arXiv Daily Digest - 2026-01-23

We study Compositional Video Understanding (CVU), where models must recognize verbs and objects and compose them to generalize to unseen combinations. We find that existing Zero-Shot Compositional Action Recognition (ZS-CAR) models fail primarily due to an overlooked failure mode: object-driven verb shortcuts. Through systematic analysis, we show that this behavior arises from two intertwined factors: severe sparsity and skewness of compositional supervision, and the asymmetric learning difficulty between verbs and objects. As training progresses, the existing ZS-CAR model increasingly ignores visual evidence and overfits to co-occurrence statistics. Consequently, the existing model does not gain the benefit of compositional recognition in unseen verb-object compositions. To address this, we propose RCORE, a simple and effective framework that enforces temporally grounded verb learning. RCORE introduces (i) a composition-aware augmentation that diversifies verb-object combinations without corrupting motion cues, and (ii) a temporal order regularization loss that penalizes shortcut behaviors by explicitly modeling temporal structure. Across two benchmarks, Sth-com and our newly constructed EK100-com, RCORE significantly improves unseen composition accuracy, reduces reliance on co-occurrence bias, and achieves consistently positive compositional gaps. Our findings reveal object-driven shortcuts as a critical limiting factor in ZS-CAR and demonstrate that addressing them is essential for robust compositional video understanding.

Discrete video VAEs underpin modern text-to-video generation and video understanding systems, yet existing tokenizers typically learn visual codebooks at a single scale with limited vocabularies and shallow language supervision, leading to poor cross-modal alignment and zero-shot transfer. We introduce PyraTok, a language-aligned pyramidal tokenizer that learns semantically structured discrete latents across multiple spatiotemporal resolutions. PyraTok builds on a pretrained video VAE and a novel Language aligned Pyramidal Quantization (LaPQ) module that discretizes encoder features at several depths using a shared large binary codebook, yielding compact yet expressive video token sequences. To tightly couple visual tokens with language, PyraTok jointly optimizes multi-scale text-guided quantization and a global autoregressive objective over the token hierarchy. Across ten benchmarks, PyraTok delivers state-of-the-art (SOTA) video reconstruction, consistently improves text-to-video quality, and sets new SOTA zero-shot performance on video segmentation, temporal action localization, and video understanding, scaling robustly to up to 4K/8K resolutions.

We introduce LLM-in-Sandbox, enabling LLMs to explore within a code sandbox (i.e., a virtual computer), to elicit general intelligence in non-code domains. We first demonstrate that strong LLMs, without additional training, exhibit generalization capabilities to leverage the code sandbox for non-code tasks. For example, LLMs spontaneously access external resources to acquire new knowledge, leverage the file system to handle long contexts, and execute scripts to satisfy formatting requirements. We further show that these agentic capabilities can be enhanced through LLM-in-Sandbox Reinforcement Learning (LLM-in-Sandbox-RL), which uses only non-agentic data to train models for sandbox exploration. Experiments demonstrate that LLM-in-Sandbox, in both training-free and post-trained settings, achieves robust generalization spanning mathematics, physics, chemistry, biomedicine, long-context understanding, and instruction following. Finally, we analyze LLM-in-Sandbox's efficiency from computational and system perspectives, and open-source it as a Python package to facilitate real-world deployment.

We propose a novel training regime termed counterfactual training that leverages counterfactual explanations to increase the explanatory capacity of models. Counterfactual explanations have emerged as a popular post-hoc explanation method for opaque machine learning models: they inform how factual inputs would need to change in order for a model to produce some desired output. To be useful in real-world decision-making systems, counterfactuals should be plausible with respect to the underlying data and actionable with respect to the feature mutability constraints. Much existing research has therefore focused on developing post-hoc methods to generate counterfactuals that meet these desiderata. In this work, we instead hold models directly accountable for the desired end goal: counterfactual training employs counterfactuals during the training phase to minimize the divergence between learned representations and plausible, actionable explanations. We demonstrate empirically and theoretically that our proposed method facilitates training models that deliver inherently desirable counterfactual explanations and additionally exhibit improved adversarial robustness.

Multimodal large language models (MLLMs) exhibit strong capabilities across diverse applications, yet remain vulnerable to adversarial perturbations that distort their feature representations and induce erroneous predictions. To address this vulnerability, we propose the Feature-space Smoothing (FS) and theoretically prove that FS offers certified robustness on the feature representations of MLLMs. Specifically, FS transforms any feature encoder into a smoothed variant that is guaranteed to maintain a certified lower bound on the feature cosine similarity between clean and adversarial representations under $\ell_2$-bounded attacks. Moreover, we indicate that the value of this Feature Cosine Similarity Bound (FCSB) derived from FS can be improved by enlarging the defined Gaussian robustness score on the vanilla encoder. Building upon this, we introduce the Purifier and Smoothness Mapper (PSM), a plug-and-play module that improves the Gaussian robustness score of MLLMs and thus enhances their certified robustness under FS, without requiring any retraining on MLLMs. We demonstrate that the FS with PSM not only provides a strong theoretical robustness guarantee but also exhibits superior empirical performance compared to adversarial training. Extensive experiments across diverse MLLMs and downstream tasks indicate the effectiveness of the FS-PSM, reducing the Attack Success Rate (ASR) of various white-box attacks from nearly 90\% to about 1\%.

This paper investigates the multi-compartment vehicle routing problem with multiple time windows (MCVRPMTW), an extension of the classical vehicle routing problem with time windows that considers vehicles equipped with multiple compartments and customers requiring service across several delivery time windows. The problem incorporates three key compartment-related features: (i) compartment flexibility in the number of compartments, (ii) item-to-compartment compatibility, and (iii) item-to-item compatibility. The problem also accommodates practical operational requirements such as driver breaks. To solve the MCVRPMTW, we develop an exact branch-and-price (B&P) algorithm in which the pricing problem is solved using a labeling algorithm. Several acceleration strategies are introduced to limit symmetry during label extensions, improve the stability of dual solutions in column generation, and enhance the branching process. To handle large-scale instances, we propose a rolling-space B&P algorithm that integrates clustering techniques into the solution framework. Extensive computational experiments on instances inspired by a real-world industrial application demonstrate the effectiveness of the proposed approach and provide useful managerial insights for practical implementation.

We propose average unfairness as a new measure of fairness in routing games, defined as the ratio between the average latency and the minimum latency experienced by users. This measure is a natural complement to two existing unfairness notions: loaded unfairness, which compares maximum and minimum latencies of routes with positive flow, and user equilibrium (UE) unfairness, which compares maximum latency with the latency of a Nash equilibrium. We show that the worst-case values of all three unfairness measures coincide and are characterized by a steepness parameter intrinsic to the latency function class. We show that average unfairness is always no greater than loaded unfairness, and the two measures are equal only when the flow is fully fair. Besides that, we offer a complete comparison of the three unfairness measures, which, to the best of our knowledge, is the first theoretical analysis in this direction. Finally, we study the constrained system optimum (CSO) problem, where one seeks to minimize total latency subject to an upper bound on unfairness. We prove that, for the same tolerance level, the optimal flow under an average unfairness constraint achieves lower total latency than any flow satisfying a loaded unfairness constraint. We show that such improvement is always strict in parallel-link networks and establish sufficient conditions for general networks. We further illustrate the latter with numerical examples. Our results provide theoretical guarantees and valuable insights for evaluating fairness-efficiency tradeoffs in network routing.

How can we use AI to discover a new state of the art for a scientific problem? Prior work in test-time scaling, such as AlphaEvolve, performs search by prompting a frozen LLM. We perform reinforcement learning at test time, so the LLM can continue to train, but now with experience specific to the test problem. This form of continual learning is quite special, because its goal is to produce one great solution rather than many good ones on average, and to solve this very problem rather than generalize to other problems. Therefore, our learning objective and search subroutine are designed to prioritize the most promising solutions. We call this method Test-Time Training to Discover (TTT-Discover). Following prior work, we focus on problems with continuous rewards. We report results for every problem we attempted, across mathematics, GPU kernel engineering, algorithm design, and biology. TTT-Discover sets the new state of the art in almost all of them: (i) Erdős' minimum overlap problem and an autocorrelation inequality; (ii) a GPUMode kernel competition (up to $2\times$ faster than prior art); (iii) past AtCoder algorithm competitions; and (iv) denoising problem in single-cell analysis. Our solutions are reviewed by experts or the organizers. All our results are achieved with an open model, OpenAI gpt-oss-120b, and can be reproduced with our publicly available code, in contrast to previous best results that required closed frontier models. Our test-time training runs are performed using Tinker, an API by Thinking Machines, with a cost of only a few hundred dollars per problem.

Uncertainty estimation in machine learning has traditionally focused on the prediction stage, aiming to quantify confidence in model outputs while treating learned representations as deterministic and reliable by default. In this work, we challenge this implicit assumption and argue that reliability should be regarded as a first-class property of learned representations themselves. We propose a principled framework for reliable representation learning that explicitly models representation-level uncertainty and leverages structural constraints as inductive biases to regularize the space of feasible representations. Our approach introduces uncertainty-aware regularization directly in the representation space, encouraging representations that are not only predictive but also stable, well-calibrated, and robust to noise and structural perturbations. Structural constraints, such as sparsity, relational structure, or feature-group dependencies, are incorporated to define meaningful geometry and reduce spurious variability in learned representations, without assuming fully correct or noise-free structure. Importantly, the proposed framework is independent of specific model architectures and can be integrated with a wide range of representation learning methods.

State-of-the-art neural theorem provers like DeepSeek-Prover-V1.5 combine large language models with reinforcement learning, achieving impressive results through sophisticated training. We ask: do these highly-trained models still benefit from simple structural guidance at inference time? We evaluate a lightweight intervention -- a fixed prompt schedule over 15 common tactic skeletons -- on the miniF2F benchmark. This simple approach yields 21.7% pass@16 compared to 15.2% for standard sampling from the same model, a 43% relative improvement using the same number of samples (k=16) and same maximum generation length (1024 tokens). Our results suggest that even capable RL-trained provers underutilize structural priors available in the tactic language, and that simple inference-time guidance remains a cheap, complementary boost.

Hybrid quantum-HPC algorithms advance research by delegating complex tasks to quantum processors and using HPC systems to orchestrate workflows and complementary computations. Sample-based quantum diagonalization (SQD) is a hybrid quantum-HPC method in which information from a molecular Hamiltonian is encoded into a quantum circuit for evaluation on a quantum computer. A set of measurements on the quantum computer yields electronic configurations that are filtered on the classical computer, which also performs diagonalization on the selected subspace and identifies configurations to be carried over to the next step in an iterative process. Diagonalization is the most demanding task for the classical computer. Previous studies used the Fugaku supercomputer and a highly scalable diagonalization code designed for CPUs. In this work, we describe our efforts to enable efficient scalable and portable diagonalization on heterogeneous systems using GPUs as the main compute engines based on the previous work. GPUs provide massive on-device thread-level parallelism that is well aligned with the algorithms used for diagonalization. We focus on the computation of ground-state energies and wavefunctions using the Davidson algorithm with a selected set of electron configurations. We describe the offload strategy, code transformations, and data-movement, with examples of measurements on the Frontier supercomputer and five other GPU accelerated systems. Our measurements show that GPUs provide an outstanding performance boost of order 100x on a per-node basis. This dramatically expedites the diagonalization step-essential for extracting ground and excited state energies-bringing the classical processing time down from hours to minutes.

Recent video generation models demonstrate remarkable ability to capture complex physical interactions and scene evolution over time. To leverage their spatiotemporal priors, robotics works have adapted video models for policy learning but introduce complexity by requiring multiple stages of post-training and new architectural components for action generation. In this work, we introduce Cosmos Policy, a simple approach for adapting a large pretrained video model (Cosmos-Predict2) into an effective robot policy through a single stage of post-training on the robot demonstration data collected on the target platform, with no architectural modifications. Cosmos Policy learns to directly generate robot actions encoded as latent frames within the video model's latent diffusion process, harnessing the model's pretrained priors and core learning algorithm to capture complex action distributions. Additionally, Cosmos Policy generates future state images and values (expected cumulative rewards), which are similarly encoded as latent frames, enabling test-time planning of action trajectories with higher likelihood of success. In our evaluations, Cosmos Policy achieves state-of-the-art performance on the LIBERO and RoboCasa simulation benchmarks (98.5% and 67.1% average success rates, respectively) and the highest average score in challenging real-world bimanual manipulation tasks, outperforming strong diffusion policies trained from scratch, video model-based policies, and state-of-the-art vision-language-action models fine-tuned on the same robot demonstrations. Furthermore, given policy rollout data, Cosmos Policy can learn from experience to refine its world model and value function and leverage model-based planning to achieve even higher success rates in challenging tasks. We release code, models, and training data at https://research.nvidia.com/labs/dir/cosmos-policy/

Keyword Spotting (KWS) systems with small footprint models deployed on edge devices face significant accuracy and robustness challenges due to domain shifts caused by varying noise and recording conditions. To address this, we propose a comprehensive framework for continual learning designed to adapt to new domains while maintaining computational efficiency. The proposed pipeline integrates a dual-input Convolutional Neural Network, utilizing both Mel Frequency Cepstral Coefficients (MFCC) and Mel-spectrogram features, supported by a multi-stage denoising process, involving discrete wavelet transform and spectral subtraction techniques, plus model and prototype update blocks. Unlike prior methods that restrict updates to specific layers, our approach updates the complete quantized model, made possible due to compact model architecture. A subset of input samples are selected during runtime using class prototypes and confidence-driven filtering, which are then pseudo-labeled and combined with rehearsal buffer for incremental model retraining. Experimental results on noisy test dataset demonstrate the framework's effectiveness, achieving 99.63\% accuracy on clean data and maintaining robust performance (exceeding 94\% accuracy) across diverse noisy environments, even at -10 dB Signal-to-Noise Ratio. The proposed framework work confirms that integrating efficient denoising with prototype-based continual learning enables KWS models to operate autonomously and robustly in resource-constrained, dynamic environments.

Modern data systems increasingly operate under conditions of persistent legal, political, and analytic disagreement. In such settings, interoperability cannot rely on shared interpretation, negotiated semantics, or centralized authority. Instead, representations must function as neutral substrates that preserve stable reference across incompatible extensions. This paper investigates the structural constraints imposed on ontological design by this requirement. Building on a neutrality framework that treats interpretive non-commitment and stability under extension as explicit design constraints, we ask what minimal ontological structure is forced if accountability relationships are to remain referable and comparable under disagreement. Minimality here is not mere parsimony: a reduction is admissible only if it does not reintroduce stability-critical distinctions as hidden roles, flags, or contextual predicates. We establish a conditional lower-bound result: any ontology capable of supporting accountability under persistent disagreement must realize at least six distinct identity-and-persistence regimes. We further show that a construction with exactly six such regimes is sufficient to satisfy the stated requirements without embedding causal or normative commitments in the substrate. The result is not a proposal for a universal ontology, but a constraint on what is possible when neutrality and stable reference are treated as non-negotiable design goals.

Melodic harmonization, the task of generating harmonic accompaniments for a given melody, remains a central challenge in computational music generation. Recent single encoder transformer approaches have framed harmonization as a masked sequence modeling problem, but existing training curricula inspired by discrete diffusion often result in weak (cross) attention between melody and harmony. This leads to limited exploitation of melodic cues, particularly in out-of-domain contexts. In this work, we introduce a training curriculum, FF (full-to-full), which keeps all harmony tokens masked for several training steps before progressively unmasking entire sequences during training to strengthen melody-harmony interactions. We systematically evaluate this approach against prior curricula across multiple experimental axes, including temporal quantization (quarter vs. sixteenth note), bar-level vs. time-signature conditioning, melody representation (full range vs. pitch class), and inference-time unmasking strategies. Models are trained on the HookTheory dataset and evaluated both in-domain and on a curated collection of jazz standards, using a comprehensive set of metrics that assess chord progression structure, harmony-melody alignment, and rhythmic coherence. Results demonstrate that the proposed FF curriculum consistently outperforms baselines in nearly all metrics, with particularly strong gains in out-of-domain evaluations where harmonic adaptability to novel melodic queues is crucial. We further find that quarter-note quantization, intertwining of bar tokens, and pitch-class melody representations are advantageous in the FF setting. Our findings highlight the importance of training curricula in enabling effective melody conditioning and suggest that full-to-full unmasking offers a robust strategy for single encoder harmonization.

Obtaining labelled ECG data for developing supervised models is challenging. Contrastive learning (CL) has emerged as a promising pretraining approach that enables effective transfer learning with limited labelled data. However, existing CL frameworks either focus solely on global context or fail to exploit ECG-specific characteristics. Furthermore, these methods rely on hard contrastive targets, which may not adequately capture the continuous nature of feature similarity in ECG signals. In this paper, we propose Beat-SSL, a contrastive learning framework that performs dual-context learning through both rhythm-level and heartbeat-level contrasting with soft targets. We evaluated our pretrained model on two downstream tasks: 1) multilabel classification for global rhythm assessment, and 2) ECG segmentation to assess its capacity to learn representations across both contexts. We conducted an ablation study and compared the best configuration with three other methods, including one ECG foundation model. Despite the foundation model's broader pretraining, Beat-SSL reached 93% of its performance in multilabel classification task and surpassed all other methods in the segmentation task by 4%.

The problem of determining the (least) fixpoint of (higher-dimensional) functions over the non-negative reals frequently occurs when dealing with systems endowed with a quantitative semantics. We focus on the situation in which the functions of interest are not known precisely but can only be approximated. As a first contribution we generalize an iteration scheme called dampened Mann iteration, recently introduced in the literature. The improved scheme relaxes previous constraints on parameter sequences, allowing learning rates to converge to zero or not converge at all. While seemingly minor, this flexibility is essential to enable the implementation of chaotic iterations, where only a subset of components is updated in each step, allowing to tackle higher-dimensional problems. Additionally, by allowing learning rates to converge to zero, we can relax conditions on the convergence speed of function approximations, making the method more adaptable to various scenarios. We also show that dampened Mann iteration applies immediately to compute the expected payoff in various probabilistic models, including simple stochastic games, not covered by previous work.

