junctional tension, and a dissipative (rate-dependent) contribution, which controls resistance to relative motion during cell rearrangements. We show that increasing the energetic component promotes migration by modifying cell shape and lowering the barrier to neighbor exchanges, whereas strengthening the dissipative component induces jamming and suppresses cell motion. Linear rheological analysis further demonstrates that, in the unjammed regime, vertex-model tissues exhibit power-law viscoelastic behavior, with adhesion modulating the power-law exponent and thereby controlling the spread of relaxation timescales. Together, these findings clarify the dual role of adhesion in governing tissue mechanics and rheology and provide a mechanistic framework for understanding the balance between fluidity and rigidity in epithelial monolayers.

New refractory alloys are being continuously designed and characterised for applications requiring good high-temperature mechanical properties and stability. Computational design from atomistic simulations is limited by interatomic potentials missing key elements, being too inaccurate, or computationally too slow for large-scale simulations. Here we present development of a refractory alloy database and two computationally efficient and general-purpose machine-learned potentials (tabGAP and NEP). We also design a cross-sampling strategy for effective sampling of training data using predictions from two potentials with completely different underlying architecture. The potentials support arbitrary alloy compositions of elements in groups four to six in the periodic table (Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W). The database is diverse yet multitargeted to enable simulations of refractory metals and alloys across different pure-metal, solid-solution, intermetallic, and glassy phases. We demonstrate the usefulness of the potentials by reproducing known pressure-, temperature-, and solute-induced phase transitions, grain boundary segregation, and simulations of radiation damage in the WTaCrVHf metallic glass.

In conventional optical Stark-shift spectroscopy, molecules are exposed to spatially homogeneous static electric fields that shift the energies of their spectral lines. These shifts are attributed to the molecular electronic properties, such as variation of dipolar moment and polarizability of the molecule associated with photo(de)excitation. In realistic environments containing structural defects and nanoscale heterogeneities, however, molecules experience internal electric fields that vary strongly on the molecular scale, rendering the standard Stark selection rules inapplicable. Here we develop an extended theory of atomic-scale Stark shift, addressing such scenarios. Specifically, we present a detailed theoretical analysis of an experimentally relevant configuration where the atomically sharp tip of a light-assisted scanning tunneling microscope is used to controllably apply inhomogeneous electrostatic fields to representative molecular dyes spanning several molecular families. We decompose the total Stark shift into linear and quadratic contributions and show that they contain different information about the molecular properties. Concretely, spatial variations of the linear Stark shift as the tip scans across the molecule enable subnanometric mapping of the charge redistribution between ground and excited electronic states, with high sensitivity to molecular composition and chemical functionalization. The quadratic Stark contribution, in contrast, reflects changes in the conventional dipolar polarizability upon excitation. Together, these results establish nanoscale Stark-shift spectroscopy as a powerful tool for resolving excited-state charge dynamics in single molecules under realistic, strongly inhomogeneous electric fields.

Van der Waals (vdW) ferroelectrics are emerging nonlinear photonic materials that combine large second-order susceptibility \c{hi}(2) with heterostructure compatibility, offering an attractive route toward miniaturized spontaneous parametric down-conversion (SPDC) sources. However, vdW SPDC sources operating under continuous irradiation in air remain limited in low brightness and poor operational stability, as oxygen and moisture exposure, together with pump-induced heating, lead to material degradation and permanent damage. Here we demonstrate an air-stable, bright SPDC source based on ferroelectric NbOI2 enabled by graphene encapsulation. Graphene provides robust environmental protection and can effectively supress pump induced degradation by enhancing heat dissipation. We report a record photon-pair generation absolute rate of 258 Hz and a normalized brightness of 19,900 Hz/(mW.mm). Leveraging this stabilized platform, we further generate polarization entangled photon pairs with 94% fidelity with respect to the maximally entangled Bell states from graphene-encapsulated 90° twisted bilayer NbOI2. Our results establish a practical and air-stable vdW ferroelectric SPDC platform that overcomes key limitations of existing vdW quantum light sources and provides a viable pathway toward scalable, integrated entangled photon sources for on chip quantum photonics.

We present a novel gradient regularization to completely eliminate the magnetic divergence error in meshless magnetohydrodynamics (MHD), which offers a high spatial resolution and conservative advantage, due to its Lagrangian nature. Comparing with the counterpart of constrained-gradient (CG) technique, we reform $\nabla \cdot \mathbf{B}=0$ by an implicit projection method to modify the magnetic-field gradients. The accuracy of modified-gradient (MG) method is verified and it achieves exact divergence-free results with round-off precision, by using tests of shock tube, 2D and 3D vortex, magneto-rotational instability, and especially, advection experiment, compared with CG method and the GIZMO code. It leads to noticeable improvement in pattern, amplitude and numerical dissipation of divergence error of magnetic field.

Contact tracing is one of the most important control measures deployed during epidemics. Relying on the identification of contacts of known infected individuals, it necessitates a network perspective. Although pairwise models have been used extensively to study contact tracing, their analysis typically depends on a decoupling assumption-most commonly that contact tracing operates on a much faster timescale than disease transmission. Furthermore, contact tracing models often assume that all infected individuals become contact tracing-triggering, which is unrealistic given partial compliance to treatment. We relax both of these restrictive assumptions and provide a full analytical characterisation of the epidemic threshold in the pairwise mean-field model. Our analysis uses a fast-variables approach that captures the rapid early stabilisation of key network quantities. Inspired by mechanisms from social adoption dynamics, we introduce triplewise contact tracing in which an infected individual can be traced not only through direct contact with a single tracing-triggering neighbor (pairwise tracing), but also indirectly when connected to two tracing-triggering nodes simultaneously. For pure pairwise and pure triplewise contact tracing, we derive analytical expressions for critical contact tracing levels and demonstrate that when many infected individuals bypass treatment, the epidemic can become uncontrollable. When both contact tracing mechanisms operate together, we map out their combined contribution and relative impact on epidemic control. This unified framework yields rigorous and tractable threshold conditions for contact tracing dynamics on networks, extending the applicability of pairwise models beyond the fast-tracing regime and providing new insight into the interplay between disease progression, partial treatment compliance, and higher-order tracing processes.

We consider the inverse problem of calculating a refracting surface that generates a prescribed irradiance distribution in the far field for a collimated incident beam. This problem can be formulated as a mass transportation problem (MTP) with a quadratic cost function. To solve this problem, we propose a version of the supporting quadric method (SQM), in which the calculation of the quadric parameters is reduced to the problem of minimizing a convex function. We obtain simple analytical expressions for the second derivatives of this function, making it possible to calculate the quadric parameters using second-order optimization methods. This allows us to refer to the proposed method as the second-order SQM. We demonstrate high efficiency of this approach by designing several optical surfaces that generate complex irradiance distributions. We also consider the application of the second-order SQM to nonimaging optics problems described by MTPs with a non-quadratic cost function.

Active topological photonic systems enable robust light control and new pathways for semiconductor lasing. However, their intrinsically non-Hermitian nature, combining gain, radiation leakage, and material loss, makes the underlying physics more complex, and prior studies have mostly focused on gain competition while the influence of loss channels is less examined. Here, we experimentally demonstrate radiative-channel-driven topological lasing in a valley photonic crystal consisting of isolated InP nanorods on an insulator, achieving room-temperature single-mode operation within an about 4 lambda scale cavity. Loss-included simulations show that material absorption and radiative leakage can be exploited to establish the lasing pathway. Local off-edge pumping provides spatial evidence of topological edge-guided lasing. Berry-curvature calculations, reflecting unit-cell symmetry breaking, verify the valley-Hall interface in a triangular lattice. Geometry and temperature tuning identify a narrow spectral window where the above-light-line edge branch matches the gain-loss balance and remains decoupled from bulk bands. These results clarify how the loss landscape governs topological lasing and establish a radiative-edge design framework for active topological photonics. The substrate-supported nanorod architecture additionally offers a scalable route for on-chip implementation.

An anomalous back discharge movement phenomenon is induced by a set of dielectric barrier corona discharges (DBCD) at unipolar half-sine voltage waveforms, where the back discharge has a time delay that relates to the applied voltage level. An ionic wind model is employed to analyze the physical behavior. Theoretical explanation and quantitative analysis are presented in this study based on abundant experimental results of 5 typical insulating materials and a FEP insulating cable. A numerical model is derived, which indicates that the back discharge can be activated under a relatively low potential voltage level in this study. The results highlight that the back discharge movement phenomenon behaves distinctly under half-sine voltage with negative polarity, yielding a significantly different partial discharge (PD) pattern with positive polarity. Besides, PD amplitude dependent on dielectric thickness is demonstrated by plotting in phase resolved partial discharge (PRPD) pattern. Furthermore, comparative experiments are conducted with respect to the variation of air gap length and dielectric geometry, manifesting different influences on PD amplitude.

Radiative-conductive systems are intrinsically nonlinear due to the quartic temperature dependence of thermal radiation. Under fixed total heating power, convexity arguments imply that nonuniform temperature distributions radiate more efficiently and therefore exhibit a lower mean temperature than their isothermal counterparts. However, this conclusion remains qualitative, and an explicit quantitative relation between temperature heterogeneity and mean temperature reduction has been lacking. Here we derive a variance-based analytical expression linking the area-averaged temperature to the corresponding isothermal equilibrium temperature in a nonuniformly heated radiative--conductive system. By integrating the governing equation and performing a systematic second-order expansion about the ambient temperature, we show that the decrease of the mean temperature relative to the isothermal equilibrium value is linearly proportional to the temperature variance, with a proportionality coefficient set solely by the ambient temperature. This result transforms the convexity-based inequality into a quantitative statistical relation within the perturbative regime and provides a physically transparent framework for describing nonlinear radiative averaging in thermally heterogeneous systems.

We propose an SIS competition model describing the propagation of conflicting rumors, such as fake news and its corrections. This simple model captures the interaction between rumor propagation and opinion dynamics, where rumors drive opinion changes and, conversely, individuals' opinions determine the infection rates of rumors. We analytically derive all steady states and their stability. These results uncover a novel coexistence mechanism. This coexistence corresponds to a scenario where belief in one rumor (e.g., fake news) paradoxically aids the spread of the opposing rumor (e.g., corrective information). Due to this mechanism, a nontrivial but realistic phenomenon occurs where a lower infection rate actually enhances the spread of a rumor. Furthermore, although the model does not explicitly incorporate majority conformity, a phenomenon where the majority gains an advantage emerges spontaneously. Consequently, even if one rumor has a higher infection rate, it may be eliminated by the other if its initial share fails to exceed a critical threshold. We analytically derive this threshold using the singular perturbation method.

State-of-the-art ensemble Kalman filtering (EnKF) algorithms require incorporating localization techniques to cope with the rank deficiency and the inherited spurious correlations in their error covariance matrices. Localization techniques are mostly ad-hoc, based on some distances between the state and observation variables, requiring demanding manual tuning. This work introduces a new ensemble filtering approach, which is inherently localized, avoiding the need for any auxiliary localization technique. Instead of explicitly applying localization on ensembles, the idea is to first localize the continuous analysis probability density function (pdf) before ensemble sampling. The localization of the analysis pdf is performed through an approximation by a product of independent marginal pdfs corresponding to small partitions of the state vector, using the variational Bayesian optimization. These marginals are then sampled following stochastic EnKF and deterministic ensemble transform Kalman filtering (ETKF) procedures, using ensembles larger than the partitions' size. The resulting filters involve the same forecast steps as their standard EnKF and ETKF counterparts but different analysis steps, iteratively adjusting the EnKF and ETKF updates of each partition based on the ensemble means of the other partitions. Numerical experiments are conducted with the Lorenz-96 model under different scenarios to demonstrate the potential of the proposed filters. The new filters' performances are comparable to those of the EnKF and ETKF with already tuned localization, both in terms of computational burden and estimation accuracy.

Objective: Real-time adaptive proton range verification systems based on produced neutrons require accurate information on their non-isotropic momentum distributions within short times, for which Monte Carlo (MC) methods are too computationally expensive. We present a surrogate model based on Fourier Neural Operators (FNO) for fast prediction of angle- and energy-resolved proton transport and neutron production within proton therapy. Approach: We treat the irradiated phantom and the proton beam's state as depth-evolving series, respectively of different materials, and of spatial, angular and energy phase space density distributions. The task is solved auto-regressively by learning changes in the distributions of protons and those of produced neutrons. For training and evaluation, two datasets of 47 MC simulations featuring different primary intensities were produced. Simulated geometries were extracted from a thoracic CT scan as series of laterally homogeneous materials. Main Results: An average relative $L^2$ discrepancy of 0.067 and 0.137 was achieved by the predicted proton and neutron distributions, respectively. This corresponded to an average gamma passing rate in the spatial distributions of 99.95$\%$ and 99.40$\%$. Training with higher primary intensities led to improvements between 12$\%$ and 30$\%$ in density metrics. Inference over depths of 40 cm at a resolution of 0.5 mm required on average 23.17 s per beam. Significance: The proposed proton beam surrogate generates accurate spatial and momentum distributions of neutrons at MC-level accuracy within seconds, while demonstrating robust generalization with respect to irradiated geometry and beam characteristics. This approach can be used for prototyping and operation of range verification systems, other tasks such as neutron dose estimation, and can be extended to include other kinds of secondary emissions.

We present a time-domain scheme for computing Casimir forces within the Maxwell stress tensor formalism, together with a specific realization using the finite-element-based discontinuous Galerkin time-domain method. The approach enables accurate evaluation of Casimir--Lifshitz interactions for a wide range of geometries and material properties at finite temperature. At the core of the method, the electromagnetic Green's tensor is expressed as the system's response to dipolar excitations, thereby recasting the Maxwell stress tensor into a set of classical scattering problems driven by electric and magnetic dipoles. We validate the approach against reference calculations of the Casimir interaction between parallel half-spaces at both zero and nonzero temperature. We further demonstrate its applicability to finite, cylindrically symmetric geometries for which closed-form solutions are unavailable, obtaining accurate agreement with asymptotic predictions based on physical considerations. These findings illustrate the method's potential for studying Casimir interactions in realistic micro- and nanoscale structures, relevant to nanodevice design and experimental settings.

We present a generative framework for inverse design of five-layer transmissive Huygens' metasurfaces (HMSs), addressing a longstanding challenge in achieving full-phase, high-efficiency unit cell designs with minimal full-wave simulations. The key to achieving this is our reliance on the field-based semianalytical (SA) scheme developed in Part I of this paper, which allows rapid and highly effective synthesis of such multilayer composites, however with limited accuracy. To overcome the prohibitive data demands of traditional pipelines, we employ Mamba, a selective state space model well suited for long-range sequence modeling as the backbone of our learning framework. A bidirectional Mamba (Bi-Mamba) forward surrogate is first trained on SA-generated data and subsequently fine-tuned with full-wave CST samples. An ablation over a 1080-sample CST pool shows that as few as 270 full-wave calibration samples suffice to reach near-CST-level agreement at a fraction of the simulation cost. An autoregressive Mamba inverse generator is subsequently trained on surrogate-augmented data, treating unit-cell synthesis as a sequential generation task. The resulting one-to-many generative model produces diverse unit cell geometries conditioned on target scattering responses. It achieves CST-validated designs with field transmission magnitude 0.9 across the full 0-$2π$ phase range at 20 GHz. Moreover, a CST-calibrated surrogate trained to accurately predict frequency responses (18-22 GHz) enables functional post-selection of inverse generated designs. Together, the hybrid SA-generative methodology in this two-part compilation establishes a scalable and data-efficient solution for multilayer HMS synthesis, with natural extensions toward broadband, oblique-incidence, and higher-dimensional electromagnetic inverse-design problems.

Radiative near-field (RNF) beamforming is vulnerable to blockages that disrupt Fresnel zones. This letter proposes an obstruction-unaware wavefront shaping strategy inspired by catastrophe optics. By superimposing a calibrated quartic phase, we generate a Pearcey-like wave packet that exhibits structural stability against perturbations. We establish a fair comparison protocol where the quartic beam is calibrated in free space to avoid exploiting obstruction knowledge. Numerical results demonstrate up to 8.5~dB SINR gain over conventional focusing for multi-user scenarios near the depth-of-focus limit. Crucially, this gain stems from improved channel conditioning under partial blockage, which mitigates the severe noise amplification inherent to zero-forcing precoding.

Femtosecond pulsed laser systems constitute powerful tools for the high-precision structuring of materials at micro/nano-scale resolutions. A critical parameter influencing the efficacy of ultrafast laser-material interactions is the laser-induced damage threshold (LIDT), which is defined as the minimum laser fluence required to induce irreversible modification to the material surface. While extensive studies have addressed single-pulse damage mechanisms, the response of thin metallic films to double-pulse femtosecond irradiation, particularly when the film thickness is of the order of the optical penetration depth, remains, generally, unexplored. In this work, we present a rigorous theoretical investigation into the spatiotemporal evolution of energy deposition, thermalization processes and optical parameter changes under double-pulse excitation conditions. The analysis considers key parameters including the inter-pulse delay and the film thickness to evaluate their influence on the LIDT for a range of technologically relevant metals: Au, Ag, Cu, Al, Ni, Ti, Cr, Pt, W, Mo and Stainless Steel (100Cr6). ). A comparative analysis highlights the potential of controlled double-pulse irradiation schemes to manipulate energy coupling efficiency, improve the spatial selectivity of laser-induced modifications and compile a comprehensive LIDT database for commonly used industrial materials. The approach is aimed to provide a robust foundation for the design and optimization of advanced laser micromachining and nanofabrication protocols across a broad spectrum of metallic systems.

Planar metasurfaces can profoundly control electromagnetic scattering. At microwave frequencies, such devices are typically implemented using multilayer cascades of patterned metallic sheets, whose design often requires time-consuming full-wave optimization. Here, we extend analytical models originally developed for sparse loaded-wire metagratings to accurately describe densely packed Jerusalem-cross meta-atoms embedded in standard printed circuit board (PCB) dielectric stacks. The model captures both near- and far-field coupling within and between layers, enabling efficient prediction of the dual-polarized response. Using this framework, we identify highly transmissive meta-atoms whose phase is controlled by the leg lengths of the Jerusalem crosses (microscopic design stage). This (phase)-(leg-length) "lookup table" allows rapid synthesis of Huygens' metasurfaces (macroscopic design stage), demonstrated through a full-wave-validated metalens exhibiting low-reflection beam manipulation. Notably, we implement a judicious scaling method to further extend the model to predict wideband meta-atom responses. In the companion paper (Part II), a hybrid machine-learning approach leverages this semianalytical framework to enhance accuracy without requiring the conventional exhaustive full-wave training, enabling ultrafast inverse design across the full parameter space. Overall, the presented methodology -- the standalone semianlytical scheme (Part I) and the machine-learning enhanced version (Part II) -- establishes an effective open-source toolkit for versatile, rapid, and highly accurate synthesis of fabrication-ready dual-polarized transmissive Huygens' meta-atoms and metasurfaces.

Label-free vibrational microscopy provides chemically specific access to cellular structure, yet quantitative volumetric chemical dynamics in living cells remain largely inaccessible, particularly on subsecond timescales relevant to intracellular transport and structural reorganization. This limitation arises because most high-speed vibrational techniques rely on raster scanning, which constrains volumetric throughput to approximately one volume per second (vps). Although mid-infrared photothermal (MIP) imaging offers a pathway toward spatially parallel chemical detection, existing implementations have remained far below video-rate volumetric operation, reflecting a fundamental trade-off between imaging speed and signal-to-noise ratio. Here, we overcome this trade-off in MIP tomography and realize video-rate volumetric chemical imaging using mid-infrared photothermal optical diffraction tomography (MIP-ODT), achieving high photothermal sensitivity while maintaining quantitative measurement fidelity. High per-angle detectability supports volumetric reconstruction without temporal averaging, yielding a signal-to-noise ratio exceeding 70 under video-rate acquisition conditions. Consequently, volumetric imaging at 19.2 vps is achieved, representing a nearly 400-fold improvement over prior implementations. Using this capability, we performed video-rate three-dimensional tracking of lipid droplets in living cells and quantified anomalous diffusion from full volumetric trajectories, revealing heterogeneous intracellular transport behaviors that are obscured in two-dimensional measurements. We further demonstrate high-speed hyperspectral volumetric chemical imaging across a 300 cm-1 spectral window within 1 s through rapid MIR wavenumber sweeping, paving the way for real-time three-dimensional organelle-specific chemical phenotyping.

We report a phase-sensitive nonlinear x-ray response in the charge-density-wave material 1T-TaS2, revealed through x-ray parametric down-conversion into the ultraviolet. Extending nonlinear x-ray wave mixing beyond conventional crystalline systems to a correlated quantum material, we employ reciprocal-lattice phase matching to isolate distinct Fourier components of the nonlinear susceptibility. By selecting a fundamental reciprocal-lattice vector and a stacking-sensitive half-integer reciprocal-lattice vector, we probe the response across the nearly commensurate and incommensurate charge-density-wave phases. Tuning the ultraviolet photon energy through Ta O-shell resonances uncovers pronounced Fourier-component and phase-dependent resonant structure, indicating that the stacking-related nonlinear susceptibility couples differently to Ta-centered resonant states than the average lattice response. Remarkably, the nonlinear signal is strongly enhanced in the nearly commensurate phase despite weaker Bragg diffraction, demonstrating that the nonlinear susceptibility provides information inaccessible to linear probes. These results establish nonlinear x-ray spectroscopy as a phase-sensitive and orbital-selective probe of electronic reconstruction in quantum materials.

A simple numerical method for loading of a relativistic Maxwellian-type distribution is proposed based on inverse transform sampling. The relativistic Maxwellian energy distribution is introduced as an alternative to the Maxwell-Jüttner distribution. The cumulative distribution of the shifted-Maxwellian energy distribution is approximated by an invertible function. Random variates of energy is transformed from uniformly distributed random variables. Then, the energy variates are converted to momentum vector variates. Numerical tests are presented to show that the present method successfully reproduce the relativistic Maxwellian energy distribution.

For window applications in hot climates, it is desirable to have windows with high visible transparency, while maintaining strong reflectance in both the ultraviolet and near infrared, to minimize unwanted heat gain. Given that more than 70% of incident solar energy is at wavelengths shorter than 1000 nm, achieving spectrally abrupt transitions from transparent to reflecting at the boundaries of the visible is essential. Such abrupt transitions at multiple wavelengths typically would require tens of dielectric layers, which is impractical for most window applications. Here, we propose and realize a structure comprising only eight planar layers that achieves sharp reflectance changes at ~390 and ~680 nm, resulting in high visible transmittance (>70%), and high near-infrared (>80%) and UV (>60%) reflectance, as well as high mid-infrared emissivity (>90%) for additional radiative cooling. We demonstrate an air temperature reduction of up to 3.8 °C with our engineered window compared to a reference window.

The inverse design of nonlocal metasurfaces requires the precise optimization of lattice geometry to engineer spatial dispersion and high-Q resonances. However, gradient-based optimization is frequently bottle-necked by the evaluation of the periodic Dyadic Green's Function (DGF), where traditional Finite Difference (FD) methods suffer from an inherent trade-off between truncation error and numerical instability near spectral singularities. In this work, we present an end-to-end Analytic Gradient Engine for 2D Bravais lattices. By mapping the spectral lattice sums of the Coupled Dipole Approximation (CDA) to the theory of Quasi-Modular Forms (QMF), we derive exact, closed-form expressions for the gradients of the interaction matrix with respect to the modular lattice parameter $τ$. Our framework explicitly handles conditionally convergent terms via regularization and addresses the non-holomorphic outlier $σ_4^{(2)}$ via a hybrid numerical strategy. We further introduce a robust evaluation scheme combining $SL(2, \mathbb{Z})$ domain reduction with automatic error certificates. Experimental validation demonstrates that our engine achieves machine-precision derivatives ($10^{-15}$) and a 6.5$\times$ speedup in optimization convergence compared to finite-difference baselines, enabling the robust design of giant anisotropy in regimes where traditional methods fail.

Parylene C thin films have significant applications in advanced packaging of microelectronics. Their thermal properties are critical for thermal management of electronic devices. However, a unified understanding of the tunable structure and the corresponding thermal conductivity is still missing. This study investigated parylene C thin films of varying thickness and post-annealing temperatures grown via thermal chemical vapor deposition. The ultralow thermal conductivity of as-deposited parylene C measured by time domain thermoreflectance (TDTR) is 0.10 W/m-K. The thermal conductivity can be tuned by post-annealing. Significant increase in thermal conductivity is observed in the annealed samples (0.18 W/m-K) which induces melting and recrystallization. The results of XRD and polarized Raman spectroscopy show that the enhanced thermal conductivity is due to improved crystalline quality and the change in chain orientations. The measured thermal conductivities of the as-deposited and annealed films are much lower than the values predicted by the Cahill minimum thermal conductivity model, which can be explained by the diffuson-mediated minimum thermal conductivity model. Parylene C is found to possess the lowest thermal conductivity among dense low-k materials. Our work provides guidance for the structural design of ultra-low thermal conductivity polymers and corresponding thermal design of electronics.

Rapid urban population growth drives car travel demand, increasing transport carbon emissions and posing a critical challenge to sustainable development. Although existing studies have demonstrated that eco-routing can reduce individual emissions, research gaps remain. On the one hand, such personal reductions have a negligible impact on overall emissions, and cannot be simply aggregated to capture the complex effects of large-scale eco-routing. On the other hand, under population growth, the long-term effectiveness of eco-routing, as well as the evolution of its efficiency and traveler route choice, remain underexplored. To address these limitations, this study proposes Time-Only and Time-Carbon user equilibrium (UE) models, integrates them with a demand forecasting method for simulating future network traffic, and designs multi-dimensional metrics to characterize urban dynamics. Using real-world road networks, commuting origin-destination (OD) demand, and population projections under various shared socioeconomic pathways (SSPs) for six representative U.S. cities as a case study, we conduct a comprehensive analysis of urban dynamics across different routing strategies and population sizes. The results reveal that while eco-routing mitigates total emissions, emissions in most cities scale superlinearly with population, a scaling order that remains invariant regardless of routing and construction strategies. Moreover, under population growth, travelers using eco-routing tend to increasingly select shorter routes, giving rise to carbon bottlenecks. A strategy of targeted capacity expansion on these critical bottlenecks (0.46% of links) significantly reduces both emissions (3%) and travel time (28%) without compromising eco-routing efficiency. This study provides a foundation for formulating low-carbon urban transport planning and emission reduction policies.

We present two independent optical methods for absolute primary thermometry using rare-earth-doped nanoparticles. Both approaches rely exclusively on the internal energy levels and population dynamics of the dopant ions, eliminating the need for external temperature references. We experimentally demonstrate the concepts by using Y$_2$O$_3$: Yb$^{3+}$/Er$^{3+}$ nanoparticles, exploiting Boltzmann distribution between individual Stark sublevels of the Er$^{3+}$ ions, emitting in the green spectral region ($\sim$550 nm) and in the near-infrared spectral region ($\sim$1600 nm). Our strategy establishes rare-earth-based luminescence thermometers as genuine absolute primary probes, conceptually comparable to Johnson noise and acoustic gas thermometers, but with the fundamental advantage of possibly being employed at the nanoscale, potentially down to the single-ion limit, with optical readout and over wide temperature ranges.

From soil to the gut, communities composed of thousands of microbes perform functions such as carbon sequestration and immune system regulation. Here, we introduce a data-driven approach that explains how community function can be traced to just a few groups of microbes or genes. In gut communities, our neural-network based clustering algorithm correctly recovers known functional groups. In the ocean metagenome, it distills ~500 gene modules down to three sparse groups highlighting survival strategies at different depths. In soils, it distills ~4400 bacterial species into two groups that enter a mathematical model of nitrate metabolism. By combining interpretable ML with strain isolation and sequencing experiments, we connect the metabolic specialization of each group to community-wide responses to perturbations. This integrated approach yields simple structure-function maps of microbiomes, allowing the discovery of molecular mechanisms underlying human and environmental health. More broadly, we illustrate how to do function-informed dimensionality reduction in biology.

The Hawking effect can be understood as a broad kinematic phenomenon associated with mode behavior near a horizon. While astrophysical black holes produce one specific realization of this radiation, this perspective inspires extensive theoretical and experimental efforts to create event horizons in diverse physical systems to observe the resulting analogue Hawking emission. One of the most successful realizations is the fiber-optical analogue, based on nonlinear quantum optics. In these notes, we introduce and motivate this system while outlining the theoretical concepts underlying the gravitational analogy. Finally, we review key experiments and discuss their impact on the field.

*The abstract of this article is too long to be included in the arXiv metadata; please see the paper for the full abstract.* ...In this paper, we introduce the detailed theory of cubic-in-magnetization magneto-optic Kerr effect (CMOKE) by deriving the magneto-optic tensor of third order in magnetization, denoted as $\bm{H}$, and comparing the strength of CMOKE for different crystal orientations theoretically and experimentally. In crystals with cubic symmetry, the tensor $\bm{H}$ is described by two independent parameters $H_{123}$ and $H_{125}$. Together with the linear magneto-optic tensor $\bm{K}$ and quadratic magento-optic tensor $\bm{G}$, the permittivity tensor is described up to third order in magnetization. We analytically describe equations of the MOKE including the contribution of QMOKE and CMOKE itself for (001)- and (111)-oriented cubic crystal structures. Those are compared to experimental measurements of two samples with an (001)- and (111)-oriented fcc Ni layer, respectively. Further, we use Yeh's 4$\times$4 transfer matrix calculus to simulate and describe the experimental measurements phenomenologically from the permittivity tensor up to third order in $\bm{M}$. We find that the MOKE anisotropy that stems from the magneto-optic tensor $\bm{H}$ described as $ΔH = H_{123}-3H_{125}$, is much more pronounced for the (111)-oriented cubic crystal structure, for which it manifests as three-fold in-plane angular dependencies of MOKE with longitudinal and also with transversal magnetization direction, respectively.