Existing approaches for watermarking AI-generated images often rely on post-hoc methods applied in pixel space, introducing computational overhead and potential visual artifacts. In this work, we explore latent space watermarking and introduce DistSeal, a unified approach for latent watermarking that works across both diffusion and autoregressive models. Our approach works by training post-hoc watermarking models in the latent space of generative models. We demonstrate that these latent watermarkers can be effectively distilled either into the generative model itself or into the latent decoder, enabling in-model watermarking. The resulting latent watermarks achieve competitive robustness while offering similar imperceptibility and up to 20x speedup compared to pixel-space baselines. Our experiments further reveal that distilling latent watermarkers outperforms distilling pixel-space ones, providing a solution that is both more efficient and more robust.

The manifold hypothesis suggests that the generalization performance of machine learning methods improves significantly when the intrinsic dimension of the input distribution's support is low. In the context of KRR, we investigate two alternative notions of intrinsic dimension. The first, denoted $d_ρ$, is the upper Minkowski dimension defined with respect to the canonical metric induced by a kernel function $K$ on a domain $Ω$. The second, denoted $d_K$, is the effective dimension, derived from the decay rate of Kolmogorov $n$-widths associated with $K$ on $Ω$. Given a probability measure $μ$ on $Ω$, we analyze the relationship between these $n$-widths and eigenvalues of the integral operator $φ\to \int_ΩK(\cdot,x)φ(x)dμ(x)$. We show that, for a fixed domain $Ω$, the Kolmogorov $n$-widths characterize the worst-case eigenvalue decay across all probability measures $μ$ supported on $Ω$. These eigenvalues are central to understanding the generalization behavior of constrained KRR, enabling us to derive an excess error bound of order $O(n^{-\frac{2+d_K}{2+2d_K} + ε})$ for any $ε> 0$, when the training set size $n$ is large. We also propose an algorithm that estimates upper bounds on the $n$-widths using only a finite sample from $μ$. For distributions close to uniform, we prove that $ε$-accurate upper bounds on all $n$-widths can be computed with high probability using at most $O\left(ε^{-d_ρ}\log\frac{1}ε\right)$ samples, with fewer required for small $n$. Finally, we compute the effective dimension $d_K$ for various fractal sets and present additional numerical experiments. Our results show that, for kernels such as the Laplace kernel, the effective dimension $d_K$ can be significantly smaller than the Minkowski dimension $d_ρ$, even though $d_K = d_ρ$ provably holds on regular domains.

The Arabic language has undergone notable transformations over time, including the emergence of new vocabulary, the obsolescence of others, and shifts in word usage. This evolution is evident in the distinction between the classical and modern Arabic eras. Although historians and linguists have partitioned Arabic literature into multiple eras, relatively little research has explored the automatic classification of Arabic texts by time period, particularly beyond the domain of poetry. This paper addresses this gap by employing neural networks and deep learning techniques to automatically classify Arabic texts into distinct eras and periods. The proposed models are evaluated using two datasets derived from two publicly available corpora, covering texts from the pre-Islamic to the modern era. The study examines class setups ranging from binary to 15-class classification and considers both predefined historical eras and custom periodizations. Results range from F1-scores of 0.83 and 0.79 on the binary-era classification task using the OpenITI and APCD datasets, respectively, to 0.20 on the 15-era classification task using OpenITI and 0.18 on the 12-era classification task using APCD.

As large language models (LLMs) become increasingly common in educational applications, there is a growing need for evidence-based methods to design and evaluate LLM prompts that produce personalized and pedagogically aligned out-puts. This study presents a generalizable, systematic approach for evaluating prompts, demonstrated through an analysis of LLM-generated follow-up questions in a structured dialogue activity. Six prompt templates were designed and tested. The templates incorporated established prompt engineering patterns, with each prompt emphasizing distinct pedagogical strategies. The prompt templates were compared through a tournament-style evaluation framework that can be adapted for other educational applications. The tournament employed the Glicko2 rating system with eight judges evaluating question pairs across three dimensions: format, dialogue support, and appropriateness for learners. Data was sourced from 120 authentic user interactions across three distinct educational deployments. Results showed that a single prompt related to strategic reading out-performed other templates with win probabilities ranging from 81% to 100% in pairwise comparisons. This prompt combined persona and context manager pat-terns and was designed to support metacognitive learning strategies such as self-directed learning. The methodology showcases how educational technology re- searchers can systematically evaluate and improve prompt designs, moving beyond ad-hoc prompt engineering toward evidence-based prompt development for educational applications.

Motivated reasoning -- the idea that individuals processing information may be motivated to reach a certain conclusion, whether it be accurate or predetermined -- has been well-explored as a human phenomenon. However, it is unclear whether base LLMs mimic these motivational changes. Replicating 4 prior political motivated reasoning studies, we find that base LLM behavior does not align with expected human behavior. Furthermore, base LLM behavior across models shares some similarities, such as smaller standard deviations and inaccurate argument strength assessments. We emphasize the importance of these findings for researchers using LLMs to automate tasks such as survey data collection and argument assessment.

Fine-tuning a task-specific multilingual large language model (LLM) involves training the model on a multilingual dataset with examples in all the required languages. Updating one or more supported languages with additional data or adding support for a new language involves retraining the model, which can be computationally inefficient and creates a severe maintenance bottleneck. Recent research on merging multilingual multitask models has shown promise in terms of improved quality, but its computational and maintenance efficiency remains unstudied. In this work, we provide the first focused analysis of this merging strategy from an efficiency perspective, evaluating it across three independent tasks. We demonstrate significant efficiency gains while maintaining parity in terms of quality: this merging approach reduces the initial training time by up to 50\%. We also demonstrate that updating an individual language and re-merging as part of model maintenance reduces training costs by more than 60\%, compared to re-training the full multilingual model. We show this on both public and proprietary industry datasets confirming that the approach works well for industrial use cases in addition to academic settings already studied in previous work.

Composed Image Retrieval (CIR) is a pivotal and complex task in multimodal understanding. Current CIR benchmarks typically feature limited query categories and fail to capture the diverse requirements of real-world scenarios. To bridge this evaluation gap, we leverage image editing to achieve precise control over modification types and content, enabling a pipeline for synthesizing queries across a broad spectrum of categories. Using this pipeline, we construct EDIR, a novel fine-grained CIR benchmark. EDIR encompasses 5,000 high-quality queries structured across five main categories and fifteen subcategories. Our comprehensive evaluation of 13 multimodal embedding models reveals a significant capability gap; even state-of-the-art models (e.g., RzenEmbed and GME) struggle to perform consistently across all subcategories, highlighting the rigorous nature of our benchmark. Through comparative analysis, we further uncover inherent limitations in existing benchmarks, such as modality biases and insufficient categorical coverage. Furthermore, an in-domain training experiment demonstrates the feasibility of our benchmark. This experiment clarifies the task challenges by distinguishing between categories that are solvable with targeted data and those that expose intrinsic limitations of current model architectures.

Imbalanced classification, where one class is observed far less frequently than the other, often causes standard training procedures to prioritize the majority class and perform poorly on rare but important cases. A classic and widely used remedy is to augment the minority class with synthetic examples, but two basic questions remain under-resolved: when does synthetic augmentation actually help, and how many synthetic samples should be generated? We develop a unified statistical framework for synthetic augmentation in imbalanced learning, studying models trained on imbalanced data augmented with synthetic minority samples and evaluated under the balanced population risk. Our theory shows that synthetic data is not always beneficial. In a ``local symmetry" regime, imbalance is not the dominant source of error near the balanced optimum, so adding synthetic samples cannot improve learning rates and can even degrade performance by amplifying generator mismatch. When augmentation can help (a ``local asymmetry" regime), the optimal synthetic size depends on generator accuracy and on whether the generator's residual mismatch is directionally aligned with the intrinsic majority-minority shift. This structure can make the best synthetic size deviate from naive full balancing, sometimes by a small refinement and sometimes substantially when generator bias is systematic. Practically, we recommend Validation-Tuned Synthetic Size (VTSS): select the synthetic size by minimizing balanced validation loss over a range centered near the fully balanced baseline, while allowing meaningful departures when the data indicate them. Simulations and a real sepsis prediction study support the theory and illustrate when synthetic augmentation helps, when it cannot, and how to tune its quantity effectively.

Executing Spiking Neural Networks (SNNs) on neuromorphic hardware poses the problem of mapping neurons to cores. SNNs operate by propagating spikes between neurons that form a graph through synapses. Neuromorphic hardware mimics them through a network-on-chip, transmitting spikes, and a mesh of cores, each managing several neurons. Its operational cost is tied to spike movement and active cores. A mapping comprises two tasks: partitioning the SNN's graph to fit inside cores and placement of each partition on the hardware mesh. Both are NP-hard problems, and as SNNs and hardware scale towards billions of neurons, they become increasingly difficult to tackle effectively. In this work, we propose to raise the abstraction of SNNs from graphs to hypergraphs, redesigning mapping techniques accordingly. The resulting model faithfully captures the replication of spikes inside cores by exposing the notion of hyperedge co-membership between neurons. We further show that the overlap and locality of hyperedges strongly correlate with high-quality mappings, making these properties instrumental in devising mapping algorithms. By exploiting them directly, grouping neurons through shared hyperedges, communication traffic and hardware resource usage can be reduced be yond what just contracting individual connections attains. To substantiate this insight, we consider several partitioning and placement algorithms, some newly devised, others adapted from literature, and compare them over progressively larger and bio-plausible SNNs. Our results show that hypergraph based techniques can achieve better mappings than the state-of-the-art at several execution time regimes. Based on these observations, we identify a promising selection of algorithms to achieve effective mappings at any scale.

Optical Character Recognition (OCR) for low-resource languages remains a significant challenge due to the scarcity of large-scale annotated training datasets. Languages such as Kashmiri, with approximately 7 million speakers and a complex Perso-Arabic script featuring unique diacritical marks, currently lack support in major OCR systems including Tesseract, TrOCR, and PaddleOCR. Manual dataset creation for such languages is prohibitively expensive, time-consuming, and error-prone, often requiring word by word transcription of printed or handwritten text. We present SynthOCR-Gen, an open-source synthetic OCR dataset generator specifically designed for low-resource languages. Our tool addresses the fundamental bottleneck in OCR development by transforming digital Unicode text corpora into ready-to-use training datasets. The system implements a comprehensive pipeline encompassing text segmentation (character, word, n-gram, sentence, and line levels), Unicode normalization with script purity enforcement, multi-font rendering with configurable distribution, and 25+ data augmentation techniques simulating real-world document degradations including rotation, blur, noise, and scanner artifacts. We demonstrate the efficacy of our approach by generating a 600,000-sample word-segmented Kashmiri OCR dataset, which we release publicly on HuggingFace. This work provides a practical pathway for bringing low-resource languages into the era of vision-language AI models, and the tool is openly available for researchers and practitioners working with underserved writing systems worldwide.

We propose a variable splitting binary tree (VSBT) model based on Bayesian context tree (BCT) models for time series segmentation. Unlike previous applications of BCT models, the tree structure in our model represents interval partitioning on the time domain. Moreover, interval partitioning is represented by recursive logistic regression models. By adjusting logistic regression coefficients, our model can represent split positions at arbitrary locations within each interval. This enables more compact tree representations. For simultaneous estimation of both split positions and tree depth, we develop an effective inference algorithm that combines local variational approximation for logistic regression with the context tree weighting (CTW) algorithm. We present numerical examples on synthetic data demonstrating the effectiveness of our model and algorithm.

Climate disinformation has become a major challenge in today digital world, especially with the rise of misleading images and videos shared widely on social media. These false claims are often convincing and difficult to detect, which can delay actions on climate change. While vision-language models (VLMs) have been used to identify visual disinformation, they rely only on the knowledge available at the time of training. This limits their ability to reason about recent events or updates. The main goal of this paper is to overcome that limitation by combining VLMs with external knowledge. By retrieving up-to-date information such as reverse image results, online fact-checks, and trusted expert content, the system can better assess whether an image and its claim are accurate, misleading, false, or unverifiable. This approach improves the model ability to handle real-world climate disinformation and supports efforts to protect public understanding of science in a rapidly changing information landscape.

Deep learning classifiers for Raman spectroscopy are increasingly reported to outperform classical chemometric approaches. However their evaluations are often conducted in isolation or compared against traditional machine learning methods or trivially adapted vision-based architectures that were not originally proposed for Raman spectroscopy. As a result, direct comparisons between existing deep learning models developed specifically for Raman spectral analysis on shared open-source datasets remain scarce. To the best of our knowledge, this study presents one of the first systematic benchmarks comparing three or more published Raman-specific deep learning classifiers across multiple open-source Raman datasets. We evaluate five representative deep learning architectures under a unified training and hyperparameter tuning protocol across three open-source Raman datasets selected to support standard evaluation, fine-tuning, and explicit distribution-shift testing. We report classification accuracies and macro-averaged F1 scores to provide a fair and reproducible comparison of deep learning models for Raman spectra based classification.

Although Mamba models greatly improve Hyperspectral Image (HSI) classification, they have critical challenges in terms defining efficient and adaptive token sequences for improve performance. This paper therefore presents CSSMamba (Clustering-guided Spatial-Spectral Mamba) framework to better address the challenges, with the following contributions. First, to achieve efficient and adaptive token sequences for improved Mamba performance, we integrate the clustering mechanism into a spatial Mamba architecture, leading to a cluster-guided spatial Mamba module (CSpaMamba) that reduces the Mamba sequence length and improves Mamba feature learning capability. Second, to improve the learning of both spatial and spectral information, we integrate the CSpaMamba module with a spectral mamba module (SpeMamba), leading to a complete clustering-guided spatial-spectral Mamba framework. Third, to further improve feature learning capability, we introduce an Attention-Driven Token Selection mechanism to optimize Mamba token sequencing. Last, to seamlessly integrate clustering into the Mamba model in a coherent manner, we design a Learnable Clustering Module that learns the cluster memberships in an adaptive manner. Experiments on the Pavia University, Indian Pines, and Liao-Ning 01 datasets demonstrate that CSSMamba achieves higher accuracy and better boundary preservation compared to state-of-the-art CNN, Transformer, and Mamba-based methods.

Large Language Models enable users to access database using natural language interfaces using tools like Text2SQL, Text2SPARQL, and Text2Cypher, which translate user questions into structured database queries. While these systems improve database accessibility, most research focuses on English with limited multilingual support. This work investigates a scalable multilingual Text2Cypher, aiming to support new languages without re-running full fine-tuning, avoiding manual hyper-parameter tuning, and maintaining performance close to joint multilingual fine-tuning. We train language-specific LoRA adapters for English, Spanish, and Turkish and combined them via uniform linear merging or learned fusion MLP with dynamic gating. Experimental results show that the fusion MLP recovers around 75\% of the accuracy gains from joint multilingual fine-tuning while requiring only a smaller subset of the data, outperforming linear merging across all three languages. This approach enables incremental language expansion to new languages by requiring only one LoRA adapter and a lightweight MLP retraining. Learned adapter fusion offers a practical alternative to expensive joint fine-tuning, balancing performance, data efficiency, and scalability for multilingual Text2Cypher task.

We introduce Neural Particle Automata (NPA), a Lagrangian generalization of Neural Cellular Automata (NCA) from static lattices to dynamic particle systems. Unlike classical Eulerian NCA where cells are pinned to pixels or voxels, NPA model each cell as a particle with a continuous position and internal state, both updated by a shared, learnable neural rule. This particle-based formulation yields clear individuation of cells, allows heterogeneous dynamics, and concentrates computation only on regions where activity is present. At the same time, particle systems pose challenges: neighborhoods are dynamic, and a naive implementation of local interactions scale quadratically with the number of particles. We address these challenges by replacing grid-based neighborhood perception with differentiable Smoothed Particle Hydrodynamics (SPH) operators backed by memory-efficient, CUDA-accelerated kernels, enabling scalable end-to-end training. Across tasks including morphogenesis, point-cloud classification, and particle-based texture synthesis, we show that NPA retain key NCA behaviors such as robustness and self-regeneration, while enabling new behaviors specific to particle systems. Together, these results position NPA as a compact neural model for learning self-organizing particle dynamics.

Clustering is a fundamental problem, aiming to partition a set of elements, like agents or data points, into clusters such that elements in the same cluster are closer to each other than to those in other clusters. In this paper, we present a new framework for studying online non-centroid clustering with delays, where elements, that arrive one at a time as points in a finite metric space, should be assigned to clusters, but assignments need not be immediate. Specifically, upon arrival, each point's location is revealed, and an online algorithm has to irrevocably assign it to an existing cluster or create a new one containing, at this moment, only this point. However, we allow decisions to be postponed at a delay cost, instead of following the more common assumption of immediate decisions upon arrival. This poses a critical challenge: the goal is to minimize both the total distance costs between points in each cluster and the overall delay costs incurred by postponing assignments. In the classic worst-case arrival model, where points arrive in an arbitrary order, no algorithm has a competitive ratio better than sublogarithmic in the number of points. To overcome this strong impossibility, we focus on a stochastic arrival model, where points' locations are drawn independently across time from an unknown and fixed probability distribution over the finite metric space. We offer hope for beyond worst-case adversaries: we devise an algorithm that is constant competitive in the sense that, as the number of points grows, the ratio between the expected overall costs of the output clustering and an optimal offline clustering is bounded by a constant.

Large language model (LLM) agents often exhibit abrupt shifts in tone and persona during extended interaction, reflecting the absence of explicit temporal structure governing agent-level state. While prior work emphasizes turn-local sentiment or static emotion classification, the role of explicit affective dynamics in shaping long-horizon agent behavior remains underexplored. This work investigates whether imposing dynamical structure on an external affective state can induce temporal coherence and controlled recovery in multi-turn dialogue. We introduce an agent-level affective subsystem that maintains a continuous Valence-Arousal-Dominance (VAD) state external to the language model and governed by first- and second-order update rules. Instantaneous affective signals are extracted using a fixed, memoryless estimator and integrated over time via exponential smoothing or momentum-based dynamics. The resulting affective state is injected back into generation without modifying model parameters. Using a fixed 25-turn dialogue protocol, we compare stateless, first-order, and second-order affective dynamics. Stateless agents fail to exhibit coherent trajectories or recovery, while state persistence enables delayed responses and reliable recovery. Second-order dynamics introduce affective inertia and hysteresis that increase with momentum, revealing a trade-off between stability and responsiveness.