Hyperentanglement offers enhanced capacity for quantum information processing and communication protocols, especially in combination with robust high-dimensional degrees of freedom such as frequency-bin encoding. Here, we present a single-pass, unfiltered, down-conversion source of hyperentangled photon pairs in polarisation and frequency-bin degrees of freedom with dynamically tunable state dimension and composition at telecom wavelengths. We achieve this by optimal tailoring of the photons' joint spectral amplitude via pump and nonlinearity shaping. Using polarisation-resolved time-of-flight spectrometry and Hong-Ou-Mandel interference, we characterise the hyperentangled states and demonstrate for the polarisation component fidelities exceeding 99% averaged over frequency bins and concurrences above 98%. The degree of spectral entanglement, quantified by the Hong-Ou-Mandel visibility, is measured as 90%, well in line with numerical simulations. This approach provides a scalable route toward high-dimensional quantum states for quantum communication and computing applications.

We advocate for a more stringent test problem for codes that aim to solve the equations of viscous hydrodynamics. Specifically, we discuss a nonuniform-density version of the common (uniform-density) Gaussian velocity shear test, where density gradients transverse to the direction of velocity shear cause the velocity profile to drift over time. By employing a nonunifom density, this test provides a test that the full viscous stress (and velocity shear) tensors are calculated correctly from the conserved variables, and checks the correctness of the fluxes and source terms calculated therefrom. In Appendix A, we present a detailed exposition of the Navier Stokes equations, particularly their fluxes and source terms in a variety of common coordinate systems.

Restaurants, cafes, and other commercial amenities are among the most visible markers of neighborhood change, yet whether their arrival drives house price appreciation or merely follows rising demand remains an open empirical question. This study investigates the causal effect of commercial entry on residential property values in Greater London. Exploiting the staggered timing of 21,189 restaurant and cafe openings across 4,835 Lower Layer Super Output Areas (LSOAs)--identified through Energy Performance Certificate records--we implement an event study design with LSOA-specific linear trends that passes the parallel trends test (F = 1.04, p = 0.384). We find that house prices rise monotonically after commercial entry, reaching +4.1% at four years post-treatment (p < 0.01). The effect is gradual and cumulative, consistent with amenity capitalisation. By matching EPC records to Google Places API price tier data at the building level, we further show that the effect is driven by upmarket commercial entry (+7.4%, clean pre-trends) rather than budget establishments (questionable pre-trends, unreliable post-treatment effect), establishing that the quality of commercial clustering--not merely its presence--drives neighborhood price dynamics. Results are robust to heterogeneity-robust estimation, alternative treatment thresholds, broader commercial category definitions, and a permutation-based placebo test.

Integrated silicon nitride (Si3N4) resonators have emerged as a leading platform for nonlinear photonics, yet generating light at wavelength in the ultraviolet (UV) has remained elusive in single-resonator systems. Here we report the first observation of fourth-harmonic generation reaching the blue and ultraviolet spectral regions in an integrated Si3N4 microring resonator. We systematically investigate the input-power dependence of the wavelength ranges supporting second-, third-, and fourth-harmonic generation, and study the input-power-dependent variation of the circulating fourth-harmonic signal in the UV. These results extend the operational bandwidth of integrated Si3N4 nonlinear photonic platforms to the lower edge of the material transparency window, enabling on-chip UV frequency conversion. Near-ultraviolet generation around 400 nm will enable on-chip excitation of defect-based quantum emitters in hexagonal boron nitride, enhance Raman spectroscopy through increased scattering cross-sections at shorter wavelengths, and support compact fluorescence-based bio-imaging platforms exploiting intrinsic cellular fluorophores.

The deployment of renewable energy technologies supposes the connection to the power grid of many new, distributed, and variable electricity production facilities. Among the investments deeply needed for a successful shift to clean energy, electricity storage systems are key to provide power reliably, continuously and economically. Here, we are concerned with the energy that must be invested and embodied in storage devices and in production oversizing to cope with natural variations of renewable electricity production, and compensate for any gap between production and consumption. We developed a model to analyze the variation of energy expenses with the location in the grid, capacity of storage and production oversizing. We apply it to a time scale of fluctuations of a few hours that can be taken care of by Li-ion batteries to calculate the optimal storage capacity and production oversizing yielding a maximum value of the ESOI ratio [Energy Stored On energy Invested] at a given satisfaction rate of customer demand. We evaluate these values for a rescaled present-time French power mix and two idealized zero-emission mixes (100% PV and 100% wind). In parallel, using a recently developed model of French transmission grid, a centrality-based analysis shows that locating storage at nodes of maximal installed power minimizes additional Joule losses. These results generalize existing grid-level energy return frameworks to incorporate storage sizing, placement, and transmission losses into a unified assessment of future power grid configurations.

Because of the special angular distribution of excited electrons by the photoelectric effect, the Gas Pixel Detector (GPD) is effective in measuring keV X-ray polarization of astrophysical events (e.g. gamma-ray bursts), by capturing ionization tracks of excited electrons as polarized images. Traditionally, the emission angles of photoelectrons are extracted from polarized images first, and statistics are then performed on these angles to infer the polarization direction and intensity. However, observation with the wide field of view requires the incident angle of X-rays not directly attainable through the traditional analysis process. In this paper, we propose using the generalized sliced Wasserstein (GSW) distance, projected by neural networks with random weights, as a completely data-driven approach to analyze X-ray polarization based on two-dimensional polarized images. We find the structures of the randomized neural networks matter when focusing on different aspects of the polarized images, and take advantage of the discrimination abilities by different neural network structures. The proposed method, named the structured GSW distance, successfully distinguishes polarized images with different configurations of incident angles and polarization directions. Furthermore, we build a simplified statistical model based on the von Mises distribution and the circular Wasserstein distance and compare the model against the proposed method, showing their high consistency. The computational method reported in this paper may benefit GPD-based polarimetry in astroparticle experiments and also pattern analysis on raw data from pixel detectors.

The stringent demands of photonic quantum computing protocols motivate photonic integrated circuit (PIC) platforms with passive optical properties such as extremely low losses and correspondingly large circuit depths, as well as active optical properties such as high reconfiguration rates, low power dissipation, and minimal crosstalk. At the same time, many quantum photonic resource state generators, such as single-photon sources and quantum memories, require operation in the visible wavelength range. These requirements make the passive optical properties of CMOS-fabricated, ultralow-loss, low-confinement silicon nitride waveguides especially attractive. However, the conventional active properties of these systems based on thermo-optic modulation are plagued by high levels of crosstalk, slow modulation rates, and high power dissipation. Although there have been recent demonstrations of CMOS-fabricated, visible wavelength, piezo-optomechanical PICs that solve the above challenges associated with implementing active functionality, these have made use of high-confinement waveguides with currently demonstrated losses of order $0.3$-$1~\mathrm{dB/cm}$, precluding circuit depths required for scalable quantum algorithms. Here, we demonstrate that combining piezo-optomechanical actuation with a low-confinement, ultra-low loss silicon nitride platform addresses the scalability challenge while enabling high-performance active functionality at visible wavelengths. This platform achieves a propagation loss $0.026~\mathrm{dB/cm}$ at $780~\mathrm{nm}$, modulation bandwidths in the MHz range, and a phase shifter voltage-length product ($V_πL$) of approximately $2.8~\mathrm{\mathrm{V}\cdot\mathrm{m}}$ and negligible hysteresis. We further demonstrate reconfigurable Mach-Zehnder interferometers based on spiral phase shifters with 0.63 dB loss per phase shifter.

Global food production must reduce environmental impact while meeting rising demand for dietary protein. Plant-based meats aim to preserve the sensory and cultural role of animal meat, while lowering greenhouse gas emissions, land use, and health risks. Advances in protein structure and flavor chemistry have improved product quality; yet, consumers continue to prioritize taste and texture over sustainability and systematic large-scale consumer surveys are scarce. It remains unclear how plant-based products rank against animal benchmarks and which product attributes most strongly influence overall liking. Here we show, in a large-scale, blinded, in-person sensory evaluation across 14 product categories, 2,684 consumers, more than 11,000 product evaluations and 800,000 data points, that plant-based products still trail animal benchmarks at the category average level, but approach parity in selected formats: Plant-based unbreaded chicken filets, chicken nuggets, and burgers achieved mean overall liking scores of 5.1, 4.9, and 5.2, differing from the animal benchmark by only 0.1, 0.2, and 0.3 points on a seven-point scale. For unbreaded chicken filets and burgers, 48% and 47% of participants rated the plant-based product the same as or better than the animal benchmark. Categories with higher sensory parity captured 5-14% market share compared with less than 1% for low-parity categories. Penalty analysis identified savoriness, aftertaste, juiciness, and tenderness as the strongest determinants of liking. These findings show that sensory parity is technically achievable, but not yet consistent across product types. By publicly sharing all data, we establish an open benchmark for alternative protein performance to democratize research and accelerate principled, data-driven innovation. All data are freely available at https://www.nectar.org/sensory-research/2025-taste-of-the-industry.

We consider the implementation of optimization techniques within the study of tectonic plate motion. Specifically, we examine the optimization underlying optAPM, a leading code for modeling absolute plate motion. We highlight that modifications in the construction of the objective function, composed of individual cost functions, can improve modelling performance. In particular, we propose a simpler and more intuitive formulation of the hotspot cost function. A key part of the new hotspot analysis is the pre-interpolation of hotspot trail data, crucial geological markers for validating absolute plate motion over O(100) Myr timescales. By reducing the propagation of modeling errors, our refined model provides more precise reconstructions of historical plate movements. Our modified hotspot modelling improves the accuracy and reliability of the optAPM outputs.

Structural and functional heterogeneity are hallmarks of cortical circuits, from broad degree distributions in the mouse connectome to diverse intrinsic neuronal timescales. Yet a mechanistic link between connectivity heterogeneity and functional diversity is lacking. To bridge this gap, we introduce a random recurrent network in which connectivity is generated by a configuration model with tunable degree heterogeneity and synaptic weights exhibiting varying levels of correlation. Using generating-functional methods, we derive a heterogeneous dynamical mean-field theory (hDMFT) with degree-conditioned stochastic dynamics. The theory shows that the interaction of partial symmetry in the weights and degree heterogeneity induces a non-Markovian memory term in the form of an emergent self-coupling whose strength scales with degree and produces a broad distribution of activity timescales. We obtain analytic stability criteria demonstrating that degree heterogeneity lowers the critical gain and localizes unstable modes onto hubs. The resulting rich dynamical landscape includes silent, chaotic, and multistable regimes, which we uncover via spectral, replica, and Lyapunov exponent analyses. We highlight the computational benefits of the observed timescale heterogeneity by revealing that, under an external input drive featuring a broadband spectrum, slow hub neurons act as integrators, demixing slow input components. Finally, instantiating the model with the empirically measured topology from the MICrONS cubic-millimeter mouse connectome explains the broad range of single-neuron timescales and their positive correlation with in-degree observed in resting-state recordings. Our results provide a mechanistic link between connectome topology, neural dynamics, and computation, identifying hubs in partially symmetric networks as a natural substrate for multiplexed processing across timescales.

Background and objective: Cell-level pathological image analysis requires working with extremely small image patches (40x40 pixels), far below standard ImageNet resolutions. It remains unclear whether modern deep learning architectures and foundation models can learn robust and scalable representations under this constraint. We systematically evaluated architectural suitability and data-scale effects for small-patch cell classification. Methods: We analyzed 303 colorectal cancer specimens with CD103/CD8 immunostaining, generating 185,432 annotated cell images. Eight task-specific architectures were trained from scratch at multiple data scales (FlagLimit: 256--16,384 samples per class), and three foundation models were evaluated via linear probing and fine-tuning after resizing inputs to 224x224 pixels. Robustness to blur was assessed using pre- and post-resize Gaussian perturbations. Results: Task-specific models improved consistently with increasing data scale, whereas foundation models saturated at moderate sample sizes. A Vision Transformer optimized for small patches (CustomViT) achieved the highest accuracy, outperforming all foundation models with substantially lower inference cost. Blur robustness was comparable across architectures, with no qualitative advantage observed for foundation models. Conclusion: For cell-level classification under extreme spatial constraints, task-specific architectures are more effective and efficient than foundation models once sufficient training data are available. Higher clean accuracy does not imply superior robustness, and large pre-trained models offer limited benefit in the small-patch regime.

\noindent\textbf{Background} Prenatal glucocorticoid exposure alters cardiac development, but whether persistent cardiac effects in childhood follow a dose-response relationship remains unknown. We recently showed that ECG foundation models detect robust cardiac differences between steroid-exposed and control children, while traditional heart rate variability metrics lose significance after covariate adjustment. Here, we investigate the dose-response dimension using complementary analytical approaches. \noindent\textbf{Methods} We studied 49 children (ages 8--15) whose mothers received betamethasone during pregnancy for multiple sclerosis: 12 low-dose ({$<$}5\,g cumulative), 13 high-dose ({$\geq$}5\,g), and 24 controls. Five-minute ECG recordings during the Trier Social Stress Test yielded 251 observations. We computed 12 multiscale complexity features and tested 11 ECG foundation model (FM) dimensions using linear mixed models, Kruskal--Wallis tests with Dunn's post-hoc comparisons, Spearman correlations, and Jonckheere--Terpstra trend tests. \noindent\textbf{Findings} The binary exposed-versus-controls comparison showed no significant complexity effects ($p>0.39$). However, dose-based analysis revealed that high-dose children exhibited significantly faster entropy rate ($h$) decay rates than low-dose children ($p=0.031$); neither sample entropy nor approximate entropy decay rates reached significance ($p=0.18$ and $p=0.12$, respectively). Effects localized to the mental arithmetic stress segment (Kruskal--Wallis $p=0.005$; Dunn's $p=0.004$). A cross-condition robustness analysis confirmed that $h$ decay rate is invariant to input signal choice and normalization ($r>0.98$), while sample and approximate entropy are not. In contrast, the 11 FM dimensions showed weak dose-response evidence: only 1 of 22 covariate-adjusted contrasts survived FDR correction, with paradoxically stronger low-dose effects. \noindent\textbf{Interpretation} The entropy rate decay rate -- uniquely robust across input conditions -- reveals a dose-dependent effect on cardiac autonomic dynamics under cognitive stress, while FM dimensions detect a dose-independent morphological ``exposure fingerprint.'' These exploratory findings suggest a two-component model of prenatal glucocorticoid cardiac programming -- ~morphological (dose-independent) and dynamical (dose-dependent)~ -- providing more complete characterization than either approach alone. Given the small sample size, these results should be considered hypothesis-generating and require replication in larger cohorts.

While EEG features differentiate Major Depressive Disorder (MDD) from healthy controls (HC), their clinical utility as biomarkers depends on a monotonic trajectory across the disease spectrum, from the acute (AC) phase to the maintenance (MA) phase and finally to the healthy baseline. However, the progression of the MA phase remains poorly understood in traditional marker analysis. Analyzing EEG data from 74 individuals (24 AC, 23 MA, and 27 HC), this study provides a comprehensive evaluation of classic ERP and resting-state indices across AC, MA, and HC groups. Our results demonstrate that almost no conventional metrics strictly satisfy the criterion of monotonic progression, likely due to profound inter-individual heterogeneity. These findings highlight the inherent limitations of group-level feature extraction and provide critical insights for developing future paradigms and algorithms to identify neurobiological markers with genuine clinical utility.

The social structures of group-living animals such as feral horses are diverse and remain insufficiently understood, even within a single species. To investigate group dynamics, aerial videos are often utilized to track individuals and analyze their movement trajectories, which are essential for evaluating inter-individual interactions and comparing social behaviors. Accurate individual tracking is therefore crucial. In multi-animal tracking, axis-aligned bounding boxes (bboxes) are widely used; however, for aerial top-view footage of entire groups, their performance degrades due to complex backgrounds, small target sizes, high animal density, and varying body orientations. To address this issue, we employ oriented bounding boxes (OBBs), which include rotation angles and reduce unnecessary background. Nevertheless, current OBB detectors such as YOLO-OBB restrict angles within a 180$^{\circ}$ range, making it impossible to distinguish head from tail and often causing sudden 180$^{\circ}$ flips across frames, which severely disrupts continuous tracking. To overcome this limitation, we propose a head-orientation estimation method that crops OBB-centered patches, applies three detectors (head, tail, and head-tail), and determines the final label through IoU-based majority voting. Experiments using 299 test images show that our method achieves 99.3% accuracy, outperforming individual models, demonstrating its effectiveness for robust OBB-based tracking.

In animal facilities, continuous surveillance of penguins is essential yet technically challenging due to their homogeneous visual characteristics, rapid and frequent posture changes, and substantial environmental noise such as water reflections. In this study, we propose a framework that enhances both detection and identification performance by integrating appearance and motion features. For detection, we adapted YOLO11 to process consecutive frames to overcome the lack of temporal consistency in single-frame detectors. This approach leverages motion cues to detect targets even when distinct visual features are obscured. Our evaluation shows that fine-tuning the model with two-frame inputs improves mAP@0.5 from 0.922 to 0.933, outperforming the baseline, and successfully recovers individuals that are indistinguishable in static images. For identification, we introduce a tracklet-based contrastive learning approach applied after tracking. Through qualitative visualization, we demonstrate that the method produces coherent feature embeddings, bringing samples from the same individual closer in the feature space, suggesting the potential for mitigating ID switching.

In population genetics, mutation rate is often treated as a homogeneous parameter across the genome. Empirical evidence, however, shows systematic variation across genomic contexts associated with chromatin organization and epigenomic features. Using gene-level de novo mutation data from Arabidopsis thaliana, we test whether chromatin features predict not only the mean per-base mutation rate but also its variability across genes. To reduce heterogeneity in selective regime, we restrict analysis to essential and lethal loci subject to strong purifying selection. Across complementary multivariable models including heteroskedasticity-robust linear regression, length-weighted regression, and Poisson generalized linear models with exposure offsets, histone marks associated with active transcription (H3K4me1, H3K4me3, H3K36ac) are consistently associated with lower mean mutation rates and substantially reduced between-gene variance. GC content shows little association with the mean once chromatin predictors are controlled but is positively associated with mutation-rate variability. Estimates of skewness and kurtosis reveal no significant higher-order structure attributable to epigenomic predictors. A standardized Tajima's $D$ statistic yields directionally consistent but statistically underpowered associations with both the mean and variance of gene-level mutation rates. These results indicate that mutation rate is systematically structured by chromatin state within functionally constrained genes and suggest that evolutionary processes may act not only on expected mutation rate but also on its variability across loci.

Estimating the effective sample size (ESS) is fundamental in Bayesian phylogenetic inference to properly account for autocorrelation in MCMC samples. While methods for continuous parameters are well established, the discrete and high-dimensional nature of treespace poses substantial challenges. Here, we compare existing tree ESS estimators with novel approaches that leverage tractable tree distributions, specifically Conditional Clade Distributions (CCDs), as well as a new probabilistic estimator based on clade frequency differences between independent chains. Using simulated chains with known ESS bounds, we assess estimator accuracy and evaluate their stability and robustness on simulated and real datasets. We further examine how multimodality in posterior distributions and poor mixing can substantially affect ESS estimates, highlighting the need for careful interpretation. Our CCD-based estimators perform comparably to existing approaches, with two methods showing lower variance by averaging across multiple estimates. In contrast, the probabilistic estimator and two previously recommended methods incur prohibitive computational costs for long chains. Together, these results provide guidance for reliable and efficient tree ESS estimation in complex phylogenetic analyses.

We propose that the jagged intelligence landscape of modern AI systems arises from a missing training signal that we call "cognitive dark matter" (CDM): brain functions that meaningfully shape behavior yet are hard to infer from behavior alone. We identify key CDM domains-metacognition, cognitive flexibility, episodic memory, lifelong learning, abductive reasoning, social and common-sense reasoning, and emotional intelligence-and present evidence that current AI benchmarks and large-scale neuroscience datasets are both heavily skewed toward already-mastered capabilities, with CDM-loaded functions largely unmeasured. We then outline a research program centered on three complementary data types designed to surface CDM for model training: (i) latent variables from large-scale cognitive models, (ii) process-tracing data such as eye-tracking and think-aloud protocols, and (iii) paired neural-behavioral data. These data will enable AI training on cognitive process rather than behavioral outcome alone, producing models with more general, less jagged intelligence. As a dual benefit, the same data will advance our understanding of human intelligence itself.

This paper addresses the estimation of a dynamic interaction network, a network of influence among individuals, under projected visual stimuli to quantify the influences of inter-individual interactions and external stimuli on collective behavior. Building upon our previously proposed network estimation model, which assumes a Boids-type model and employs a sparse regression framework to infer inter-individual influence networks from trajectory data, we extend the formulation by introducing a stimulus term. This enables the model to capture how individuals react to and propagate externally projected visual stimuli within the group. The resulting framework allows simultaneous estimation of inter-individual and stimulus-related interaction strengths. We also introduce entropy-based indices to capture the possible biases of individuals' influence. Our experiments with fish schools under projector-based visual stimuli demonstrate the effectiveness of the proposed indices in quantifying schooling behavior and identifying influential individuals within the group, serving as the basis for real-time, interpretable metrics of collective dynamics.

The Bayesian brain hypothesis has been a leading theory in understanding perceptual decision-making under uncertainty. While extensive psychophysical evidence supports the notion of the brain performing Bayesian computations, how uncertainty information is encoded in sensory neural populations remains elusive. Specifically, two competing hypotheses propose that early sensory populations encode either the likelihood function (exemplified by probabilistic population codes) or the posterior distribution (exemplified by neural sampling codes) over the stimulus, with the key distinction lying in whether stimulus priors would modulate the neural responses. However, experimentally differentiating these two hypotheses has remained challenging, as it is unclear what task design would effectively distinguish the two. In this work, we present an information-theoretic framework for optimizing the task stimulus distribution that would maximally differentiate competing probabilistic neural codes. To quantify how distinguishable the two probabilistic coding hypotheses are under a given task design, we derive the information gap--the expected performance difference when likelihood versus posterior decoders are applied to neural populations--by evaluating the Kullback-Leibler divergence between the true posterior and a task-marginalized surrogate posterior. Through extensive simulations, we demonstrate that the information gap accurately predicts decoder performance differences across diverse task settings. Critically, maximizing the information gap yields stimulus distributions that optimally differentiate likelihood and posterior coding hypotheses. Our framework enables principled, theory-driven experimental designs with maximal discriminative power to differentiate probabilistic neural codes, advancing our understanding of how neural populations represent and process sensory uncertainty.

The brain achieves stability and plasticity in a topologically complex, shifting world through Metric-Topology Factorization (MTF), separating discrete topological indexing for context selection from continuous metric condensation for local inference. Semantically rich environments defy single globally contractive geometries, causing obstructions under shifts, so intelligence factorizes these: the hippocampus provides sparse signatures indexing manifold identity, while the neocortex untangles geometry hierarchically. In the ventral stream, a dynamic-programming-like process quotients symmetries (e.g., translation, scale), transforming non-convex sensory mazes into separable bowls. Offline replay and consolidation amortize transformations for rapid task switching. Dreaming in REM involves stochastic hippocampal traversal to expose and regularize latent structures. Consciousness arises from resolving topological uncertainty into stable embeddings, with awareness for unamortized states. Evolutionarily, transitions like sensorimotor control to language expand topological complexity, demanding advanced indexing-metric separation. Intelligence emerges via recalibrating context-specific geometries, converting global navigation into local dynamics, not deeper search.

Low Earth Orbit (LEO) satellite constellations in the 6G era are evolving into intelligent in-orbit computational platforms, forming Space Computing Power Networks (SCPNs) to deliver global-scale computing services. However, the intensive computation within SCPN incurs a significant ``unseen cost'': the frequent charge-discharge cycles accelerate the physical degradation of satellites' life-limiting and high-cost batteries, thereby threatening the long-term operational viability of such a system. Existing approaches, often relying on indirect metrics like Depth of Discharge (DoD) and neglecting the complex, nonlinear degradation process of battery aging, fail to accurately quantify this cost. To address this, we introduce a high-fidelity, physics-driven model that quantitatively links computational workload parameters to the nonlinear battery degradation. Building on this model, we formulate a degradation-aware scheduling problem and analyze heuristic policies across different energy regimes. Simulations reveal that the optimal strategy should be adaptive: in solar-rich conditions, a myopic policy maximizing instantaneous solar utilization is superior, whereas under energy scarcity, a reactive policy leveraging real-time battery state significantly extends lifetime.

Low Earth orbit (LEO) satellite constellations have become a critical enabler for global coverage, utilizing numerous satellites orbiting Earth at high speeds. By decomposing complex network services into lightweight service functions, network function virtualization (NFV) transforms global network services into diverse service function chains (SFCs), coordinated by resource-constrained LEOs. However, the dynamic topology of satellite networks, marked by highly variable inter-satellite link delays, poses significant challenges for designing efficient routing strategies that ensure reliable and low-latency communication. Many existing routing methods suffer from poor scalability and degraded performance, limiting their practical implementation. To address these challenges, this paper proposes a novel SFC routing approach that leverages the statistical properties of network link states to mitigate instability caused by instantaneous modeling in dynamic satellite networks. Through comprehensive simulations on end-to-end shortest-path propagation delays in LEO networks, we identify and validate the statistical stability of multi-hop routes. Building on this insight, we introduce the Stability-Aware Multi-Stage Graph Routing (SA-MSGR) algorithm, which incorporates pre-computed average delays into a multi-stage graph optimization framework. Extensive simulations demonstrate the superior performance of SA-MSGR, achieving significantly lower and more predictable end-to-end SFC delays compared to representative baseline strategies.

The Unscented Kalman Filter (UKF) is a ubiquitous tool for nonlinear state estimation; however, its performance is limited by the static parameterization of the Unscented Transform (UT). Conventional weighting schemes, governed by fixed scaling parameters, assume implicit Gaussianity and fail to adapt to time-varying dynamics or heavy-tailed measurement noise. This work introduces the Meta-Adaptive UKF (MA-UKF), a framework that reformulates sigma-point weight synthesis as a hyperparameter optimization problem addressed via memory-augmented meta-learning. Unlike standard adaptive filters that rely on instantaneous heuristic corrections, our approach employs a Recurrent Context Encoder to compress the history of measurement innovations into a compact latent embedding. This embedding informs a policy network that dynamically synthesizes the mean and covariance weights of the sigma points at each time step, effectively governing the filter's trust in the prediction versus the measurement. By optimizing the system end-to-end through the filter's recursive logic, the MA-UKF learns to maximize tracking accuracy while maintaining estimation consistency. Numerical benchmarks on maneuvering targets demonstrate that the MA-UKF significantly outperforms standard baselines, exhibiting superior robustness to non-Gaussian glint noise and effective generalization to out-of-distribution (OOD) dynamic regimes unseen during training.

The KID_READOUT board, developed for the CONCERTO millimeter-wave astronomy instrument, implements FPGA-based digital frequency multiplexing to read out large arrays of Microwave Kinetic Inductance Detectors (MKIDs). The complexity of the implemented multirate DSP chain, which combines tones synthesis, interpolation, digital frequency translation, polyphase filter-bank (PFB) channelization, and digital down-conversion (DDC), makes analytical performance optimization difficult. To address this, we developed a cycle-and bitaccurate, Python-based digital twin of the FPGA readout firmware DSP chain. Using this model, we identified the origin of previously measured and unexplained spurs in the readout channels, tracing them to periodicity mismatches between the excitation and analysis paths and to insufficient suppression of negative-frequency components by the DDC filters. Based on these insights, we implemented a mitigation strategy that aligns the periodicities and improves the DDC filter characteristics, effectively eliminating the spurs with a minor increase in FPGA resource usage.

Edema is a potential indicator of underlying pathological changes. However, its low-contrast signature is often masked in conventional B-mode imaging by strong scatterers, making reliable detection challenging. Ultrasound (US) provides a non-invasive, non-ionizing, and cost-efficient imaging option that is widely used. Conventional techniques, which rely on beamforming, often lack sufficient physical interpretability. Quantitative US (QUS) can estimate physical properties such as the speed of sound (SoS) and density by solving a physics-based inverse problem directly on the measured US wavefields, i.e., the raw per-element channel data (CD), to recover their spatial distribution. However, state-of-the-art physics-based inversion methods, including full waveform inversion (FWI) and model-based quantitative radar and US (MB-QRUS), are computationally intensive and susceptible to local minima, which constrains their clinical utility. We introduce deep unfolded FWI (DUFWI), a physics-faithful unfolded iterative inversion method that exhibits FWI-like refinement behavior while learning the update rule from data, requiring only a small number of iterations for real-time SoS reconstruction. Across both simulated datasets and hardware measurements acquired with a Verasonics US system, the DUFWI significantly outperforms classical FWI and MB-QRUS in reconstruction quality while maintaining high computational efficiency. These results demonstrate real-time edema diagnosis in both simulation and hardware experiments, with phantom-based validation using cylindrical rods, supporting practical deployment under typical US imaging setting.

The cross-domain oceanic connectivity ranging from underwater to the sky has become increasingly indispensable for a plethora of data-consuming maritime applications, such as maritime meteorological monitoring and offshore exploration. However, broadband implementations can be severely hindered by the isolation from terrestrial networks, limited satellite resources, and the fundamental inability of radio waves to bridge the water-air interface at high rates. To this end, this paper introduces an optical network bridging underwater, air and near space, which features a number of cooperative low-altitude platforms (LAPs), serving as compute-capable, sensing-aware, and mission-adaptive agents. The network architecture consists of three scenario-specific segments, i.e., water-air direct link, low-altitude mesh network, and the near-space access network. With coordinate sensing and intelligent control, the system tightly couples beam tracking and resource optimization, enabling resilient networking under high mobility and harsh maritime dynamics. Furthermore, we review enabling technologies spanning from water-air channel modeling, adaptive beam alignment under sea-surface perturbations, to swarm-intelligence networking for decentralized control, integrated pose-topology planning, and optical Integrated sensing and communication (ISAC) for near-space target detection and beam alignment. Finally, open issues are also highlighted, constituting a clear roadmap toward scalable, secure, and ultra-broadband oceanic optical networks.

The integration of artificial intelligence as a design tool for metasurfaces, and the implementation of a deep-learning model pose a challenge in the development of an automated solution due to high resources requirements. The presented work introduces a network-layer solution to configure such environment for end user objectives, and for an underlying physical-layer technology. An architecture is developed to design an anomalous reflector by employing the Redhat Openshift AI (RHOAI) technology to support an automated machine learning operations (MLOps) framework in smart radio environments. This entails the design of lossless impenetrable metasurfaces characterized by a scalar surface impedance for an optimal anomalous reflection, achieved by optimizing the number of the Floquet modes through the utilization of a local power conservation constraint qualified as a fitness function. The metasurfaces design process is implemented by using a conditional generative adversarial network (cGAN). An extended cGAN with a surrogate model assists in a high-quality freeform metasurfaces design, where it introduces a swift simulation tool for the metasurfaces design process and analysis of the far-field model. The paper focuses on the challenges of building such a system, and potential abstraction layers. The training accuracy value of the proposed model demonstrates the feasibility and benefits of deploying in containerized environment of Red Hat Openshift in comparison with other deployments of ResNet-50 reported in literature.

In cell-free massive MIMO, centralized precoding is {theoretically known} to {remarkably} outperform its distributed counterparts, albeit {with} high implementation complexity. However, this letter highlights a practical limitation {often overlooked:} {widely used closed-form} centralized {precoders} are typically derived under a sum-power constraint, which often demands unrealistic power allocation that exceeds hardware capabilities. {When two simple heuristics (global power scaling and local normalization) are applied to enforce the per-AP instantaneous power constraint}, the centralized performance superiority disappears, making distributed precoding {a robust option}.

Continuous efforts have been devoted to integrate millimeter wave (mmWave) and terahertz (THz) bands into future communication standards in order to overcome the bandwidth shortage problem and achieve high data rates, primarily through developing accompanying technologies that can overcome the severe propagation loss and blockage associated with increased carrier frequency. One of the most notable accompanying technologies is reconfigurable intelligent surface (RIS), which uses a large number of low-cost passive reflecting elements to reconfigure the propagation environments for improved communication performance and coverage. Despite its numerous benefits, RIS can make channel estimation more difficult due to its lack of radio frequency (RF) chains that can perform baseband signal processing. In addition, the cascaded channel structure of RIS-aided communication systems, which differs from that in conventional systems, brings about significant challenges in both channel estimation and beamforming. In this paper, we propose the joint channel estimation and beamforming optimization algorithm for RIS-aided multiple-input multipleoutput (MIMO) communication systems. By carefully exploiting the angular sparsity of mmWave/THz channels, our proposed algorithm successfully designs the RIS matrices that not only facilitate the channel estimation process but also achieve the passive beamforming gain through increased channel capacity. Simulation results demonstrate that our proposed algorithm provides the systems of interest with significant improvement in spectral efficiency.

Testing Ultra-Wideband (UWB) systems is challenging, as multiple devices need to coordinate over lossy links and the systems' behavior is influenced by timing, synchronization, and environmental factors. Traditional testing is often insufficient to capture these complex interactions, highlighting the need for an overarching testbed infrastructure that can manage devices, control the environment, and make measurements and test scenarios repeatable. In this work, we present a highly automated testbed architecture built on Robot Operating System Version 2, integrating device management with environmental control and measurement systems. It includes an optical reference system, a controllable Autonomous Guided Vehicle to position devices within the environment, and time synchronization via Network Time Protocol (NTP). The testbed achieves a Root Mean Squared Error of 4.8 mm for positioning repeatability and 0.493$°$ for the orientation, and our NTP-based synchronization approach achieves a timing accuracy of below 1 ms. All testbed functionality can be controlled remotely through simple Python scripts to allow automated orchestration tasks such as conducting complex measurement scenarios. We demonstrate this with a measurement campaign on UWB localization, showing how it enables repeatable, observable, and fully controlled wireless experiments.

Integrated sensing and communication (ISAC) techniques can leverage existing, wide-coverage communication networks to perform sensing tasks, enabling large-scale and low-cost target sensing. However, the inherent randomness of communication data payloads introduces undesired sidelobes in the ambiguity function that may degrade target detection and parameter estimation performance. This paper develops a communication-centric ISAC framework that is standards-compliant and compatible with existing devices. Specifically, we propose a low-complexity constellation selection scheme over a finite, off-the-shelf alphabet, achieving an efficient sensing-communication trade-off without custom waveforms or frame-structure changes. To this end, we analyze two classical sensing receivers including matched filtering (MF) and reciprocal filtering (RF) for ranging measurements, and derive closed-form sensing laws that link constellation statistics to sensing performance. Under any finite-alphabet constellation combination, MF sidelobes depend on the weighted sum of the kurtosis values of the per-subcarrier constellations, while RF noise enhancement depends on the inverse second moment of the transmit symbol, providing a tractable expression for tuning the sensing-communication trade-off. The analysis extends to multi-symbol coherent integration and achieves the expected processing gain. We prove that in flat-fading channels, any Pareto-optimal solution activates no more than three constellations. For frequency-selective channels, a bilevel algorithm with closed-form inner updates attains near-optimal performance while sharply reducing computational complexity. We validate the entire theoretical pipeline with numerical simulations as well as experimental results.

A transmit pinching-antenna system (T-PASS) framework is proposed, in which a single pinched waveguide is employed to jointly serve one wired user equipment (UE) and multiple wireless UEs. The signal radiated by the pinching antennas (PAs) is used to serve the wireless UEs, whereas the residual guided signal at the waveguide termination is used to serve the wired UE. To facilitate T-PASS transmission and mitigate inter-user interference, a hybrid non-orthogonal multiple access (NOMA) scheme is introduced. Wireless UEs are scheduled by time-division multiple access (TDMA), and, in each slot, the scheduled wireless UE is paired with the wired UE through power-domain NOMA. Within this framework, the PA positions, PA radiation coefficients, power allocation, and TDMA time-slot allocation are jointly optimized to maximize a weighted sum rate (WSR). i) For the two-user case with one wired UE and one wireless UE, the optimal PA position and successive interference cancellation (SIC) decoding order are derived. Closed-form optimal power allocation is obtained, and a near-optimal PA radiation coefficient is determined through a low-complexity one-dimensional search. ii) For the multiuser case with one wired UE and multiple wireless UEs, four protocols with different PA-position and PA-radiation configurations are proposed. For each protocol, a low-complexity element-wise alternating optimization algorithm is developed to optimize the PA positions and radiation coefficients, while closed-form solutions are derived for the optimal power allocation and time-slot allocation. Numerical results are presented to show that: i) under typical T-PASS configurations, the wired UE is selected as the strong user in the optimal SIC decoding order; ii) the proposed T-PASS framework achieves a significantly higher WSR than conventional wireless-only PASS.

Hybrid Automatic Repeat Request (HARQ) schemes typically allocate all available resources to retransmit failed packets to ensure reliability. However, under stringent delay constraints, these schemes often exhibit low spectral efficiency and increased transmission latency. To address these challenges, this paper proposes an efficient Non-Orthogonal HARQ with Chase Combining (N-HARQ-CC) transmission strategy. Specifically, the proposed approach allocates a larger portion of retransmission resources to new data packets, reserving only a small fraction for retransmitting previously erroneous packets. This is based on the observation that only a small number of information bits are typically incorrect, enabling surplus communication resources to be utilized for transmitting new messages. The N-HARQ-CC scheme retransmits the same redundant version of a failed packet and employs Maximum Ratio Combining (MRC) for decoding. To minimize complex packet scheduling and decoding complexity, the proposed scheme limits superposition to at most two messages per transmission round. At the receiver, Successive Interference Cancellation (SIC) is used to decouple the superimposed messages. The proposed N-HARQ-CC system was implemented using GNU Radio and USRP platforms for validation. Compared to conventional Type-I HARQ and HARQ-CC schemes, the proposed scheme achieves a significant improvement in spectral efficiency of approximately 0.5 bps/Hz, aligning with the low-latency requirements of 6G networks.

The rapid deployment of large-scale low Earth orbit (LEO) satellite constellations has positioned direct-to-handset (D2H) communications as a key enabler of future non-terrestrial networks. However, the limited link budget of handheld devices makes broadband service delivery challenging, and multi-satellite cooperative transmission is often required to provide sufficient power gain. In practice, such cooperation is severely hindered by asynchronous reception across satellites. This paper analyzes the received-signal model under the 3rd Generation Partnership Project (3GPP) transmitter structure and shows that satellite-dependent propagation delays prevent simultaneous timing alignment for multiple user terminals (UTs). This timing mismatch induces severe inter-carrier interference (ICI) and inter-symbol interference (ISI), even from the intended signals, which fundamentally constrains the achievable cooperative gain. To address this issue, we propose a timing-aware satellite association strategy that enables cooperation only with satellites expected to satisfy a UT-side timing tolerance, thereby avoiding dominant asynchronous interference by design. Simulation results demonstrate that the proposed strategy improves throughput performance compared to single-satellite transmission and fully connected multi-satellite baselines.

Orthogonal time frequency space (OTFS)/vector OFDM (VOFDM) is widely regarded as a promising waveform for next-generation mobile communications. However, its spectral characteristics are not yet fully understood. The bandwidth allocation scheme, which is crucial for OTFS's integration into practical wireless standards, also remains unexplored. In this paper, we investigate the spectral characteristics of OTFS signals by analyzing their power spectral density (PSD). We demonstrate that the PSD of discrete-time OTFS signals is periodic with a period of $\frac{1}{MT_s}$, where $M$ is the size of the time/Doppler domain in OTFS, a.k.a., the vector size in VOFDM, and $T_s$ is the sampling interval length of digital to analog converter (DAC), resulting in $M$ identical spectral components within the spectral range $[-\frac{1}{2T_s}, \frac{1}{2T_s})$ of the continuous-time OTFS signal. The periodicity makes bandwidth allocation for OTFS/VOFDM signaling substantially challenging. Furthermore, we establish a relationship between the PSD of OFDM signals and that of OTFS signals, revealing that, when the information symbols are independent, the PSD of OTFS signals is equal to the sum of the PSDs of the component-expanded OFDM (CEP-OFDM) signals. Lastly, we derive a relationship between the information symbols and the corresponding OTFS spectrum, and based on which, we propose a null-space-based linear precoding (NSLP) method for OTFS signals to enable flexible bandwidth allocation. Numerical results validate our analytical results regarding the PSD of OTFS signals and show the effectiveness of our proposed NSLP method in tailoring the spectrum of OTFS signals.

In massive multiple-input multiple-output (MIMO) systems, achieving high spectral efficiency (SE) often requires advanced precoding algorithms whose complexity scales rapidly with the number of antennas, limiting practical deployment. In this paper, we develop a scalable and computationally efficient generalized power iteration precoding (GPIP) framework for massive MIMO systems under both perfect and imperfect channel state information at the transmitter (CSIT). By exploiting the low-dimensional subspace property of optimal precoders, we reformulate the high-dimensional beamforming problem into a lower-dimensional weight optimization that scales with the number of users rather than antennas. We further extend this framework to the imperfect CSIT scenario by showing that stationary solutions reside in a combined subspace spanned by the estimated channel and error covariance matrices, enabling a robust design via low-rank approximation. To reduce computational cost, we leverage the Sherman-Morrison formula to simplify matrix inversions. Moreover, interpreting the GPIP update as a projected preconditioned gradient ascent method, we establish convergence guarantees and develop a stable and monotonic algorithm using a backtracking line search. Numerical results demonstrate that the proposed methods achieve the highest SE performance compared to state-of-the-art linear precoders with significantly reduced complexity and convergence, highlighting their suitability for large-scale MIMO systems.

Sensing is an integral part of 6G and beyond systems, providing exceptional environmental perception along with communication. RF-based sensing often relies on simplified geometric assumptions (e.g., point scatterers or planar surfaces) to model specular multipath and keep inference tractable. However, such representations are not physically informative and fail to accurately capture extended objects with complex shapes and properties. This paper presents a probabilistic occupancy grid framework for radio-based simultaneous localization and mapping (SLAM), jointly reconstructing geometric structures and their radio-related properties. The proposed occupancy grid map representation is integrated into a multipath-based SLAM (MP-SLAM) formulation to enable simultaneous mobile-agent localization and environment mapping using multipath measurements. To connect RF measurements with the grid map, a surface model is employed to describe candidate reflection paths, while occupancy grid cell states capture measurement uncertainties and fine--grained geometric details. Object RF-related properties are modeled via reflection coefficients. The proposed framework offers a principled, proof-of-concept approach to physically interpretable radio-based mapping, and simulation results demonstrate accurate reconstruction of geometry and material properties, as well as high-accuracy localization. In addition, the results highlight the potential to use prior occupancy maps obtained from other radio devices or complementary sensors for subsequent map extension and refinement.

Cyber-security systems often operate in resource-constrained environments, such as edge environments and real-time monitoring systems, where model size and inference time are crucial. A light-weight intrusion detection framework is proposed that utilizes the Kolmogorov-Arnold Network (KAN) to capture complex features in the data, with the efficiency of decoupled knowledge distillation (DKD) training approach. A high-capacity KAN network is first trained to detect attacks performed on the test bed. This model then serves as a teacher to guide a much smaller multilayer perceptron (MLP) student model via DKD. The resulting DKD-MLP model contains only 2,522 and 1,622 parameters for WADI and SWaT datasets, which are significantly smaller than the number of parameters of the KAN teacher model. This is highly appropriate for deployment in resource-constrained devices with limited computational resources. Despite its low size, the student model maintains a high performance. Our approach demonstrate the practicality of using KAN as a knowledge-rich teacher to train much smaller student models, without considerable drop in accuracy in intrusion detection frameworks. We have validated our approach on two publicly available datasets. We report F1-score improvements of 4.18% on WADI and 3.07% on SWaT when using the DKD-MLP model, compared to the bare student model. The implementation of this paper is available on our GitHub repository.

Detecting speech from biosignals is gaining increasing attention due to the potential to develop human-computer interfaces that are noise-robust, privacy-preserving, and scalable for both clinical applications and daily use. However, most existing approaches remain limited by insufficient wearability and the lack of edge-processing capabilities, which are essential for minimally obtrusive, responsive, and private assistive technologies. In this work, we present SilentWear, a fully wearable, textile-based neck interface for EMG signal acquisition and processing. Powered by BioGAP-Ultra, the system enables end-to-end data acquisition from 14 differential channels and on-device speech recognition. SilentWear is coupled with SpeechNet, a lightweight 15k-parameter CNN architecture specifically tailored for EMG-based speech decoding, achieving an average cross-validated accuracy of 84.8$\pm$4.6% and 77.5$\pm$6.6% for vocalized and silent speech, respectively, over eight representative human-machine interaction commands collected over multiple days. We evaluate robustness to repositioning induced by multi-day use. In an inter-session setting, the system achieves average accuracies of 71.1$\pm$8.3% and 59.3\pm2.2% for vocalized and silent speech, respectively. To mitigate performance degradation due to repositioning, we propose an incremental fine-tuning strategy, demonstrating more than 10% accuracy recovery with less than 10 minutes of additional user data. Finally, we demonstrate end-to-end real-time on-device speech recognition on a commercial multi-core microcontroller unit (MCU), achieving an energy consumption of 63.9$μ$J per inference with a latency of 2.47 ms. With a total power consumption of 20.5mW for acquisition, inference, and wireless transmission of results, SilentWear enables continuous operation for more than 27 hours.

We investigate the effects, stability, and nonlinear dynamics of ghost scalar matter modeled as a field with a negative kinetic term confined within the cores of neutron stars. To this end, we analyze static configurations of the coupled Einstein-Euler-(ghost, complex) Klein-Gordon system and then we perform fully dynamical numerical evolutions of illustrative cases. Our results demonstrate that neutron stars can gravitationally confine a finite amount of ghost matter and support continuous families of equilibrium solutions, indicating that these configurations are not the result of fine tuning. We analyze the properties of the final states and find that the neutron star undergoes a persistent pulse-like oscillatory motion. In particular, we explicitly compute the frequency synchronization between the stellar fluid oscillation modes and those of the ghost scalar sector.

Hybrid continuous-variable (CV)-discrete-variable (DV) quantum systems present a promising direction for quantum computing by combining the high dimensional encoding capabilities of qumodes with the control offered by DV qubits on the coupled qumodes. There have been exciting recent progresses on hybrid CV-DV quantum computing, including variational algorithms, error correction, compiler-level optimizations for Hamiltonian simulation, etc. However, there is a lack of a standardized CV-DV benchmark suite for assessing various emerging hardware platforms and evaluating software optimizations on hybrid CV-DV circuits. In this work, we introduce a simulation and benchmarking framework for hybrid CV-DV circuits, implemented using Bosonic Qiskit-a tool specifically designed to model CV-DV systems, along with QuTip for functional correctness verification. We construct and characterize representative CV-DV benchmarks, including cat state generation, GKP state generation, CV-DV state transfers, hybrid quantum Fourier transform, variational quantum algorithms, Hamiltonian simulation, and Shor's algorithm. To assess circuit complexity and scalability, we define a feature map organized into two categories: general features (e.g., qubit/qumode count, gate counts) and CV-DV-specific features (e.g., Wigner negativity, energy, truncation cost). These metrics enable evaluation of both classical simulability and hardware resource requirements. Our results, including one benchmark on real hardware, demonstrate that hybrid CV-DV architectures are not only viable but well-suited for a range of computational tasks, from optimization to Hamiltonian simulation. This framework lays the groundwork for systematic evaluation and future development of hybrid quantum systems.

As quantum computing hardware rapidly advances, objectively evaluating the capabilities and error rates of new processors remains a critical challenge for the field. A clear and realistic understanding of current quantum performance is essential to guide research priorities and drive meaningful progress. In this work, we apply and extend a protocol-based benchmarking methodology (presented in arXiv:2505.12441) that utilizes well-defined quantumness thresholds. By evaluating performance at protocol level rather then the gate level, this approach provides a transparent and intuitive assessment of whether specific quantum processors, or isolated sub-chips within them, can demonstrate a practical quantum advantage. To illustrate the utility of this method, we compare two generations of IBM quantum computers: the older Eagle architecture and the newer Heron architecture. Our findings reveal the genuine operational strengths and limitations of these devices, demonstrating substantial performance improvements in the newer Heron generation.

Hermiticity is usually treated as a foundational axiom of quantum mechanics, guaranteeing real spectra and unitary time evolution. In this work we argue that Hermiticity is more naturally understood as a symmetry law arising from the global conservation of an inner product current. We show that in spacetimes admitting complete Cauchy surfaces without boundary flux this conservation reduces to the familiar Hermiticity condition of the canonical inner product. However, in the presence of causal horizons, most strikingly in black hole geometries, this conservation law becomes obstructed for restricted observers. Tracing over inaccessible degrees of freedom then inevitably yields completely positive trace preserving dynamics with an effective non-Hermitian generator. Using quantum thermodynamics and the monotonicity of relative entropy, we demonstrate that the generalized second law may be reinterpreted as an entropy balance that compensates precisely for the flux of inner product charge through the horizon. The structure of Einstein equations, through the Bianchi identity and the Raychaudhuri focusing equation, provides the geometric mechanism underlying this balance. We also show that black hole ringdown can serve as a realistic observational probe of this idea and may provide quantitative upper bounds on the strength of horizon-induced inner product flux. In this way gravity, entropy production, and effective non-Hermiticity are unified under a single structural principle, with Hermiticity emerging as the special case of globally conserved inner product symmetry.

We investigate the propagation of two-point density correlations in the one-dimensional Bose-Hubbard Hamiltonian in the thermodynamic limit in terms of the correlation transport distance (CTD), an experimentally measurable magnitude that characterizes the spatial spreading of correlations in time. We confirm that the integrable limits of the model exhibit CTD ballistic growth, while the onset of the chaotic phase leads to the emergence of a pronounced sub-ballistic regime, in agreement with previous results for finite systems. By a meticulous analysis of the spatio-temporal correlation profiles, we show that the correlation front nonetheless propagates ballistically for all interaction strengths, and that the chaos-induced slowdown of the CTD originates from the emergence of long-time distance-dependent correlation tails, together with an enhanced decay of the correlation front amplitude. Our results thus provide a detailed characterization of correlation transport that goes beyond a simple light-cone picture.

This paper focuses on error thresholds for Pauli channels. We numerically compute lower bounds for the thresholds using the analytic framework of coset weight enumerators pioneered by DiVincenzo, Shor and Smolin in 1998. In particular, we study potential non-additivity of a variety of small stabilizer codes and their concatenations, and report several new concatenated stabilizer codes of small length that show significant non-additivity. We also give a closed form expression of coset weight enumerators of concatenated phase and bit flip repetition codes. Using insights from this formalism, we estimate the threshold for concatenated repetition codes of large lengths. Finally, for several concatenations of small stabilizer codes we optimize for channels which lead to maximal non-additivity at the hashing point of the corresponding channel. We supplement these results with a discussion on the performance of various stabilizer codes from the perspective of the non-additivity and threshold problem. We report both positive and negative results, and highlight some counterintuitive observations, to support subsequent work on lower bounds for error thresholds.

Quantum simulations on current NISQ hardware are limited by its noisy nature, making efficient quantum error mitigation methods highly demanded. In this paper we introduce a novel mitigation scheme, applicable to arbitrary quantum simulations of time-dependent Hamiltonian dynamics on NISQ devices. The scheme uses a polynomial subset of extended qubit Bogoliubov-Born-Green-Kirkwood-Yvon (BBGKY) hierarchy equations as a sampling criterion of possible mitigated candidates for the quantum observables. We show that for favorable Hamiltonians the polynomial subset of BBGKY hierarchy equations leads to a polynomial overhead in both classical and quantum resources. We employ the method to mitigate simulations of the chiral magnetic effect (CME), a chiral feature of the Schwinger model. We empirically show the effectiveness of our scheme at recovering the real-time dynamics of the CME from noisy quantum simulations of the Schwinger model, for a range of different parameter values of the model. We numerically demonstrate a systematic reduction of quantum noise, together with an increasing noise reduction capability as the amount of BBGKY constraints grows.

The modified Langevin noise formalism (MLNF) models the interaction of the quantized electromagnetic field with an arbitrary lossy magneto-dielectric object placed in vacuum using three types of non-interacting bosonic polaritons: scattering, electric, and magnetic. These respectively represent free-space photons scattered by the object, and photons radiated by quantized electric and magnetic dipolar sources embedded within its volume. Recently [A. Ciattoni, Phys. Rev. A 110, 013707 (2024)], this formalism was justified from the canonical quantization of macroscopic electromagnetism (CQME) [Philbin, New J. Phys. 12, 123008 (2010)] in the Heisenberg picture. This was achieved by identifying the polariton operators within the formal solution of the macroscopic Maxwell equations, assuming they obey bosonic commutation relations to retrieve the canonical ones, and showing they diagonalize the CQME Hamiltonian. However, the explicit functional dependence of these polaritons on the underlying canonical field operators remained undetermined. In this paper, we derive the exact analytical expressions for the polariton operators in terms of the canonical CQME field operators. Using these mappings, we provide a direct and rigorous derivation of the MLNF from the canonical theory in the Schrödinger picture. Our derivation is structured in three foundational steps: 1) adopting the derived analytical expressions as the constitutive definitions of the polariton operators; 2) mathematically proving that these operators are strictly bosonic as a direct consequence of the canonical commutation relations; and 3) demonstrating that they exactly diagonalize the macroscopic CQME Hamiltonian.

We investigate the intrinsic ambiguity in the definition of correlation functions arising from the inevitable invasiveness of quantum measurements. While algebraic correlations defined as expectation values of products of observables are widely used, their relationship to operational ones defined through actual measurement procedures remain unclear. We demonstrate that the differences among various definitions of correlation functions and those among their underlying (quasi-)joint probability distributions are bounded above by a quantitative measure of measurement invasiveness. We further obtain a lower bound on the discrepancy among operational and algebraic (quasi-)joint probability distributions, providing a new form of the uncertainty relation. In addition, we identify an equivalence condition under which operational and algebraic correlations coincide. As an application, we analyze the quantum violation of the Leggett-Garg inequality and clarify the structural origin of the equivalence among different approaches to observing the violation, including sequential projective measurements and weak-measurement. Our results provide an operational foundation for the commonly used algebraic concepts of quantum theory.

Single-minus tree-level $n$-graviton scattering amplitudes are revisited. Often presumed to vanish, they are shown here to be nonvanishing for certain "half-collinear" configurations existing in Klein space or for complexified momenta. A Berends-Giele recursion relation for these amplitudes is derived and solved in a form involving a sum over trees. In a restricted kinematic decay region, this solution simplifies significantly to an $(n{-}2)$-fold product of soft factors. It is further shown in this region that, combined with suitable analyticity assumptions, the $n$-graviton amplitude is generated by a recursive $\mathcal{L}w_{1+\infty}$ Ward identity with the three-graviton amplitude as a seed.

Magic states, by allowing non-Clifford gates through gate teleportation, are important building blocks of fault-tolerant quantum computation. Magic state distillation protocols aim to create clean copies of magic states from many noisier copies. However, the prevailing protocols require substantial qubit overhead. We present a distillation protocol based on permutation-invariant gnu codes, as small as two qubits. The two-qubit protocol achieves a 0.5 error threshold and 1/2 distillation rate, surpassing prior schemes for comparable codes. Our protocol furthermore distils magic states with arbitrary magic by varying the position of the ideal input states on the Bloch sphere. We achieve this by departing from the usual magic state distillation formalism, allowing the use of non-Clifford gates in the distillation protocol, and allowing the form of the output state to differ from the input state. Our protocol is compatible for use in tandem with existing magic state distillation protocols to enhance their performance.

Wave-optics effects in strongly lensed gravitational waves (GWs) provide a new interferometric probe of dark matter substructure. We compute the full diffraction integral for GWs propagating through statistically generated cold dark matter subhalo populations and quantify the resulting frequency-dependent amplification in the Laser Interferometer Space Antenna (LISA) band. We show that realistic galaxy-scale lenses generically produce percent-level amplitude and phase distortions in strongly magnified images, primarily induced by subhalos in the mass range $10^4$-$10^7\,M_{\odot}$. These signatures arise naturally within the standard cold dark matter paradigm and should be detectable in high signal-to-noise LISA events. Strongly lensed GWs thus offer a direct and complementary window on dark matter structure at subgalactic mass scales inaccessible to electromagnetic measurements.

We study the occurrence of long-lived metastable states in the 4f$^{13}$5d6s electron configuration of Yb$^+$. By optical pumping of a single trapped ion on the $^2F^\text{o}_{7/2}\rightarrow (7/2,0)_{7/2}$ transition at 377.5 nm, we prepare a wide range of metastable electronic states. We use a co-trapped control ion to sympathetically cool the spectroscopy ion, allowing us to accurately time its subsequent decay. We record a strong decay signal corresponding to a lifetime of 0.92(8) s, a weaker decay signal with lifetime 9.8(+2.9, -2.0) s, and find evidence for a much longer lifetime, $>$ 30 s. We identify the metastable states with these lifetimes qualitatively, and corroborate our results with atomic structure calculations that support the observed lifetimes and decay paths. These long-lived states provide new opportunities in qubit and qudit state detection and optical clocks.

All utility-scale quantum computers will require some form of Quantum Error Correction in which logical qubits are encoded in a larger number of physical qubits. One promising encoding is known as the colour code which has broad applicability across all qubit types and can decisively reduce the overhead of certain logical operations when compared to other two-dimensional topological codes such as the surface code. However, whereas the surface code decoding problem can be solved exactly in polynomial time by finding minimum weight matchings in a graph, prior to this work, it was not known whether exact and efficient colour code decoding was possible. Optimism in this area, stemming from the colour code's significant structure and well understood similarities to the surface code, fanned this uncertainty. In this paper we resolve this, proving that exact decoding of the colour code is NP-hard -- that is, there does not exist a polynomial time algorithm unless P=NP. This highlights a notable contrast to some of the colour code's key competitors, such as the surface code, and motivates continued work in the narrower space of heuristic and approximate algorithms for fast, accurate and scalable colour code decoding.

We investigate the relativistic scattering of spinning black holes using modern amplitude methods within a heavy-mass effective field theory formalism at third post-Minkowskian order. Using a systematic self-force expansion up to first order in the mass ratio, the gravitational amplitude and the associated eikonal-like phase are computed for a spin-aligned binary system comprising a heavy and a light black hole up to fifth order in the total spin and up to quadratic order in the spin of the light black hole. We also consider the resummation of the heavy black hole's spin in both the probe limit and the radiation-reaction sector, and verify that the resulting phase displays the characteristic ring singularity features associated with the Kerr metric.

Coherent control of motional degrees of freedom of ultracold atoms in optical lattices offers a promising route towards programmable quantum dynamics with massive particles. We propose and analyze a scheme for implementing coherent rearrangement of ultracold atoms, corresponding to arbitrary unitary transformations on single-particle motional states. Exploiting an analogy between dynamics in optical superlattices and discrete linear optics, we employ the Clements scheme to systematically construct any global $N$-dimensional single-particle unitary from tunneling and phase shifts in arrays of double wells. Tunneling is controlled globally, while local operations are achieved through site-resolved potential shifts. We numerically investigate the susceptibility of the scheme to intensity noise and addressing crosstalk. We identify key subroutines enabled by this unitary construction, including the Discrete Fourier Transform and the implementation of non-native Hamiltonians. Extending the scheme to two dimensions enables all-to-all atomic rearrangement with a circuit depth that scales sublinearly with the atom number, providing a high-density and highly scalable approach to atom rearrangement.