Computing the conditional mode of a distribution, better known as the $\mathit{maximum\ a\ posteriori}$ (MAP) assignment, is a fundamental task in probabilistic inference. However, MAP estimation is generally intractable, and remains hard even under many common structural constraints and approximation schemes. We introduce $\mathit{probably\ approximately\ correct}$ (PAC) algorithms for MAP inference that provide provably optimal solutions under variable and fixed computational budgets. We characterize tractability conditions for PAC-MAP using information theoretic measures that can be estimated from finite samples. Our PAC-MAP solvers are efficiently implemented using probabilistic circuits with appropriate architectures. The randomization strategies we develop can be used either as standalone MAP inference techniques or to improve on popular heuristics, fortifying their solutions with rigorous guarantees. Experiments confirm the benefits of our method in a range of benchmarks.

Implementing privacy by design (PbD) according to the General Data Protection Regulation (GDPR) is met with a growing number of requirements engineering (RE) approaches. However, the question of which RE method for PbD fits best the goals of organisations remains a challenge. We report our endeavor to close this gap by synthesizing a goal-centric approach for PbD methods assessment. We used literature review, interviews, and validation with practitioners to achieve the goal of our study. As practitioners do not approach PbD systematically, we suggest that RE methods for PbD should be assessed against organisational goals, rather than process characteristics only. We hope that, when further developed, the goal-centric approach could support the development, selection, and tailoring of RE practices for PbD.

Industrial Cyber-Physical Systems (CPS) are sensitive infrastructure from both safety and economics perspectives, making their reliability critically important. Machine Learning (ML), specifically deep learning, is increasingly integrated in industrial CPS, but the inherent complexity of ML models results in non-transparent operation. Rigorous evaluation is needed to prevent models from exhibiting unexpected behaviour on future, unseen data. Explainable AI (XAI) can be used to uncover model reasoning, allowing a more extensive analysis of behaviour. We apply XAI to to improve predictive performance of ML models intended for industrial CPS. We analyse the effects of components from time-series data decomposition on model predictions using SHAP values. Through this method, we observe evidence on the lack of sufficient contextual information during model training. By increasing the window size of data instances, informed by the XAI findings, we are able to improve model performance.

Foundation Models (FMs) have demonstrated strong generalization across diverse vision tasks. However, their deployment in federated settings is hindered by high computational demands, substantial communication overhead, and significant inference costs. We propose DSFedMed, a dual-scale federated framework that enables mutual knowledge distillation between a centralized foundation model and lightweight client models for medical image segmentation. To support knowledge distillation, a set of high-quality medical images is generated to replace real public datasets, and a learnability-guided sample selection strategy is proposed to enhance efficiency and effectiveness in dual-scale distillation. This mutual distillation enables the foundation model to transfer general knowledge to lightweight clients, while also incorporating client-specific insights to refine the foundation model. Evaluations on five medical imaging segmentation datasets show that DSFedMed achieves an average 2 percent improvement in Dice score while reducing communication costs and inference time by nearly 90 percent compared to existing federated foundation model baselines. These results demonstrate significant efficiency gains and scalability for resource-limited federated deployments.

We study Constrained Online Convex Optimization (COCO), where a learner chooses actions iteratively, observes both unanticipated convex loss and convex constraint, and accumulates loss while incurring penalties for constraint violations. We introduce CLASP (Convex Losses And Squared Penalties), an algorithm that minimizes cumulative loss together with squared constraint violations. Our analysis departs from prior work by fully leveraging the firm non-expansiveness of convex projectors, a proof strategy not previously applied in this setting. For convex losses, CLASP achieves regret $O\left(T^{\max\{β,1-β\}}\right)$ and cumulative squared penalty $O\left(T^{1-β}\right)$ for any $β\in (0,1)$. Most importantly, for strongly convex problems, CLASP provides the first logarithmic guarantees on both regret and cumulative squared penalty. In the strongly convex case, the regret is upper bounded by $O( \log T )$ and the cumulative squared penalty is also upper bounded by $O( \log T )$.

Deep neural networks (DNNs) typically involve a large number of parameters and are trained to achieve zero or near-zero training error. Despite such interpolation, they often exhibit strong generalization performance on unseen data, a phenomenon that has motivated extensive theoretical investigations. Comforting results show that interpolation indeed may not affect the minimax rate of convergence under the squared error loss. In the mean time, DNNs are well known to be highly vulnerable to adversarial perturbations in future inputs. A natural question then arises: Can interpolation also escape from suboptimal performance under a future $X$-attack? In this paper, we investigate the adversarial robustness of interpolating estimators in a framework of nonparametric regression. A finding is that interpolating estimators must be suboptimal even under a subtle future $X$-attack, and achieving perfect fitting can substantially damage their robustness. An interesting phenomenon in the high interpolation regime, which we term the curse of simple size, is also revealed and discussed. Numerical experiments support our theoretical findings.

Designing faster algorithms for solving Mixed-Integer Linear Programming (MILP) problems is highly desired across numerous practical domains, as a vast array of complex real-world challenges can be effectively modeled as MILP formulations. Solving these problems typically employs the branch-and-bound algorithm, the core of which can be conceived as searching for a path of nodes (or sub-problems) that contains the optimal solution to the original MILP problem. Traditional approaches to finding this path rely heavily on hand-crafted, intuition-based heuristic strategies, which often suffer from unstable and unpredictable performance across different MILP problem instances. To address this limitation, we introduce DeepBound, a deep learning-based node selection algorithm that automates the learning of such human intuition from data. The core of DeepBound lies in learning to prioritize nodes containing the optimal solution, thereby improving solving efficiency. DeepBound introduces a multi-level feature fusion network to capture the node representations. To tackle the inherent node imbalance in branch-and-bound trees, DeepBound employs a pairwise training paradigm that enhances the model's ability to discriminate between nodes. Extensive experiments on three NP-hard MILP benchmarks demonstrate that DeepBound achieves superior solving efficiency over conventional heuristic rules and existing learning-based approaches, obtaining optimal feasible solutions with significantly reduced computation time. Moreover, DeepBound demonstrates strong generalization capability on large and complex instances. The analysis of its learned features reveals that the method can automatically discover more flexible and robust feature selection, which may effectively improve and potentially replace human-designed heuristic rules.

Accurate prediction of crop above-ground biomass (AGB) under water stress is critical for monitoring crop productivity, guiding irrigation, and supporting climate-resilient agriculture. Data-driven models scale well but often lack interpretability and degrade under distribution shift, whereas process-based crop models (e.g. DSSAT, APSIM, LINTUL5) require extensive calibration and are difficult to deploy over large spatial domains. To address these limitations, we propose AgriPINN, a process-informed neural network that integrates a biophysical crop-growth differential equation as a differentiable constraint within a deep learning backbone. This design encourages physiologically consistent biomass dynamics under water-stress conditions while preserving model scalability for spatially distributed AGB prediction. AgriPINN recovers latent physiological variables, including leaf area index (LAI), absorbed photosynthetically active radiation (PAR), radiation use efficiency (RUE), and water-stress factors, without requiring direct supervision. We pretrain AgriPINN on 60 years of historical data across 397 regions in Germany and fine-tune it on three years of field experiments under controlled water treatments. Results show that AgriPINN consistently outperforms state-of-the-art deep-learning baselines (ConvLSTM-ViT, SLTF, CNN-Transformer) and the process-based LINTUL5 model in terms of accuracy (RMSE reductions up to $43\%$) and computational efficiency. By combining the scalability of deep learning with the biophysical rigor of process-based modeling, AgriPINN provides a robust and interpretable framework for spatio-temporal AGB prediction, offering practical value for planning of irrigation infrastructure, yield forecasting, and climate-adaptation planning.

In constrained stochastic optimization, one naturally expects that imposing a stricter feasible set does not increase the statistical risk of an estimator defined by projection onto that set. In this paper, we show that this intuition can fail even in canonical settings. We study the Gaussian sequence model, a deliberately austere test best, where for a compact, convex set $Θ\subset \mathbb{R}^d$ one observes \[ Y = θ^\star + σZ, \qquad Z \sim N(0, I_d), \] and seeks to estimate an unknown parameter $θ^\star \in Θ$. The natural estimator is the least squares estimator (LSE), which coincides with the Euclidean projection of $Y$ onto $Θ$. We construct an explicit example exhibiting \emph{risk reversal}: for sufficiently large noise, there exist nested compact convex sets $Θ_S \subset Θ_L$ and a parameter $θ^\star \in Θ_S$ such that the LSE constrained to $Θ_S$ has strictly larger risk than the LSE constrained to $Θ_L$. We further show that this phenomenon can persist at the level of worst-case risk, with the supremum risk over the smaller constraint set exceeding that over the larger one. We clarify this behavior by contrasting noise regimes. In the vanishing-noise limit, the risk admits a first-order expansion governed by the statistical dimension of the tangent cone at $θ^\star$, and tighter constraints uniformly reduce risk. In contrast, in the diverging-noise regime, the risk is determined by global geometric interactions between the constraint set and random noise directions. Here, the embedding of $Θ_S$ within $Θ_L$ can reverse the risk ordering. These results reveal a previously unrecognized failure mode of projection-based estimators: in sufficiently noisy settings, tightening a constraint can paradoxically degrade statistical performance.

Large Language Models (LLMs) can aid synthesis planning in chemistry, but standard prompting methods often yield hallucinated or outdated suggestions. We study LLM interactions with a reaction knowledge graph by casting reaction path retrieval as a Text2Cypher (natural language to graph query) generation problem, and define single- and multi-step retrieval tasks. We compare zero-shot prompting to one-shot variants using static, random, and embedding-based exemplar selection, and assess a checklist-driven validator/corrector loop. To evaluate our framework, we consider query validity and retrieval accuracy. We find that one-shot prompting with aligned exemplars consistently performs best. Our checklist-style self-correction loop mainly improves executability in zero-shot settings and offers limited additional retrieval gains once a good exemplar is present. We provide a reproducible Text2Cypher evaluation setup to facilitate further work on KG-grounded LLMs for synthesis planning. Code is available at https://github.com/Intelligent-molecular-systems/KG-LLM-Synthesis-Retrieval.

Refusal behavior in aligned LLMs is often viewed as model-specific, yet we hypothesize it stems from a universal, low-dimensional semantic circuit shared across models. To test this, we introduce Trajectory Replay via Concept-Basis Reconstruction, a framework that transfers refusal interventions from donor to target models, spanning diverse architectures (e.g., Dense to MoE) and training regimes, without using target-side refusal supervision. By aligning layers via concept fingerprints and reconstructing refusal directions using a shared ``recipe'' of concept atoms, we map the donor's ablation trajectory into the target's semantic space. To preserve capabilities, we introduce a weight-SVD stability guard that projects interventions away from high-variance weight subspaces to prevent collateral damage. Our evaluation across 8 model pairs (including GPT-OSS-20B and GLM-4) confirms that these transferred recipes consistently attenuate refusal while maintaining performance, providing strong evidence for the semantic universality of safety alignment.

High-performance attention kernels are essential for Large Language Models. This paper presents analysis of CuTile-based Flash Attention memory behavior and a technique to improve its cache performance. In particular, our analysis on the NVIDIA GB10 (Grace Blackwell) identifies the main cause of L2 cache miss. Leveraging this insight, we introduce a new programming technique called Sawtooth Wavefront Reordering that reduces L2 misses. We validate it in both CUDA and CuTile, observing 50\% or greater reduction in L2 misses and up to 60\% increase in throughput on GB10.

With the increasing flexibilization of processes, determining robust scheduling decisions has become an important goal. Traditionally, the flexibility index has been used to identify safe operating schedules by approximating the admissible uncertainty region using simple admissible uncertainty sets, such as hypercubes. Presently, available contextual information, such as forecasts, has not been considered to define the admissible uncertainty set when determining the flexibility index. We propose the conditional flexibility index (CFI), which extends the traditional flexibility index in two ways: by learning the parametrized admissible uncertainty set from historical data and by using contextual information to make the admissible uncertainty set conditional. This is achieved using a normalizing flow that learns a bijective mapping from a Gaussian base distribution to the data distribution. The admissible latent uncertainty set is constructed as a hypersphere in the latent space and mapped to the data space. By incorporating contextual information, the CFI provides a more informative estimate of flexibility by defining admissible uncertainty sets in regions that are more likely to be relevant under given conditions. Using an illustrative example, we show that no general statement can be made about data-driven admissible uncertainty sets outperforming simple sets, or conditional sets outperforming unconditional ones. However, both data-driven and conditional admissible uncertainty sets ensure that only regions of the uncertain parameter space containing realizations are considered. We apply the CFI to a security-constrained unit commitment example and demonstrate that the CFI can improve scheduling quality by incorporating temporal information.

The rise of live streaming has transformed online interaction, enabling massive real-time engagement but also exposing platforms to complex risks such as scams and coordinated malicious behaviors. Detecting these risks is challenging because harmful actions often accumulate gradually and recur across seemingly unrelated streams. To address this, we propose CS-VAR (Cross-Session Evidence-Aware Retrieval-Augmented Detector) for live streaming risk assessment. In CS-VAR, a lightweight, domain-specific model performs fast session-level risk inference, guided during training by a Large Language Model (LLM) that reasons over retrieved cross-session behavioral evidence and transfers its local-to-global insights to the small model. This design enables the small model to recognize recurring patterns across streams, perform structured risk assessment, and maintain efficiency for real-time deployment. Extensive offline experiments on large-scale industrial datasets, combined with online validation, demonstrate the state-of-the-art performance of CS-VAR. Furthermore, CS-VAR provides interpretable, localized signals that effectively empower real-world moderation for live streaming.

This paper presents Mecellem models, a framework for developing specialized language models for the Turkish legal domain through domain adaptation strategies. We make two contributions: (1)Encoder Model Pre-trained from Scratch: ModernBERT-based bidirectional encoders pre-trained on a Turkish-dominant corpus of 112.7 billion tokens. We implement a checkpoint selection strategy that evaluates downstream retrieval performance throughout training, revealing that optimal checkpoints achieve best retrieval scores before pre-training loss reaches its minimum. Our encoder models achieve top-3 rankings on the Turkish retrieval leaderboard, with smaller models (155M parameters) achieving comparable performance to larger reference models (307M-567M parameters). Our approach achieves 92.36% production efficiency compared to state-of-the-art models (embeddinggemma-300m: 100.00%, BAAI/bge-m3: 99.54%, newmindai/bge-m3-stsb: 94.38%), ranking fourth overall despite requiring less computational resources. SOTA models rely on multi-stage, computationally intensive training pipelines, making our single-stage pre-training followed by efficient post-training approach a cost-effective alternative; (2)Decoder Model with Continual Pre-training (CPT): Qwen3-1.7B and Qwen3-4B models adapted to Turkish legal domain through controlled curriculum learning. Four-phase CPT with optimal sample ratios enables gradual transition from general language knowledge to specialized legal terminology and long-context reasoning. This approach achieves 36.2% perplexity reduction on Turkish legal text, demonstrating domain adaptation gains.

Current Earth observation foundation models are architecturally rigid, struggle with heterogeneous sensors and are constrained to fixed patch sizes. This limits their deployment in real-world scenarios requiring flexible computeaccuracy trade-offs. We propose THOR, a "computeadaptive" foundation model that solves both input heterogeneity and deployment rigidity. THOR is the first architecture to unify data from Copernicus Sentinel-1, -2, and -3 (OLCI & SLSTR) satellites, processing their native 10 m to 1000 m resolutions in a single model. We pre-train THOR with a novel randomized patch and input image size strategy. This allows a single set of pre-trained weights to be deployed at inference with any patch size, enabling a dynamic trade-off between computational cost and feature resolution without retraining. We pre-train THOR on THOR Pretrain, a new, large-scale multi-sensor dataset and demonstrate state-of-the-art performance on downstream benchmarks, particularly in data-limited regimes like the PANGAEA 10% split, validating that THOR's flexible feature generation excels for diverse climate and society applications.

The representation of requirements plays a critical role in the accuracy of requirements inspection. While visual representations, such as UML diagrams, are widely used alongside text-based requirements, their effectiveness in supporting inspection is still debated. Cognitive abilities, such as working memory and mental rotation skills, may also influence inspection accuracy. This study aims to evaluate whether the use of UML sequence diagrams alongside text-based requirements improves the accuracy of requirements inspection compared to text-based requirements alone and to explore whether cognitive abilities are associated with differences in performance across the two treatments (text vs text with UML support). We conducted a crossover experiment with 38 participants to assess the accuracy of requirements inspection under the two treatments in terms of issues found and justifications provided. Linear mixed-effects and generalized linear models were used to analyse the effects of treatment, period, sequence, and cognitive abilities. The results indicate a significant three-way interaction between representation type, working memory capacity, and mental rotation ability. This finding suggests that the effectiveness of UML support is not uniform across individuals: participants with high scores in both cognitive abilities experienced reduced performance when using UML for violation detection. Conversely, the same cognitive profile was associated with improved justification accuracy under UML-aided inspection, indicating that higher cognitive abilities may support deeper reasoning processes when dealing with multi-modal information, i.e., diagrams and text.