We construct nonlinear wave operators and prove small-data asymptotic completeness for the Maxwell--Higgs system on the domain of outer communications of every four-dimensional subextremal Kerr black hole $(\mathcal D_{M,a},g_{M,a})$ with $M>0$ and $|a|<M$, for gauge-invariant nonnegative scalar potentials $P$ satisfying Assumption~\ref{asumsiP} with mass parameter $m^{2}\ge0$. The massless case $m=0$ is unconditional on the full subextremal range. For $m^{2}>0$ the same conclusions follow assuming the massive linear package $\Lin_{k}^{(m)}$ for the linear comparison system (in particular, no exponentially growing modes); this fails for an open set of masses due to superradiant instability \cite{ShlapentokhRothmanKGKerr}. We work in the radiative (charge-free) regime; stationary Coulomb (Kerr--Newman) modes are treated separately. Asymptotic states are described by gauge-covariant radiation fields on $\mathcal I^{\pm}\cup\mathcal H^{\pm}$ (and, when $m>0$, an additional timelike/Dollard channel), yielding a gauge-invariant nonlinear scattering map on the residual-gauge quotient. The scattering map is a small-data bijection, is Fréchet differentiable at $0$ with derivative equal to linear Kerr scattering, admits a quadratic (Born) expansion with an $O(\|U\|^{3})$ remainder in the natural asymptotic topology, and is real-analytic for analytic $P$. The nonlinear argument is presented as a transfer principle from a black-box linear estimate package for inhomogeneous Klein--Gordon and charge-free Maxwell fields, verified here in the massless Kerr case (and proved self-contained in Schwarzschild).

Despite the robust security guarantees of Quantum Key Distribution (QKD), its practical deployment is significantly challenged by the dynamic nature of quantum channels and the complexity of real-time parameter optimization. In this paper, we propose OptiQKD, a protocol-agnostic machine learning framework specifically engineered to maximize the Secure Key Rate (SKR) and minimize the Quantum Bit Error Rate (QBER) for the BB84, E91, and COW protocols. OptiQKD integrates Temporal Convolutional Networks (TCNs) for high-accuracy and short-horizon forecasting of channel-state fluctuations with a Reinforcement Learning (RL) controller for autonomous and real-time parameter selection. This optimization stack is strictly constrained by standard composable-security assumptions to ensure that performance gains do not compromise the underlying quantum security. We evaluate the framework by simulating critical environmental stressors, including depolarizing and amplitude-damping noise, under realistic device constraints, including channel loss, detector efficiency, and dark counts. Our results demonstrate substantial protocol-agnostic improvements: the median SKR increases by 20--30%, while the median QBER is reduced from 3.0% to 1.5% through predictive state optimization. These findings establish that OptiQKD provides an efficient, security-preserving mechanism for dynamic parameter tuning, paving the way for more resilient and high-throughput practical QKD deployments.

This paper proposes a distributed computing framework for solving the Lindblad master equation in large-dimensional cavity QED systems. By leveraging the sparsity of the jump operator and combining this approach with the Cannon algorithm, the computational complexity of non-unitary terms is reduced from $O(MN^3)$ to $O(MN)$. For unitary terms, a combination of Taylor series approximation and the Cannon algorithm enables distributed matrix exponentiation, though scalability is limited by cross-processor communication. The proposed dynamic subspace construction method further reduces the Hamiltonian dimension: when $n_{\text{at}}=10$, the dimension is reduced to $5.63\%$ of the full Hamiltonian, with a memory footprint of only $0.32\%$. Results show that this framework significantly accelerates non-unitary evolution, providing a feasible solution for simulating large-scale open quantum systems where the number of dissipative channels $M$ is much larger than the Hamiltonian dimension $N$.

In the excursion-set formalism, the mass distribution of primordial black holes (PBHs) is derived from the first-passage time of a random walk describing the density contrast as the coarse-graining scale varies. We address two recent criticisms that have been raised about this approach. First, it was argued that the random walks are subject to colored (i.e. correlated over time) noise, making the first-passage-time problem cumbersome. We show that this arises from an incorrect separation of drift and noise when sampling on the Hubble-crossing surface: if Fourier modes are uncorrelated, the noise is strictly white. Moreover, sampling along the Hubble-crossing surface precludes using the density dispersion as a time variable, explaining the reported pathologies. Sampling instead on a synchronous surface removes both issues. This requires solving a first-passage-time problem with a moving barrier, for which we provide an efficient numerical framework. Second, it was suggested that cloud-in-cloud (i.e. that large black holes may engulf smaller ones) is irrelevant for PBHs and that the excursion set is therefore not needed. While valid for widely separated scales, this statement fails for broad power spectra with enhanced continua of modes. We further show that Press-Schechter estimates neglecting boundary evolution can break down even without cloud-in-cloud effects. Our results establish the robustness and necessity of the excursion-set formalism in realistic PBH formation scenarios.

We study slow-roll inflation driven by a scalar field non-minimally coupled to gravity within the effective framework of Loop Quantum Cosmology (LQC), including inverse-volume corrections. We consider two physically motivated classes of potentials, a Higgs-like quartic potential $V\proptoφ^{4}$ and string-inspired fractional monomial potentials $V\proptoφ^{p}$ with $p<1$. Working at first order in the slow-roll expansion, we derive analytic expressions for the inflationary observables, namely the scalar spectral index $n_s$, the tensor-to-scalar ratio $r$, and the running $α_s\equiv dn_s/d\ln k$, and then solve the corrected background dynamics numerically to obtain quantitative predictions. Confronting these results with current observational constraints from Planck 2018 and ACT DR6, we find that the model can lie within the allowed region of the $(n_s,r,α_s)$ parameter space, including a mild preference for slightly larger $n_s$, as suggested by recent ground-based measurements. We also compute the probability of achieving sufficient slow-roll inflation in this setting. Although effective LQC replaces the initial singularity with a nonsingular quantum bounce, the likelihood of a sufficiently long inflationary phase depends on the pre-inflationary dynamics and on the inflaton potential. Using the canonical Liouville measure on the effective phase space, we determine the fraction of post-bounce trajectories that yield sufficient inflation and find that the non-minimal coupling parameter $ξ$ substantially enlarges the phase-space volume of favorable initial conditions relative to the minimally coupled case, exhibiting an attractor-like enhancement that saturates at large $ξ$.

Can we infer whether all of physical reality began to exist? Several novel results are offered suggesting a negative verdict. First, a common strategy for defending a cosmic beginning involves showing that individual beginningless cosmological models are implausible. This strategy is shown to make an elementary error in confirmation theory. Second, two necessary (but not necessarily sufficient) conditions are offered for a cosmic beginning. Third, three extensions are offered to the Malament-Manchak theorems. The three extensions show that in almost all classical spacetimes, observers cannot collect sufficient data to determine whether the application conditions for the classic singularity theorems are satisfied or whether their spacetime satisfies the two necessary conditions for a cosmic beginning. Lastly, a reply is offered to the objection that the skeptical consequences of the three extensions can be overcome with induction. Importantly, all past singular dust FLRW spacetimes have observationally indistinguishable counterparts which, while sharing a number of important local properties, either do not include a singularity to the past of every point or else do not have the sort of time ordering intuitively required for a cosmic beginning.

Atom loss is a major error source in neutral-atom quantum computers, accounting for over 40% of the total physical errors in recent experiments. Unlike Pauli errors, atom loss poses significant challenges for both syndrome extraction and decoding due to its nonlinearity and correlated nature. Current syndrome extraction circuits either require additional physical overhead or do not provide optimal loss tolerance. On the decoding side, existing methods are either computationally inefficient, achieve suboptimal logical error rates, or rely on machine learning without provable guarantees. To address these challenges, we propose the Pauli Envelope framework. This framework constructs a Pauli envelope that bounds the effect of atom loss while remaining low weight and efficiently computable. Guided by this framework, we first design a new atom-replenishing syndrome extraction circuit, the Mid-SWAP syndrome extraction, that reduces error propagation with no additional space-time cost. We then propose an optimal decoder for Mid-SWAP syndrome extraction: the Envelope-MLE decoder formulated as an MILP that achieves optimal effective code distance dloss ~ d for atom-loss errors. Inspired by the exclusivity constraint of the optimal decoder, we also propose an Envelope-Matching decoder to approximately enforce the exclusivity constraint within the MWPM framework. This decoder achieves d_loss ~ 2d/3, surpassing the previous best algorithmic decoder, which achieves dloss ~ d/2 even with an MILP formulation. Circuit-level simulations demonstrate that our approach attains up to 40% higher thresholds and 30% higher effective distances compared with existing algorithmic decoders and syndrome extraction circuits in the loss-dominated regime. On recent experimental data, our Envelope-MLE decoder improves the error suppression factor of a hybrid MLE--machine-learning decoder from 2.14 to 2.24.

The property of non-stabiliserness, or ``magic'', is of interest in quantum computing due to its role in developing fault-tolerant quantum algorithms with genuine computational advantage over classical counterparts. There has been much interest in quantifying magic in various physical systems, in order to probe how to produce and enhance it. The production of magic has previously been quantified in gluon and graviton scattering, in the so-called helicity basis relating particle spins with momentum directions. For a basis-independent statement, one should instead use the recently developed concept of non-local non-stabiliserness, and our aim in this paper is to derive how this varies for gluon and graviton scattering processes. Our results show that, for many initial states, including those produced with polarised beams, the helicity basis coincides with a basis in which the non-local magic is manifest, providing a physical motivation for using the helicity basis to study quantum information quantities. However, this property breaks upon adding additional operators to the Yang-Mills Lagrangian, as would be the case in new physics scenarios.

To overcome the physical limitations of scaling monolithic quantum computers, distributed quantum computing (DQC) interconnects multiple smaller-scale quantum processing units (QPUs) to form a quantum network. However, this approach introduces a critical challenge, namely the high cost of quantum communication between remote QPUs incurred by quantum state teleportation and quantum gate teleportation. To minimize this communication overhead, DQC compilers must strategically partition quantum circuits by mapping logical qubits to distributed physical QPUs. Static graph partitioning methods are fundamentally ill-equipped for this task as they ignore execution dynamics and underlying network topology, while metaheuristics require substantial computational runtime. In this work, we propose a heuristic based on beam search to solve the circuit partitioning problem. Our time-aware algorithm incrementally constructs a low-cost sequence of qubit assignments across successive time steps to minimize overall communication overhead. The time and space complexities of the proposed algorithm scale quadratically with the number of qubits and linearly with circuit depth, offering a significant computational speedup over common metaheuristics. We demonstrate that our proposed algorithm consistently achieves significantly lower communication costs than static baselines across varying circuit sizes, depths, and network topologies, providing an efficient compilation tool for near-term distributed quantum hardware.

We study quantum entanglement in a system comprising two quantum dots interconnected through the short topological superconducting nanowire, which hosts overlapping boundary Majorana modes. Inspecting the fermionic negativity, we analyze the variation of entanglement against the position of the energy levels of quantum dots and their hybridization with the topological superconducting nanowire. In the absence of electron correlations, the optimal entanglement occurs when the energy levels coincide with the zero-energy Majorana modes, whereas upon increasing the hybridizations, the entanglement is gradually suppressed. Such monotonous behavior is no longer valid when the quantum dot levels are detuned from the zero-energy. Under these circumstances, the quantum dots become maximally entangled for a certain optimal hybridization. Moreover, we study the thermal concurrence to explore the entanglement properties at finite temperatures. We also compute the quantum mutual information and propose recipes for robust finite-temperature entanglement transmission via Majorana modes.

As global demand for food and clean energy intensifies, agrivoltaic systems have emerged as a vital solution for land-use optimization. However, current designs overwhelmingly treat incident light as a classical photon flux, overlooking the quantum mechanical nature of photosynthetic energy transfer. We introduce spectral bath engineering-the strategic spectral filtering of sunlight through semi-transparent organic photovoltaic (OPV) panels to exploit non-Markovian quantum coherence in biological light-harvesting. Using Process Tensor HOPS (PT-HOPS) and Spectrally Bundled Dissipators (SBD) to simulate the Fenna-Matthews-Olson complex, we demonstrate that selective filtering at vibronic resonance wavelengths (750nm and 820nm) enhances the electron transport rate (ETR) by 25% relative to standard Markovian models. This quantum advantage is driven by vibronic resonance-assisted transport, which extends coherence lifetimes by 20% to 50% and nearly doubles pairwise concurrence (89%). Multi-objective Pareto optimization identifies OPV configurations reaching 18.8% power conversion efficiency while sustaining an 80.5% system ETR, potentially generating an additional USD 470 to 3000 $ha^{-1}$$yr^{-1}$ in revenue. Environmental simulations across nine climate zones, including sub-Saharan Africa, confirm persistent ETR enhancements of 18% to 24%. Finally, eco-design analysis using quantum reactivity descriptors ensures that these technological gains are achieved using sustainable, biodegradable materials. By bridging quantum biology and renewable energy engineering, this work provides a quantitative blueprint for next-generation agrivoltaic materials that co-optimize agricultural productivity and energy yield.

The Steiner Tree Problem (STP) is a well-known NP-hard combinatorial optimization problem, which has wide applications in network design, integrated circuit layout, bioinformatics, and other fields. However, traditional algorithms often struggle to balance efficiency and solution quality when dealing with large-scale STP instances. In this paper, we propose a new quantum annealing-based algorithm for solving the STP: we first model the STP into a quadratic unconstrained binary optimization (QUBO) form suitable for quantum annealing, then design a corresponding encoding strategy, and finally verify the algorithm through experimental tests. The results show that our quantum annealing-based method can obtain high-quality solutions with relatively low computational overhead for moderate-scale STP instances, providing a new feasible path for handling this intractable combinatorial optimization problem.

It is mandatory to know how to operationally define and translate a reference frame into mathematics, in order that a physical interpretation of theory calculations in terms of observational data is possible. The situation is particularly challenging for gauge systems such as General Relativity where spacetime coordinates are subject to spacetime diffeomorphisms considered as gauge transformations turning coordinates into non-observables. This motivates the idea of operationally defined (material) reference frames which specify coordinates in terms of matter or geometry reference fields leading to the concept of relational observables, relational reference frames and gauge reduction. Upon quantisation, all fields become operator valued distributions. Now new conceptual and technical questions arise such as: Should one reduce before or after quantisation and how are the reference fields quantised respectively in either route? Is a reference frame itself subject to quantisation and how are different quantum reference frames related? How does the gauge reduction fit into this, i.e. how can it be that a certain reference field is considered a non-observable in one reference frame and an observable in another which upon quantisation even displays fluctuations? How precisely are gauge dependent fields interpreted in terms of the relational observables in a given reference frame? What is the relative dynamics, e.g. how exactly are physical Hamiltonians of two relational reference frames related? The present conceptual work addresses these and related questions in a non-perturbative field theory context of sufficient generality to cover General Relativity coupled to standard matter. A central role is played by the concept of the relational reference frame transformation (RRFT) for which a general formula is derived and its properties are explored.

Physical quantum systems are generically coupled to an environment, resulting in open system dynamics. A typical approach to simulating this dynamics is to propagate the density matrix of the system via the Lindblad master equation. This approach is numerically challenging due to the size of the density matrix, which has led to the development of quantum jump methods, which unravel the density matrix into an ensemble of state vectors. These methods utilize a stochastic sampling of the quantum jump times, which becomes inefficent for weakly dissipative dynamics, in which jumps are rare events. Here, we propose the deterministic quantum jump (DQJ) method, which we show to outperform standard quantum jump methods in the weakly dissipative regime, by removing the error of stochastic sampling. We describe the methodology at the single-jump and two-jump level, reconstructing the density matrix at the corresponding level. We demonstrate the performance of the method for two examples, the dissipative transverse-field Ising model, and the dissipative Kerr oscillator. Given that quantum technologies such as quantum computing have weakly dissipative quantum dynamics as their central focus, we propose this method to be utilized in that context, for exploring and understanding quantum technology platforms.

We develop a systematic framework for formulating and solving the conditions that lead to separability in stationary, axisymmetric spacetimes in the presence of matter fields. Guided by Carter's metric form, we introduce a general stationary, axisymmetric metric ansatz that allows for a transparent separation of radial and angular variables. This construction yields a broad family of stationary rotating solutions admitting separability structures. To illustrate the applicability of the formalism, we explicitly construct several examples, including a rotating black hole with a global monopole supported by anisotropic matter, as well as a new class of rotating wormhole geometries.

A new phenomenological dark energy model, originally associated to the large-scale structure formation and considered as a solution to the fine-tuning and coincidence problems related to the cosmological constant, was analyzed within the framework of General Relativity in a Friedman-Robertson-Walker spacetime and its model parameters were estimated using cosmic chronometers and recent DESI data. It turns out that the proposed model can serve as an alternative evolving dark energy model with a novel equation of state function, apart from other popular propositions in the literature. Due to the form of this phenomenological energy density ansatz, which starts to rise with the nonlinear structure growth in the universe and falls with the domination of cosmic voids, we prefer to call it structure-induced dark energy. Observational constraints show that it is not only a suitable solution for the fundamental problems such as coincidence or fine-tuning problems, it gives flexibility, when considering the cosmic tensions and presents a new perspective on the evolving dark energy models.

Quantum Singular Value Transformation (QSVT) provides a unified framework for applying polynomial functions to the singular values of a block-encoded matrix. QSVT prepares a state proportional to $\bA^{-1}\bb$ with circuit depth $O(d\cdot\mathrm{polylog}(N))$, where $d$ is the polynomial degree of the $1/x$ approximation and $N$ is the size of $\bA$. Current polynomial approximation methods are over the continuous interval $[a,1]$, giving $d = O(\sqrt{\kap}\log(1/\varepsilon))$, and make no use of any properties of $\bA$. We observe here that QSVT solution accuracy depends only on the polynomial accuracy at the eigenvalues of $\bA$. When all $N$ eigenvalues are known exactly, a pure spectral polynomial $p_{S}$ can interpolate $1/x$ at these eigenvalues and achieve unit fidelity at reduced degree. But its practical applicability is limited. To address this, we propose a spectral correction that exploits prior knowledge of $K$ eigenvalues of $\bA$. Given any base polynomial $p_0$, such as Remez, of degree $d_0$, a $K\times K$ linear system enforces exact interpolation of $1/x$ only at these $K$ eigenvalues without increasing $d_0$. The spectrally corrected polynomial $p_{SC}$ preserves the continuous error profile between eigenvalues and inherits the parity of $p_0$. QSVT experiments on the 1D Poisson equation demonstrate up to a $5\times$ reduction in circuit depth relative to the base polynomial, at unit fidelity and improved compliance error. The correction is agnostic to the choice of base polynomial and robust to eigenvalue perturbations up to $10\%$ relative error. Extension to the 2D Poisson equation suggests that correcting a small fraction of the spectrum may suffice to achieve fidelity above $0.999$.

We present a semiclassical phase-space method to calculate thermal and ground states of large interacting spin systems. To this end, we extend the recently developed truncated Wigner approximation for spins (TWA) to the imaginary time, termed iTWA. The evolution of the canonical density matrix in imaginary time is mapped to a partial differential equation of its Wigner function. Truncation at the Fokker-Planck level leads to a set of stochastic differential equations, which can be efficiently simulated. We show that the iTWA can provide very good approximations to the ground state of a random and in general frustrated anti-ferromagnetic Ising Hamiltonian on a 3-regular graph, for which finding the exact ground state and approximations to it beyond a certain accuracy is NP hard. Furthermore in order to assess the ability of the method to properly account for leading-order quantum effects, we analyze the ground-state quantum phase transition of the nearest-neighbor, transverse-field Ising model in one and two spatial dimensions, finding very good agreement with the exact behaviour. The critical behavior obtained in iTWA follows the quantum-classical correspondence.

This study explores singularity-free solutions to the static, spherical symmetric Einstein equations with the standard Schwarzschild solution as a boundary condition. Imposing the absence of curvature singularities and requiring differentiability of the time component of the metric leads to a sign change across the horizon, violating the Principle of Equivalence locally. We find a solution within the event horizon with a simple ``cosmological constant'' stress-energy tensor. Considering the impact of sign change to a compact stellar remnant, modeled by an incompressible perfect fluid obeying the Tolman-Oppenheimer-Volkoff equation, we rediscover the same geometry, indicating both mathematical and physical feasibility of the model. We also find a new theoretical limit M/R=3/8, which is lower than the Buchdahl limit of M/R=4/9 for the density of a perfect fluid that will recede behind an event horizon. The equation of state is discussed, and we propose that the final state is described by a Higgs-like free scalar field.

Quantum technologies rely on high-quality resource states, such as maximally entangled or private states, which are indispensable for quantum communication and cryptography. In practice, however, these states are inevitably degraded by noise. Distillation protocols aim to recover high-resource states from multiple imperfect copies, and while stabilizer-based methods have demonstrated high performance in entanglement purification, they have yet to be established for broader tasks such as secret-key distillation. This work introduces a unified framework for stabilizer-based resource distillation in systems of prime local dimension. By formulating stabilizer routines as quantum channels and deriving closed-form expressions for their output, we enable the application of stabilizer operations to general input states and diverse distillation objectives. We identify key invariances in resource measures, such as coherent and private information, and demonstrate how they can be leveraged to significantly reduce the numerical complexity of channel optimization. To illustrate the framework's versatility, we introduce several protocols: gF-IMAX for general fidelity optimization, and (S)CI-IMAX and (S)PI-IMAX for optimizing (smooth) coherent and private information in both asymptotic and one-shot regimes. Our numerical results confirm that these protocols effectively tailor stabilizer channels to specific operational tasks, establishing them as a robust and flexible tool for quantum resource distillation.

In this work, we construct a locally inertial reference system adapted to a geodesic observer in stationary, axisymmetric dust solutions of the Einstein equations employed as effective models of a portion of a galactic disc. To ensure a consistent spatial orientation among different local observers, we also introduce the radially locked reference system, in which one spatial axis is aligned with the radial direction defined by null geodesics passing through the galactic center. Within this framework, we analyze how the dust configuration is described by such observers by computing the frequency shift of photons exchanged between pairs of dust geodesics. Building on this construction, we outline a procedure to reconstruct spectroscopic and astrometric relative velocities with respect to locally inertial observers, providing a coherent foundation for the study of galactic kinematics in a fully general relativistic context.

Following any quantum information processing protocol, it is essential to reset a mixed state of a many-body interacting spin-network to the computational-zero pure state. This task is challenging, both theoretically and experimentally, because of the quantum correlations. There is currently no effective cooling strategy for both high and low temperatures in such networks. Here we put forth a universal cooling strategy for multi-spin interacting networks. The strategy is based on the collective coupling of the system to an ancilla spin that intermittently dumps part of its entropy into an ultracold bath. Yet this strategy should overcome the symmetry-imposed correlations that impede the cooling. To avoid the prohibitive complexity of computing the dynamics, we resort to graph analysis of the network. %To approach the desired state, We show that a unique choice of alternating, non-commuting system-ancilla interaction Hamiltonians exists that breaks the symmetry constraints and allows the network to approach the desired pure state. We illustrate this universal purification strategy in diverse experimental settings.

The diversity of quantum correlations -- discord, entanglement, steering, and Bell nonlocality -- disappears at the observable second-moment kinematic level. By treating state purity as a finite resource, we introduce a local-unitary-invariant budget split of symmetrised second moments into local and nonlocal sectors that maps quantum systems onto compact, two-dimensional, hole-free manifolds. The topology of these manifolds is governed by state purity and time-reversal symmetry. This dimensional reduction reveals a deep structural link: exceeding classical capacity limits forces the activation of time-asymmetric generators, guaranteeing non-positive partial transpose entanglement. For two qubits, the geometry is analytically solvable. A single boundary elegantly isolates classical correlations, while nested regions physically dictate entanglement, steering, Bell nonlocality, and bounds on non-stabiliser magic. Beyond two qubits, dimensional capacity bottlenecks enforce these universal kinematic limits on correlation structures. Because this macroscopic representation is completely determined by global and marginal purities, it bypasses the exponential scaling of full-state tomography. By coarse-graining over gauge-like first moments, the budget geometry acts as a thermodynamic phase diagram, exposing both the static hierarchy of quantum resources and their dynamic redistribution under decoherence.

Federated Learning (FL) enables privacy-preserving distributed model training, yet remains vulnerable to gradient inversion and model leakage attacks. Quantum communication has been proposed to provide information-theoretic security for parameter aggregation. However, practical deployment is severely constrained by communication overhead and quantum channel noise. In this work, we present a systematic quantitative study of communication--convergence--noise trade-offs in Quantum Federated Learning (QFL). We introduce two complementary strategies to reduce quantum transmissions: (1) structured parameter reduction based on light-cone feature selection in parametrized quantum circuits, and (2) a Hybrid QFL architecture that dynamically switches from centralized to decentralized aggregation during training. We derive explicit communication cost formulas and show that Hybrid QFL reduces quantum transmissions from $3NMP$ per round to $\{3t + 2(T - t)\}NMP$, achieving substantial savings while preserving near-centralized convergence. We further analyze robustness under depolarizing noise and show that decentralized aggregation is more noise-resilient because it transmits fewer qubits per round. Finally, we evaluate the effectiveness of Steane code-based quantum error correction under high-noise regimes. Our results provide an integrated design framework for communication-efficient and noise-aware QFL, clarifying practical trade-offs necessary for scalable quantum-secure distributed learning.

We analyse the dynamics of a qubit coupled to a dissipative impurity by comparing local and global derivation schemes of a Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) master equation within the Born-Markov and full secular (FS) approximations. We show that the local approach correctly captures a crossover in the dynamics of the qubit coherence, while the FS approximation restricts the validity of the global approach to regimes with well-separated energy scales. Our results clarify the domains of validity of the two approaches and show that the local scheme provides a better GKSL description of the qubit dynamics in the experimentally relevant parameter regime.

Amplitude encoding of real-world data on quantum computers is often the workflow bottleneck: direct amplitude encoding scales poorly with input size and can offset any speedups in subsequent processing. Fourier-based sparse amplitude encoding lowers cost by retaining only a small subset of dominant coefficients, but its fixed, non-adaptive basis leads to significant information loss. In this work, we replace the Fourier transform with the adaptive interpolating quantum transform (AIQT) in the sparse amplitude encoding workflow. The AIQT learns a data-adapted basis that concentrates information into a small number of coefficients. Consequently, at matched sparsity, the AIQT retains more information and achieves lower reconstruction error compared to the Fourier baseline. On financial time-series data, the AIQT reduces reconstruction error by 40% relative to the Fourier baseline, and on image datasets the reduction is up to 50% at the same sparsity level, with nearly identical encoding gate cost. Crucially, the approach preserves the efficiency of Fourier-based methods: the AIQT is built on the structure of the quantum Fourier transform circuit. Its gate count scales quadratically with the number of qubits, while classical evaluation can be carried out in quasilinear time. In addition, the AIQT is trained without labels and does not require sampling from quantum hardware or a simulator, removing a major bottleneck in data-driven amplitude-encoding methods.

We implement a variational quantum algorithm for Gibbs state preparation of a transverse-field Ising model on IonQ's quantum computers. To this end, we train the variational parameters via classical simulation and perform state tomography on the quantum devices to evaluate the fidelity of the prepared Gibbs state. As a main result, we find that fidelity decreases (non-monotonically) as a function of the inverse temperature $β$ of the system. Fidelity also decreases as a function of the size of the system. Interestingly, we find that a Gibbs state prepared for a specified $β$ is a better representative of a Gibbs state prepared for a $\textit{lower}$ $β$; or in other words, thermal fluctuations in the quantum hardware lead to digital heating, that is, an increase in the temperature of the prepared Gibbs state above what was intended.

Computing the vacuum and energy spectrum in non-Abelian, interacting lattice gauge theories remains an open challenge, in part because approximating the continuum limit requires large lattices and huge Hilbert spaces. To address this difficulty with near-term quantum computing devices, we adapt the variational quantum eigensolver to non-Abelian gauge theories. We outline scaling advantages when using a spin-network basis to simulate the gauge-invariant Hilbert space and develop a systematic state preparation ansatz that creates gauge-invariant excitations while alleviating the barren plateau problem. We illustrate our method in the context of SU(2) Yang-Mills theory by testing it on a minimal toy model consisting of a single vertex in 3+1 dimensions. In this toy model, simulations allow us to investigate the impact of noise expected in current quantum devices.

We investigate the principles of quantum field theory using a stiff de Sitter space. We demonstrate that a non-unitary Lagrangian on a Euclidean AdS geometry can produce the perturbative expansion of late-time correlation functions to all orders. This discovery greatly simplifies perturbative computations while also allowing us to prove fundamental features of these correlators, which are part of a Euclidean CFT. This allows us to construct an OPE expansion, limit the operator spectrum, and deduce the analytic structure of the spectral density that captures the conformal partial wave expansion of a late-time four-point function. In general, the standard CFT concept of unitarity does not apply to dimensions and OPE coefficients. Rather, the positivity of the spectral density represents the unitarity of the de Sitter theory. This assertion is non-perturbative and does not depend on the use of Euclidean AdS Lagrangians. In a scalar theory, we compute tree-level and entire one-loop-resummed exchange diagrams to demonstrate and verify these characteristics. In the spectrum density, an exchanged particle shows up as a resonant characteristic that may be helpful in experimental searches.

Fault-tolerant quantum computation (FTQC) requires fast and accurate decoding of quantum errors, which is often formulated as a minimum-weight perfect matching (MWPM) problem. A determinant-based approach has been proposed as a promising method to surpass the conventional polynomial runtime of MWPM decoding via the blossom algorithm, asymptotically achieving polylogarithmic parallel runtime. However, the existing approach requires an impractically large bit length to represent intermediate values during the computation of the matrix determinant; moreover, when implemented on a finite-bit machine, the algorithm cannot detect overflow, and therefore, the mathematical correctness of such algorithms cannot be guaranteed. In this work, we address these issues by presenting a polylog-time MWPM decoder that detects overflow in finite-bit representations by employing an algebraic framework over a truncated polynomial ring. Within this framework, all arithmetic operations are implemented using bitwise XOR and shift operations, enabling efficient and hardware-friendly implementation. Furthermore, with algorithmic optimizations tailored to the structure of the determinant-based approach, we reduce the arithmetic bit length required to represent intermediate values in the determinant computation by more than $99.9\%$, while preserving its polylogarithmic runtime scaling. These results open the possibility of a proof-of-principle demonstration of the polylog-time MPWM decoding in the early FTQC regime.

Recently, an exact rotating black hole solution in a parity-violating theory of gravity was obtained via a conformal transformation of the Kerr solution in general relativity, with parity-violating effects encoded in the conformal factor. We study the quasinormal modes (QNMs) of a test scalar field minimally coupled to gravity on this conformal Kerr background, treating the parity-violating effects perturbatively while allowing for arbitrary black hole spin, from the non-rotating case to the near-extremal regime. For low spin, we derive a perturbative formula for the QNM frequencies that includes the leading-order parity-violating correction. For high spin, particularly in the near-extremal regime, we find sizable deviations from the Kerr QNM frequencies. Our results point to a new avenue for probing parity-violating physics in the strong-gravity regime through black hole QNMs.

Given the rapid advances in quantum computing hardware, establishing systematic strategies for verifying the correctness of quantum computations has become increasingly important. Exploiting the fact that the axial anomaly in gauge theories is exact to all orders in perturbation theory, we propose the axial anomaly as a nontrivial benchmark for quantum simulations of lattice gauge theories. We simulate anomalous axial-charge production in ${\mathbb Z}_N$ lattice gauge theories on the trapped-ion quantum computer ``Reimei''. After taking the U(1), infinitesimal time, and infinite volume limits, we successfully reproduce the anomaly coefficient within statistical uncertainties, even without error mitigation. Our results demonstrate that the axial anomaly can be simulated on current quantum computers and serves as a verification test of quantum computations.

An explicit correspondence is established between the Extended Uncertainty Principle (EUP) and the metric function by directly relating the radiation temperature function modified by EUP to the modified spacetime metric. Utilising this modified metric, we subsequently derive the corresponding thermodynamic quantities of the black hole, and calculate the photon sphere radius and the size of the black hole shadow. The results of the study indicate that, in comparison with Schwarzschild black holes, the position of the event horizon remains constant under EUP modifications. However, the photon sphere radius increases with growing EUP parameters, while the shadow size decreases with increasing parameters, demonstrating that EUP induces optical shift phenomena. By comparing with observations of the galactic centre black hole $\text{Sgr}{\text{A}^{*}}$ from the Event Horizon Telescope, new constraints are established on EUP parameters.

Topological superconductivity could host quasiparticles that are key candidates for fault-tolerant quantum computation due to their immunity to noise as they obey non-Abelian exchange statistics. For example, in the case of Majorana Zero Modes (MZM), braiding enables two topologically protected quantum gates. While their direct manipulation in solid-state systems remains experimentally challenging, digital emulation of MZM behavior has provided insight as well as a deeper understanding of controlling these topological quantum systems. This emulation is typically accomplished by mapping the topological and trivial phases of a Majorana system to ferromagnetic and paramagnetic Hamiltonians of a spin-glass model. This approach usually relies on adiabatic evolution of superconducting Hamiltonians, which require circuits with very large depths. In this work, we present a resource-efficient method to emulate MZM braiding in a trijunction geometry using a quantum processor. We introduce direct braiding operators which simulate the evolution more efficiently, reducing the quantum gate overhead. We then further generalize this method to emulate braiding operations in extended trijunction architectures based on Kitaev chains.

Quantum transducers are critical for quantum interconnect, enabling coherent signal transfer across disparate frequency domains. Beyond material and device advances, protocol design has become a powerful means to improve transduction. We introduce a variational quantum transduction (VQT) framework that employs variational tools from near-term quantum computing to systematically optimize protocol performance. As a variational quantum circuit framework, VQT is not plagued by known training issues such as barren plateau, because a small-scale problem is sufficient for substantial advantage and training only needs to be done once to configure a VQT system. Maximizing the quantum information rate within this framework yields protocols that surpass all known schemes in their respective classes. For non-adaptive protocols, VQT exceeds the performance envelopes of Gottesman-Kitaev-Preskill (GKP)-based and entanglement-assisted approaches. In the adaptive setting, VQT provides only a marginal improvement over Gaussian feedforward strategies, indicating that Gaussian adaptive transduction is already close to optimal. With increasingly universal quantum control, VQT provides a systematic path toward optimal quantum transduction.

Quantum crosstalk poses a major challenge to scaling up quantum computations as its strength is typically unknown and its effect accumulates exponentially as system size grows. Here, we show that many-body robust control can be utilized to suppress unwanted couplings during multi-qubit gate operations and state preparation. By combining tensor network simulations with the GRAPE algorithm, and leveraging an efficient random sampling over noise ensembles, our method overcomes the exponential scaling of the Hilbert space. We demonstrate its effectiveness for designing control solutions for high-fidelity implementations of parallel X gates and parallel CNOT on a chain of 50 qubits, and for realizing a 30-qubit GHZ state and the ground state of a 20-qubit Heisenberg model. In the presence of many-body quantum crosstalk due to parasitic interaction between neighboring qubits, robust control results in order-of magnitude improvement in fidelity for large system sizes. These findings pave the way for more reliable operations on near-term quantum processors.

We introduce Monarq, a unified quantum data processing framework that combines QCrank encoding with the EHands protocol for polynomial transformations, and demonstrate its implementation on noisy intermediate-scale quantum (NISQ) hardware. This framework provides fundamental quantum building blocks for signal and image processing tasks, including convolution, discrete-time Fourier transform (DFT), squared gradient computation, and edge detection, serving as a reference for a broad class of data processing applications on near-term quantum devices.

One of the primary challenges of photonic quantum information processing is the on-demand preparation of multiple single-photon-level quantum states from probabilistic photon pair sources. Motivated by recent developments in frequency-bin-encoded photonic quantum information processing, here we consider active time multiplexing to generate n-photon states, where n single photons with n distinct frequencies occupy the same spatiotemporal mode. We devise an approach that uses optical quantum memories to manipulate the temporal mode of heralded single photons and an array of fiber Bragg grating reflectors to jointly manipulate the frequency and temporal modes of the photons, overlapping n photons in n separate frequency bins into a single spatiotemporal mode. We calculate multiphoton state generation rates that, accounting for loss, are realistically achievable with commercially available hardware. Using only a single free-space switchable delay loop for an optical quantum memory, this scheme could feasibly produce 8-photon states at an average rate of 1 kHz.

A recently proposed generalized entropy by Luciano and Saridakis extends the standard Boltzmann-Gibbs and Bekenstein-Hawking framework through a microscopically motivated construction involving two independent entropic exponents. When applied within the gravity-thermodynamics correspondence, this entropy leads to a modified cosmological dynamics that can be interpreted as an effective dark energy sector of entropic origin, while recovering $Λ$CDM in appropriate limits. In this work, we perform the first observational confrontation of the resulting entropic cosmology at the background level. Focusing on the case $α_δ=0$, we constrain the model using Cosmic Chronometers, Pantheon$^+$ Type Ia supernovae calibrated with SH0ES, BAO measurements from DESI DR2 and compressed Planck 2018 CMB information. We find that the model yields a statistically robust fit to the combined data sets and can simultaneously satisfy Pantheon$^+$, SH0ES and CMB shift-parameter constraints, unlike $Λ$CDM. Although the entropic parameters remain close to their standard values, the $Λ$CDM limit is excluded at the $2σ$ level within the restricted parameter space considered. These results indicate that the Luciano-Saridakis entropic cosmology offers a viable extension of the standard model with the potential to alleviate the Hubble tension at the background level.

Inspired by the formalism that relates the star-exponential with the quantum propagator for bosonic systems, in this work we introduce the analogous extension for the fermionic case. In particular, we analyse the problem of calculating the star-exponential (i.e., the symbol of the evolution operator) for Fermi systems within the deformation quantization program. Grassmann variables and coherent states are considered in order to obtain a closed-form expression for the fermionic star-exponential in terms of its associated propagator. As a primary application, a fermionic version of the Feynman-Kac formula is derived within this formalism, thus allowing a straightforward calculation of the ground state energy in phase space. Finally, the method is validated by successfully applying it to the simple harmonic and driven Fermi oscillators, for which the results developed here provide a powerful alternative computational tool for the study of fermionic systems.

The Sleeping Beauty Problem (SBP) is a long-standing puzzle in classical probability theory and has been used to challenge the Many-Worlds Interpretation (MWI) of quantum mechanics, since both involve objective determinacy combined with subjective uncertainty about certain events. A common concern is that MWI yields a different answer to the quantum version of SBP than the widely supported Thirder position in the classical case. We argue that this concern is unwarranted. We show that in both the quantum and classical versions of SBP, the correct credence is given by the Halfer position. In the quantum (MWI) SBP, we show that if no unjustified renormalization is introduced, the correct credence is 1/2. We then extend this result to the classical SBP by refuting four major arguments for 1/3. First, we reject the Proportion Argument by distinguishing event weight from probability. Second, we rebut Elga's Variant Argument by extending an earlier critique that identifies the implicit introduction of additional information; we further clarify this point by constructing a new variant and explaining why the Principle of Indifference is inapplicable, drawing an analogy with a mistake by d'Alembert in the history of probability theory. Third, we identify a flaw in the Technicolor Beauty Variant Argument, which arises from treating overlapping events as disjoint. Finally, we argue that causal decision theory is inappropriate for SBP, rendering the Thirders vulnerable to a Dutch Book. Our results support the consistency of MWI under the challenge posed by SBP and suggest that the dominant position on SBP needs careful reconsideration.

Transversal gates are the simplest form of fault-tolerant gates and are relatively easy to implement in practice. Yet designing codes that support useful transversal operations -- especially non-Clifford or addressable gates -- remains difficult within the stabilizer formalism or CSS constructions alone. We show that these limitations can be overcome using tensor-network frameworks such as the quantum lego formalism, where transversal gates naturally appear as global or localized symmetries. Within the quantum lego formalism, small codes carrying desirable symmetries can be "glued" into larger ones, with operator-flow rules guiding how logical symmetries are preserved. This approach enables the systematic construction of codes with addressable transversal single- and multi-qubit gates targeting specific logical qubits regardless of whether the gate is Clifford or not. As a proof of principle, we build new finite-rate code families that support strongly transversal $T$, $CCZ$, $SH$, and Gottesman's $K_3$ gates, structures that are challenging to realize with conventional methods. We further construct holographic and fractal-like codes that admit addressable transversal inter-, meso-, and intra-block $T$, $CS$, and $C^\ell Z$ gates. As a corollary, we demonstrate that the heterogeneous holographic Steane-Reed-Muller black hole code also supports fully addressable transversal inter- and intra-block $CZ$ gates, significantly lowering the overhead for universal fault-tolerant computation.

While Quantum Key Distribution (QKD) provides information-theoretic security, the transition from theory to physical hardware introduces side-channel vulnerabilities that traditional error metrics often fail to characterize. This paper presents a high-fidelity simulation framework for intrusion detection in decoy-state QKD, modeled as a minimax game between a learning-based defender and a physically constrained, adaptive adversary. The defender utilizes block-level telemetry (comprising decoy-state residuals, timing-histogram moments, and detector imbalances) to trigger alarms that gate key distillation . Unlike heuristic thresholds, our optimization objective is strictly operational: missed detections are penalized based on the resulting degradation of the finite-key secret fraction calculated via three-intensity decoy estimators and entropy-accumulation (EAT) penalties. The emulated adversary performs an automated search over time-shift, detector-blinding, photon number splitting (PNS), and Trojan-horse families, subject to hardware-limited feasibility bands. Concurrently, the defender co-trains one-class and temporal detectors (LSTM/TCN) using hard-negative mining to minimize the missed-attack rate at a calibrated false-alarm rate ($\text{FAR}$). Under adaptive attack scenarios, the system preserves $82\text{--}92\%$ of the honest finite-key rate while discarding only approximately $1.2\%$ of traffic, representing a net gain of $+20\text{--}35$ percentage points in usable secret bits over non-adversarial baselines. These results demonstrate that optimizing detection directly for secret-bit retention provides a robust, physically grounded layer of defense against adaptive side-channel strategies in practical QKD deployments.

We experimentally demonstrate that a hybrid quantum-classical algorithm can outperform purely classical, off-the-shelf selected configuration interaction methods. First, we construct a class of local Hamiltonian problems with sparse ground states, and show that representative classical heuristics fail to find the ground state of a specific 49-qubit instance. Next, we show that the sample-based Krylov quantum diagonalization algorithm, run on an IBM Heron R3 processor, succeeds at the same task. This algorithm uses quantum samples from a grid of time-evolved quantum states, and offers provable convergence guarantees for sparse ground state problems with guiding states. While the problem is also solvable classically using two iterative solvers that we designed specifically to target our Hamiltonian construction, this work resolves the previously open question of whether a sample-based quantum diagonalization algorithm can outperform standard selected configuration interaction heuristics.

We study the Hawking temperature, geodesic motion, and observable signatures of the accelerating Bertotti-Robinson (BR) spacetime, a vacuum black-hole solution deformed by a uniform magnetic field $B$ and an acceleration parameter $α$. In the timelike sector, we derive the effective potential for massive particles, determine the specific energy and angular momentum for equatorial circular orbits, and determine how $(B,α)$ shifts the ISCO; we also illustrate representative trajectories of massive particles. We then compute the radial and latitudinal epicyclic frequencies for small perturbations about circular orbits, quantifying how the magnetic field and acceleration modify local radial and vertical stability. In the null sector, we derive the photon effective potential and obtain analytical expressions for the photon-sphere radius, critical impact parameter, and shadow radius, complemented by photon trajectories, the effective radial force, and the Lyapunov exponent controlling the instability of circular null orbits; we also provide parameter-space maps for the photon sphere and shadow. Finally, we obtain the energy emission rate emitted from the black hole, showing how the acceleration parameter and the magnetic field affect this.

We revisit the spectral statistics of the C$_3$--symmetric billiard introduced by Dembowski [Phys. Rev. E, R4516 (2000)], which exhibits both GOE and GUE statistics depending on the symmetry block. Using high--precision Beyn's contour--integral method for the nonlinear Fredholm eigenvalue problem with built-in separation of irreducible subspaces, we compute 2.8x10$^5$ eigenvalues in each symmetry subspace, enabling statistically meaningful comparisons with random matrix theory. The improved spectra reveal clear GOE--GUE correspondence and resolve previously observed deviations in long--range spectral correlations. Furthermore, we analyze phase--space eigenstate localization through the distribution of entropy localization measures, which, for chaotic states follow a Beta distribution whose standard deviation decays as a power--law with energy, consistent with the onset of quantum ergodicity as described by Schnirelman's theorem.

Mixed-state phases have recently attracted significant attention as a generalization beyond their pure-state counterparts. Prominent examples include mixed-state symmetry-protected topological (mSPT) phases and the strong-to-weak symmetry breaking (SWSSB) phases. It has been shown recently that mSPT phases admit a holographic dual description in terms of higher-order subsystem SPT phases. In this work, we investigate the mixed-state phases obtained by tracing out the bulk degrees of freedom of higher-order subsystem SPT phases protected by non-invertible symmetries. We find that the resulting mixed states exhibit the coexistence of the symmetry-protected topological order and SWSSB. We also use the interface as a probe to characterize the mixed state phases, and specifically, when there is no local modification to preserve the symmetries across the interface, the two sides of the interface are in distinct phases.

Critical quantum metrology exploits the hypersensitivity of quantum systems near phase transitions to achieve enhanced precision in parameter estimation. While single-parameter estimation near critical points is well established, the simultaneous estimation of multiple parameters, which is essential for practical sensing applications, remains challenging. This difficulty arises from sloppiness, a phenomenon that typically renders the QFIM singular or nearly singular. In this work, we demonstrate that multiparameter critical metrology is not only feasible but can also retain divergent precision scaling, provided one accepts a trade-off in the scaling exponent. Using the GS of the single-cavity DM, we show that two Hamiltonian parameters can be simultaneously estimated with a scalar variance bound scaling as the square root of the critical parameter. This overcomes the inherent sloppiness by leveraging higher-order contributions to the QFIM. To recover the optimal quadratic scaling, we introduce the DD with photon hopping. In this model, a triple point in the phase diagram enables the simultaneous closure of two excitation gaps, which effectively increases the rank of the QFIM and restores the ideal single-parameter scaling for specific parameter pairs. Furthermore, we extend our to a lossy scenario. We prove that the SSs of both the DM and the DD still support multiparameter estimation with diverging precision, exhibiting linear scaling in the DM and quadratic scaling near the triple point in the DD. Finally, we establish a connection between the derived critical scalings and the fundamental state preparation time, providing a unified framework to operationally compare different sensing strategies. Our results demonstrate that critical quantum metrology can be made robust against dissipation, paving the way for practical quantum sensors operating near phase transitions.

The precision era of multi-messenger astronomy, together with modern astrophysical, cosmological, and gravitational wave observations, increasingly points toward the existence of a ``dark" sector that cannot be explained within the framework of the Standard Model of particle physics and General Relativity. In this review, we explore extensions of standard physics and examine how observational data can be used to probe new particles and interactions. We consider a wide range of scales, from Solar System tests to galactic and cosmological observations, and investigate both conventional dark matter candidates, such as weakly interacting massive particles, and alternative scenarios including ultralight fields and primordial black holes. We discuss constraints derived from compact objects such as neutron stars, black holes, pulsars, and magnetars observations as well as from high-energy astrophysical phenomena. In addition, we analyze extensions of General Relativity involving additional scalar fields and their impact on gravitational wave signals and stochastic backgrounds from primordial black holes. We also study the capture and accumulation of dark matter in compact objects, which can alter properties such as mass, radius, and tidal deformability, and consider scenarios in which dark matter decays into Standard Model particles. While current observations already place significant limits on dark matter and modified-gravity models, upcoming experiments and observatories are expected to further probe or discover such new physics by improving constraints on particle masses and interaction strengths.

Stage-IV photometric galaxy surveys are designed to measure the position and shapes of billions of galaxies. Their aim is to characterise the large-scale distribution of matter in the Universe using galaxy clustering and weak gravitational lensing. As a byproduct, stage-IV surveys are expected to detect more than a hundred thousand strong gravitational lenses. In this article, we propose the use of weak-lensing perturbations to strong lenses, specifically their line-of-sight (LOS) shear, as a cosmological probe. This new observable allows us to define three new correlation functions: the LOS shear with itself, with galaxy positions, and with galaxy shapes, thereby promoting the standard $3\times 2$pt correlation method to a $6\times 2$pt scheme. We design estimators for these new correlation functions and determine their expectation values as a function of the matter power spectrum. We then derive the analytical expression for the full covariance matrix of the $6\times 2$pt correlation scheme. Considering various scenarios for the stage-IV strong-lensing samples, we demonstrate that the cosmological information carried by the LOS shear of strong lenses will be detectable with a very high signal-to-noise ratio, even in the most pessimistic of cases. Strong lenses are thus extremely promising cosmological probes, whose synergy with galaxy positions and shapes should also contribute to mitigating systematics in stage-IV surveys.

We consider a metric-affine formulation of Chern-Simons modified gravity in 2 + 1 dimensions. The theory is built requiring projective invariance, and the structure of the equations is analyzed using a decomposition in terms of scalar, vectorial, and purely tensorial objects. This approach allows us to implement a perturbative approach to study the corrections that emerge around a given background solution, for which we consider a BTZ charged, non-rotating metric. We show that conditions on model parameters are necessary to keep perturbations under control, yielding a rotating solution with a constant angular momentum and magnetic field at the horizon, and a smooth decay further away. We comment on the possibility of going beyond the leading order in perturbations and on its dynamical implications.

This work investigates the cosmology of a thick brane within the context of $f(\mathbb{Q})$ gravity, an extension of symmetric teleparallelism. Using a five-dimensional Friedmann-Lemaître-Robertson-Walker metric, we solve the field equations to obtain dynamic solutions for the scale factor. We demonstrate that the effective cosmological constant on the brane naturally emerges as a function of the extra dimension $Λ(y)$, being both generated and confined by the curved geometry of the bulk. We analyze two distinct regimes: Randall-Sundrum-type thin brane and thick brane through the Sine-Gordon model. Our model reproduces accelerated expansion solutions without requiring the introduction of a fundamental cosmological constant on the brane, showing that the cosmic acceleration emerges as a consequence of the brane's embedding and the gravitational dynamics in the bulk . The variation of the $c_i$ parameters in symmetric teleparallelism enables different cosmological scenarios, including de Sitter-type expansion, contraction, and oscillatory solutions. The results indicate that the brane's position in the bulk determines its cosmology, providing a geometric explanation for the smallness of the observed cosmological constant.

We introduce a quantifier of phase-space complexity for discrete-variable (DV) quantum systems. Motivated by a recent framework developed for continuous-variable systems, we construct a complexity measure of quantum states based on the Husimi Q-function defined over spin coherent states. The quantifier combines into a single scalar quantity two complementary information-theoretic quantities, the Wehrl entropy, which captures phase-space spread, and the Fisher information, which captures localization. We derive fundamental properties of this measure, including its invariance under SU(2) displacements. The complexity is normalized such that coherent states have unit complexity, while the completely mixed state has zero complexity, a feature distinct from the continuous-variable case. We provide analytic expressions for several relevant families of states, including Gibbs and Dicke states, and perform a numerical analysis of spin-squeezed states, NOON states, and randomly generated states. Numerical results reveal a monotonic, but not deterministic, relationship between complexity and purity, leading us to conjecture that maximal complexity is attained by pure states, thereby connecting the problem to the optimization of Wehrl entropy via Majorana constellations. Finally, we extend the framework to quantum channels, defining measures for both the generation and breaking of complexity. We analyze the performance of common unitary gates and the amplitude damping channel, showing that while low-dimensional systems can achieve maximal complexity via spin squeezing or NOON states, this becomes impossible in higher dimensions. These results highlight dimension-dependent limitations in the generation of phase-space complexity and establish a unified phase-space approach to complexity across both continuous and discrete variable regimes.

We investigate an approach to determine the correct Poisson brackets of fields restricted to codimension 2 and 3 surfaces in 4D gravity, which are of great potential use in holographic setups and discretisation. Employing a specific BF-BB type parametrisation of gravity which relaxes Plebanski's simplicity constraints, we find that gravity in 4 dimensions carries Chern-Simons like phase spaces in codimension 2 and Kac-Moody algebras in codimension 3. The necessary gauge algebra in this context shows that the appropriate generalisation of the double $\mathcal{D}\mathfrak{so}(1,2)$ of 3D gravity is the Maxwell algebra, $\mathfrak{g}=\mathfrak{so}(1,3)\ltimes(\mathbb{R}^{1,3}\tilde\oplus \mathfrak{so}(1,3)^\ast)$. This realises the corner Poisson bracket of the spin connection for the first time and shows it is off-shell commutative, while the corner metric is noncommutative.

Typical gravitational production of relics from amplification of inflationary perturbations assumes Bunch-Davies initial conditions, i.e. a vacuum with initially no particles. In this paper we investigate the impact of non Bunch-Davies initial conditions to the final abundance of relics, with particular attention to the parameter space where the total dark matter abundance is reproduced. We present a general framework for any initial condition, through which we show their non-trivial effect on both spectrum and late-time abundance. We argue that for particles whose source of conformal symmetry breaking comes only from a mass term (spin-1/2 fermions and conformally coupled scalars), the choice of initial conditions has little impact on the mass range relevant to dark matter. For other particles, e.g. the longitudinal mode of spin-1, we see a large deviation from the standard computation. We exemplify and quantify our results with an initial thermal state and a two-stage inflation scenario, highlighting that the total dark matter can be obtained for a wide range of masses.

We study gravimetry with bosonic trapped atoms in the presence of random spatial inhomogeneity. The errors resulting from a random, shot-to-shot fluctuating spatial inhomogeneity are quantum non-Markovian. We show that in a system with $L>2$ modes (i.e., trapping sites), these errors can be post-corrected using a Bayesian inference. The post-correction is done via in situ measurements of the errors and refining the data-processing according to the measured error. We define an effective Fisher information $F_{\text{eff}}$ for such measurements with a Bayesian post-correction and show that the Cramer-Rao bound for the final precision is $\frac{1}{\sqrt{F_{\text{eff}}}}$. Exploring the scaling of the effective Fisher information with the number of atoms $N$, we show that it saturates to a constant when there are too many sources of error and too few modes. That is, with $\ell$ independent sources of error, we show that the effective Fisher information scales as $F_{\text{eff}} \sim \frac{N^2}{a+bN^2}$ for constants $a, b>0$ when the number of modes is small: $L<\ell+2$, even after maximization over the Hilbert space. With larger number of modes, $L\geq \ell+2$, we show that the effective Fisher information has a Heisenberg scaling $F_{\text{eff}}= O(N^2)$ when optimized over the Hilbert space. Finally, we study the density of the effective Fisher information in the Hilbert space and show that when $L\geq \ell+2$, almost any Haar random state has a Heisenberg scaling, i.e., $F_{\text{eff}}=O(N^2)$. Based on these results, we develop a Loschmidt echo-like experimental sequence for error mitigated gravimetry and gradiometry and discuss potential implementations. Finally, we argue that the effective Fisher information can be interpreted as the Fisher information corresponding to an equivalent non-Hertimitian evolution.

The non-linearly coupled Einstein-matter field equations predict the existence of shadows with well-defined boundaries around black holes. We prove that, in spherically symmetric hairy black-hole spacetimes whose matter fields satisfy the weak energy condition, the radii of these shadows are bounded from below by the dimensionless relation $r_{\text{sh}}/r_{\text{H}}\geq 3\sqrt{3}/2$, where $r_{\text{H}}$ is the horizon radius of the central hairy black hole. The characteristic shadow of the (bald) Schwarzschild black-hole spacetime saturates the analytically derived lower bound.

In this paper, we study the $D\to3$ limit of Gauss-Bonnet gravity with quintessential matter, obtaining exact solutions that extend the BTZ metric through higher-curvature terms and quintessence coupling. The solutions exhibit a single event horizon whose radius decreases with increasing quintessence parameter $ω_q$, while developing a curvature singularity at the origin for non-vanishing quintessence. The geodesic analysis reveals stable circular photon orbits exist exclusively for phantom-like quintessence ($ω_q < -1$). Thermodynamically, the system is stable, since the specific heat is positive, and with evaporation it evolves to stable remnants whose characteristic size decreases as $ω_q$ increases, with complete evaporation prevented by quintessence effects. Furthermore, we find that all physical quantities intrinsically depend on the parameter $α$ of the Gauss-Bonnet extension.These results demonstrate the profound influence of quintessential matter on both geometric and thermodynamic properties of (2+1)-dimensional black holes, offering new perspectives on gravitational theories in lower dimensions and black hole final states.

We will address the existence of a new symmetry for an imperfect fluid by introducing local four-velocity gauge-like transformations for the case when there is vorticity. A similar tetrad formulation as to the Einstein-Maxwell spacetimes formalism presented in previous manuscripts will be developed in this manuscript for the imperfect fluids. The four-velocity curl and the metric tensor will be invariant under these kind of four-velocity gauge-like local transformations. While the Einstein-Maxwell stress-energy tensor is locally gauge invariant under electromagnetic gauge transformations, the perfect fluid stress-energy tensor will not be invariant under four-velocity gauge-like local transformations. We will dedicate our analysis to the imperfect fluid stress-energy tensor that will be invariant under local four-velocity gauge-like transformations when additional transformations are introduced for several variables included in the stress-energy tensor itself. We will also pay special attention to the construction of a vorticity stress-energy tensor invariant under local four-velocity gauge-like transformations. An application on neutron stars will be developed in order to show the simplifications brought about by these new tetrads.

We investigate the steady, axisymmetric, force-free magnetosphere of Kerr-Sen black hole (BH) within the framework of the Einstein-Maxwell-dilaton-axion (EMDA) theory. By perturbatively solving the nonlinear Grad-Shafranov (GS) equation, we determine the magnetic field configuration and quantify the influence of the dilaton parameter $r_2$ on the energy extraction rate and radiative efficiency. Our results show that both the energy extraction power and the radiative efficiency increase with $r_2$, exceeding those of the standard Kerr BH, whereas the extraction efficiency remain consistent with the Kerr case. In addition, we perform $χ^2$ statistical analysis using observational data from six binary BH systems, which indicates that the Kerr BH currently provides a better fit for bulk Lorentz factors $Γ= 2$ and $5$.

We investigate a possible resolution of the dark energy problem within a pair-universe framework, in which the Universe emerges as an entangled pair of time-reversed sectors. In this setting, a global zero-energy condition allows vacuum energy contributions from the two sectors to cancel, alleviating the need for extreme fine-tuning. We propose that the observed dark energy does not originate from vacuum fluctuations but instead arises as an effective entanglement energy between the visible universe and its mirror counterpart. Treating the cosmological constant as an integration constant fixed by boundary conditions rather than a fundamental parameter, we show that the cosmological equations can be formulated without explicitly introducing vacuum energy. By imposing physically motivated boundary conditions at the cosmological event horizon, we obtain an integration constant consistent with the observed dark energy density. The parallel mirror world scenario thus provides a unified framework that may simultaneously explain the origins of dark energy and dark matter.

This research applies the generalized technique of gravitational decoupling to the Bardeen black hole, producing novel black hole solutions in the context of Rastall theory. We proceed by decomposition of the field equations corresponding to an additional matter source into two sets, for further considerations. The metric functions of the Bardeen black hole are adopted to specify the first set. The second one, which is subject to an extra source, is resolved considering a linear equation of state of matter. Through the integration of the solutions of these sets, we develop two expanded models and conduct an in-depth analysis of their distinct physical characteristics, governed by specific parameters. We investigate thermodynamic quantities like density, anisotropic pressure, energy bounds, asymptotic flatness, and thermodynamical properties like the Hawking temperature, entropy, and specific heat, etc. Both models are asymptotically flat but violate the energy bounds. Furthermore, the density, radial pressure, and Hawking temperature demonstrate consistent and acceptable behavior. Ultimately, the thermodynamic stability is affirmed through the analysis of specific heat and the Hessian matrix.