Modern foundational Multimodal Large Language Models (MLLMs) and video world models have advanced significantly in mathematical, common-sense, and visual reasoning, but their grasp of the underlying physics remains underexplored. Existing benchmarks attempting to measure this matter rely on synthetic, Visual Question Answer templates or focus on perceptual video quality that is tangential to measuring how well the video abides by physical laws. To address this fragmentation, we introduce PhysicsMind, a unified benchmark with both real and simulation environments that evaluates law-consistent reasoning and generation over three canonical principles: Center of Mass, Lever Equilibrium, and Newton's First Law. PhysicsMind comprises two main tasks: i) VQA tasks, testing whether models can reason and determine physical quantities and values from images or short videos, and ii) Video Generation(VG) tasks, evaluating if predicted motion trajectories obey the same center-of-mass, torque, and inertial constraints as the ground truth. A broad range of recent models and video generation models is evaluated on PhysicsMind and found to rely on appearance heuristics while often violating basic mechanics. These gaps indicate that current scaling and training are still insufficient for robust physical understanding, underscoring PhysicsMind as a focused testbed for physics-aware multimodal models. Our data will be released upon acceptance.

Parkour tasks for quadrupeds have emerged as a promising benchmark for agile locomotion. While human athletes can effectively perceive environmental characteristics to select appropriate footholds for obstacle traversal, endowing legged robots with similar perceptual reasoning remains a significant challenge. Existing methods often rely on hierarchical controllers that follow pre-computed footholds, thereby constraining the robot's real-time adaptability and the exploratory potential of reinforcement learning. To overcome these challenges, we present PUMA, an end-to-end learning framework that integrates visual perception and foothold priors into a single-stage training process. This approach leverages terrain features to estimate egocentric polar foothold priors, composed of relative distance and heading, guiding the robot in active posture adaptation for parkour tasks. Extensive experiments conducted in simulation and real-world environments across various discrete complex terrains, demonstrate PUMA's exceptional agility and robustness in challenging scenarios.

We introduce $k$-lazyGD, an online learning algorithm that bridges the gap between greedy Online Gradient Descent (OGD, for $k=1$) and lazy GD/dual-averaging (for $k=T$), creating a spectrum between reactive and stable updates. We analyze this spectrum in Smoothed Online Convex Optimization (SOCO), where the learner incurs both hitting and movement costs. Our main contribution is establishing that laziness is possible without sacrificing hitting performance: we prove that $k$-lazyGD achieves the optimal dynamic regret $\mathcal{O}(\sqrt{(P_T+1)T})$ for any laziness slack $k$ up to $Θ(\sqrt{T/P_T})$, where $P_T$ is the comparator path length. This result formally connects the allowable laziness to the comparator's shifts, showing that $k$-lazyGD can retain the inherently small movements of lazy methods without compromising tracking ability. We base our analysis on the Follow the Regularized Leader (FTRL) framework, and derive a matching lower bound. Since the slack depends on $P_T$, an ensemble of learners with various slacks is used, yielding a method that is provably stable when it can be, and agile when it must be.

Healthcare visitation patterns are influenced by a complex interplay of hospital attributes, population socioeconomics, and spatial factors. However, existing research often adopts a fragmented approach, examining these determinants in isolation. This study addresses this gap by integrating hospital capacities, occupancy rates, reputation, and popularity with population SES and spatial mobility patterns to predict visitation flows and analyze influencing factors. Utilizing four years of SafeGraph mobility data and user experience data from Google Maps Reviews, five flow prediction models, Naive Regression, Gradient Boosting, Multilayer Perceptrons (MLPs), Deep Gravity, and Heterogeneous Graph Neural Networks (HGNN),were trained and applied to simulate visitation flows in Houston, Texas, U.S. The Shapley additive explanation (SHAP) analysis and the Partial Dependence Plot (PDP) method were employed to examine the combined impacts of different factors on visitation patterns. The findings reveal that Deep Gravity outperformed other models. Hospital capacities, ICU occupancy rates, ratings, and popularity significantly influence visitation patterns, with their effects varying across different travel distances. Short-distance visits are primarily driven by convenience, whereas long-distance visits are influenced by hospital ratings. White-majority areas exhibited lower sensitivity to hospital ratings for short-distance visits, while Asian populations and those with higher education levels prioritized hospital rating in their visitation decisions. SES further influence these patterns, as areas with higher proportions of Hispanic, Black, under-18, and over-65 populations tend to have more frequent hospital visits, potentially reflecting greater healthcare needs or limited access to alternative medical services.

The Decision Transformer (DT) has established a powerful sequence modeling approach to offline reinforcement learning. It conditions its action predictions on Return-to-Go (RTG), using it both to distinguish trajectory quality during training and to guide action generation at inference. In this work, we identify a critical redundancy in this design: feeding the entire sequence of RTGs into the Transformer is theoretically unnecessary, as only the most recent RTG affects action prediction. We show that this redundancy can impair DT's performance through experiments. To resolve this, we propose the Decoupled DT (DDT). DDT simplifies the architecture by processing only observation and action sequences through the Transformer, using the latest RTG to guide the action prediction. This streamlined approach not only improves performance but also reduces computational cost. Our experiments show that DDT significantly outperforms DT and establishes competitive performance against state-of-the-art DT variants across multiple offline RL tasks.

Planetary surfaces are typically analyzed using high-level semantic concepts in natural language, yet vast orbital image archives remain organized at the pixel level. This mismatch limits scalable, open-ended exploration of planetary surfaces. Here we present MarScope, a planetary-scale vision-language framework enabling natural language-driven, label-free mapping of Martian landforms. MarScope aligns planetary images and text in a shared semantic space, trained on over 200,000 curated image-text pairs. This framework transforms global geomorphic mapping on Mars by replacing pre-defined classifications with flexible semantic retrieval, enabling arbitrary user queries across the entire planet in 5 seconds with F1 scores up to 0.978. Applications further show that it extends beyond morphological classification to facilitate process-oriented analysis and similarity-based geomorphological mapping at a planetary scale. MarScope establishes a new paradigm where natural language serves as a direct interface for scientific discovery over massive geospatial datasets.

Text-Based Person Search (TBPS) holds unique value in real-world surveillance bridging visual perception and language understanding, yet current paradigms utilizing pre-training models often fail to transfer effectively to complex open-world scenarios. The reliance on "Passive Observation" leads to multifaceted spurious correlations and spatial semantic misalignment, causing a lack of robustness against distribution shifts. To fundamentally resolve these defects, this paper proposes ICON (Invariant Counterfactual Optimization with Neuro-symbolic priors), a framework integrating causal and topological priors. First, we introduce Rule-Guided Spatial Intervention to strictly penalize sensitivity to bounding box noise, forcibly severing location shortcuts to achieve geometric invariance. Second, Counterfactual Context Disentanglement is implemented via semantic-driven background transplantation, compelling the model to ignore background interference for environmental independence. Then, we employ Saliency-Driven Semantic Regularization with adaptive masking to resolve local saliency bias and guarantee holistic completeness. Finally, Neuro-Symbolic Topological Alignment utilizes neuro-symbolic priors to constrain feature matching, ensuring activated regions are topologically consistent with human structural logic. Experimental results demonstrate that ICON not only maintains leading performance on standard benchmarks but also exhibits exceptional robustness against occlusion, background interference, and localization noise. This approach effectively advances the field by shifting from fitting statistical co-occurrences to learning causal invariance.

Model merging (MM) offers an efficient mechanism for integrating multiple specialized models without access to original training data or costly retraining. While MM has demonstrated success in domains like computer vision, its role in recommender systems (RSs) remains largely unexplored. Recently, Generative Recommendation (GR) has emerged as a new paradigm in RSs, characterized by rapidly growing model scales and substantial computational costs, making MM particularly appealing for cost-sensitive deployment scenarios. In this work, we present the first systematic study of MM in GR through a contextual lens. We focus on a fundamental yet underexplored challenge in real-world: how to merge generative recommenders specialized to different real-world contexts, arising from temporal evolving user behaviors and heterogeneous application domains. To this end, we propose a unified framework MMGRid, a structured contextual grid of GR checkpoints that organizes models trained under diverse contexts induced by temporal evolution and domain diversity. All checkpoints are derived from a shared base LLM but fine-tuned on context-specific data, forming a realistic and controlled model space for systematically analyzing MM across GR paradigms and merging algorithms. Our investigation reveals several key insights. First, training GR models from LLMs can introduce parameter conflicts during merging due to token distribution shifts and objective disparities; such conflicts can be alleviated by disentangling task-aware and context-specific parameter changes via base model replacement. Second, incremental training across contexts induces recency bias, which can be effectively balanced through weighted contextual merging. Notably, we observe that optimal merging weights correlate with context-dependent interaction characteristics, offering practical guidance for weight selection in real-world deployments.

Deep neural network models degrade significantly in the long-tailed data distribution, with the overall training data dominated by a small set of classes in the head, and the tail classes obtaining less training examples. Addressing the imbalance in the classes, attention in the related literature was given mainly to the adjustments carried out in the decision space in terms of either corrections performed at the logit level in order to compensate class-prior bias, with the least attention to the optimization process resulting from the adjustments introduced through the differences in the confidences among the samples. In the current study, we present the design of a class and confidence-aware re-weighting scheme for long-tailed learning. This scheme is purely based upon the loss level and has a complementary nature to the existing methods performing the adjustment of the logits. In the practical implementation stage of the proposed scheme, we use an Ω(p_t, f_c) function. This function enables the modulation of the contribution towards the training task based upon the confidence value of the prediction, as well as the relative frequency of the corresponding class. Our observations in the experiments are corroborated by significant experimental results performed on the CIFAR-100-LT, ImageNet-LT, and iNaturalist2018 datasets under various values of imbalance factors that clearly authenticate the theoretical discussions above.

We propose a novel first-order method for non-convex optimization of the form $\max_{\bm{w}\in\mathbb{R}^d}\mathbb{E}_{\bm{x}\sim\mathcal{D}}[f_{\bm{w}}(\bm{x})]$, termed Progressive Power Homotopy (Prog-PowerHP). The method applies stochastic gradient ascent to a surrogate objective obtained by first performing a power transformation and then Gaussian smoothing, $F_{N,σ}(\bmμ):=\mathbb{E}_{\bm{w}\sim\mathcal{N}(\bmμ,σ^2I_d),\bm{x}\sim\mathcal{D}}[e^{Nf_w(\bm{x})}]$, while progressively increasing the power parameter $N$ and decreasing the smoothing scale $σ$ along the optimization trajectory. We prove that, under mild regularity conditions, Prog-PowerHP converges to a small neighborhood of the global optimum with an iteration complexity scaling nearly as $O(d^2\varepsilon^{-2})$. Empirically, Prog-PowerHP demonstrates clear advantages in phase retrieval when the samples-to-dimension ratio approaches the information-theoretic limit, and in training two-layer neural networks in under-parameterized regimes. These results suggest that Prog-PowerHP is particularly effective for navigating cluttered non-convex landscapes where standard first-order methods struggle.

Robots that follow natural-language instructions often either plan at a high level using hand-designed interfaces or rely on large end-to-end models that are difficult to deploy for real-time control. We propose TeNet (Text-to-Network), a framework for instantiating compact, task-specific robot policies directly from natural language descriptions. TeNet conditions a hypernetwork on text embeddings produced by a pretrained large language model (LLM) to generate a fully executable policy, which then operates solely on low-dimensional state inputs at high control frequencies. By using the language only once at the policy instantiation time, TeNet inherits the general knowledge and paraphrasing robustness of pretrained LLMs while remaining lightweight and efficient at execution time. To improve generalization, we optionally ground language in behavior during training by aligning text embeddings with demonstrated actions, while requiring no demonstrations at inference time. Experiments on MuJoCo and Meta-World benchmarks show that TeNet produces policies that are orders of magnitude smaller than sequence-based baselines, while achieving strong performance in both multi-task and meta-learning settings and supporting high-frequency control. These results show that text-conditioned hypernetworks offer a practical way to build compact, language-driven controllers for ressource-constrained robot control tasks with real-time requirements.

Non-invasive decoding of imagined speech remains challenging due to weak, distributed signals and limited labeled data. Our paper introduces an image-based approach that transforms magnetoencephalography (MEG) signals into time-frequency representations compatible with pretrained vision models. MEG data from 21 participants performing imagined speech tasks were projected into three spatial scalogram mixtures via a learnable sensor-space convolution, producing compact image-like inputs for ImageNet-pretrained vision architectures. These models outperformed classical and non-pretrained models, achieving up to 90.4% balanced accuracy for imagery vs. silence, 81.0% vs. silent reading, and 60.6% for vowel decoding. Cross-subject evaluation confirmed that pretrained models capture shared neural representations, and temporal analyses localized discriminative information to imagery-locked intervals. These findings show that pretrained vision models applied to image-based MEG representations can effectively capture the structure of imagined speech in non-invasive neural signals.

In this paper we propose the Iterative Amortized Hierarchical Variational Autoencoder (IA-HVAE), which expands on amortized inference with a hybrid scheme containing an initial amortized guess and iterative refinement with decoder gradients. We achieve this by creating a linearly separable decoder in a transform domain (e.g. Fourier space), enabling real-time applications with very high model depths. The architectural change leads to a 35x speed-up for iterative inference with respect to the traditional HVAE. We show that our hybrid approach outperforms fully amortized and fully iterative equivalents in accuracy and speed respectively. Moreover, the IAHVAE shows improved reconstruction quality over a vanilla HVAE in inverse problems such as deblurring and denoising.

Diffusion-based language models (DLLMs) offer non-sequential, block-wise generation and richer data reuse compared to autoregressive (AR) models, but existing code DLLMs still lag behind strong AR baselines under comparable budgets. We revisit this setting in a controlled study and introduce Stable-DiffCoder, a block diffusion code model that reuses the Seed-Coder architecture, data, and training pipeline. To enable efficient knowledge learning and stable training, we incorporate a block diffusion continual pretraining (CPT) stage enhanced by a tailored warmup and block-wise clipped noise schedule. Under the same data and architecture, Stable-DiffCoder overall outperforms its AR counterpart on a broad suite of code benchmarks. Moreover, relying only on the CPT and supervised fine-tuning stages, Stable-DiffCoder achieves stronger performance than a wide range of \~8B ARs and DLLMs, demonstrating that diffusion-based training can improve code modeling quality beyond AR training alone. Moreover, diffusion-based any-order modeling improves structured code modeling for editing and reasoning, and through data augmentation, benefits low-resource coding languages.

Foundation models leverage large-scale pretraining to capture extensive knowledge, demonstrating generalization in a wide range of language tasks. By comparison, vision foundation models (VFMs) often exhibit uneven improvements across downstream tasks, despite substantial computational investment. We postulate that this limitation arises from a mismatch between pretraining objectives and the demands of downstream vision-and-imaging tasks. Pretraining strategies like masked image reconstruction or contrastive learning shape representations for tasks such as recovery of generic visual patterns or global semantic structures, which may not align with the task-specific requirements of downstream applications including segmentation, classification, or image synthesis. To investigate this in a concrete real-world clinical area, we assess two VFMs, a reconstruction-focused MAE-based model (ProFound) and a contrastive-learning-based model (ProViCNet), on five prostate multiparametric MR imaging tasks, examining how such task alignment influences transfer performance, i.e., from pretraining to fine-tuning. Our findings indicate that better alignment between pretraining and downstream tasks, measured by simple divergence metrics such as maximum-mean-discrepancy (MMD) between the same features before and after fine-tuning, correlates with greater performance improvements and faster convergence, emphasizing the importance of designing and analyzing pretraining objectives with downstream applicability in mind.

Code completion has become a central task, gaining significant attention with the rise of large language model (LLM)-based tools in software engineering. Although recent advances have greatly improved LLMs' code completion abilities, evaluation methods have not advanced equally. Most current benchmarks focus solely on functional correctness of code completions based on given context, overlooking models' ability to follow user instructions during completion-a common scenario in LLM-assisted programming. To address this limitation, we present the first instruction-guided code completion benchmark, Controllable Code Completion Benchmark (C3-Bench), comprising 2,195 carefully designed completion tasks. Through comprehensive evaluation of over 40 mainstream LLMs across C3-Bench and conventional benchmarks, we reveal substantial gaps in instruction-following capabilities between open-source and advanced proprietary models during code completion tasks. Moreover, we develop a straightforward data synthesis pipeline that leverages Qwen2.5-Coder to generate high-quality instruction-completion pairs for supervised fine-tuning (SFT). The resulting model, Qwen2.5-Coder-C3, achieves state-of-the-art performance on C3-Bench. Our findings provide valuable insights for enhancing LLMs' code completion and instruction-following capabilities, establishing new directions for future research in code LLMs. To facilitate reproducibility and foster further research in code LLMs, we open-source all code, datasets, and models.

The development of native computer-use agents (CUA) represents a significant leap in multimodal AI. However, their potential is currently bottlenecked by the constraints of static data scaling. Existing paradigms relying primarily on passive imitation of static datasets struggle to capture the intricate causal dynamics inherent in long-horizon computer tasks. In this work, we introduce EvoCUA, a native computer use agentic model. Unlike static imitation, EvoCUA integrates data generation and policy optimization into a self-sustaining evolutionary cycle. To mitigate data scarcity, we develop a verifiable synthesis engine that autonomously generates diverse tasks coupled with executable validators. To enable large-scale experience acquisition, we design a scalable infrastructure orchestrating tens of thousands of asynchronous sandbox rollouts. Building on these massive trajectories, we propose an iterative evolving learning strategy to efficiently internalize this experience. This mechanism dynamically regulates policy updates by identifying capability boundaries -- reinforcing successful routines while transforming failure trajectories into rich supervision through error analysis and self-correction. Empirical evaluations on the OSWorld benchmark demonstrate that EvoCUA achieves a success rate of 56.7%, establishing a new open-source state-of-the-art. Notably, EvoCUA significantly outperforms the previous best open-source model, OpenCUA-72B (45.0%), and surpasses leading closed-weights models such as UI-TARS-2 (53.1%). Crucially, our results underscore the generalizability of this approach: the evolving paradigm driven by learning from experience yields consistent performance gains across foundation models of varying scales, establishing a robust and scalable path for advancing native agent capabilities.