With the advent of quantum simulators, exploring exotic collective phenomena in lattice models with local symmetries and unconventional geometries is at reach of near-term experiments. Motivated by recent progress in this direction, we study a $\mathbb{Z}_2$ lattice gauge theory defined on a multi-graph with links that can be visualized as great circles of a spherical shell hosting the $\mathbb{Z}_2$ gauge fields. Elementary Wilson loops along pairs of these bonds allow to identify a dynamical gauge-invariant flux, responsible for Aharonov-Bohm-like interference effects in the tunneling dynamics of charged matter residing on the vertices. Focusing on an odd number of links, we show that this leads to state-dependent tunneling amplitudes underlying a phenomenon analogous to the Peierls instability. We find inhomogeneous phases in which an ordered pattern of the gauge fluxes spontaneously breaks translational invariance, and intertwines with a bond order wave for the gauge-invariant kinetic matter operators. Long-range order is shown to coexist with symmetry protected topological order, which survives the quantum fluctuations of the gauge flux induced by an external electric field. Doping the system above half filling leads to the formation of topological soliton/anti-soliton pairs interpolating between different inhomogeneous orderings of the gauge fluxes. By performining a detailed analysis based on matrix product states, we prove that charge deconfinement emerges as a consequence of charge-fractionalization. Quasiparticles carrying fractional charge and bound at the soliton centers can be arbitrarily separated without feeling a confining force, in spite of the long-range attractive interactions set by the small electric field on the individual integer charges.

Single photons enable the distribution of quantum information over large distances and thus play a major role in quantum technologies such as communication and computing. Solid-state emitters are practical and efficient sources of single photons that can be manufactured in large numbers. When combined with a spin, the resulting spin-photon interfaces can store quantum states for extended periods and serve as the basis for quantum networks and repeaters. Among the many host materials explored over the past few decades, silicon stands out for its advanced nanofabrication, the maturity of its integrated photonics and microelectronics, and its high isotopic purity, which leads to exceptionally long spin coherence. These properties position silicon single-photon emitters and spin-photon interfaces among the most promising hardware platforms for implementing quantum networks and distributed quantum information processors. This review summarizes the current state of the art and open challenges towards coherent single-photon sources and scalable spin-photon interfaces based on color centers and erbium dopants in nanophotonic silicon structures.

Among the most compelling cold dark matter candidates, the axion has recently been subject to a wide range of astrophysical studies aiming to constraints its properties. We present updated bounds on the isocurvature fraction, $f_{\rm{iso}}$, which parameterizes the contribution of isocurvature perturbations induced by post-inflationary produced axion-like particles (ALPs) to the ordinary power spectrum. We use new simulations based on the Sherwood-Relics suite to fit high-resolution Lyman-$α$ forest flux power spectrum data. With the published noise model of the Lyman-$α$ forest data, we find a tentative detection of $f_{\rm{iso}}$ = ${0.0064^{+0.0012}_{-0.0014}}$ (68% C.L), after accounting for the degenerate effect of IGM thermal evolution. With a more conservative modelling of the residual noise in the data, the upper bound is weakened to $f_{\rm{iso}}< 0.0084$ (95% C.L), which translates into an ALP temperature-independent mass $m_a > 1.73 \times 10^{-18}$eV. Our constraints are stronger than bounds derived from large-scale structure probes at higher and lower redshifts and are competitive with those derived from UV luminosity function data. Interestingly, the best current Lyman-$α$ forest data prefers a non-zero contribution from isocurvature modes.

We study light spin-1 gauge bosons coupled to electroweak-anomalous currents. For generic charge assignments, anomaly cancellation requires new fermions (anomalons) that are chiral under the new abelian symmetry and carry electroweak charges. If their masses arise from the breaking of the new gauge symmetry, integrating them out generates Wess-Zumino interactions fixed by mixed-anomaly matching, providing the infrared description of the theory. We classify minimal anomalon spectra, derive the corresponding effective interactions, and combine experimental constraints with finite-naturalness considerations to bound the UV completion scale. Motivated by recent NA62 and Belle II results, we then develop a unified phenomenological framework for the missing-energy signatures of these anomalous light vectors, focusing on scenarios where the new vector decays predominantly into neutrinos so that the leading probes are rare processes with invisible final states. As applications, we survey current and projected searches across flavour and electroweak observables, including $K\toπE_{\rm miss}$, $B\to K^{(*)}E_{\rm miss}$, and $Z\toγE_{\rm miss}$, and discuss their interplay with direct searches for anomalons.

This work investigates the influence of Lorentz symmetry breaking, introduced by an aether-like field $α_φ$, on the Casimir effect within a five-dimensional flat spacetime. By considering a quasiperiodic condition regulated by the parameter $β$ and an extra dimension compactified at scale $b$, we derive closed-form expressions for the Casimir energy and the resulting force between two parallel plates under Neumann boundary conditions. Our results demonstrate that $β$ acts as a crucial control parameter, enabling a continuous transition between attractive and repulsive regimes, with a characteristic symmetry around $β= 0.5$. We show that the Lorentz-violating parameter $α_φ$ functions as an enhancement factor, significantly amplifying the vacuum interaction, while the geometric ratio $a/b$ proves decisive for system stabilization. Specifically, we find that the high-compactification regime leads to a plateau in the Casimir force, effectively stabilizing the interaction. Furthermore, we analyze the mass spectrum of the field, recovering standard geometric forms in the massless limit and demonstrating that while light fields ($M \ll 1$) exhibit subtle quadratic corrections, heavy fields ($M \gg 1$) lead to an exponential suppression of the Casimir effect. The interplay between Lorentz violation and extra-dimensional compactification provides a rich mechanism with potential applications in the modulation of vacuum-induced interactions at micro and nano scales.

Lorentz invariance is a fundamental symmetry underlying both the Standard Model of particle physics and General Relativity. Testing its validity provides a direct means of searching for new physics emerging near the Planck scale. A search for isotropic Lorentz invariance violation with 1.4 years of atmospheric neutrino data collected by a partial configuration of the KM3NeT/ORCA detector comprising six detection units is presented. No evidence for such violation is found; thus, competitive limits are set on a subset of isotropic Lorentz invariance violating coefficients, which complement and extend existing experimental constraints.

We develop a quantum kinetic theory for QCD, which incorporates all leading order collision terms. At lowest order in gradient expansion, it reproduces the spin-averaged Boltzmann equation with both elastic and inelastic collisions. At next order in gradient expansion, the solution to the quantum kinetic equations give spin polarization of on-shell quarks and gluons in quark-gluon plasma when the gradients are of hydrodynamic ones. A power counting in the coupling shows the spin polarization behaves differently in vortical and non-vortical gradients: the former is free of collisional contribution to leading order, while the latter contains a collisional contribution at parametrically the same order as the free theory counterpart. We also find the inelastic collision in a spin basis provides a possible mechanism for conversion between spin and orbital angular momentum.

The forward--backward asymmetry ($A_{FB}$) in light-quark production is a sensitive probe of the electroweak sector and potential flavour-dependent BSM effects. We present a study of $e^+e^- \rightarrow s\bar{s}$ at $\sqrt{s}=250$ GeV at future linear colliders, using full ILD simulation and reconstruction tools for ILC and LCF@CERN. We assess the impact of particle identification on charge reconstruction and $A_{FB}$ extraction, considering software improvements using Comprehensive PID (CPID) for optimal $dE/dx$ usage, as well as hardware scenarios including cluster counting ($dN/dx$) and ideal TPC performance. Statistical precision gains are evaluated, with corrections for charge misidentification and acceptance applied. Results indicate that precise $A_{FB}^{s\bar{s}}$ measurements are feasible and that advanced PID is key to maximising sensitivity to electroweak and new-physics effects.

We present a unified framework for heavy tetraquark and pentaquark systems within the quark-diquark effective mass formalism, extending its baryon-calibrated construction to multiquark states without introducing sector-dependent parameters. Intra-diquark color-spin correlations are encoded in effective diquark masses fixed from baryon spectroscopy, while the inter-cluster chromomagnetic scale, independently determined from vector-pseudoscalar meson splittings, is propagated unchanged to exotic configurations, ensuring residual one-gluon exchange dynamics only between composite color sources. Within this framework, we compute the complete spectra for both $\bar{\mathbf{3}}_c\otimes\mathbf{3}_c$ and $\mathbf{6}_c\otimes\bar{\mathbf{6}}_c$ configurations in tetraquarks, whereas the pentaquark analysis focuses on the dominant $\bar{\mathbf{3}}_c\otimes\bar{\mathbf{3}}_c\otimes\bar{\mathbf{3}}_c$ clustering. Heavy-quark spin symmetry and flavor-symmetry breaking across light, charm, and bottom sectors emerge naturally through the explicit $1/(m_{D_1}m_{D_2})$ scaling of the calibrated couplings. The resulting spectra exhibit a coherent dynamical hierarchy spanning baryons and multiquark states. Established exotic candidates are reproduced within hadronic uncertainties, while the unified calibration enables quantitative predictive control across flavor sectors. The framework thus provides a parameter-economical, systematically constrained baseline with unified dynamical consistency for heavy multiquark spectroscopy.

One of the three common experimental channels searching for charged lepton flavor violation (CLFV) via $μ-e$ conversion involves first capturing the muon on a nuclear target. This channel provides a unique opportunity to constrain four-point quark-lepton operators that may contribute to CLFV, in addition to the electromagnetic operators that can also be accessed in the other channels. We investigate the leading Lorentz- and CPT-violating contributions for nuclear $μ-e$ conversion experiments within the Standard-Model Extension and obtain the first bounds on the relevant quark-lepton operators using the results of the SINDRUM II experiment; we also consider possibilities for improved constraints from upcoming experiments.

In this letter, we explore the phenomenological impact of inflationary gravitational particle production in the physics of Dark Matter (DM). Large-scale DM fluctuations generated during inflation behave as gravitational particles upon their post-inflationary horizon reentry and alter the conventional Boltzmann dynamics of DM with a non-conserving source term, thereby producing significant phenomenological consequences. Within this framework, we analyze four distinct types of DM classified according to their production mechanisms. Dark matter may be completely non-interacting with the thermal bath, behaving as Inert Dark Matter. Alternatively, depending on the strength of its interactions with bath particles, DM may exhibit WIMPy, UFOy, or FIMPy behavior, sharing characteristics with their conventional counterparts. The late-time enhancement of the DM number density, driven by the successive horizon reentry of gravitationally produced low-momentum modes, enlarges the viable parameter space for both thermal and non-thermal DM scenarios. Remarkably, this expanded parameter space remains consistent with current constraints from $ΔN_{\rm eff}$ and Lyman-$α$ bound.

Many new physics models, such as the Sequential Standard Model, Grand Unified Theories, models of extra dimensions, or models like leptoquarks or vector-like leptons, predict heavy mediators at the TeV energy scale. We present recent results of such searches in leptonic final states obtained using data recorded by the CMS experiment during Run-II of the LHC.

The BESIII experiment is currently the world's only electron-positron collider operating in the tau-charm physical energy region. Since starting data taking in 2009, BESIII has accumulated the world's largest data set in the center-of-mass energy range of 1.84-4.95 GeV, including approximately 10 billion $J/ψ$ events and 3 billion $ψ(3686)$ events, together with extensive data on open-charm hadron pair production near threshold regions. These unique datasets, characterized by high statistics and low background, provide unprecedented experimental conditions for studying light baryon spectroscopy. This article systematically reviews the progress made by BESIII in baryon spectroscopy, with a focus on recent breakthrough achievements, including the discovery of excited nucleon states, $Λ$ hyperon states, $Σ$ hyperon states, $Ξ$ hyperon states and $Ω^{-}$ hyperon states. These results expand the spectrum of baryon excited states and provide crucial experimental support for understanding non-perturbative QCD and resolving the ``missing baryon resonances'' problem.

We report a study of the semileptonic decays $D^0 \rightarrow \bar{K}^0π^-\ell^+ν_{\ell}$ (where $\ell=e,~μ$) based on a sample of $20.3~\mathrm{fb}^{-1}$ of $e^+e^-$ annihilation data collected at a center-of-mass energy of 3.773~GeV with the BESIII detector at the BEPCII collider. Based on an investigation of the decay dynamics in $D^0 \rightarrow \bar{K}^0π^-\ell^+ν_{\ell}$ decays, a $\mathcal{D}-$wave component of $D^0\rightarrow K_2^*(1430)^-\ell^+ν_{\ell}$ is observed for the first time with a statistical significance of $8.0σ$, in addition to the dominant $K^*(892)^-$ and $\mathcal{S}-$wave components. The $\mathcal{D}-$wave component is determined to account for $(0.092 \pm 0.028_{\rm stat} \pm 0.018_{\rm syst})\%$ of the total decay rate. The branching fractions of the dominant $K^*(892)^-$ components are measured as $\mathcal{B}(D^0\rightarrow K^{*}(892)^-e^+ν_{e}) = (2.043 \pm 0.018_{\rm stat} \pm 0.012_{\rm syst})\%$ and $\mathcal{B}(D^0\rightarrow K^{*}(892)^-μ^+ν_μ) = (1.964 \pm 0.018_{\rm stat} \pm 0.012_{\rm syst})\%$, which are the most precise measurements to date and represent significant improvements over the previous world averages. The hadronic form-factor parameters are measured to be $r_{V} = V(0)/A_1(0) = 1.444 \pm 0.026_{\rm stat} \pm 0.010_{\rm syst}$, $r_{2} = A_2(0)/A_1(0) = 0.752 \pm 0.020_{\rm stat} \pm 0.004_{\rm syst}$, and $A_1(0)=0.618\pm0.002_{\rm stat} \pm0.004_{\rm syst}$, where $V(0)$ is the vector form factor and $A_{1,2}(0)$ are the axial-vector form factors evaluated at $q^2=0$. This is the most precise determination of the form-factor parameters to date measured in a $D\rightarrow K^*(892)$ transition. In addition, we report the first model-independent measurement of the $\mathcal{S}-$wave phase shift in the hadronic $\bar{K}^0π^-$ system.

This paper evaluates the radiation tolerance of commercial off-the-shelf (COTS) electronics components for use in the Thin Gap Chamber (TGC) frontend electronics of the ATLAS experiment at the High-Luminosity LHC (HL-LHC). The ATLAS experiment has accumulated more than 450 fb^-1 of data as of 2025. Its luminosity upgrade, the HL-LHC scheduled to begin operation in 2030, will deliver 3000-4000 fb^-1 over ten years and lead to substantially higher radiation levels in detector electronics. The radiation levels for the TGC frontend electronics are estimated to be 4.1-7.3 Gy in terms of Total Ionizing Dose (TID) and 1.1-2.2 x 10^11 n_1MeV cm^-2 in terms of Non-Ionizing Energy Loss (NIEL). To evaluate component suitability under these conditions, TID tests were conducted using Cobalt-60 gamma rays at Nagoya University, and NIEL tests were performed with the Tandem Accelerator at Kobe University. Various COTS components, including SFP+ optical transceivers, clock jitter cleaners, optical fibers, voltage references, operational amplifiers, analog-to-digital converters, digital-to-analog converters, SD cards, flash memories, and low-dropout regulators, were tested and evaluated against the required radiation levels. The results demonstrate that all evaluated components meet the TID and NIEL tolerance requirements for application in the TGC frontend electronics at the HL-LHC.

We prove, under suitable global assumptions, that the only heterotic horizons with closed 3-form field strength that preserve strictly 6 supersymmetries have spatial horizon section diffeomorphic to $SU(3)$, up to identifications with the action of a discrete group. Under similar assumptions, which include the compactness of the transverse space, we demonstrate that there are no heterotic AdS$_3$ solutions that preserve 6 supersymmetries. The proof is based on a topological argument. We also re-examine the conditions required for the existence of such backgrounds that preserve 4 supersymmetries focusing on those that admit an additional $\oplus^2\mathfrak{u}(1)$ symmetry. We provide some additional explanation for the existence of solutions and point out the similarities that these conditions have with those that have recently emerged in the classification of compact strong 6-dimensional Calabi-Yau manifolds with torsion.

Within the framework of SU(2) chiral perturbation theory, we derive the general solution of the QCD $θ$-vacuum for an arbitrary vacuum phase, explicitly incorporating isospin-breaking effects from the light quark mass difference, and compute the temperature dependence of the topological susceptibility, higher-order cumulants, and the domain wall tension up to next-to-leading order. We find that the topological susceptibility agrees with lattice data at low temperatures but deviates at higher temperatures as expected from the breakdown of the chiral expansion; moreover, we demonstrate that the normalized fourth-order cumulant and the domain wall tension decrease monotonically with increasing temperature, while the normalized sixth-order cumulant exhibits the opposite behavior. These results extend earlier analyses by showing how isospin breaking reshapes the full hierarchy of topological charge cumulants and the dynamics of $θ$-vacuum domain walls, thereby offering new theoretical input on the $θ$-vacuum properties, which are relevant for axion-related effective theories in hot QCD matter.

The CMS muon system is undergoing substantial upgrades to meet the challenges of the High-Luminosity LHC (HL-LHC), including the installation of the new Muon Endcap 0 (ME0) detector. Large-scale production started in 2024. ME0 is a six-layer station designed to extend pseudo-rapidity coverage to |η| = 2.8 from the previous maximum of |η| = 2.4, enhancing sensitivity to forward physics processes. Each endcap will host 18 ME0 stacks, with each stack comprising six triple-layer gas electron multiplier (GEM) chambers. The system adds up to six additional hits per track, which significantly improves muon identification, spatial resolution, and robust track reconstruction at the first trigger level. Chamber production and quality control across multiple international sites ensure scalability and timely delivery. The ME0 design incorporates lessons learned from earlier GEM deployments, with improvements in electronics robustness, grounding, and segmentation to withstand high background rates and minimize damage from discharges. This contribution provides a comprehensive overview of the ME0 detector concept, assembly strategy, quality assurance procedures, current production status, and its pivotal role in strengthening CMS muon reconstruction during HL-LHC operations.

We present a dynamical picture of kink-anti-kink scattering in a pair of special, Frankensteinian potentials made of piece-wise quadratic and linear pieces. Specifically, we focus on models that support kinks without skin and core regions. We propose an intuitive interpretation for these models as being essentially free massive theories with a built-in particle-pair like production mechanism that enters into the dynamics above certain field-value thresholds. We present results concerning the kink's characteristics depending on these thresholds and the distribution of bouncing windows. We show that the second model exhibits a phase-transition-like property in which the nature of collisions switches from disintegration into a massive wave to production of oscillons for large segments of initial velocities when the field threshold is low enough.

For over two decades, the gravity sector of the Standard-Model Extension (SME) has served as a phenomenological framework for testing spacetime symmetry breaking in the presence of gravity. During this time, various theoretical features have been examined in greater detail and some refinements have been made. In particular, differences between spontaneous and explicit breaking of diffeomorphisms, local translations, and local Lorentz transformations in Riemann-Cartan geometry, as well as their corresponding consistency issues with geometric and mathematical identities, have been probed more deeply. This has led to a modified version of the SME being developed that is suitable for investigating explicit breaking in gravity theories, which can be used as well to search for new geometries that go beyond Riemann-Cartan. A selective overview of some of these features is presented here.

We report a complete computation of next-to-next-to-leading order (NNLO) helicity and transversity Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) splitting functions, in both space-like and time-like kinematics. These results are obtained from the next-to-next-to-next-to-leading order (N$^3$LO) twist-2 matching of polarized transverse-momentum-dependent (TMD) parton distribution and fragmentation functions, including helicity, quark transversity, and linearly polarized gluons. We compare our results with existing calculations in the literature and discuss both agreements and discrepancies. Our results provide all perturbative ingredients required for the computation of N$^3$LO differential cross sections below the resolution scale $q_{T\mathrm{cut}}$ in transverse-momentum subtraction and enable next-to-next-to-next-to-next-to-leading logarithmic (N$^4$LL) resummation of $q_T$ observables in the Sudakov region. We further determine the small-$x$ structure of the polarized matching coefficients through N$^3$LO. These fixed-order results furnish the data for future small-$x$ resummation in polarized TMD factorization, where high-energy logarithms and Sudakov logarithms become simultaneously relevant. Establishing a consistent joint treatment of polarized small-$x$ evolution and transverse-momentum resummation remains an important open direction toward uniform precision in spin-dependent phenomenology. Our results provide essential theoretical input for precision spin physics at the forthcoming Electron-Ion Collider.

The contribution of the weak interaction to the hyperfine splitting of the ground state of muonium is investigated. The amplitudes of the one- and two-quantum exchange determined by the Z and W bosons are calculated. One-loop corrections in the photon and Z boson propagators and their contribution to the hyperfine structure of the spectrum are obtained.

We investigate the mass and strong decay properties of the $Ω(2012)$ resonance using QCD sum rules, assuming it to be an S-wave $Ξ(1530)K$ molecular pentaquark state with $IJ^{P} = 0\frac{3}{2}^{-}$. A unified interpolating current is constructed, and the two-point and three-point correlation functions are calculated up to dimension-10 condensates in the OPE series. The negative-parity contribution is isolated by employing parity-projected sum rules. The two-body strong decays to $Ξ^0 K^-$ and $Ξ^- K^0$ are studied via their three-point correlation functions. Our analysis yields a mass of $2.00 \pm 0.15~\mathrm{GeV}$ and a total two-body decay width of $Γ= 3.97^{+8.31}_{-1.92}~\mathrm{MeV}$ for the $Ξ(1530)K$ molecular state. The ratios of branching fractions are obtained as $\mathcal{R}^{Ξ^- K^0}_{Ξ^0 K^-} = 0.85$ and $\mathcal{R}^{ΞπK}_{ΞK} = 0.61$. These results are compatible well with the experimental data for the $Ω(2012)$ and support its interpretation as a $Ξ(1530)K$ molecular pentaquark state.

We study the indivisibility of a quantum-corrected AdS black hole involving two kinds of global monopoles, regular or phantom ones to declare the effects from the quantum fluctuation, monopole factors and the negative cosmological constant on the evolution of black holes. We focus our attention on the possibility that this kind of black holes break into their own two parts because of the second law of thermodynamics. We derive and calculate the entropies of the initial black hole and the broken parts respectively and the entropy difference relates to the black holes structure including the quantum-gravity oscillation, global monopoles and the influence from AdS environment. It is found that the two kinds of global monopoles both keep the black holes intact instead of splitting because the entropy difference is negative. The considerable quantum fluctuation compels the defference to be positive and the fragmentation of the black holes may happen in the case that the mass of one fragmented black hole is tiny and the other one's mass is huge. The AdS radius can increase the whole value of change in entropy and the difference near the midpoint of mass ratio will become positive, so the large enough radius may lead the isolated black hole to become two parts nearly equal in the mass.

Leptogenesis from low-energy CP violation provides a vital link between neutrino physics and the observed baryon asymmetry of the universe. However, this connection is typically obscured by unknown high-energy parameters. In this work, we investigate thermal leptogenesis in the Minimal Left-Right Symmetric Model with generalized parity as the left-right symmetry, where the hermiticity of the Dirac neutrino coupling allows the right-handed mixing matrix $V_\mathrm{R}$ to be determined with minimal assumptions. We show that for a real $V_\mathrm{R}$, these conditions favor CP-conserving Majorana phases, leaving the Dirac CP-violating phase ($δ$) as the sole source of asymmetry. By numerically exploring all four leptogenesis scenarios, we demonstrate that $δ$ alone can generate the observed baryon asymmetry with the correct sign within specific regions of the parameter space. The results exhibit a high sensitivity to the neutrino mass ordering and the lightest neutrino mass, providing a stringent, testable framework for future experimental measurements of the CP phase and neutrino mass scale.

For twenty years, a persistent discrepancy between experimental measurements and theoretical calculations of the muon anomalous magnetic moment have provided tantalising hints of new physics. In recent years, improvements to the experimental precision have appeared to make the tension stronger and stronger. However, at the same time, our lattice calculation overturned the theoretical consensus, completely eliminating the tension. I will present the latest results from the Budapest-Marseille-Wuppertal (BMW) and DMZ collaborations, with a hybrid determination of the hadronic vacuum polarisation contribution to a precision of 0.45%

The Collins-Soper (CS) kernel may be obtained through the TMD soft function by formulating the Wilson line in terms of 1-dimensional auxiliary fermion fields on the lattice. Our computation takes place in the region of the lattice that corresponds to the "spacelike" region in Minkowski space, i.e., Collins' scheme. We explore two methods for obtaining the CS kernel. The "ratio method"; which would allow us to obtain the soft function as well as the CS kernel. And the "double ratio"; which allows us to achieve a high degree of statistical precision, but only produces the CS kernel. The matching of our result to Minkowski space is achieved through the mapping of the complex auxiliary field directional vector to the Wilson line rapidity. We present a preliminary extraction of the CS kernel using the "double ratio", and discuss the methodology employed.

We investigate the $T_{cc}$ tetraquark, treating it as a bound state of a heavy diquark and a light antidiquark. Using the Silvestre-Brac potential and solving the Schrödinger equation via the Gaussian Expansion Method, we find that the excitation energy between the heavy diquark and light antidiquark is unexpectedly larger than that between the two light anti-quarks within the anti-diquark -- contrary to the naive expectation where the former is smaller than the latter. We trace this inversion of the mass hierarchy to the centrifugal force acting on the light degree of freedom. Applying the same framework to other systems ($T_{bb}, Λ_b, Λ_c$) yields qualitatively identical behavior, demonstrating the robustness of the mechanism. These results provide new insights into diquark dynamics and the mass structure of exotic hadrons.

The spatial extent of the proton is a key factor in nuclear physics. Different measurement techniques probe different aspects of the proton, yielding different radii. The mass and charge radii depend on the parton and quark distributions respectively, while the mechanical radius depends on the mass/energy distribution. Here, we probe the spatial distribution of a new proton characteristic, studying the distribution of baryon number within the proton. We investigate the baryon number distribution by studying four exclusive meson production channels arising from photon-proton collisions ($γp \rightarrow p ρ^0$, $γp \rightarrow p ω$, $γp \rightarrow n π^+$, and $γp \rightarrow p π^0$). The two-dimensional transverse sizes of the interacting systems are extracted by analyzing the transverse momentum, $p_T$, dependence of the meson production cross section, using Fourier-Bessel transformations. We find that baryon number is confined to a transverse radius of $0.33 - 0.53$~fm. In comparison, the transverse radius of the proton charge and mass distributions are considerably larger, at least 0.67~fm. The baryon number is concentrated in the center of the proton.

In this contribution, we present a recently introduced approach [BorkLeeOnishchenko2025] to the calculation of slightly off-shell dual conformal integrals based on the method of regions with regularization preserving dual conformal invariance (DCI). Unlike conventional dimensional regularization, which breaks DCI, our approach uses a combination of dimensional and analytic regularizations specifically designed to retain DCI throughout the calculation. Our approach drastically simplifies the computation of slightly off-shell dual conformal integrals. For the two-loop five-point DCI integrals we find that with DCI-preserving regularization, the contributions of all regions can be expressed in terms of $Γ$-functions, resulting in a remarkably compact final expression in terms of logarithms of cross-ratios only. This is in sharp contrast to conventional approach which yields complex polylogarithmic expressions [Belitsky&Smirnov2025]. We argue that a similar approach might be useful also for non-DCI integrals.

A study of angular correlations inside jets induced by gluon polarization is performed using proton-proton collisions at a center-of-mass energy of $\sqrt{s}$ = 13.6 TeV. The data correspond to an integrated luminosity of 34.7 fb$^{-1}$, collected in 2022 with the CMS detector at the LHC. The details of the parton shower are investigated using jets reconstructed with the anti-$k_\mathrm{T}$ algorithm and subsequently declustered with the Cambridge$-$Aachen algorithm. A novel analysis technique is developed to identify characteristic features of the jet substructure and to select intermediate gluon splittings into quark-antiquark pairs. An observable sensitive to gluon polarization in the parton shower is measured and compared with PYTHIA 8 and HERWIG 7 model predictions, with and without angular correlations induced by the gluon spin. The results are consistent with models that incorporate gluon polarization and strongly disfavor those that neglect them.

In this talk, we use several examples to elaborate on how a recently proposed algorithm can turn non-trivial Feynman integrals into an $\varepsilon $-factorised manner, regardless of their hidden geometric essence. In particular, some extra details about three-loop banana integrals with unequal-mass configuration are provided.

In the ALP model, lepton flavor violation (LFV) can arise from off-diagonal ALP-lepton couplings ($g_{a\ell_i\ell_j}$), which are absent in the Standard Model. We focus on ALP production via gluon fusion ($pp \to a$) at the Large Hadron Collider (LHC), which dominates the ALP mass range of 5-1000 GeV due to its high cross section and manageable backgrounds. We are interested in the decay of an axion into an electron and a muon, two oppositely charged leptons of different flavors. After background suppression, we obtain the sensitivity to ALP in this mass range, finding significantly improved limits for $5 < m_a < 1000$ GeV, where SM backgrounds are suppressed. Our research is complementary with relevant studies.

Quark and lepton hierarchies are organized by rational powers of a single parameter in units of one ninth. We show that this ``flavor in ninths'' structure points to a discrete $\mathbb{Z}_{18}$ gauge origin of Froggatt--Nielsen symmetry, whose $\mathbb{Z}_9$ subgroup controls the flavor lattice. Identifying the flavon with the Peccei--Quinn field, the same symmetry stabilizes the QCD axion, enforces $N_{\rm DW}=1$, and predicts $E/N=8/3$ (or $2$ with light higgsinos). The lowest Planck-suppressed operator appears at dimension eighteen, naturally solving the axion quality problem. For $f_a\sim(5$--$8)\times10^{11}$ GeV the axion accounts for dark matter and lies within near-term haloscope reach.

The HOMFLY-PT and Kauffman polynomials are related to each other for special classes of knots constructed by full twists and Jucys-Murphy twists. The conditions for this relation are articulated in terms of characters of the Birman-Murakami-Wenzl algebra. The latter are the coefficients in the expansion of the Kauffman polynomial involving the quantum dimensions of SO(N + 1). This expansion allows to prove the conjectural 1-1 correspondence between the HOMFLY-PT/Kauffman relation and the Harer-Zagier (HZ) factorisability for a large family of 3-strand knots. However, explicit counterexamples with 4-strands negate one side of the conjecture, i.e. the HOMFLY-PT/Kauffman relation only implies HZ factorisability for knots with braid index four or higher.

We study the one-dimensional Generalized Dirac Oscillator (GDO) under Doubly Special Relativity (DSR) kinematics. The GDO extends the Dirac oscillator by replacing the linear non-minimal coupling with a general interaction function $f(x)$, thereby generating broad families of exactly solvable relativistic models and, for suitable complex choices of $f(x)$, entering the domain of $η$-pseudo-Hermitian and $\mathcal{PT}$-symmetric dynamics with real spectra. We present a review of the factorization (supersymmetric) structure that decouples the GDO into partner Schrödinger-like Hamiltonians, and we clarify how pseudo-Hermiticity and $\mathcal{PT}$ symmetry provide consistent inner products and reality conditions for the spatial spectrum. We then embed these results into two representative DSR prescriptions: the Magueijo--Smolin (MS) and the Amelino--Camelia (AC) frameworks. In this approach, the spatial problem yields a real set $\{ε_n\}$, while DSR deforms the algebraic reconstruction map between $ε_n$ and the relativistic energies $E_n$. The MS model induces a branch-asymmetric deformation through an energy-dependent effective mass, whereas the AC model introduces a characteristic criticality through a momentum-sector deformation, resulting in an admissibility requirement of the form $ε_n<4k^2$ in the leading-order realization adopted here. As an explicit illustration, we treat a pseudo-Hermitian complexified Morse interaction, discuss the interplay between the intrinsic Morse finiteness of bound states and DSR-induced truncations, and analyze the massless limit ($m=0$), where MS collapses to the undeformed energy map while AC remains deformed.

We present a potential improvement over the standard method developed to determine antineutrino directionality in inverse-beta-decay detectors. The previously developed method for quantifying directionality in monolithic and segmented detectors may be ambiguous in methodology. In this paper, we present a new directionality algorithm and include error analysis. We have developed a new algorithm based on a measure of ``distance'' between two matrices. We report findings for our research in reactor-antineutrino directionality, and emphasize that the algorithm has broad applications whenever one desires computationally efficient 2D pattern-matching. We treat data from detector segments in the form of a matrix. The validation of our algorithm boils down to comparing a Monte Carlo generated ``empirical'' data set to a simulated data set. The empirical data set is generated for a particular orientation of the neutrino beam. We identify an optimal segmentation scale in the low-count regime. We also discuss the shortcomings of the conventional method and how this knowledge can be applied to segmented detectors, hybrid designs, and generalized validation, agnostic to the physics of detector design.

We formulate the infrared sector of asymptotically flat quantum gravity in terms of asymptotic configurations accessible to external observers. Starting from the Regge-Teitelboim Hamiltonian that generates physical evolution in the presence of gravitational constraints, we perform a Born-Oppenheimer reduction separating slow asymptotic data from fast bulk gravitational fluctuations. We show that integrating out the fast sector induces a functional Berry connection over the space of asymptotic charges, so that the effective infrared dynamics is governed by parallel transport on this charge space. In this framework, infrared gravitational states are naturally organized into superselection sectors labelled by the holonomy of the induced connection, and the reduced density matrix obtained after tracing over ultraviolet bulk modes acquires a geometric contribution. This provides an effective geometric description of the asymptotic quantum gravitational sector, where quantization arises as a global consistency condition under adiabatic transport rather than as a spectral property of local bulk operators.

We propose a unified model that simultaneously addresses the hierarchy problem and the strong CP problem by considering the most simple fundamental composite Higgs setup and showing that it can feature a viable emergent axion that could be accessible at collider experiments. To maintain a low decay constant, and therefore address the hierarchy problem, while simultaneously avoiding experimental bounds on the axion couplings, we increase the axion mass via additional small instanton contributions coming from a new hidden gauge sector with a confinement scale larger than $Λ_{\text{QCD}}$. Specifically, the axion will be identified with the CP-odd scalar singlet contained in the $SU(4)/Sp(4)$ coset of the minimal fundamental composite Higgs model. Both the SM color group and the additional hidden sector group are then embedded into a larger non-Abelian \textit{grandcolor} group, such that the topological angles of the two sectors are guaranteed to agree at tree-level. Beyond that, we show that radiative corrections and other CP-violating sources can be controlled. After examining the field content and the gauge structure of the model, we analyze the pNGB spectrum, potential, and couplings, as well as the resulting phenomenology. We focus in particular on the axion potential, to understand under which conditions the CP-odd scalar singlet can solve the strong CP problem while maintaining a naturally low compositeness scale, identifying interesting viable parameter space for an axion in the GeV range.

Dark matter direct detection experiments involving electron recoils are beginning to test highly-predictive, thermal-relic milestones for sub-GeV dark matter models. Due to the Lee-Weinberg bound, thermal dark matter candidates in this mass range necessarily require comparably-light mediator particles to achieve a suitably large annihilation cross section. Here we present new thermal-relic milestones for sub-GeV dark matter candidates that couple to vector mediators. In these models, the mediators are massive gauge bosons of anomaly-free abelian extensions to the Standard Model, including the dark photon, gauged $L_i - L_j, B-L$, and $B-3L_i$ models, where $B$ is the baryon number, $L$ is the lepton number, and $i,j$ index the lepton families. Since the same interactions that govern cosmological production also govern electron scattering, the targets we present are firmly predictive and allow for these models to be robustly discovered or falsified. Furthermore, since the mediators we study exhaust the minimal anomaly-free U(1) extensions to the Standard Model, our results offer a complete list of predictive milestones for sub-GeV dark matter coupled to vector mediators.

Most of the known non-invertible symmetries of quantum field theories in three and four spacetime dimensions act invertibly on local operators. An exception is coset symmetries, which can be constructed from gauging a non-normal subgroup of an invertible symmetry. In this paper, we study the action of a general finite non-invertible symmetry on local operators in four dimensions. We show that non-invertible symmetries without topological line operators necessarily act invertibly on local operators. Using this result, we argue that the action of a general non-invertible symmetry in 3+1d on local operators can be decomposed into the invertible action of some operators composed with the action of a gauging interface. We use this result to study when such a symmetry is anomaly-free. We find a necessary condition for a finite non-invertible symmetry in 3+1d to be anomaly-free, and show that anomaly-free non-invertible symmetries without topological line operators are non-intrinsically non-invertible.

Axions and other very weakly interacting slim particles (WISPs), with masses below 1 GeV, arise naturally in many extensions of the Standard Model of particle physics. In particular, they could offer a new framework to explain the nature of dark matter and may help address a range of puzzling observations in astrophysics and particle physics. This review provides an overview of ongoing WISP searches and outlines the prospects for the next decade, spanning their theoretical motivation, indirect signatures in astrophysical observations, and dedicated laboratory experiments. It is based on the work carried on by the EU-funded COST Action ``Cosmic WISPers in the Dark Universe: Theory, astrophysics, and experiments'' (CA21106, https://www.cost.eu/actions/CA21106). This network plays a key role in coordinating and supporting WISP searches across Europe, while also contributing to the development of a roadmap aimed at securing European leadership in this research area. It is emphasized that Europe is currently pursuing a rich, diverse, and cost-effective experimental program, with the potential to deliver one or more transformative discoveries.

The cosmohedron was recently proposed as a polytope underlying the cosmological wavefunction for $\text{Tr}(Φ^3)$ theory.Its faces were conjectured to be in bijection with Matryoshkas, which are obtained from a subdivision of a polygon by sequentially wrapping groups of polygons into larger polygons. In this paper we prove the correctness of this construction, and elucidate its combinatorial structure. Cosmohedra generalize to a wider class of $\mathcal{X}$ in $Y$ polytopes, where we chisel a polytope from the family $\mathcal{X}$ at each vertex of a polytope $Y$. We sketch a new application of these chiseled polytopes to the physics of ultraviolet divergences in loop-integrated Feynman amplitudes.

We present a comprehensive reassessment of perturbative unitarity bounds in the dimension-six Standard Model Effective Field Theory, exploiting a new formalism based on spinor-helicity techniques to derive partial-wave unitarity bounds for generic $N \to M$ scattering amplitudes. We find that, in several cases, these theoretical constraints are already competitive with, or even stronger than, the corresponding experimental bounds for energy scales above a few TeV. This is especially the case for four-fermion operators under realistic flavor assumptions, where unitarity bounds can be further strengthened by exploiting sum rules.

This paper provides a detailed analysis of the associated production of the Higgs boson with a single top quark ($tH$) in proton-proton collisions at $\sqrt{s} = 13~\mathrm{TeV}$ and $14~\mathrm{TeV}$. Based on the ATLAS search, we have employed innovative modeling approaches to improve the sensitivity to new physics and improve Standard Model constraints. The major goals are to optimize the data selection, statistical error estimation, determination of physical limits on the top quark Yukawa coupling ($κ_t$), and future experimental projections for HL-LHC. Simulations with MadGraph5aMC@NLO at LO+MLM give cross-sections of $σ_{tHq}$ and $σ_{tWh}$ in SM, scaled by K-factors to simulate NLO accuracy. For inverted $κ_t = -1$ (ITC), positive enhancements are seen, consistent with constructive interference. Kinematic distributions ($H_T$, $p_T(h)$, $η[j]$, $ΔR[t,h]$) are checked against ATLAS expectations, confirming the optimized approach to maximize signal extraction and suppress systematics.

We propose a lattice-ready entanglement observable for QCD hadrons: the vacuum-subtracted radius flow of the ball Rényi entropy, $\mathfrak{s}_n(R;h)\equiv R\,\partial_RΔS_n(B_R;h)$, defined via the Euclidean replica cut-and-glue construction in a rest-frame momentum-projected one-hadron state, with spin averaging performed at the level of the final flow. In the continuum, varying $R$ at fixed shape is equivalent to a Weyl rescaling, so the flow is trace selected and admits a surface-plus-remainder organization on the entangling sphere. We use this to formulate a lattice stability test of boundary dominance: fit the measured flow on local $R$ windows to a low-curvature remainder plus a small template basis built from hadronic gravitational form factors (GFFs). The two endpoint templates are the spin-0/trace shape $\mathfrak{t}_h^{(0)}(R)=R^3ρ_S(R)$ constructed from $A^S(t)$ and a spin-2/TT proxy $\mathfrak{t}_h^{(2)}(R)=R^3ρ_A(R)$ constructed from $A(t)$, together with the mixed family $\mathfrak{t}_h^{\rm mix}(R;c_0,c_2)=c_0\mathfrak{t}_h^{(0)}(R)+c_2\mathfrak{t}_h^{(2)}(R)$. A soft-wall AdS/QCD appendix shows that the pole-subtracted integrated trace--energy correlator closes on this same $\{A^S,A\}$ basis and supplies a model-dependent benchmark ratio for $c_0/c_2$; for lattice comparison the coefficients are left free and extracted from data. For representative nucleon dipole inputs, the pure endpoints predict distinct single-extremum scales, $R_{\rm EE}^{(0)}\sim0.84~\mathrm{fm}$ and $R_{\rm EE}^{(2)}\sim0.43~\mathrm{fm}$, enabling discrimination among scalar control, spin-2 control, and genuine mixing through the turning-point location, the sign change of the slope across it, and the fitted ratio of template weights.

The scattering amplitudes for colored scalars can be calculated using the so-called curve integral formula, relying on simple combinatorics. It introduces a set of global Schwinger parameters for all Feynman diagrams that contribute to an amplitude. We extend this construction to non-orientable surfaces by making use of the quasi-cluster algebras defined for non-orientable surfaces. We embed the non-orientable surface in a doubled orientable surface, and project the appropriate features onto the non-orientable surface. The curve integral formula can also be thought of as the high-tension limit of an appropriate string amplitude. As a check of our construction, we take a superstring amplitude with the Möbius strip topology and take its field theory limit to obtain the same Feynman diagrams as in the corresponding curve integral. Our construction can be generalized to arbitrary higher genus non-orientable surfaces. To illustrate this, we list the possible curves and their dual momenta for a two-loop non-orientable surface, and construct the surface Symanzik polynomials using the surface generalization of spanning trees.

This paper provides the first algebraic characterization of an algebra of cohomological Hecke operators associated with modifications of coherent sheaves on a smooth surface $X$ along a fixed proper curve $Z \subset X$ (possibly singular and reducible), establishing a direct connection with Yangians. It is based on the theory of equivariant nilpotent cohomological Hall algebras $\mathbf{HA}^T_{X,Z}$, developed by the same authors. More precisely, let $X$ be a resolution of a Kleinian singularity (for example, $X = T^\ast\mathbb{P}^1$) and let $Z$ be the exceptional divisor. One of the main results of this paper is an explicit isomorphism $\mathbf{HA}^T_{X,Z} \simeq \mathbb{Y}^+_\infty$, where $\mathbb{Y}^+_\infty$ is a completed, nonstandard, positive half of the affine Yangian $\mathbb{Y}(\mathfrak{g})$ of the corresponding affine ADE Lie algebra $\mathfrak{g}$. Furthermore, the generators of $\mathbf{HA}^T_{X,Z}$--given by fundamental classes of substacks of zero-dimensional sheaves and of pushforwards of line bundles on $Z$--are expressed explicitly in terms of Yangian generators. Our main tools, which may be of independent interest, are: (i) a `continuity' theorem describing the behavior of cohomological Hall algebras of objects in the heart of $t$-structures $τ_n$ when the sequence $(τ_n)_n$ converges, in an appropriate sense, to a fixed $t$-structure $τ_\infty$; (ii) the definition of a multi-parameter Yangian $\mathbb{Y}_Q$ for an arbitrary quiver $Q$, given by generators and relations; (iii) a theorem relating the algebraic action of the braid group $B_Q$ on the Yangian $\mathbb{Y}_Q$ to the action of $B_Q$ on the equivariant 2-dimensional cohomological Hall algebra $\mathbf{HA}^T_Q$ of $Q$, where the latter can be described in terms of derived reflection functors of the bounded derived category of modules over the preprojective algebra of $Q$.

A complete picture of the gas thermodynamics around galaxies is imprinted on the cosmic microwave background (CMB). Indeed, the thermal, kinematic, and relativistic Sunyaev-Zel'dovich effects (tSZ, kSZ, rSZ) measure the gas density, temperature, pressure of baryonic feedback and bulk velocity around galaxies, along with the gravitational potential it sits in. This full thermodynamic picture promises to constrain galaxy formation models and gas related uncertainties in the impact on galaxy lensing. Recent kSZ measurements around galaxies suggest that the gas may be more extended than anticipated, pointing to powerful feedback processes and large baryonic corrections to lensing. How robust are these conclusions about the galaxy-halo connection, including satellite fraction and high-mass outliers, or to 2D projection effects and large-scale velocity modes? In this paper, we give estimates for these effects using a simulated sample of DESI-like luminous red galaxies within the IllustrisTNG hydrodynamical simulation and the Abacus N-body simulation. We show that analyzing projected 2D profiles can lead to biases when computing quantities like the gas fraction. We also find that in the absence of spatial filtering, the 2-halo term is non-negligible for kSZ even at the smaller radii where the 1-halo term dominates. We show that a 1% uncertainty in the satellite fraction of galaxies can propagate into uncertainties of $\pm$1%, $\pm$3% and $\pm$5% in the 1-halo terms of the kSZ, tSZ, and rSZ signals, respectively. We show that masking the 2% most massive objects in the sample reduces the profile amplitudes by up to 10%, 40%, and 75% for the kSZ, tSZ, and rSZ signals, respectively. Finally, we show that naive simulations of the kSZ effect can be biased by an artificial Doppler term, which is automatically removed when high-pass or compensated aperture filtering is applied.

The large population of broad-line Active Galactic Nuclei (AGN) observed with the James Webb Space Telescope (JWST) at $z \gtrsim 4$ opens a new window onto the black hole-galaxy connection in the first Gyr of cosmic history. We use the JADES survey-level dataset and develop a forward-modeling Bayesian framework that explicitly accounts for broad H$α$ detectability, ensuring that selection effects are incorporated into the likelihood function. With this approach, we constrain the black hole-stellar mass ($M_{\mathrm{BH}}$-$M_\star$) relation to be $\log M_{\rm BH} = -4.06^{+0.50}_{-0.51} + 1.17^{+0.06}_{-0.06}\,\log M_\star$, with an intrinsic orthogonal scatter of $σ_{\rm int} = 0.63^{+0.14}_{-0.11}$ dex. The slope and normalization are consistent with local determinations, indicating that the average scaling was already established by $z \sim 4$-6. This suggests that the primary evolution of the relation occurs in its dispersion rather than in its mean normalization. In contrast, the substantially larger intrinsic scatter relative to the nearby Universe reveals a wider diversity of black hole-galaxy growth histories, likely driven by bursty accretion, delayed feedback, and differences in merger or seeding histories. Future JWST samples will be crucial to test whether this increased scatter is a persistent feature of the high-redshift Universe.

We present an automated framework to search for optical counterparts of LIGO-Virgo-KAGRA (LVK) gravitational wave (GW) superevents using public Zwicky Transient Facility (ZTF) alerts processed through the ALeRCE broker. The goal is to filter and identify optical transients potentially associated with binary black hole (BBH) mergers during the LVK O4a and O4b observing runs. Using the Automatic Learning for the Rapid Classification of Events (ALeRCE) infrastructure, we spatially query ZTF alerts within GW localization regions and apply machine learning classifiers, host-galaxy crossmatching, and temporal cuts within 200 days post-merger to isolate plausible candidates. Our search yielded one candidate in O4a and four in O4b, several consistent with the supernova or tidal disruption event regime. This work demonstrates that public alert brokers can establish a robust baseline for systematic searches for electromagnetic counterparts to GW superevents in current and future observing runs. Our algorithm provides a systematic approach to search for BBH counterparts for all significant LVK GW superevents using survey telescope alerts. The search, together with the accompanying analysis, demonstrates the significance of the counterpart candidates, with one candidate ultimately identified as a transient event consistent with a Bowen fluorescence flare in a now discarded active galactic nucleus (AGN).

We present an alternative measurement of the Baryon Acoustic Oscillation (BAO) using Legendre multipole representation of the Ly$α$ forest correlation functions from the second data release (DR2) of the Dark Energy Spectroscopic Instrument survey. Compressing the auto- and cross-correlation functions into Legendre multipoles yields a positive-definite covariance matrix without any smoothing -- unlike the baseline DR2 analysis -- thanks to a significantly reduced data vector size. We introduce the statistical corrections required to debias the finite-sample covariance matrix estimate and demonstrate that monopole and quadrupole terms for both auto- and cross-correlations can be used even when the correlation functions are distorted by continuum errors and contaminated by metals. This formalism has slightly diminished the constraining power of the BAO scale, while considerably weakening constraints on nuisance parameters. We measure the isotropic BAO scale with $0.96\%$ precision at $z_\mathrm{eff}=2.35$, the Hubble parameter $H(z_\mathrm{eff})=(238.7\pm3.4)~(147.09~\mathrm{Mpc}/r_d) ~\mathrm{km~s}^{-1}~\text{Mpc}^{-1}$, and the transverse comoving distance $D_M(z_\mathrm{eff})=(5.79 \pm 0.10)~(r_d/147.09~\mathrm{Mpc})$~Gpc for a given value of the sound horizon ($r_d$). Our BAO results are entirely consistent with the baseline DR2 analysis.

Strong gravitational lensing provides an independent and powerful probe of cosmic expansion by directly linking observables to cosmological distances. Upcoming surveys such as LSST will discover large number of galaxy-galaxy strong lensing systems, offering a new route to precise cosmological constraints. In this paper, we propose a Fisher-like sensitivity factor to map how the cosmological information of strong-lensing distances changes across the lens-source redshift plane. Applying such factor to the distance ratio $D_{ls}/D_s$, the time-delay distance $D_{Δt}$, and the double-source-plane ratio, we determine the ``sensitivity valleys'' where an observable becomes insensitive to a given parameter. The realistically simulated LSST lens population, which largely lies outside the distance-ratio valleys, covers the most sensitive region for $(w_0,w_a)$ parameter space. We then develop a new hierarchical framework, which could calibrate the redshift evolution of lens mass-density slopes and constrain cosmological parameters simultaneously. Focusing on the LSST mock data, we demonstrate that ignoring mass-profile evolution can bias $Ω_m$ by up to $\sim 10σ$, while modeling the lens evolution could perfectly recovers the fiducial cosmology and yield stringent cosmological constraints (e.g., $ΔΩ_m \simeq 0.01$ and $Δw \simeq 0.1$ for $\sim 10^4$ lenses).

We present a novel method for the tomographic reconstruction of the bias-weighted mean electron pressure $\langle bP_e \rangle$ and star formation rate density $\langle bρ_{\mathrm{SFR}} \rangle$, by simultaneously modelling the contribution from the thermal Sunyaev-Zel'dovich (tSZ) effect and the Cosmic Infrared Background (CIB) to the cross-correlation between photometric galaxy samples and multi-frequency Cosmic Microwave Background (CMB) maps. The resulting measurements are independent of the galaxy clustering properties and robust against cross-contamination between tSZ and CIB. Applying this method to publicly available data, we reconstruct the cosmic evolution of $\langle bP_e \rangle$ and $\langle bρ_{\mathrm{SFR}} \rangle$ out to $z\sim1$, making our measurements publicly available. Our measurements of both quantities are broadly compatible with predictions from the fiducial FLAMINGO hydrodynamical simulation, although we observe a lower gas pressure at low redshifts, in agreement with other measurements.

Since 2019, three scintillator detectors of the EEE collaboration have been continuously measuring cosmic muon rates at 78.9°N at the Ny-Ålesund Research Station (Svalbard). The resulting six-year time series reveals a pronounced annual modulation, driven primarily by seasonal atmospheric variations. Utilizing routine radiosonde profiles collected above the same site, we applied several established techniques --along with a tailored analysis approach-- to investigate the relationship between muon rate and atmospheric temperature. The temperature-corrected muon-rates are analysed using the Lomb-Scargle periodogram technique in order to investigate the presence of remaining periodic structures. Finally, the temperature corrections coefficients of our analysis are compared with measurements in other stations located at lower latitudes.

We present the first spatially resolved at $\sim20$ pc scale application of AGN-specific metallicity diagnostics for the nearby Compton-thick Seyfert 2 galaxy Mrk 573 ($z = 0.017$). We use Hubble Space Telescope narrow-band imaging, MUSE integral-field spectroscopy and apply AGN strong-line metallicity diagnostics based on [O III], [S II], H$β$, H$α$, and [N II] emission lines. We construct maps of $12 + \log$(O/H) for two different metallicity calibrations and two different N/O-O/H scaling relations out to $\sim1$ kpc and down to $\sim20$ pc scales. Our analysis reveals metallicity enhancement in AGN-dominated regions, with oxygen abundances reaching up to few times Solar. The metallicity shows a patchy spatial distribution, varying on $\sim100$ pc scales, appears to trace the high Seyfert/LINER index (SLI) value regions and the VLA 6 cm jet/radio lobe emission. These spatial correspondences and the lack of evidence for star formation in the bicone region suggest that the enrichment originates from metals transported from the nuclear AGN regions by winds, outflows, or jets.

The supernova remnant (SNR) G315.4$-$2.3 (MSH 14$-$63 or RCW 86) exhibits strong emission across the electromagnetic spectrum. Radio polarization observations probe magnetic fields and will help to understand the evolution of the SNR. We aim to investigate the radio spectrum and magnetic field properties of the SNR. We observed G315.4$-$2.3 using the Australia Telescope Compact Array (ATCA), covering the frequency range of 1.1-3.1 GHz. We performed rotation measure (RM) synthesis on the $Q$ and $U$ frequency cubes to obtain polarized intensity and RM. The regular component of the line-of-sight magnetic field was estimated from RM. The fractional polarization versus wavelength squared was used to constrain the properties of the turbulent magnetic field. We obtained image cubes of Stokes $I$, $Q$, and $U$, along with images of polarized intensity $P$, RM, and fractional polarization $p$. The radio spectra are very similar for different areas of the SNR. The foreground RM was estimated to be 55 rad m$^{-2}$, and the internal RM of most SNR areas is less than about 50 rad m$^{-2}$. The regular magnetic field along the line of sight was estimated to be about 1.4 $μ$G in the southwest, much smaller than the total magnetic field. For most parts of the southwest and northeast, $p$ is less than 8% and is nearly constant with $λ^2$. We estimated the ratio of turbulent to regular magnetic field to be larger than about 3. The scale of the turbulent magnetic field for some area in the northwest might be smaller than about 0.4 pc. The radio characteristics, including spectrum and turbulent magnetic field, are very similar in the northeast and southwest, even though the evolution is quite different for these two regions based on the current models. These should be taken into account for future modeling of the evolution of the SNR.

Gas accretion process can fuel both star formation and black hole activity, playing a critical role in galaxy evolution. The counter-rotating structures are believed to originate from gas accretion, serving as an ideal laboratory for studying its impact on galaxy evolution. Based on the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey, we built a sample of 147 galaxies with counter-rotating stellar disks (CRDs). This is the largest CRD sample to date, accounting for $\sim$1.5% of the MaNGA survey. For a subset of 138 CRDs, global stellar mass ($M_\ast$) and star formation rate (SFR) were measured in reference. We constructed a control sample with similar $M_\ast$ and SFR but lacking counter-rotating structures. The CRDs relatively exhibit more bulge-dominated morphology, lower molecular gas mass fraction and reside in less dense environment, supporting the hypothesis that they primarily originate from gas accretion. We classified 96 out of 138 CRDs into four types based on their stellar and gas kinematics following the criteria from Bao et al. (2022). There are two additional CRD types: 8 CRDs show misalignment between both stellar disks and gas disk, indicating multiple gas accretion events with differing angular momentum directions; 34 CRDs lack ionized gas emission, showing the highest $M_\ast$ among all the CRD types, which may represent a final stage of CRD evolution. We compared the radial gradients of gas-phase metallicity and stellar population properties between CRD types, and found that the impact of gas accretion on galaxy evolution primarily depends on the abundance of pre-existing gas in progenitors.

This work presents a description of the scientific goals and objectives of OPTIMus (OPTical, Infrared, Millimeter survey of massive star-forming regions), a survey of massive star-forming regions in the optical, infrared, and millimeter wavelengths. The survey is aimed at constructing a comprehensive characterization of the multicomponent and structurally complex interstellar medium in the vicinity of young massive stars, combining both observational and theoretical aspects. Using multi-wavelength observational data, we will reconstruct the three-dimensional structure and determine the physical parameters of HII regions, photodissociation regions, and the surrounding molecular clouds. The paper describes the observational data obtained with the BTA 6-m and Zeiss-1000 telescopes of the Special Astrophysical Observatory of the Russian Academy of Sciences, the 2.5-m telescope of the Caucasian Mountain Observatory of the Sternberg Astronomical Institute of Moscow State University, and the 20-m telescope of the Onsala Space Observatory.

Galaxy morphology encodes key information about formation and evolution. Large imaging surveys require automated, reproducible methods beyond visual inspection. Non--parametric indices provide an useful framework, but their performance must be quantitatively assessed. We present a homogeneous catalog of non--parametric morphological indices for DECaLS galaxies with effective radii larger than 2 arcsec. Our goal is to evaluate the reliability of indices in separating spirals and ellipticals, test their consistency with existing classification schemes, and establish their applicability for the upcoming surveys focused in the southern hemisphere. We developed galmex, a modular Python package for preprocessing images and measuring a variety of non--parametric indices. Using bona-fide spirals and ellipticals as control samples, we assessed the discriminatory power of each index, and compared them with CNN-based T-Types and Galaxy Zoo DECaLS labels. We use the indices as input for a Light Gradient Boosted Machine (LightGBM) to obtain probabilistic classifications. Concentration is the most reliable parameter from the Concentratiom + Asymmetry + Smoothness system (CAS), while asymmetry--based indices (A and S) are limited to detecting disturbed morphologies. MEGG indices (M20, Entropy, Gini, G2) provide stronger separation and trace a gradient with T--Type. By using a simple binary (0/1) label for ellipticals/spirals, classifiers trained on non--parametric indices achieve high accuracy and well--calibrated probabilities, dominated by entropy, concentration, and Gini. We release the first public catalog of CA[A_S]S+MEGG indices for DECaLS, together with galmex. We combine the non-parametric indices with machine learning framework to derive spiral/elliptical separation for galaxies below z~0.15 through a probabilistic approach.

Ridges in galaxy density fields measured by photometric surveys are 2D projections of filaments in the cosmic web, and so should lens light from background galaxies. We report on a detection of this effect in Dark Energy Survey Year 3 data at high significance, though not independently of galaxy-galaxy lensing. We describe improvements to the existing subspace-constrained mean shift algorithm to locate these ridges efficiently at scale, and examine the dependence of the signal in simulations on cosmological and algorithmic parameters. We find that it depends primarily on $S_8=σ_8 \left( Ω_m / 0.3 \right)^{1/2}$, and discuss improvements to our methodology that would be needed to allow precision parameter estimation.

Sagittarius A* (Sgr A*) is a compact radio source at the Galactic center. Observations have confirmed that its mass is approximately (4.1)*10$^{6}$ M$_{\odot}$, and Sgr A* is generally believed to be powered by gas accretion onto a supermassive black hole. Multifrequency radio observations of the pulsar J1745-2900, about 0.12 pc away from Sgr A*, reveal an unusually large Faraday rotation. Combined with X-ray observations, this indicates that there is a strong magnetic field (greater than 8 mG) leading to a low $β$ plasma at large scales.We show that the gas starts to be captured by the black hole below tens of thousands of the Schwarzschild radii $r_S$, where the gas pressure starts to dominate. Assuming that the accretion rate along magnetic fields at large scales decreases with the distance to the black hole following a power law, it is shown that, with an accretion disk below tens of $r_S$, as revealed with the EHT observations, there should be a supersonic wind above such a small accretion disk, and the accretion flow may be convection-dominated from tens of $r_S$ to tens of thousands of $r_S$. Detailed modeling is warranted.