Membership Inference Attacks (MIAs) are currently a dominant approach for evaluating privacy in machine learning applications. Despite their significance in identifying records belonging to the training dataset, several concerns remain unexplored, particularly with regard to tabular data. In this paper, first, we provide an extensive review and analysis of MIAs considering two main learning paradigms: centralized and federated learning. We extend and refine the taxonomy for both. Second, we demonstrate the efficacy of MIAs in tabular data using several attack strategies, also including defenses. Furthermore, in a federated learning scenario, we consider the threat posed by an outsider adversary, which is often neglected. Third, we demonstrate the high vulnerability of single-outs (records with a unique signature) to MIAs. Lastly, we explore how MIAs transfer across model architectures. Our results point towards a general poor performance of these attacks in tabular data which contrasts with previous state-of-the-art. Notably, even attacks with limited attack performance can still successfully expose a large portion of single-outs. Moreover, our findings suggest that using different surrogate models makes MIAs more effective.

Dance-to-music generation aims to generate music that is aligned with dance movements. Existing approaches typically rely on body motion features extracted from a single human dancer and limited dance-to-music datasets, which restrict their performance and applicability to real-world scenarios involving multiple dancers and non-human dancers. In this paper, we propose PF-D2M, a universal diffusion-based dance-to-music generation model that incorporates visual features extracted from dance videos. PF-D2M is trained with a progressive training strategy that effectively addresses data scarcity and generalization challenges. Both objective and subjective evaluations show that PF-D2M achieves state-of-the-art performance in dance-music alignment and music quality.

Inference in large-scale AI models is typically performed on dense parameter matrices, leading to inference cost and system complexity that scale unsustainably with model size. This limitation does not arise from insufficient model capacity, but from treating post-training inference systems as monolithic operators while ignoring internal structures formed during learning. We show that gradient update events in large models are highly localized and selective, leaving many parameter dependencies statistically indistinguishable from their initialization distribution after training. As a result, post-training inference systems are structurally non-uniform and inherently decomposable. Based on this observation, we introduce a post-training statistical criterion and a structural annealing procedure that removes unsupported dependencies and reveals stable, independent substructures. This work establishes a post-training, model-agnostic structural view of inference systems and enables structured, parallel inference without modifying model functionality or interfaces.

This study evaluates AV-HuBERT's perceptual bio-fidelity by benchmarking its response to incongruent audiovisual stimuli (McGurk effect) against human observers (N=44). Results reveal a striking quantitative isomorphism: AI and humans exhibited nearly identical auditory dominance rates (32.0% vs. 31.8%), suggesting the model captures biological thresholds for auditory resistance. However, AV-HuBERT showed a deterministic bias toward phonetic fusion (68.0%), significantly exceeding human rates (47.7%). While humans displayed perceptual stochasticity and diverse error profiles, the model remained strictly categorical. Findings suggest that current self-supervised architectures mimic multisensory outcomes but lack the neural variability inherent to human speech perception.

Background: Coronary angiography (CAG) is a cornerstone imaging modality for assessing coronary artery disease and guiding interventional treatment decisions. However, in real-world clinical settings, angiographic images are often characterized by complex lesion morphology, severe class imbalance, label uncertainty, and limited computational resources, posing substantial challenges to conventional deep learning approaches in terms of robustness and generalization.Methods: The proposed framework is built upon a pretrained convolutional neural network to construct a lightweight hybrid neural representation. A selective neural plasticity training strategy is introduced to enable efficient parameter adaptation. Furthermore, a brain-inspired attention-modulated loss function, combining Focal Loss with label smoothing, is employed to enhance sensitivity to hard samples and uncertain annotations. Class-imbalance-aware sampling and cosine annealing with warm restarts are adopted to mimic rhythmic regulation and attention allocation mechanisms observed in biological neural systems.Results: Experimental results demonstrate that the proposed lightweight brain-inspired model achieves strong and stable performance in binary coronary angiography classification, yielding competitive accuracy, recall, F1-score, and AUC metrics while maintaining high computational efficiency.Conclusion: This study validates the effectiveness of brain-inspired learning mechanisms in lightweight medical image analysis and provides a biologically plausible and deployable solution for intelligent clinical decision support under limited computational resources.

X-ray dark-field radiography provides complementary diagnostic information to conventional attenuation imaging by visualizing microstructural tissue changes through small-angle scattering. However, the limited availability of such data poses challenges for developing robust deep learning models. In this work, we present the first framework for generating dark-field images directly from standard attenuation chest X-rays using an Uncertainty-Guided Progressive Generative Adversarial Network. The model incorporates both aleatoric and epistemic uncertainty to improve interpretability and reliability. Experiments demonstrate high structural fidelity of the generated images, with consistent improvement of quantitative metrics across stages. Furthermore, out-of-distribution evaluation confirms that the proposed model generalizes well. Our results indicate that uncertainty-guided generative modeling enables realistic dark-field image synthesis and provides a reliable foundation for future clinical applications.

Introduction: Clinical text classification using natural language processing (NLP) models requires adequate training data to achieve optimal performance. For that, 200-500 documents are typically annotated. The number is constrained by time and costs and lacks justification of the sample size requirements and their relationship to text vocabulary properties. Methods: Using the publicly available MIMIC-III dataset containing hospital discharge notes with ICD-9 diagnoses as labels, we employed pre-trained BERT embeddings followed by Random Forest classifiers to identify 10 randomly selected diagnoses, varying training corpus sizes from 100 to 10,000 documents, and analyzed vocabulary properties by identifying strong and noisy predictive words through Lasso logistic regression on bag-of-words embeddings. Results: Learning curves varied significantly across the 10 classification tasks despite identical preprocessing and algorithms, with 600 documents sufficient to achieve 95% of the performance attainable with 10,000 documents for all tasks. Vocabulary analysis revealed that more strong predictors and fewer noisy predictors were associated with steeper learning curves, where every 100 additional noisy words decreased accuracy by approximately 0.02 while 100 additional strong predictors increased maximum accuracy by approximately 0.04.

This study investigates whether professional translators can reliably identify short stories generated in Italian by artificial intelligence (AI) without prior specialized training. Sixty-nine translators took part in an in-person experiment, where they assessed three anonymized short stories - two written by ChatGPT-4o and one by a human author. For each story, participants rated the likelihood of AI authorship and provided justifications for their choices. While average results were inconclusive, a statistically significant subset (16.2%) successfully distinguished the synthetic texts from the human text, suggesting that their judgements were informed by analytical skill rather than chance. However, a nearly equal number misclassified the texts in the opposite direction, often relying on subjective impressions rather than objective markers, possibly reflecting a reader preference for AI-generated texts. Low burstiness and narrative contradiction emerged as the most reliable indicators of synthetic authorship, with unexpected calques, semantic loans and syntactic transfer from English also reported. In contrast, features such as grammatical accuracy and emotional tone frequently led to misclassification. These findings raise questions about the role and scope of synthetic-text editing in professional contexts.

This paper introduces the Generative Application Firewall (GAF), a new architectural layer for securing LLM applications. Existing defenses -- prompt filters, guardrails, and data-masking -- remain fragmented; GAF unifies them into a single enforcement point, much like a WAF coordinates defenses for web traffic, while also covering autonomous agents and their tool interactions.

The rapid spread of multimodal fake news poses a serious societal threat, as its evolving nature and reliance on timely factual details challenge existing detection methods. Dynamic Retrieval-Augmented Generation provides a promising solution by triggering keyword-based retrieval and incorporating external knowledge, thus enabling both efficient and accurate evidence selection. However, it still faces challenges in addressing issues such as redundant retrieval, coarse similarity, and irrelevant evidence when applied to deceptive content. In this paper, we propose ExDR, an Explanation-driven Dynamic Retrieval-Augmented Generation framework for Multimodal Fake News Detection. Our framework systematically leverages model-generated explanations in both the retrieval triggering and evidence retrieval modules. It assesses triggering confidence from three complementary dimensions, constructs entity-aware indices by fusing deceptive entities, and retrieves contrastive evidence based on deception-specific features to challenge the initial claim and enhance the final prediction. Experiments on two benchmark datasets, AMG and MR2, demonstrate that ExDR consistently outperforms previous methods in retrieval triggering accuracy, retrieval quality, and overall detection performance, highlighting its effectiveness and generalization capability.

Adaptive traffic signal control (TSC) has demonstrated strong effectiveness in managing dynamic traffic flows. However, conventional methods often struggle when unforeseen traffic incidents occur (e.g., accidents and road maintenance), which typically require labor-intensive and inefficient manual interventions by traffic police officers. Large Language Models (LLMs) appear to be a promising solution thanks to their remarkable reasoning and generalization capabilities. Nevertheless, existing works often propose to replace existing TSC systems with LLM-based systems, which can be (i) unreliable due to the inherent hallucinations of LLMs and (ii) costly due to the need for system replacement. To address the issues of existing works, we propose a hierarchical framework that augments existing TSC systems with LLMs, whereby a virtual traffic police agent at the upper level dynamically fine-tunes selected parameters of signal controllers at the lower level in response to real-time traffic incidents. To enhance domain-specific reliability in response to unforeseen traffic incidents, we devise a self-refined traffic language retrieval system (TLRS), whereby retrieval-augmented generation is employed to draw knowledge from a tailored traffic language database that encompasses traffic conditions and controller operation principles. Moreover, we devise an LLM-based verifier to update the TLRS continuously over the reasoning process. Our results show that LLMs can serve as trustworthy virtual traffic police officers that can adapt conventional TSC methods to unforeseen traffic incidents with significantly improved operational efficiency and reliability.

We introduce a lightweight experimentation pipeline designed to lower the barrier for applying machine learning (ML) methods for classifying images in ecological research. We enable ecologists to experiment with ML models independently, thus they can move beyond off-the-shelf models and generate insights tailored to local datasets and specific classification tasks and target variables. Our tool combines a simple command-line interface for preprocessing, training, and evaluation with a graphical interface for annotation, error analysis, and model comparison. This design enables ecologists to build and iterate on compact, task-specific classifiers without requiring advanced ML expertise. As a proof of concept, we apply the pipeline to classify red deer (Cervus elaphus) by age and sex from 3392 camera trap images collected in the Veldenstein Forest, Germany. Using 4352 cropped images containing individual deer labeled by experts, we trained and evaluated multiple backbone architectures with a wide variety of parameters and data augmentation strategies. Our best-performing models achieved 90.77% accuracy for age classification and 96.15% for sex classification. These results demonstrate that reliable demographic classification is feasible even with limited data to answer narrow, well-defined ecological problems. More broadly, the framework provides ecologists with an accessible tool for developing ML models tailored to specific research questions, paving the way for broader adoption of ML in wildlife monitoring and demographic analysis.

Large Language Models (LLM) benchmarks tell us when models fail, but not why they fail. A wrong answer on a reasoning dataset may stem from formatting issues, calculation errors, or dataset noise rather than weak reasoning. Without disentangling such causes, benchmarks remain incomplete and cannot reliably guide model improvement. We introduce ErrorMap, the first method to chart the sources of LLM failure. It extracts a model's unique "failure signature", clarifies what benchmarks measure, and broadens error identification to reduce blind spots. This helps developers debug models, aligns benchmark goals with outcomes, and supports informed model selection. ErrorMap works on any model or dataset with the same logic. Applying our method to 35 datasets and 83 models we generate ErrorAtlas, a taxonomy of model errors, revealing recurring failure patterns. ErrorAtlas highlights error types that are currently underexplored in LLM research, such as omissions of required details in the output and question misinterpretation. By shifting focus from where models succeed to why they fail, ErrorMap and ErrorAtlas enable advanced evaluation - one that exposes hidden weaknesses and directs progress. Unlike success, typically measured by task-level metrics, our approach introduces a deeper evaluation layer that can be applied globally across models and tasks, offering richer insights into model behavior and limitations. We make the taxonomy and code publicly available with plans to periodically update ErrorAtlas as new benchmarks and models emerge.

A convolutional neural network (CNN) is a deep learning algorithm that has been specifically designed for computer vision applications. The CNNs proved successful in handling the increasing amount of data in many computer vision problems, where classical machine learning algorithms were insufficient. Flowers have many uses in our daily lives, from decorating to making medicines to detoxifying the environment. Identifying flower types requires expert knowledge. However, accessing experts at any time and in any location may not always be feasible. In this study a mobile application based on CNNs was developed to recognize different types of flowers to provide non-specialists with quick and easy access to information about flower types. The study employed three distinct CNN models, namely MobileNet, DenseNet121, and Xception, to determine the most suitable model for the mobile application. The classification performances of the models were evaluated by training them with seven different optimization algorithms. The DenseNet-121 architecture, which uses the stochastic gradient descent (SGD) optimization algorithm, was the most successful, achieving 95.84 % accuracy, 96.00% precision, recall, and F1-score. This result shows that CNNs can be used for flower classification in mobile applications.

An increasing body of work has leveraged multilingual language models for Natural Language Generation tasks such as summarization. A major empirical bottleneck in this area is the shortage of accurate and robust evaluation metrics for many languages, which hinders progress. Recent studies suggest that multilingual language models often use English as an internal pivot language, and that misalignment with this pivot can lead to degraded downstream performance. Motivated by the hypothesis that this mismatch could also apply to multilingual neural metrics, we ask whether steering their activations toward an English pivot can improve correlation with human judgments. We experiment with encoder- and decoder-based metrics and find that test-time intervention methods are effective across the board, increasing metric effectiveness for diverse languages.

Recent advances in Deep Research Agents (DRAs) are transforming automated knowledge discovery and problem-solving. While the majority of existing efforts focus on enhancing policy capabilities via post-training, we propose an alternative paradigm: self-evolving the agent's ability by iteratively verifying the policy model's outputs, guided by meticulously crafted rubrics. This approach gives rise to the inference-time scaling of verification, wherein an agent self-improves by evaluating its generated answers to produce iterative feedback and refinements. We derive the rubrics based on an automatically constructed DRA Failure Taxonomy, which systematically classifies agent failures into five major categories and thirteen sub-categories. We present DeepVerifier, a rubrics-based outcome reward verifier that leverages the asymmetry of verification and outperforms vanilla agent-as-judge and LLM judge baselines by 12%-48% in meta-evaluation F1 score. To enable practical self-evolution, DeepVerifier integrates as a plug-and-play module during test-time inference. The verifier produces detailed rubric-based feedback, which is fed back to the agent for iterative bootstrapping, refining responses without additional training. This test-time scaling delivers 8%-11% accuracy gains on challenging subsets of GAIA and XBench-DeepResearch when powered by capable closed-source LLMs. Finally, to support open-source advancement, we release DeepVerifier-4K, a curated supervised fine-tuning dataset of 4,646 high-quality agent steps focused on DRA verification. These examples emphasize reflection and self-critique, enabling open models to develop robust verification capabilities.

We introduce the AlgebraicStatistics section of the OSCAR computer algebra system. We give an overview of its extensible design and highlight its features including serialization of data types for sharing results and creating databases, and state-of-the-art implicitization algorithms.

Autonomous Unmanned Underwater Vehicles (UUVs) enable military and civilian covert operations in coastal areas without relying on support vessels or Global Navigation Satellite Systems (GNSS). Such operations are critical when surface access is not possible and stealthy navigation is required in restricted environments such as protected zones or dangerous areas under access ban. GNSS denied navigation is then essential to maintaining concealment as surfacing could expose UUVs to detection. To ensure a precise fleet positioning a constellation of beacons deployed by aerial or surface drones establish a synthetic landmark network that will guide the fleet of UUVs along an optimized path from the continental shelf to the goal on the shore. These beacons either submerged or floating emit acoustic signals for UUV localisation and navigation. A hierarchical planner generates an adaptive route for the drones executing primitive actions while continuously monitoring and replanning as needed to maintain trajectory accuracy.

While Large Language Models (LLMs) are aligned to mitigate risks, their safety guardrails remain fragile against jailbreak attacks. This reveals limited understanding of components governing safety. Existing methods rely on local, greedy attribution that assumes independent component contributions. However, they overlook the cooperative interactions between different components in LLMs, such as attention heads, which jointly contribute to safety mechanisms. We propose \textbf{G}lobal \textbf{O}ptimization for \textbf{S}afety \textbf{V}ector Extraction (GOSV), a framework that identifies safety-critical attention heads through global optimization over all heads simultaneously. We employ two complementary activation repatching strategies: Harmful Patching and Zero Ablation. These strategies identify two spatially distinct sets of safety vectors with consistently low overlap, termed Malicious Injection Vectors and Safety Suppression Vectors, demonstrating that aligned LLMs maintain separate functional pathways for safety purposes. Through systematic analyses, we find that complete safety breakdown occurs when approximately 30\% of total heads are repatched across all models. Building on these insights, we develop a novel inference-time white-box jailbreak method that exploits the identified safety vectors through activation repatching. Our attack substantially outperforms existing white-box attacks across all test models, providing strong evidence for the effectiveness of the proposed GOSV framework on LLM safety interpretability.

Chronic diseases have become the leading cause of death worldwide, a challenge intensified by strained medical resources and an aging population. Individually, patients often struggle to interpret early signs of deterioration or maintain adherence to care plans. In this paper, we introduce VitalDiagnosis, an LLM-driven ecosystem designed to shift chronic disease management from passive monitoring to proactive, interactive engagement. By integrating continuous data from wearable devices with the reasoning capabilities of LLMs, the system addresses both acute health anomalies and routine adherence. It analyzes triggers through context-aware inquiries, produces provisional insights within a collaborative patient-clinician workflow, and offers personalized guidance. This approach aims to promote a more proactive and cooperative care paradigm, with the potential to enhance patient self-management and reduce avoidable clinical workload.

Many theorists of creativity maintain that intentional agency is a necessary condition of creativity. We argue that this requirement, which we call the Intentional Agency Condition (IAC), should be rejected as a general condition of creativity, while retaining its relevance in specific contexts. We show that recent advances in generative AI have rendered the IAC increasingly problematic, both descriptively and functionally. We offer two reasons for abandoning it at the general level. First, we present corpus evidence indicating that authors and journalists are increasingly comfortable ascribing creativity to generative AI, despite its lack of intentional agency. This development places pressure on the linguistic intuitions that have traditionally been taken to support the IAC. Second, drawing on the method of conceptual engineering, we argue that the IAC no longer fulfils its core social function. Rather than facilitating the identification and encouragement of reliable sources of novel and valuable products, it now feeds into biases that distort our assessments of AI-generated outputs. We therefore propose replacing the IAC with a consistency requirement, according to which creativity tracks the reliable generation of novel and valuable products. Nonetheless, we explain why the IAC should be retained in specific local domains.