We present a timing analysis of the high-cadence optical light curve of the high-redshift flat-spectrum radio quasar PKS\,0805$-$07 obtained during \textit{TESS} Sector~34 (MJD $59230.90$--$59239.90$). We search for short-timescale quasi-periodic oscillations (QPOs) using complementary time-series techniques, including the Lomb--Scargle periodogram (LSP) and the weighted wavelet $Z$-transform (WWZ), and evaluate their significance against red-noise variability using Monte Carlo simulations. The LSP reveals a dominant modulation at $f \approx 0.597\,\mathrm{d^{-1}}$ ($P \approx 1.7\,\mathrm{d}$) exceeding the $99.99\%$ confidence level, while the WWZ independently recovers a consistent timescale at the $\sim 99.9\%$ level and shows that the signal is temporally localized rather than persistent across the full light curve. The modulation spans $\sim 5$ coherent cycles, indicating a transient quasi-periodic feature. We discuss possible physical interpretations of the detected modulation. In a disk-based scenario, orbital motion of a hotspot near the innermost stable circular orbit implies a black hole mass of $M_{\rm BH}\sim7.2\times10^{8}\,M_\odot$, consistent with typical FSRQ values. Alternatively, magnetohydrodynamic kink instabilities in the relativistic jet can naturally produce day-scale variability for standard blazar parameters and account for the transient character of the signal. We conclude that the observed modulation is consistent with a compact, short-lived structure embedded within stochastic jet variability.

We study the impact of supermassive black hole (SMBH) growth, $\langle \dot{M}_\mathrm{SMBH}\rangle$, major and minor galaxy mergers, and gas processes, on the average gas metallicity of galaxies, with the aim to uncover which of these processes drive the scatter in the gas metallicity-stellar mass relation (MZR) at different redshifts in nodes, filaments and voids. At $z=5$, minor mergers produce the largest differential in $\log[Z_g/Z_\odot]$ for all environments, where the node population displays a maximum $0.38$ dex increase in the average $\log[Z_g/Z_\odot]$ compared to non-merging galaxies. The node population also displays a consistent $0.1$ dex reduction in $δ\log[Z_g/Z_{\odot}]$ across all redshifts, whilst filament and void galaxies show a lower magnitude of reduction. Major mergers show little influence on these same properties. This suggests minor mergers regulate metallicity and contribute to over galaxy mass growth concurrently, accelerating chemical evolution post merger. Between $z=1-3$, a high $\langle \dot{M}_\mathrm{SMBH}\rangle$ leads to a reduction in $δ\log[Z_g/Z_{\odot}]$ for all environments. Here, node galaxies show the largest reduction of approximately $0.25$ dex, suggesting that metal-rich outflows strongly drive the MZR at intermediate times. Finally, galaxies with low $M_{gas}/{M_{tot}}$ show increased $δ\log[Z_g/Z_{\odot}]$ across all redshifts and environments, again a $0.25$ dex maximum for node galaxies. These galaxies also spike in $δ\log[Z_g/Z_{\odot}]$ at late times, below $z=1$. At this time, galaxies in the nodes show negative $\langle \dot{M}_\mathrm{gas} \rangle$ whilst also showing the largest $δ\log[Z_g/Z_{\odot}]$ values we observe of $0.2$ dex, suggesting the importance of the balance between gas accretion and starvation in driving MZR scatter at low redshifts.

Modeling inviscid compressible flows with shocks and vortex dominated dynamics remains challenging for particle grid methods due to moving discontinuities, cell crossing noise, and quadrature degradation under strong deformation. Building on a FLIP/APIC framework with vorticity aware tensor artificial viscosity, we identify a long time RTI failure mode: particle depletion at spike heads degrades quadrature and particle grid coupling, producing nonphysical, void-like dents. Standard mitigations (CPDI lite and subcell-jittered seeding) reduce but do not eliminate this artifact. We therefore add two sampling-aware controls: (i) conservative split resampling that replenishes depleted cells while exactly conserving mass, momentum, and internal energy; and (ii) a soft-switch that attenuates only the APIC affine term when local support is insufficient. Tests on the Sod shock tube and single/multi mode RTI show that the method removes spike head voids in long-time RTI while preserving vortex roll up, and matches reference Euler growth metrics

We measure galaxy stellar mass functions (SMFs) for nine of the most massive galaxy clusters in the local universe ($0.07 < z < 0.11$) using deep and complete spectroscopy from the MAssive Cluster Survey with Hectospec (MACH). We construct the cluster SMFs down to $\log(M_*/M_\odot) \gtrsim 8.5$. For comparison, we measure the SMF for field galaxies, complete to $\log(M_*/M_\odot) \gtrsim 10.5$, based on Sloan Digital Sky Survey (SDSS) spectroscopy over the same redshift range. The mean MACH SMF shows a shape similar to that of the field SMF but with a significantly higher amplitude at $\log(M_*/M_\odot) < 11.4$. At $\log(M_*/M_\odot) > 11.4$, the MACH SMF shows a clear excess, indicating the contribution of massive galaxies, including Brightest Cluster Galaxies (BCGs). Based on homogeneous MACH spectroscopy, we compare SMF shapes for quiescent and star-forming members as a function of cluster-centric distance. The quiescent SMFs display a curved shape with a peak at $\log(M_*/M_\odot) \approx 10.5$; the star-forming SMFs decline monotonically with increasing stellar mass. We further compare the mean MACH SMF with SMFs derived from similarly massive clusters in the IllustrisTNG-300 simulations. The shape of the observed and simulated SMFs agree well overall. However, the MACH clusters contain roughly a factor of two more galaxies at $9.0 < \log(M_*/M_\odot) < 10.5$. These results demonstrate that constructing cluster SMFs from complete spectroscopic samples can test simulations and provide powerful constraints on galaxy formation and evolution in dense environments.

We investigate the accretion dynamics of the black hole X-ray binary Swift J1727.8-1613 during its $2023-2024$ discovery outburst that lasted for $\sim10$ months. Insight-HXMT monitored the rising phase of the outburst of Swift J1727.8-1613 roughly continuously from 2023 Aug 25 to 2023 Oct 05. Strong signatures of type-C Quasi-Periodic Oscillations (QPOs) are observed during this phase of the outburst. In our recent paper, nature of the QPOs are studied with the propagating oscillatory shock (POS) model. In this paper, we report on the observation of both positive (or hard) and negative (or soft) time-lags in the $4-10$ keV (LE), $10-30$ keV (ME), and $30 -150$ keV (HE) bands, computed with respect to the $2-4$ keV reference band. We detect a clear transition from hard to soft lags as the outburst evolves. We show the evolution of QPOs and associated time-lags between different X-ray energy bands, correlated with changes in the QPO frequency, spectral state, and the size of the Comptonizing region. Our analysis reveals strong anti-correlations between the time-lags and both QPO frequency and photon index, and a strong positive correlation with the shock location. These evolving lag characteristics and their correlations provide crucial insights into the changing accretion geometry and the interplay of radiative processes, further supporting dynamic models like the POS in explaining the coupled spectro-temporal evolution in black hole X-ray binaries.

Determining precise stellar ages and masses for evolved giants is crucial for Galactic archaeology but challenged by spectral degeneracies. Gaia's low-resolution XP spectra offer a unique opportunity to infer these parameters on a massive scale using data-driven methods. We extend a transformer-based astronomical foundation model to evolved stars, establishing a unified framework to simultaneously predict atmospheric parameters ($T_{\mathrm{eff}}$, $\log g$, $[\mathrm{M}/\mathrm{H}]$) and evolutionary labels (mass, age) with physical consistency. Treating spectra as token sequences, we integrated mass and age into the model's vocabulary. The model is trained on Gaia XP spectra cross-matched with the APOGEE DR17 DistMass catalog. Our generative approach enables flexible input handling, including spectral inpainting and parameter-to-spectrum generation. On an independent test set, the model achieves a prediction scatter of $σ\approx 0.114 \, M_{\odot}$ for mass and $σ\approx 1.334$ Gyr for age. Beyond numerical accuracy, it successfully reproduces the giant branch's mass-luminosity relation and autonomously disentangles interstellar extinction from intrinsic temperature variations without explicit physical priors. It also robustly recovers missing spectral data and estimates reliable uncertainties. Validating that foundation models can internalize stellar physics from data, this physically-aware, probabilistic framework offers a powerful tool for unraveling Milky Way history using large-scale spectroscopic surveys.

Neutrino astronomy provides another window to exploring the Universe, exemplified by the detection of a megaelectronvolt neutrino burst from the core-collapse supernova (CCSN) SN~1987A (refs.~\citenum{hir+87,bio+87}). Commonly discussed theories suggest that some CCSNe could produce neutrinos with energies a thousand times more than those of SN~1987A \cite{tm18}, which has been probed with new-generation facilities \cite{abb+12,aar+15,abb+23}. The interaction of SN ejecta with a dense circumstellar medium (CSM) or a jet, launched in a CCSN, being choked in the stellar envelope of the progenitor or an outside CSM are both well-accepted scenarios for the high-energy neutrino production. Here we report the detection of a high-energy neutrino flare at a 3.9$σ$ significance from SN~2017hcd, made by our analysis of the public track-like neutrino data taken by the IceCube Neutrino Observatory \cite{IceCube17}. A Type IIn SN with optical emissions arising from the ejecta--CSM interaction, SN~2017hcd's neutrino flare lasted $\sim$1--2 month, with its central time $\sim$14-day prior to the SN's optical discovery time. Its estimated isotropic neutrino energy (all flavors) is approximately two orders of magnitude higher than the energy ($\sim 10^{50}$\,erg) carried in the SN's ejecta, too high to be explained with the ejecta--CSM scenario. Thus, a choked jet may be the source of the neutrino flare.

Cosmological reionization was a highly out-of-equilibrium event that affected every parcel of the intergalactic medium, making it a candidate for astrophysical generation of intergalactic magnetic fields. During reionization, the first stars and galaxies ionized the surrounding, largely neutral, medium in ever-expanding envelopes. Photoionization from sources on one side of the front, combined with the quadrupolar angular dependence of the photoionization cross section, leads to an anisotropic electron velocity distribution. We investigate instabilities in these reionization fronts as a mechanism to generate seed magnetic fields. The Weibel instability has the potential to create a magnetic field from these anisotropies. We calculate the magnitude of the isotropic and anisotropic distribution within a simulated reionization front. We find that the fractional anisotropy can grow to $6\times 10^{-3}$ toward the middle of the ionization front. We show that the linear growth timescale of the Weibel instability is fast compared to the crossing time of the ionization front ($\sim 2\times 10^5$ seconds). We briefly speculate on the possible non-linear evolution of the instability and the implications for cosmological magnetogenesis.

Modern datasets have the precision necessary to uncover new information by including higher-order, non-Gaussian information into cosmological inference. The marked power spectrum offers access to such information while preserving the structure of two-point correlators. This approach to higher-order statistics has the advantage that many modeling questions can directly benefit from progress already made in standard cosmological analyses using the power spectrum and correlation function, while increasing the data vector size negligibly and retaining much of the degeneracy-breaking power of the bispectrum. In this work, we first restructure the marked power spectrum to isolate its higher-order information and demonstrate its ability to break parameter degeneracies. We then investigate the effect of survey geometry on the marked power spectrum and find that a treatment similar to that of the power spectrum is sufficient. Additionally, we investigate the perturbative modeling and covariance structure of the marked power spectrum, shedding light on its degeneracy breaking power and cross-covariance with the power spectrum. Finally, we demonstrate that the cosmology dependence of the marked power spectrum is smooth, indicating that cosmological inference is possible by modeling the cosmology dependence through interpolation rather than analytical modeling.

Accretion is the primary driver of protostellar evolution, regulating mass assembly and shaping the physical and chemical environments of young stellar objects. Quantifying accretion in the Class 0 protostellar phase is particularly important, yet remains observationally challenging due to high extinction toward the central protostars. In this paper, we present JWST NIRSpec and MIRI/MRS IFU data towards the Class 0 protostar L1527 IRS. We extract one-dimensional spectra and find emission from atomic and molecular hydrogen, water, OH, and several ionic species. The atomic hydrogen lines, Br$α$, Pf$α$, and Pf$γ$ are the most critical to this study since they can be used as accretion diagnostics. The existence of these atomic hydrogen lines viewed in scattered light indicates that accretion is likely occurring magnetospherically rather than through a boundary layer. Moment 0 emission maps show that the hydrogen emission is co-spatial with the scattered light continuum with a strong east-west asymmetry which is not due to outflow shocks. We additionally present moment 0 maps of other detected species and discuss their emission morphology. By primarily analyzing the Br$α$ line, the strongest of our detected atomic hydrogen lines, we characterize the accretion onto L1527 IRS by estimating the accretion luminosity to be $0.4~\text{L}_\odot$ and the accretion rate to be around $1\times10^{-7}~ \text{M}_\odot \text{yr}^{-1}$. We lastly discuss the implications of our results with respect to both non-steady and asymmetric accretion possibly occurring in L1527 IRS.

We report the discovery of a Lyman-alpha emitter (LAE) protocluster at z = 4.90 in the COSMOS field, comprising four distinct overdensity peaks spanning ~65 x 36 cMpc^2, with the primary concentration exhibiting a 4-fold surface density enhancement relative to the field within a 1.5 proper Mpc (pMpc) radius. Using SILVERRUSH narrowband survey data combined with JWST COSMOS-Web imaging, we perform a first systematic rest-frame optical and UV morphological comparison of protocluster versus field LAEs at this redshift using JWST NIRCam rest-frame UV (F150W, ~2540 Angstrom) and optical (F277W, ~4700 Angstrom) imaging. Sersic profile fitting for 16 protocluster members and 23 field LAEs reveals a significant size difference: protocluster LAEs are ~40% larger in rest-optical (median R_e = 0.81+0.26/-0.04 kpc vs. 0.58+0.11/-0.04 kpc, p = 0.041) with no significant difference in rest-UV (p = 0.51) or Sersic index. At fixed stellar mass, protocluster LAEs are offset by +0.12 dex (~31%) in rest-optical size from the field size-mass relation (68% CI: [+0.08, +0.21]; Mann-Whitney p = 0.033), with 75% exhibiting positive size residuals compared to 44% of field LAEs. This wavelength-dependent environmental signature suggests that protocluster environments at z ~ 5 preferentially affect extended stellar populations, possibly through tidal interactions or an earlier onset of star formation in the dense environment, with no significant environmental difference detected in rest-UV sizes, providing observational evidence for environmental influences on the structure of LAEs during the early build-up phase of cosmic star formation.

We test whether parsec-scale stellar systems in the Milky Way follow the galactic radial acceleration relation (RAR) or the baryonic TullyFisher relation (BTFR). We analyse 5646 Gaia DR3 open clusters from the Hunt \& Reffert catalogue. Observed accelerations are derived from velocity dispersions and characteristic radii, and baryonic accelerations from stellar masses and characterisitc radii. The clusters are placed on the RAR and BTFR planes and compared with Newtonian and MOND expectations. Approximately 90 per cent of open clusters (those with $N_\star \leq 250$) lie close to the RAR, albeit with significant scatter. In a first-of-its-kind test, a smaller fiducial sample is consistent with a best-fitting acceleration scale $g_\dagger \approx 1.2 \times 10^{-10}\ \mathrm{m\,s^{-2}} \pm 0.5$ dex, compatible with canonical MOND values. More massive clusters approach the Newtonian virial expectation. No correlations are found between RAR residuals and galactocentric radii, distance to the Galactic disk midplane, age, or morphology. Tidal effects and unresolved binaries are insufficient to reproduce the observations without fine-tuning. Interpreted within a MOND framework, the alignment of most open clusters with the RAR and BTFR suggests that low-acceleration dynamics operate on parsec scales within the Milky Way. This implies that the Galactic gravitational field is not smooth on these scales and may include regions where the total gravitational acceleration falls below $a_0$, partially mitigating the external field effect, thereby motivating higher-resolution modelling of the Galactic potential and informing other small-scale gravity tests within the Galaxy.

In this work, we present comprehensive observational constraints on the time-varying vacuum $Λ(t)$CDM cosmology using the latest baryon acoustic oscillation (BAO) data from Dark Energy Spectroscopic Instrument (DESI) Data Release 2 measurements in combination with cosmic chronometer $H(z)$ data. We explicitly quantify the impact of vacuum dynamics on the expansion history, the total effective equation of state parameter of the unified cosmic fluid, the effective dark energy equation of state parameter, and the deceleration parameter in the spatially flat $Λ(t)$CDM model. We perform a full Markov Chain Monte Carlo (MCMC) analysis and statistical model comparison, providing a consistent assessment of the $Λ(t)$CDM model relative to the standard $Λ$CDM framework. Our results demonstrate that $H(z)$ and BAO observations strongly constrain deviations from the $Λ$CDM model, driving the vacuum dynamics parameter toward $α\simeq 0$, while significantly reducing parameter degeneracies and alleviating the Hubble tension.

We report the discovery of significant X-ray polarization from the dynamically confirmed black hole X-ray binary (BHXB) GS 1354-64 during its 2025-2026 outburst, obtained with the Imaging X-ray Polarimetry Explorer (IXPE). The observation, obtained shortly after a bright X-ray flare, captures the source in an intermediate state following a stalled (failed) state transition. We discover significant 2-8 keV polarization at the ~4% level with high statistical support--14-sigma significance from frequentist analysis and log Bayes Factor 283+/-1 in a Bayesian framework--measuring PD 4.0+/-0.2% and PA-1+/-2 degrees (90% credible interval). The PD exhibits a statistically significant increasing trend with energy--the strongest such increase yet observed by IXPE in a BHXB--going from 2.1+/-0.3% in the 2-3 keV band to 11+/-3% in the 6.5-7 keV band, while the PA appears stable across both energy and time to within statistical uncertainties. Timing analysis of the IXPE data reveals a ~5 Hz Type-C quasi-periodic oscillation. IXPE + NuSTAR spectropolarimetric modeling suggests that the data can be described by polarized thermal disk and Comptonized components with PAs differing by ~90 degrees, or by a dominant Comptonized polarized component whose effective PD increases across the IXPE bandpass--the inferred component-level polarization levels are therefore model-dependent. In either picture, GS 1354-64 retains a strong coronal component during the transitional period observed by IXPE. These results illustrate how X-ray polarimetry can provide a sensitive diagnostic of the accretion state and geometry in black hole X-ray binary accretion flows, exploring a liminal phase at the cusp of state transition.

Constraining stellar models using asteroseismic and spectroscopic observations is a powerful method for precisely determining the fundamental properties of stars in different kinematic components of our galaxy. We use spectroscopy and individual oscillation mode frequencies to perform a homogeneous modeling study of eight evolved metal-poor stars enhanced in $α$-elements. We compare a full treatment of $α$-enhancement against an ad hoc correction to the total metallicity and show that the stellar properties inferred from asteroseismic modeling using both sets of models agree with each other. Additionally, we find that the uncertainties on stellar parameters derived from the both $α$-enhanced modeling methods are comparable. This is in qualitative disagreement with existing works showing red-giant ages constrained by only the global asteroseismic parameters to depend strongly on the opacities and abundances assumed in 1D modeling. We also show that the observed frequency of maximum oscillation power ($ν_{\text{max}}$) is larger than the value predicted from applying the $ν_{\text{max}}$ scaling relation to the masses, radii, and temperatures inferred from the detailed modeling. This discrepancy is pronounced at low metallicities, consistent with recent findings indicating a breakdown of the $ν_{\text{max}}$ scaling relation for metal-poor stars. Understanding the extent to which the $ν_{\text{max}}$ scaling relation fails for low-metallicity solar-like oscillators through detailed modeling will enable more accurate mass and age determinations for hundreds of giant stars in the Galactic Halo for which only global asteroseismic parameters are available.

Analyses of baryon acoustic oscillations (BAO) commonly employ template-based methods to extract compressed parameters from the clustering of dark-matter tracers, which are then interpreted in terms of ratios of the sound-horizon scale and cosmological distances relative to a fiducial cosmology. A small mismatch between the sound-horizon scale derived from the standard analytic formulation (integral over the sound speed) and the effective scale imprinted in clustered matter can, however, introduce a systematic bias in cosmological inference. We extend previous work to a broader class of cosmological models, quantify this bias for surveys with DESI-like precision, and propose strategies to correct for the effect. We find that the induced bias becomes a significant fraction of the statistical uncertainty for deviations from the fiducial cosmology, at the level of $|ΔΩ_m| = 0.03$ and $|ΔN_\mathrm{eff}| = 0.3$, and for very precise data corresponding to a forecasted Year-5 DESI survey (or other stage IV dark energy galaxy surveys). We present several ways to correct for this effect, suitable for a variety of applications. We therefore recommend that analyses exploring such parameter regimes either apply the proposed corrections or include an appropriate systematic error budget.

The classical picture that planet formation occurs in protoplanetary disks that are isolated from their environment is undergoing a major shift toward a more connected picture. An increasing amount of evolved disks are found to be actively interacting with their environment, often showing various types of spiral structures. In this work, we aim to investigate if these spirals can be a direct result of ongoing late infall using the grid-based 3D hydrodynamics code FARGO3D. We perform a detailed analysis of the spiral properties and appearance in scattered light and CO line emission using the radiative transfer code RADMC3D. In scattered light, we find both well-defined spirals with few arms (m=2) and more flocculent structures: The gradual accretion of gas remnants after a major accretion event has the most success in the former, whereas active accretion via streamers favors the latter. The m=2 spirals we find have a very low pattern speed, making them easily discernible from spirals caused by a perturber. We also find spiral patterns in the $^{12}$CO residual motions, but their morphology does not match the one found in scattered light. The disk perturbations are strongest in the upper layers (z>4H), which is reflected by the reduced amplitude of the residual motions in the more optically thin $^{13}$CO emission. Moreover, we find that the formation of m=2 spirals is not promoted in disks with lower mass, despite being more susceptible to deeper kinematic perturbations. While the late-infall streamers impact planet formation directly through the delivery of fresh material, we show that the midplane remains unperturbed unless the infalling mass is of the same order of magnitude as the disk mass. Planet formation can therefore only be impacted by late infall through secondary mechanisms that lead to dust trapping or the generation of turbulence starting from surface-level perturbations.

Galaxy clusters offer powerful insights into the large-scale structure of the Universe and the physics of baryons in hot state. Their scientific exploitation, however, hinges on our ability to accurately measure key thermodynamic properties. In this work, we aim to assess the reliability of current analysis techniques in reconstructing these properties, with particular focus on samples similar to those observed in the Cluster HEritage project with XMM-Newton (CHEX-MATE). We develop a suite of dedicated end-to-end simulations of CHEX-MATE-like clusters selected from large scale hydrodynamical simulations, and processed through a newly developed realistic XMM-Newton simulator. We apply a full X-ray data analysis pipeline to the mock datasets, including imaging, spectral fitting, and profile reconstruction. The gas density profiles can be robustly recovered across a wide radial range, when using azimuthal mean surface brightness profiles. Our reconstruction techniques are able to reproduce the intrinsic density profile with the correct scatter, with deviations of at most 10% between 0.1 and 1xR500c. The gas mass is reconstructed with better than 1% accuracy. Accurate measurement of temperature profiles is more challenging and possibly subject to biases, particularly in the presence of azimuthal variations and multi-temperature gas along the line of sight, which dominate over projection effects. Our results highlight the need for caution in interpreting cluster temperature measurements and underscore the value of tailored mock observations for understanding observational systematics. These findings also suggest that biases in X-ray temperature measurements may alter the interpretation of the thermodynamical state of the intra-cluster medium, an outlook particularly relevant in light of recent low velocity measurements from the XRISM mission.

Galaxy mergers have long been proposed as a mechanism for funneling gas toward galactic centres, potentially triggering accretion onto supermassive black holes (SMBHs) and igniting active galactic nuclei (AGN). While simulations often support this scenario, observational studies have yielded conflicting results regarding the AGN-merger connection. In this study, we analyze 31 galaxies from cosmological zoom-in simulations spanning redshifts $0.5 < z < 3$. We identify mergers using detailed merger trees based on six-dimensional dark matter particle information and identify AGN activity through SMBH accretion histories. To bridge the gap between simulations and observations, we generate mock JWST-like images and extract non-parametric morphological parameters. Employing a $k$-nearest neighbours (KNN) classifier in a five-dimensional space (four morphological parameters and redshift), we identify mergers in the mock-observed dataset. Our analysis reveals a statistically significant enhancement of AGN activity in merging systems, particularly at lower redshifts ($0.5 < z < 0.9$), where central gas reservoirs are more depleted. This supports the view that mergers contribute more significantly to AGN triggering in environments with low internal gas reservoirs, while their impact may be less pronounced in gas-rich systems. However, when relying solely on morphological classifications from mock observations, the observed AGN-merger connection weakens, especially at higher redshifts. This underscores the challenges in detecting merger-induced AGN activity observationally and highlights the importance of combining simulations with realistic mock observations to fully understand the AGN-merger relationship.

We present the first systematic census of quenching mechanisms using kinematic asymmetries in a large sample of $\sim$6,700 galaxies from the MaNGA survey, providing a unified view of what halts star formation in the local Universe ($z<0.2$). We quantify stellar and nebular gas disturbances through the higher-order terms of a Fourier series expansion. These asymmetries serve as powerful diagnostics, as different quenching mechanisms leave distinct kinematic signatures on gas and stars. Our analysis reveals that the most effective quenching pathways leave minimal kinematic imprints by the time galaxies are fully quenched. This "kinematic regularity" points toward slow-acting processes (>3 Gyr) such as starvation and maintenance feedback. A striking finding emerges from our mass-matched analysis: quenched symmetric satellites are significantly more compact than their asymmetric counterparts ($3.4σ$), a trend that is even more pronounced for symmetric centrals ($12.3σ$). Our results suggest that environment drives the dominant satellite quenching pathway through rapid gas stripping followed by long-term starvation. These compact, kinematically undisturbed satellites (the most representative case within our sample) have undergone intense gas stripping and central compaction, creating bulge-like structures with old, metal-rich stellar populations. Combined with halo gas cut-off and the prevention of cosmological accretion due to starvation, this creates an irreversible quenching path. Conversely, the larger sizes of disturbed, quenched centrals are consistent with merger-driven growth. Internal processes, likely driven by the AGN cycle over 1-3 Gyr that prevents hot halo gas cooling, sustain quenching maintenance in this population. The absence of asymmetric satellites in the star-forming regime suggests environmental quenching operates without significant kinematic perturbation.

We present a JWST analysis of intracluster light (ICL) in XLSSC 122 at z = 1.98, currently the most distant known strong lensing galaxy cluster with an evolved member population. Using deep JWST imaging complemented by HST data and careful control of systematics, we robustly detect diffuse emission extending to several hundred kpc from the brightest cluster galaxy (BCG) down to about 29 mag arcsec^-2. Multi component PSF convolved Sersic modeling separates the surface brightness profiles into three components: a BCG core, a BCG envelope, and an ICL component, with stable Sersic indices across wavelengths. Nearly flat color profiles indicate minimal radial variation in the stellar populations of the BCG envelope and the ICL. The median ICL fraction measured across seven bands is about 17 percent, demonstrating that the buildup of intracluster stars in massive halos was already well underway by z about 2. The ICL fraction peaks near 5000 Angstrom in the rest frame, resembling the behavior observed in dynamically active clusters. We also detect a southern excess of ICL relative to the best fit Sersic model and quantify it using wavelet based modeling, providing additional support that this system is dynamically active. The BCG + ICL light distribution and strong lensing mass map show strong morphological agreement within about 100 kpc. These findings establish the ICL as an early forming and dynamically informative component of massive halos.

Dedispersion is the computational process of correcting for the frequency-dependent time delay affecting a radio signal that propagates through the interstellar and intergalactic media. It is a crucial component of transient search pipelines that maximises the signal-to-noise ratio, especially when targeting highly dispersed signals: for instance, pulsar emissions making their way through a dense cloud of ionised gas, and fast radio bursts travelling cosmological distances. This paper introduces Streaming high Time-Resolution Imaging DEdispersion (STRIDE), a novel dedispersion algorithm to generate per-pixel dedispersed time series from high time and frequency resolution interferometric images. Unlike straightforward approaches to image dedispersion, STRIDE does not involve expensive manipulation of the input data layout, such as explicitly building dynamic spectra or shifting images. Furthermore, it is the first dedispersion algorithm to partition a dispersive sweep over the time dimension, in addition to frequency. As a consequence, images corresponding to the entire time span of the target dispersive delay are not required all at once. Instead, the algorithm works with an arbitrarily-sized subset of images at a time, adopting an incremental, streaming-based approach to dedispersion. In evaluating STRIDE on the presented test case, it is shown that the minimum memory requirement is reduced by 97.9%, going from 684.5 GB to 14.4 GB. As current and future generations of widefield interferometers increasingly turn to imaging techniques for detection and localisation of radio transients, STRIDE positions itself as a strong alternative to traditional dedispersion methodologies. It arguably is the only viable option for imaging-based searches with low-frequency instruments such as the Murchison Widefield Array (MWA) and low-frequency Square Kilometre Array (SKA-Low).

We present data from the ALMA Central Molecular Zone Exploration Survey (ACES) Large Program, which provides broad spectral-line and 3 mm continuum coverage of the Central Molecular Zone (CMZ) at a spatial resolution of 0.1 pc. The survey delivers homogeneous, wide-field mosaics that enable direct comparisons of the physical and chemical conditions across diverse environments in the Galactic center. In this data release paper, we present the CS(2-1), SO(2_3-1_2), CH3CHO(5_1,4-4_1,3), HC3N(11-10), and H40a lines observed simultaneously within two broad spectral windows. These lines reveal pronounced spatial and chemical variations across the CMZ, tracing distinct components of molecular gas, shock-affected regions, and ionized structures. The high angular resolution and multi-line capability of the ACES dataset make it a powerful resource for future studies of gas dynamics, star formation activity, and the physical connection between the CMZ and Sgr A*.