Motivated by the remarkable progress of large language models (LLMs) in objective tasks like mathematics and coding, there is growing interest in their potential to simulate human behavior--a capability with profound implications for transforming social science research and customer-centric business insights. However, LLMs often lack a nuanced understanding of human cognition and behavior, limiting their effectiveness in social simulation and personalized applications. We posit that this limitation stems from a fundamental misalignment: standard LLM pretraining on vast, uncontextualized web data does not capture the continuous, situated context of an individual's decisions, thoughts, and behaviors over time. To bridge this gap, we introduce HumanLLM, a foundation model designed for personalized understanding and simulation of individuals. We first construct the Cognitive Genome Dataset, a large-scale corpus curated from real-world user data on platforms like Reddit, Twitter, Blogger, and Amazon. Through a rigorous, multi-stage pipeline involving data filtering, synthesis, and quality control, we automatically extract over 5.5 million user logs to distill rich profiles, behaviors, and thinking patterns. We then formulate diverse learning tasks and perform supervised fine-tuning to empower the model to predict a wide range of individualized human behaviors, thoughts, and experiences. Comprehensive evaluations demonstrate that HumanLLM achieves superior performance in predicting user actions and inner thoughts, more accurately mimics user writing styles and preferences, and generates more authentic user profiles compared to base models. Furthermore, HumanLLM shows significant gains on out-of-domain social intelligence benchmarks, indicating enhanced generalization.

AI agents are rapidly advancing from passive language models to autonomous systems executing complex, multi-step tasks. Yet their overconfidence in failure remains a fundamental barrier to deployment in high-stakes settings. Existing calibration methods, built for static single-turn outputs, cannot address the unique challenges of agentic systems, such as compounding errors along trajectories, uncertainty from external tools, and opaque failure modes. To address these challenges, we introduce, for the first time, the problem of Agentic Confidence Calibration and propose Holistic Trajectory Calibration (HTC), a novel diagnostic framework that extracts rich process-level features ranging from macro dynamics to micro stability across an agent's entire trajectory. Powered by a simple, interpretable model, HTC consistently surpasses strong baselines in both calibration and discrimination, across eight benchmarks, multiple LLMs, and diverse agent frameworks. Beyond performance, HTC delivers three essential advances: it provides interpretability by revealing the signals behind failure, enables transferability by applying across domains without retraining, and achieves generalization through a General Agent Calibrator (GAC) that achieves the best calibration (lowest ECE) on the out-of-domain GAIA benchmark. Together, these contributions establish a new process-centric paradigm for confidence calibration, providing a framework for diagnosing and enhancing the reliability of AI agents.

Monolithic Firmware is widespread. Unsurprisingly, fuzz testing firmware is an active research field with new advances addressing the unique challenges in the domain. However, understanding and evaluating improvements by deriving metrics such as code coverage and unique crashes are problematic, leading to a desire for a reliable bug-based benchmark. To address the need, we design and build FirmReBugger, a holistic framework for fairly assessing monolithic firmware fuzzers with a realistic, diverse, bug-based benchmark. FirmReBugger proposes using bug oracles--C syntax expressions of bug descriptors--with an interpreter to automate analysis and accurately report on bugs discovered, discriminating between states of detected, triggered, reached and not reached. Importantly, our idea of benchmarking does not modify the target binary and simply replays fuzzing seeds to isolate the benchmark implementation from the fuzzer while providing a simple means to extend with new bug oracles. Further, analyzing fuzzing roadblocks, we created FirmBench, a set of diverse, real-world binary targets with 313 software bug oracles. Incorporating our analysis of roadblocks challenging monolithic firmware fuzzing, the bench provides for rapid evaluation of future advances. We implement FirmReBugger in a FuzzBench-for-Firmware type service and use FirmBench to evaluate 9 state-of-the art monolithic firmware fuzzers in the style of a reproducibility study, using a 10 CPU-year effort, to report our findings.

With the rapid advancement and strong generalization capabilities of large language models (LLMs), they have been increasingly incorporated into the active learning pipelines as annotators to reduce annotation costs. However, considering the annotation quality, labels generated by LLMs often fall short of real-world applicability. To address this, we propose a novel active learning framework, Mixture of LLMs in the Loop Active Learning, replacing human annotators with labels generated through a Mixture-of-LLMs-based annotation model, aimed at enhancing LLM-based annotation robustness by aggregating the strengths of multiple LLMs. To further mitigate the impact of the noisy labels, we introduce annotation discrepancy and negative learning to identify the unreliable annotations and enhance learning effectiveness. Extensive experiments demonstrate that our framework achieves performance comparable to human annotation and consistently outperforms single-LLM baselines and other LLM-ensemble-based approaches. Moreover, our framework is built on lightweight LLMs, enabling it to operate fully on local machines in real-world applications.

Drug-drug interaction (DDI) prediction is central to drug discovery and clinical development, particularly in the context of increasingly prevalent polypharmacy. Although existing computational methods achieve strong performance on standard benchmarks, they often fail to generalize to realistic deployment scenarios, where most candidate drug pairs involve previously unseen drugs and validated interactions are scarce. We demonstrate that proximity in the embedding spaces of prevailing molecule-centric DDI models does not reliably correspond to interaction labels, and that simply scaling up model capacity therefore fails to improve generalization. To address these limitations, we propose GenRel-DDI, a generalizable relation learning framework that reformulates DDI prediction as a relation-centric learning problem, in which interaction representations are learned independently of drug identities. This relation-level abstraction enables the capture of transferable interaction patterns that generalize to unseen drugs and novel drug pairs. Extensive experiments across multiple benchmark demonstrate that GenRel-DDI consistently and significantly outperforms state-of-the-art methods, with particularly large gains on strict entity-disjoint evaluations, highlighting the effectiveness and practical utility of relation learning for robust DDI prediction. The code is available at https://github.com/SZU-ADDG/GenRel-DDI.

Deploying reinforcement learning in the real world remains challenging due to sample inefficiency, sparse rewards, and noisy visual observations. Prior work leverages demonstrations and human feedback to improve learning efficiency and robustness. However, offline-to-online methods need large datasets and can be unstable, while VLA-assisted RL relies on large-scale pretraining and fine-tuning. As a result, a low-cost real-world RL method with minimal data requirements has yet to emerge. We introduce \textbf{SigEnt-SAC}, an off-policy actor-critic method that learns from scratch using a single expert trajectory. Our key design is a sigmoid-bounded entropy term that prevents negative-entropy-driven optimization toward out-of-distribution actions and reduces Q-function oscillations. We benchmark SigEnt-SAC on D4RL tasks against representative baselines. Experiments show that SigEnt-SAC substantially alleviates Q-function oscillations and reaches a 100\% success rate faster than prior methods. Finally, we validate SigEnt-SAC on four real-world robotic tasks across multiple embodiments, where agents learn from raw images and sparse rewards; results demonstrate that SigEnt-SAC can learn successful policies with only a small number of real-world interactions, suggesting a low-cost and practical pathway for real-world RL deployment.

This paper presents FeTal-SAM, a novel adaptation of the Segment Anything Model (SAM) tailored for fetal brain MRI segmentation. Traditional deep learning methods often require large annotated datasets for a fixed set of labels, making them inflexible when clinical or research needs change. By integrating atlas-based prompts and foundation-model principles, FeTal-SAM addresses two key limitations in fetal brain MRI segmentation: (1) the need to retrain models for varying label definitions, and (2) the lack of insight into whether segmentations are driven by genuine image contrast or by learned spatial priors. We leverage multi-atlas registration to generate spatially aligned label templates that serve as dense prompts, alongside a bounding-box prompt, for SAM's segmentation decoder. This strategy enables binary segmentation on a per-structure basis, which is subsequently fused to reconstruct the full 3D segmentation volumes. Evaluations on two datasets, the dHCP dataset and an in-house dataset demonstrate FeTal-SAM's robust performance across gestational ages. Notably, it achieves Dice scores comparable to state-of-the-art baselines which were trained for each dataset and label definition for well-contrasted structures like cortical plate and cerebellum, while maintaining the flexibility to segment any user-specified anatomy. Although slightly lower accuracy is observed for subtle, low-contrast structures (e.g., hippocampus, amygdala), our results highlight FeTal-SAM's potential to serve as a general-purpose segmentation model without exhaustive retraining. This method thus constitutes a promising step toward clinically adaptable fetal brain MRI analysis tools.

Large language models are increasingly used to represent human opinions, values, or beliefs, and their steerability towards these ideals is an active area of research. Existing work focuses predominantly on aligning marginal response distributions, treating each survey item independently. While essential, this may overlook deeper latent structures that characterise real populations and underpin cultural values theories. We propose a framework for evaluating the representativeness of aligned models through multivariate correlation patterns in addition to marginal distributions. We show the value of our evaluation scheme by comparing two model steering techniques (persona prompting and demographic fine-tuning) and evaluating them against human responses from the World Values Survey. While the demographically fine-tuned model better approximates marginal response distributions than persona prompting, both techniques fail to fully capture the gold standard correlation patterns. We conclude that representativeness is a distinct aspect of value alignment and an evaluation focused on marginals can mask structural failures, leading to overly optimistic conclusions about model capabilities.

High-dimensional malware datasets often exhibit feature redundancy, instability, and scalability limitations, which hinder the effectiveness and interpretability of machine learning-based malware detection systems. Although feature selection is commonly employed to mitigate these issues, many existing approaches lack robustness when applied to large-scale and heterogeneous malware data. To address this gap, this paper proposes CAFE-GB (Chunk-wise Aggregated Feature Estimation using Gradient Boosting), a scalable feature selection framework designed to produce stable and globally consistent feature rankings for high-dimensional malware detection. CAFE-GB partitions training data into overlapping chunks, estimates local feature importance using gradient boosting models, and aggregates these estimates to derive a robust global ranking. Feature budget selection is performed separately through a systematic k-selection and stability analysis to balance detection performance and robustness. The proposed framework is evaluated on two large-scale malware datasets: BODMAS and CIC-AndMal2020, representing large and diverse malware feature spaces. Experimental results show that classifiers trained on CAFE-GB -selected features achieve performance parity with full-feature baselines across multiple metrics, including Accuracy, F1-score, MCC, ROC-AUC, and PR-AUC, while reducing feature dimensionality by more than 95\%. Paired Wilcoxon signed-rank tests confirm that this reduction does not introduce statistically significant performance degradation. Additional analyses demonstrate low inter-feature redundancy and improved interpretability through SHAP-based explanations. Runtime and memory profiling further indicate reduced downstream classification overhead. Overall, CAFE-GB provides a stable, interpretable, and scalable feature selection strategy for large-scale malware detection.

Tabular data is a fundamental form of data structure. The evolution of table analysis tools reflects humanity's continuous progress in data acquisition, management, and processing. The dynamic changes in table columns arise from technological advancements, changing needs, data integration, etc. However, the standard process of training AI models on tables with fixed columns and then performing inference is not suitable for handling dynamically changed tables. Therefore, new methods are needed for efficiently handling such tables in an unsupervised manner. In this paper, we introduce a new task, Tabular Incremental Inference (TabII), which aims to enable trained models to incorporate new columns during the inference stage, enhancing the practicality of AI models in scenarios where tables are dynamically changed. Furthermore, we demonstrate that this new task can be framed as an optimization problem based on the information bottleneck theory, which emphasizes that the key to an ideal tabular incremental inference approach lies in minimizing mutual information between tabular data and representation while maximizing between representation and task labels. Under this guidance, we design a TabII method with Large Language Model placeholders and Pretrained TabAdapter to provide external knowledge and Incremental Sample Condensation blocks to condense the task-relevant information given by incremental column attributes. Experimental results across eight public datasets show that TabII effectively utilizes incremental attributes, achieving state-of-the-art performance.

In the realm of medical report generation (MRG), the integration of natural language processing has emerged as a vital tool to alleviate the workload of radiologists. Despite the impressive capabilities demonstrated by large vision language models (LVLMs) in understanding natural language, their susceptibility to generating plausible yet inaccurate claims, known as ``hallucinations'', raises concerns-especially in the nuanced and critical field of medical. In this work, we introduce a framework, \textbf{K}nowledge-\textbf{E}nhanced with Fine-Grained \textbf{R}einforced Rewards \textbf{M}edical Report Generation (KERM), to tackle the issue. Our approach refines the input to the LVLM by first utilizing MedCLIP for knowledge retrieval, incorporating relevant lesion fact sentences from a curated knowledge corpus. We then introduce a novel purification module to ensure the retrieved knowledge is contextually relevant to the patient's clinical context. Subsequently, we employ fine-grained rewards to guide these models in generating highly supportive and clinically relevant descriptions, ensuring the alignment of model's outputs with desired behaviors. Experimental results on IU-Xray and MIMIC-CXR datasets validate the effectiveness of our approach in mitigating hallucinations and enhancing report quality.

Dynamic multi-product delivery environments demand rapid coordination of part completion and product-level kitting within hybrid processing and assembly systems to satisfy strict hierarchical supply constraints. The flexible assembly flow shop scheduling problem formally defines dependencies for multi-stage kitting, yet dynamic variants make designing integrated scheduling rules under multi-level time coupling highly challenging. Existing automated heuristic design methods, particularly genetic programming constrained to fixed terminal symbol sets, struggle to capture and leverage dynamic uncertainties and hierarchical dependency information under transient decision states. This study develops an LLM-assisted Dynamic Rule Design framework (LLM4DRD) that automatically evolves integrated online scheduling rules adapted to scheduling features. Firstly, multi-stage processing and assembly supply decisions are transformed into feasible directed edge orderings based on heterogeneous graph. Then, an elite knowledge guided initialization embeds advanced design expertise into initial rules to enhance initial quality. Additionally, a dual-expert mechanism is introduced in which LLM-A evolutionary code to generate candidate rules and LLM-S conducts scheduling evaluation, while dynamic feature-fitting rule evolution combined with hybrid evaluation enables continuous improvement and extracts adaptive rules with strong generalization capability. A series of experiments are conducted to validate the effectiveness of the method. The average tardiness of LLM4DRD is 3.17-12.39% higher than state-of-the-art methods in 20 practical instances used for training and testing, respectively. In 24 scenarios with different resource configurations, order loads, and disturbance levels totaling 480 instances, it achieves 11.10% higher performance than the second best competitor, exhibiting excellent robustness.

The combination of verifiable languages and LLMs has significantly influenced both the mathematical and computer science communities because it provides a rigorous foundation for theorem proving. Recent advancements in the field provide foundation models and sophisticated agentic systems pushing the boundaries of formal mathematical reasoning to approach the natural language capability of LLMs. However, little attention has been given to the formal physics reasoning, which also heavily relies on similar problem-solving and theorem-proving frameworks. To solve this problem, this paper presents, to the best of our knowledge, the first approach to enhance formal theorem proving in the physics domain. We compose a dedicated dataset PhysLeanData for the task. It is composed of theorems sampled from PhysLean and data generated by a conjecture-based formal data generation pipeline. In the training pipeline, we leverage DeepSeek-Prover-V2-7B, a strong open-source mathematical theorem prover, and apply Reinforcement Learning with Verifiable Rewards (RLVR) to train our model PhysProver. Comprehensive experiments demonstrate that, using only $\sim$5K training samples, PhysProver achieves an overall 2.4\% improvement in multiple sub-domains. Furthermore, after formal physics training, we observe 1.3\% gains on the MiniF2F-Test benchmark, which indicates non-trivial generalization beyond physics domains and enhancement for formal math capability as well. The results highlight the effectiveness and efficiency of our approach, which provides a paradigm for extending formal provers outside mathematical domains. To foster further research, we will release both our dataset and model to the community.

An essential technique for diagnosing brain disorders is electrophysiological source imaging (ESI). While model-based optimization and deep learning methods have achieved promising results in this field, the accurate selection and refinement of features remains a central challenge for precise ESI. This paper proposes FAIR-ESI, a novel framework that adaptively refines feature importance across different views, including FFT-based spectral feature refinement, weighted temporal feature refinement, and self-attention-based patch-wise feature refinement. Extensive experiments on two simulation datasets with diverse configurations and two real-world clinical datasets validate our framework's efficacy, highlighting its potential to advance brain disorder diagnosis and offer new insights into brain function.

Diffusion models have emerged as a powerful approach for multimodal motion planning in autonomous driving. However, their practical deployment is typically hindered by the inherent difficulty in enforcing vehicle dynamics and a critical reliance on accurate predictions of other agents, making them prone to safety issues under uncertain interactions. To address these limitations, we introduce DualShield, a planning and control framework that leverages Hamilton-Jacobi (HJ) reachability value functions in a dual capacity. First, the value functions act as proactive guidance, steering the diffusion denoising process towards safe and dynamically feasible regions. Second, they form a reactive safety shield using control barrier-value functions (CBVFs) to modify the executed actions and ensure safety. This dual mechanism preserves the rich exploration capabilities of diffusion models while providing principled safety assurance under uncertain and even adversarial interactions. Simulations in challenging unprotected U-turn scenarios demonstrate that DualShield significantly improves both safety and task efficiency compared to leading methods from different planning paradigms under uncertainty.

While Text-to-SQL remains the dominant approach for database interaction, real-world analytics increasingly require the flexibility of general-purpose programming languages such as Python or Pandas to manage file-based data and complex analytical workflows. Despite this growing need, the reliability of Text-to-Python in core data retrieval remains underexplored relative to the mature SQL ecosystem. To address this gap, we introduce BIRD-Python, a benchmark designed for cross-paradigm evaluation. We systematically refined the original dataset to reduce annotation noise and align execution semantics, thereby establishing a consistent and standardized baseline for comparison. Our analysis reveals a fundamental paradigmatic divergence: whereas SQL leverages implicit DBMS behaviors through its declarative structure, Python requires explicit procedural logic, making it highly sensitive to underspecified user intent. To mitigate this challenge, we propose the Logic Completion Framework (LCF), which resolves ambiguity by incorporating latent domain knowledge into the generation process. Experimental results show that (1) performance differences primarily stem from missing domain context rather than inherent limitations in code generation, and (2) when these gaps are addressed, Text-to-Python achieves performance parity with Text-to-SQL. These findings establish Python as a viable foundation for analytical agents-provided that systems effectively ground ambiguous natural language inputs in executable logical specifications. Resources are available at https://anonymous.4open.science/r/Bird-Python-43B7/.

The performance of modern AI systems is fundamentally constrained by the quality of their underlying kernels, which translate high-level algorithmic semantics into low-level hardware operations. Achieving near-optimal kernels requires expert-level understanding of hardware architectures and programming models, making kernel engineering a critical but notoriously time-consuming and non-scalable process. Recent advances in large language models (LLMs) and LLM-based agents have opened new possibilities for automating kernel generation and optimization. LLMs are well-suited to compress expert-level kernel knowledge that is difficult to formalize, while agentic systems further enable scalable optimization by casting kernel development as an iterative, feedback-driven loop. Rapid progress has been made in this area. However, the field remains fragmented, lacking a systematic perspective for LLM-driven kernel generation. This survey addresses this gap by providing a structured overview of existing approaches, spanning LLM-based approaches and agentic optimization workflows, and systematically compiling the datasets and benchmarks that underpin learning and evaluation in this domain. Moreover, key open challenges and future research directions are further outlined, aiming to establish a comprehensive reference for the next generation of automated kernel optimization. To keep track of this field, we maintain an open-source GitHub repository at https://github.com/flagos-ai/awesome-LLM-driven-kernel-generation.

Long-form video understanding remains a fundamental challenge for current Video Large Language Models. Most existing models rely on static reasoning over uniformly sampled frames, which weakens temporal localization and leads to substantial information loss in long videos. Agentic tools such as temporal retrieval, spatial zoom, and temporal zoom offer a natural way to overcome these limitations by enabling adaptive exploration of key moments. However, constructing agentic video understanding data requires models that already possess strong long-form video comprehension, creating a circular dependency. We address this challenge with VideoThinker, an agentic Video Large Language Model trained entirely on synthetic tool interaction trajectories. Our key idea is to convert videos into rich captions and employ a powerful agentic language model to generate multi-step tool use sequences in caption space. These trajectories are subsequently grounded back to video by replacing captions with the corresponding frames, yielding a large-scale interleaved video and tool reasoning dataset without requiring any long-form understanding from the underlying model. Training on this synthetic agentic dataset equips VideoThinker with dynamic reasoning capabilities, adaptive temporal exploration, and multi-step tool use. Remarkably, VideoThinker significantly outperforms both caption-only language model agents and strong video model baselines across long-video benchmarks, demonstrating the effectiveness of tool augmented synthetic data and adaptive retrieval and zoom reasoning for long-form video understanding.

Graph Neural Networks (GNNs) unlock new ways of learning from graph-structured data, proving highly effective in capturing complex relationships and patterns. Federated GNNs (FGNNs) have emerged as a prominent distributed learning paradigm for training GNNs over decentralized data. However, FGNNs face two significant challenges: high communication overhead from multiple rounds of parameter exchanges and non-IID data characteristics across clients. To address these issues, we introduce CeFGC, a novel FGNN paradigm that facilitates efficient GNN training over non-IID data by limiting communication between the server and clients to three rounds only. The core idea of CeFGC is to leverage generative diffusion models to minimize direct client-server communication. Each client trains a generative diffusion model that captures its local graph distribution and shares this model with the server, which then redistributes it back to all clients. Using these generative models, clients generate synthetic graphs combined with their local graphs to train local GNN models. Finally, clients upload their model weights to the server for aggregation into a global GNN model. We theoretically analyze the I/O complexity of communication volume to show that CeFGC reduces to a constant of three communication rounds only. Extensive experiments on several real graph datasets demonstrate the effectiveness and efficiency of CeFGC against state-of-the-art competitors, reflecting our superior performance on non-IID graphs by aligning local and global model objectives and enriching the training set with diverse graphs.

Understanding what users like is relatively straightforward; understanding what users dislike, however, remains a challenging and underexplored problem. Research into users' negative preferences has gained increasing importance in modern recommendation systems. Numerous platforms have introduced explicit negative feedback mechanisms and leverage such signals to refine their recommendation models. Beyond traditional business metrics, user experience-driven metrics, such as negative feedback rates, have become critical indicators for evaluating system performance. However, most existing approaches primarily use negative feedback as an auxiliary signal to enhance positive recommendations, paying little attention to directly modeling negative interests, which can be highly valuable in offline applications. Moreover, due to the inherent sparsity of negative feedback data, models often suffer from context understanding biases induced by positive feedback dominance. To address these challenges, we propose the first large language model framework for negative feedback modeling with special designed context-discerning modules. We use semantic ID Representation to replace text-based item descriptions and introduce an item-level alignment task that enhances the LLM's understanding of the semantic context behind negative feedback. Furthermore, we design a Progressive GRPO training paradigm that enables the model to dynamically balance the positive and negative behavioral context utilization. Besides, our investigation further reveals a fundamental misalignment between the conventional next-negative-item prediction objective and users' true negative preferences, which is heavily influenced by the system's recommendation order. To mitigate this, we propose a novel reward function and evaluation metric grounded in multi-day future negative feedback and their collaborative signals.

Dynamic Flexible Job Shop Scheduling (DFJSS) is a complex combinatorial optimisation problem that requires simultaneous machine assignment and operation sequencing decisions in dynamic production environments. Genetic Programming (GP) has been widely applied to automatically evolve scheduling rules for DFJSS. However, existing studies typically train and test GP-evolved rules on DFJSS instances of the same type, which differ only by random seeds rather than by structural characteristics, leaving their cross-type generalisation ability largely unexplored. To address this gap, this paper systematically investigates the generalisation ability of GP-evolved scheduling rules under diverse DFJSS conditions. A series of experiments are conducted across multiple dimensions, including problem scale (i.e., the number of machines and jobs), key job shop parameters (e.g., utilisation level), and data distributions, to analyse how these factors influence GP performance on unseen instance types. The results show that good generalisation occurs when the training instances contain more jobs than the test instances while keeping the number of machines fixed, and when both training and test instances have similar scales or job shop parameters. Further analysis reveals that the number and distribution of decision points in DFJSS instances play a crucial role in explaining these performance differences. Similar decision point distributions lead to better generalisation, whereas significant discrepancies result in a marked degradation of performance. Overall, this study provides new insights into the generalisation ability of GP in DFJSS and highlights the necessity of evolving more generalisable GP rules capable of handling heterogeneous DFJSS instances effectively.

Although artificial intelligence (AI) has become deeply integrated into various stages of the research workflow and achieved remarkable advancements, academic rebuttal remains a significant and underexplored challenge. This is because rebuttal is a complex process of strategic communication under severe information asymmetry rather than a simple technical debate. Consequently, current approaches struggle as they largely imitate surface-level linguistics, missing the essential element of perspective-taking required for effective persuasion. In this paper, we introduce RebuttalAgent, the first framework to ground academic rebuttal in Theory of Mind (ToM), operationalized through a ToM-Strategy-Response (TSR) pipeline that models reviewer mental state, formulates persuasion strategy, and generates strategy-grounded response. To train our agent, we construct RebuttalBench, a large-scale dataset synthesized via a novel critique-and-refine approach. Our training process consists of two stages, beginning with a supervised fine-tuning phase to equip the agent with ToM-based analysis and strategic planning capabilities, followed by a reinforcement learning phase leveraging the self-reward mechanism for scalable self-improvement. For reliable and efficient automated evaluation, we further develop Rebuttal-RM, a specialized evaluator trained on over 100K samples of multi-source rebuttal data, which achieves scoring consistency with human preferences surpassing powerful judge GPT-4.1. Extensive experiments show RebuttalAgent significantly outperforms the base model by an average of 18.3% on automated metrics, while also outperforming advanced proprietary models across both automated and human evaluations. Disclaimer: the generated rebuttal content is for reference only to inspire authors and assist in drafting. It is not intended to replace the author's own critical analysis and response.

We propose Zero-Error Horizon (ZEH) for trustworthy LLMs, which represents the maximum range that a model can solve without any errors. While ZEH itself is simple, we demonstrate that evaluating the ZEH of state-of-the-art LLMs yields abundant insights. For example, by evaluating the ZEH of GPT-5.2, we found that GPT-5.2 cannot even compute the parity of a short string like 11000, and GPT-5.2 cannot determine whether the parentheses in ((((()))))) are balanced. This is surprising given the excellent capabilities of GPT-5.2. The fact that LLMs make mistakes on such simple problems serves as an important lesson when applying LLMs to safety-critical domains. By applying ZEH to Qwen2.5 and conducting detailed analysis, we found that while ZEH correlates with accuracy, the detailed behaviors differ, and ZEH provides clues about the emergence of algorithmic capabilities. Finally, while computing ZEH incurs significant computational cost, we discuss how to mitigate this cost by achieving up to one order of magnitude speedup using tree structures and online softmax.

We present FlexLLM, a composable High-Level Synthesis (HLS) library for rapid development of domain-specific LLM accelerators. FlexLLM exposes key architectural degrees of freedom for stage-customized inference, enabling hybrid designs that tailor temporal reuse and spatial dataflow differently for prefill and decode, and provides a comprehensive quantization suite to support accurate low-bit deployment. Using FlexLLM, we build a complete inference system for the Llama-3.2 1B model in under two months with only 1K lines of code. The system includes: (1) a stage-customized accelerator with hardware-efficient quantization (12.68 WikiText-2 PPL) surpassing SpinQuant baseline, and (2) a Hierarchical Memory Transformer (HMT) plug-in for efficient long-context processing. On the AMD U280 FPGA at 16nm, the accelerator achieves 1.29$\times$ end-to-end speedup, 1.64$\times$ higher decode throughput, and 3.14$\times$ better energy efficiency than an NVIDIA A100 GPU (7nm) running BF16 inference; projected results on the V80 FPGA at 7nm reach 4.71$\times$, 6.55$\times$, and 4.13$\times$, respectively. In long-context scenarios, integrating the HMT plug-in reduces prefill latency by 23.23$\times$ and extends the context window by 64$\times$, delivering 1.10$\times$/4.86$\times$ lower end-to-end latency and 5.21$\times$/6.27$\times$ higher energy efficiency on the U280/V80 compared to the A100 baseline. FlexLLM thus bridges algorithmic innovation in LLM inference and high-performance accelerators with minimal manual effort.

Recent advances in LLM-based Text-to-SQL have achieved remarkable gains on public benchmarks such as BIRD and Spider. Yet, these systems struggle to scale in realistic enterprise settings with large, complex schemas, diverse SQL dialects, and expensive multi-step reasoning. Emerging agentic approaches show potential for adaptive reasoning but often suffer from inefficiency and instability-repeating interactions with databases, producing inconsistent outputs, and occasionally failing to generate valid answers. To address these challenges, we introduce Agent Semantic Memory (AgentSM), an agentic framework for Text-to-SQL that builds and leverages interpretable semantic memory. Instead of relying on raw scratchpads or vector retrieval, AgentSM captures prior execution traces-or synthesizes curated ones-as structured programs that directly guide future reasoning. This design enables systematic reuse of reasoning paths, which allows agents to scale to larger schemas, more complex questions, and longer trajectories efficiently and reliably. Compared to state-of-the-art systems, AgentSM achieves higher efficiency by reducing average token usage and trajectory length by 25% and 35%, respectively, on the Spider 2.0 benchmark. It also improves execution accuracy, reaching a state-of-the-art accuracy of 44.8% on the Spider 2.0 Lite benchmark.

Role-play prompting is known to steer the behavior of language models by injecting a persona into the prompt, improving their zero-shot reasoning capabilities. However, such improvements are inconsistent across different tasks or instances. This inconsistency suggests that zero-shot and role-play prompting may offer complementary strengths rather than one being universally superior. Building on this insight, we propose Persona Switch, a novel decoding method that dynamically combines the benefits of both prompting strategies. Our method proceeds step-by-step, selecting the better output between zero-shot and role-play prompting at each step by comparing their output confidence, as measured by the logit gap. Experiments with widely-used LLMs demonstrate that Persona Switch consistently outperforms competitive baselines, achieving up to 5.13% accuracy improvement. Furthermore, we show that output confidence serves as an informative measure for selecting the more reliable output.

As frontier AI models are deployed globally, it is essential that their behaviour remains safe and reliable across diverse linguistic and cultural contexts. To examine how current model safeguards hold up in such settings, participants from the International Network for Advanced AI Measurement, Evaluation and Science, including representatives from Singapore, Japan, Australia, Canada, the EU, France, Kenya, South Korea and the UK conducted a joint multilingual evaluation exercise. Led by Singapore AISI, two open-weight models were tested across ten languages spanning high and low resourced groups: Cantonese English, Farsi, French, Japanese, Korean, Kiswahili, Malay, Mandarin Chinese and Telugu. Over 6,000 newly translated prompts were evaluated across five harm categories (privacy, non-violent crime, violent crime, intellectual property and jailbreak robustness), using both LLM-as-a-judge and human annotation. The exercise shows how safety behaviours can vary across languages. These include differences in safeguard robustness across languages and harm types and variation in evaluator reliability (LLM-as-judge vs. human review). Further, it also generated methodological insights for improving multilingual safety evaluations, such as the need for culturally contextualised translations, stress-tested evaluator prompts and clearer human annotation guidelines. This work represents an initial step toward a shared framework for multilingual safety testing of advanced AI systems and calls for continued collaboration with the wider research community and industry.

Although AI agents have demonstrated impressive capabilities in long-horizon reasoning, their reliability is severely hampered by the ``Spiral of Hallucination,'' where early epistemic errors propagate irreversibly. Existing methods face a dilemma: uncertainty quantification (UQ) methods typically act as passive sensors, only diagnosing risks without addressing them, while self-reflection mechanisms suffer from continuous or aimless corrections. To bridge this gap, we propose a unified Dual-Process Agentic UQ (AUQ) framework that transforms verbalized uncertainty into active, bi-directional control signals. Our architecture comprises two complementary mechanisms: System 1 (Uncertainty-Aware Memory, UAM), which implicitly propagates verbalized confidence and semantic explanations to prevent blind decision-making; and System 2 (Uncertainty-Aware Reflection, UAR), which utilizes these explanations as rational cues to trigger targeted inference-time resolution only when necessary. This enables the agent to balance efficient execution and deep deliberation dynamically. Extensive experiments on closed-loop benchmarks and open-ended deep research tasks demonstrate that our training-free approach achieves superior performance and trajectory-level calibration. We believe this principled framework AUQ represents a significant step towards reliable agents.

The rapid advancement of Multimodal Large Language Models (MLLMs) has introduced complex security challenges, particularly at the intersection of textual and visual safety. While existing schemes have explored the security vulnerabilities of MLLMs, the investigation into their visual safety boundaries remains insufficient. In this paper, we propose Beyond Visual Safety (BVS), a novel image-text pair jailbreaking framework specifically designed to probe the visual safety boundaries of MLLMs. BVS employs a "reconstruction-then-generation" strategy, leveraging neutralized visual splicing and inductive recomposition to decouple malicious intent from raw inputs, thereby leading MLLMs to be induced into generating harmful images. Experimental results demonstrate that BVS achieves a remarkable jailbreak success rate of 98.21\% against GPT-5 (12 January 2026 release). Our findings expose critical vulnerabilities in the visual safety alignment of current MLLMs.

As digital threats continue to grow, organizations must find ways to enhance security while protecting user privacy. This paper explores how artificial intelligence (AI) plays a crucial role in achieving this balance. AI technologies can improve security by detecting threats, monitoring systems, and automating responses. However, using AI also raises privacy concerns that need careful consideration.We examine real-world examples from the healthcare sector to illustrate how organizations can implement AI solutions that strengthen security without compromising patient privacy. Additionally, we discuss the importance of creating transparent AI systems and adhering to privacy regulations.Ultimately, this paper provides insights and recommendations for integrating AI into healthcare security practices, helping organizations navigate the challenges of modern management while keeping patient data safe.

While Large Language Models (LLMs) show remarkable capabilities, their unreliability remains a critical barrier to deployment in high-stakes domains. This survey charts a functional evolution in addressing this challenge: the evolution of uncertainty from a passive diagnostic metric to an active control signal guiding real-time model behavior. We demonstrate how uncertainty is leveraged as an active control signal across three frontiers: in \textbf{advanced reasoning} to optimize computation and trigger self-correction; in \textbf{autonomous agents} to govern metacognitive decisions about tool use and information seeking; and in \textbf{reinforcement learning} to mitigate reward hacking and enable self-improvement via intrinsic rewards. By grounding these advancements in emerging theoretical frameworks like Bayesian methods and Conformal Prediction, we provide a unified perspective on this transformative trend. This survey provides a comprehensive overview, critical analysis, and practical design patterns, arguing that mastering the new trend of uncertainty is essential for building the next generation of scalable, reliable, and trustworthy AI.

Active learning (AL) strategies aim to train high-performance models with minimal labeling efforts, only selecting the most informative instances for annotation. Current approaches to evaluating data informativeness predominantly focus on the data's distribution or intrinsic information content and do not directly correlate with downstream task performance, such as mean average precision (mAP) in object detection. Thus, we propose Performance-guided (i.e. mAP-guided) Reinforced Active Learning for Object Detection (MGRAL), a novel approach that leverages the concept of expected model output changes as informativeness. To address the combinatorial explosion challenge of batch sample selection and the non-differentiable correlation between model performance and selected batches, MGRAL skillfully employs a reinforcement learning-based sampling agent that optimizes selection using policy gradient with mAP improvement as reward. Moreover, to reduce the computational overhead of mAP estimation with unlabeled samples, MGRAL utilizes an unsupervised way with fast look-up tables, ensuring feasible deployment. We evaluate MGRAL's active learning performance on detection tasks over PASCAL VOC and COCO benchmarks. Our approach demonstrates the highest AL curve with convincing visualizations, establishing a new paradigm in reinforcement learning-driven active object detection.

Trigger-Action Programming (TAP) platforms such as IFTTT and Zapier enable Web of Things (WoT) automation by composing event-driven rules across heterogeneous services. A TAP applet links a trigger to an action and must bind trigger outputs (ingredients) to action inputs (fields) to be executable. Prior work largely treats TAP as service-level prediction from natural language, which often yields non-executable applets that still require manual configuration. We study the function-level configuration problem: generating complete applets with correct ingredient-to-field bindings. We propose FARM (Field-Aware Resolution Model), a two-stage architecture for automated applet generation with full configuration. Stage 1 trains contrastive dual encoders with selective layer freezing over schema-enriched representations, retrieving candidates from 1,724 trigger functions and 1,287 action functions (2.2M possible trigger-action pairs). Stage 2 performs selection and configuration using an LLM-based multi-agent pipeline. It includes intent analysis, trigger selection, action selection via cross-schema scoring, and configuration verification. Agents coordinate through shared state and agreement-based selection. FARM achieves 81% joint accuracy on Gold (62% Noisy, 70% One-shot) at the function level, where both trigger and action functions must match the ground truth. For comparison with service-level baselines, we map functions to their parent services and evaluate at the service level. FARM reaches 81% joint accuracy and improves over TARGE by 23 percentage points. FARM also generates ingredient-to-field bindings, producing executable automation configurations.

Model editing updates a pre-trained LLM with new facts or rules without re-training, while preserving unrelated behavior. In real deployment, edits arrive as long streams, and existing editors often face a plasticity-stability dilemma: locate-then-edit "hard writes" can accumulate interference over time, while null-space-style "hard preservation" preserves only what is explicitly constrained, so past edits can be overwritten and unconstrained behaviors may deviate, degrading general capabilities in the many-edits regime. We propose RLSEdit, a recursive least-squares editor for long sequential editing. RLSEdit formulates editing as an online quadratic optimization with soft constraints, minimizing a cumulative key-value fitting objective with two regularizers that control for both deviation from the pre-trained weights and from a designated anchor mapping. The resulting update admits an efficient online recursion via the Woodbury identity, with per-edit cost independent of history length and scaling only with the current edit size. We further provide deviation bounds and an asymptotic characterization of the adherence-preservation trade-off in the many-edits regime. Experiments on multiple model families demonstrate stable scaling to 10K edits, outperforming strong baselines in both edit success and holistic stability -- crucially retaining early edits, and preserving general capabilities on GLUE and held-out reasoning/code benchmarks.

The rapid rise of autonomous AI systems and advancements in agent capabilities are introducing new risks due to reduced oversight of real-world interactions. Yet agent testing remains nascent and is still a developing science. As AI agents begin to be deployed globally, it is important that they handle different languages and cultures accurately and securely. To address this, participants from The International Network for Advanced AI Measurement, Evaluation and Science, including representatives from Singapore, Japan, Australia, Canada, the European Commission, France, Kenya, South Korea, and the United Kingdom have come together to align approaches to agentic evaluations. This is the third exercise, building on insights from two earlier joint testing exercises conducted by the Network in November 2024 and February 2025. The objective is to further refine best practices for testing advanced AI systems. The exercise was split into two strands: (1) common risks, including leakage of sensitive information and fraud, led by Singapore AISI; and (2) cybersecurity, led by UK AISI. A mix of open and closed-weight models were evaluated against tasks from various public agentic benchmarks. Given the nascency of agentic testing, our primary focus was on understanding methodological issues in conducting such tests, rather than examining test results or model capabilities. This collaboration marks an important step forward as participants work together to advance the science of agentic evaluations.

Retrieval-augmented generation (RAG) systems integrate document retrieval with large language models and have been widely adopted. However, in privacy-related scenarios, RAG introduces a new privacy risk: adversaries can issue carefully crafted queries to exfiltrate sensitive content from the underlying corpus gradually. Although recent studies have demonstrated multi-turn extraction attacks, they rely on heuristics and fail to perform long-term extraction planning. To address these limitations, we formulate the RAG extraction attack as an adaptive stochastic coverage problem (ASCP). In ASCP, each query is treated as a probabilistic action that aims to maximize conditional marginal gain (CMG), enabling principled long-term planning under uncertainty. However, integrating ASCP with practical RAG attack faces three key challenges: unobservable CMG, intractability in the action space, and feasibility constraints. To overcome these challenges, we maintain a global attacker-side state to guide the attack. Building on this idea, we introduce RAGCRAWLER, which builds a knowledge graph to represent revealed information, uses this global state to estimate CMG, and plans queries in semantic space that target unretrieved regions. In comprehensive experiments across diverse RAG architectures and datasets, our proposed method, RAGCRAWLER, consistently outperforms all baselines. It achieves up to 84.4% corpus coverage within a fixed query budget and deliver an average improvement of 20.7% over the top-performing baseline. It also maintains high semantic fidelity and strong content reconstruction accuracy with low attack cost. Crucially, RAGCRAWLER proves its robustness by maintaining effectiveness against advanced RAG systems employing query rewriting and multi-query retrieval strategies. Our work reveals significant security gaps and highlights the pressing need for stronger safeguards for RAG.

Deploying Audio-Language Models (Audio-LLMs) on edge infrastructure exposes a persistent tension between perception depth and computational efficiency. Lightweight local models tend to produce passive perception - generic summaries that miss the subtle evidence required for multi-step audio reasoning - while indiscriminate cloud offloading incurs unacceptable latency, bandwidth cost, and privacy risk. We propose CoFi-Agent (Tool-Augmented Coarse-to-Fine Agent), a hybrid architecture targeting edge servers and gateways. It performs fast local perception and triggers conditional forensic refinement only when uncertainty is detected. CoFi-Agent runs an initial single-pass on a local 7B Audio-LLM, then a cloud controller gates difficult cases and issues lightweight plans for on-device tools such as temporal re-listening and local ASR. On the MMAR benchmark, CoFi-Agent improves accuracy from 27.20% to 53.60%, while achieving a better accuracy-efficiency trade-off than an always-on investigation pipeline. Overall, CoFi-Agent bridges the perception gap via tool-enabled, conditional edge-cloud collaboration under practical system constraints.

Patients are increasingly using large language models (LLMs) to seek answers to their healthcare-related questions. However, benchmarking efforts in LLMs for question answering often focus on medical exam questions, which differ significantly in style and content from the questions patients actually raise in real life. To bridge this gap, we sourced data from Google's People Also Ask feature by querying the top 200 prescribed medications in the United States, curating a dataset of medical questions people commonly ask. A considerable portion of the collected questions contains incorrect assumptions and dangerous intentions. We demonstrate that the emergence of these corrupted questions is not uniformly random and depends heavily on the degree of incorrectness in the history of questions that led to their appearance. Current LLMs that perform strongly on other benchmarks struggle to identify incorrect assumptions in everyday questions.

Contemporary sequential recommendation methods are becoming more complex, shifting from classification to a diffusion-guided generative paradigm. However, the quality of guidance in the form of user information is often compromised by missing data in the observed sequences, leading to suboptimal generation quality. Existing methods address this by removing locally similar items, but overlook ``critical turning points'' in user interest, which are crucial for accurately predicting subsequent user intent. To address this, we propose a novel Counterfactual Attention Regulation Diffusion model (CARD), which focuses on amplifying the signal from key interest-turning-point items while concurrently identifying and suppressing noise within the user sequence. CARD consists of (1) a Dual-side Thompson Sampling method to identify sequences undergoing significant interest shift, and (2) a counterfactual attention mechanism for these sequences to quantify the importance of each item. In this manner, CARD provides the diffusion model with a high-quality guidance signal composed of dynamically re-weighted interaction vectors to enable effective generation. Experiments show our method works well on real-world data without being computationally expensive. Our code is available at https://github.com/yanqilong3321/CARD.

Designing inclusive cycling infrastructure requires balancing competing needs of diverse user groups, yet designers often struggle to anticipate how different cyclists experience the same street. We investigate how persona-based multi-agent evaluation can support inclusive design by making experiential conflicts explicit. We present StreetDesignAI, an interactive system that enables designers to (1) ground evaluation in street context through imagery and map data, (2) receive parallel feedback from cyclist personas spanning confident to cautious users, and (3) iteratively modify designs while surfacing conflicts across perspectives. A within-subjects study with 26 transportation professionals demonstrates that structured multi-perspective feedback significantly improves designers' understanding of diverse user perspectives, ability to identify persona needs, and confidence in translating them into design decisions, with higher satisfaction and stronger intention for professional adoption. Qualitative findings reveal how conflict surfacing transforms design exploration from single-perspective optimization toward deliberate trade-off reasoning. We discuss implications for AI tools that scaffold inclusive design through disagreement as an interaction primitive.

Transformer-based models, despite their promise for long-term time series forecasting (LTSF), suffer from an inherent low-pass filtering effect that limits their effectiveness. This issue arises due to undifferentiated propagation of frequency components across layers, causing a progressive attenuation of high-frequency information crucial for capturing fine-grained temporal variations. To address this limitation, we propose Dualformer, a principled dual-domain framework that rethinks frequency modeling from a layer-wise perspective. Dualformer introduces three key components: (1) a dual-branch architecture that concurrently models complementary temporal patterns in both time and frequency domains; (2) a hierarchical frequency sampling module that allocates distinct frequency bands to different layers, preserving high-frequency details in lower layers while modeling low-frequency trends in deeper layers; and (3) a periodicity-aware weighting mechanism that dynamically balances contributions from the dual branches based on the harmonic energy ratio of inputs, supported theoretically by a derived lower bound. This design enables structured frequency modeling and adaptive integration of time-frequency features, effectively preserving high-frequency information and enhancing generalization. Extensive experiments conducted on eight widely used benchmarks demonstrate Dualformer's robustness and superior performance, particularly on heterogeneous or weakly periodic data. Our code is publicly available at https://github.com/Akira-221/Dualformer.

The recent surge in popularity of Nano-Banana and Seedream 4.0 underscores the community's strong interest in multi-image composition tasks. Compared to single-image editing, multi-image composition presents significantly greater challenges in terms of consistency and quality, yet existing models have not disclosed specific methodological details for achieving high-quality fusion. Through statistical analysis, we identify Human-Object Interaction (HOI) as the most sought-after category by the community. We therefore systematically analyze and implement a state-of-the-art solution for multi-image composition with a primary focus on HOI-centric tasks. We present Skywork UniPic 3.0, a unified multimodal framework that integrates single-image editing and multi-image composition. Our model supports an arbitrary (1~6) number and resolution of input images, as well as arbitrary output resolutions (within a total pixel budget of 1024x1024). To address the challenges of multi-image composition, we design a comprehensive data collection, filtering, and synthesis pipeline, achieving strong performance with only 700K high-quality training samples. Furthermore, we introduce a novel training paradigm that formulates multi-image composition as a sequence-modeling problem, transforming conditional generation into unified sequence synthesis. To accelerate inference, we integrate trajectory mapping and distribution matching into the post-training stage, enabling the model to produce high-fidelity samples in just 8 steps and achieve a 12.5x speedup over standard synthesis sampling. Skywork UniPic 3.0 achieves state-of-the-art performance on single-image editing benchmark and surpasses both Nano-Banana and Seedream 4.0 on multi-image composition benchmark, thereby validating the effectiveness of our data pipeline and training paradigm. Code, models and dataset are publicly available.

Realistic network traffic simulation is critical for evaluating intrusion detection systems, stress-testing network protocols, and constructing high-fidelity environments for cybersecurity training. While attack traffic can often be layered into training environments using red-teaming or replay methods, generating authentic benign background traffic remains a core challenge -- particularly in simulating the complex temporal and communication dynamics of real-world networks. This paper introduces TempoNet, a novel generative model that combines multi-task learning with multi-mark temporal point processes to jointly model inter-arrival times and all packet- and flow-header fields. TempoNet captures fine-grained timing patterns and higher-order correlations such as host-pair behavior and seasonal trends, addressing key limitations of GAN-, LLM-, and Bayesian-based methods that fail to reproduce structured temporal variation. TempoNet produces temporally consistent, high-fidelity traces, validated on real-world datasets. Furthermore, we show that intrusion detection models trained on TempoNet-generated background traffic perform comparably to those trained on real data, validating its utility for real-world security applications.

Knowledge distillation (KD) transfers knowledge from large teacher models to compact student models, enabling efficient deployment on resource constrained devices. While diverse KD methods, including response based, feature based, and relation based approaches, capture different aspects of teacher knowledge, integrating multiple methods or knowledge sources is promising but often hampered by complex implementation, inflexible combinations, and catastrophic forgetting, which limits practical effectiveness. This work proposes SMSKD (Sequential Multi Stage Knowledge Distillation), a flexible framework that sequentially integrates heterogeneous KD methods. At each stage, the student is trained with a specific distillation method, while a frozen reference model from the previous stage anchors learned knowledge to mitigate forgetting. In addition, we introduce an adaptive weighting mechanism based on the teacher true class probability (TCP) that dynamically adjusts the reference loss per sample to balance knowledge retention and integration. By design, SMSKD supports arbitrary method combinations and stage counts with negligible computational overhead. Extensive experiments show that SMSKD consistently improves student accuracy across diverse teacher student architectures and method combinations, outperforming existing baselines. Ablation studies confirm that stage wise distillation and reference model supervision are primary contributors to performance gains, with TCP based adaptive weighting providing complementary benefits. Overall, SMSKD is a practical and resource efficient solution for integrating heterogeneous KD methods.

Real-time understanding of long video streams remains challenging for multimodal large language models (VLMs) due to redundant frame processing and rapid forgetting of past context. Existing streaming systems rely on fixed-interval decoding or cache pruning, which either produce repetitive outputs or discard crucial temporal information. We introduce Event-VStream, an event-aware framework that represents continuous video as a sequence of discrete, semantically coherent events. Our system detects meaningful state transitions by integrating motion, semantic, and predictive cues, and triggers language generation only at those boundaries. Each event embedding is consolidated into a persistent memory bank, enabling long-horizon reasoning while maintaining low latency. Across OVOBench-Realtime, and long-form Ego4D evaluations, Event-VStream achieves competitive performance. It improves over a VideoLLM-Online-8B baseline by +10.4 points on OVOBench-Realtime, achieves performance close to Flash-VStream-7B despite using only a general-purpose LLaMA-3-8B text backbone, and maintains around 70% GPT-5 win rate on 2-hour Ego4D streams.

Hallucinations in Large Language Models (LLMs) -- generations that are plausible but factually unfaithful -- remain a critical barrier to high-stakes deployment. Current detection methods typically rely on computationally expensive external retrieval loops or opaque black-box LLM judges requiring 70B+ parameters. In this work, we introduce [Model Name], a hybrid detection framework that combines neuroscience-inspired signal design with supervised machine learning. We extract interpretable signals grounded in Predictive Coding (quantifying surprise against internal priors) and the Information Bottleneck (measuring signal retention under perturbation). Through systematic ablation, we demonstrate three key enhancements: Entity-Focused Uptake (concentrating on high-value tokens), Context Adherence (measuring grounding strength), and Falsifiability Score (detecting confident but contradictory claims). Evaluating on HaluBench (n=200, perfectly balanced), our theory-guided baseline achieves 0.8017 AUROC. BASE supervised models reach 0.8274 AUROC, while IMPROVED features boost performance to 0.8669 AUROC (4.95% gain), demonstrating consistent improvements across architectures. This competitive performance is achieved while using 75x less training data than Lynx (200 vs 15,000 samples), 1000x faster inference (5ms vs 5s), and remaining fully interpretable. Crucially, we report a negative result: the Rationalization signal fails to distinguish hallucinations, suggesting that LLMs generate coherent reasoning for false premises ("Sycophancy"). This work demonstrates that domain knowledge encoded in signal architecture provides superior data efficiency compared to scaling LLM judges, achieving strong performance with lightweight (less than 1M parameter), explainable models suitable for production deployment.

Large-scale language models (LLMs) often offer clinical judgments based on incomplete information, increasing the risk of misdiagnosis. Existing studies have primarily evaluated confidence in single-turn, static settings, overlooking the coupling between confidence and correctness as clinical evidence accumulates during real consultations, which limits their support for reliable decision-making. We propose the first benchmark for assessing confidence in multi-turn interaction during realistic medical consultations. Our benchmark unifies three types of medical data for open-ended diagnostic generation and introduces an information sufficiency gradient to characterize the confidence-correctness dynamics as evidence increases. We implement and compare 27 representative methods on this benchmark; two key insights emerge: (1) medical data amplifies the inherent limitations of token-level and consistency-level confidence methods, and (2) medical reasoning must be evaluated for both diagnostic accuracy and information completeness. Based on these insights, we present MedConf, an evidence-grounded linguistic self-assessment framework that constructs symptom profiles via retrieval-augmented generation, aligns patient information with supporting, missing, and contradictory relations, and aggregates them into an interpretable confidence estimate through weighted integration. Across two LLMs and three medical datasets, MedConf consistently outperforms state-of-the-art methods on both AUROC and Pearson correlation coefficient metrics, maintaining stable performance under conditions of information insufficiency and multimorbidity. These results demonstrate that information adequacy is a key determinant of credible medical confidence modeling, providing a new pathway toward building more reliable and interpretable large medical models.

Detecting machine failures promptly is of utmost importance in industry for maintaining efficiency and minimizing downtime. This paper introduces a failure detection algorithm based on quantum computing and a statistical change-point detection approach. Our method leverages the potential of projected quantum feature maps to enhance the precision of anomaly detection in machine monitoring systems. We empirically validate our approach on benchmark multi-dimensional time series datasets as well as on a real-world dataset comprising IoT sensor readings from operational machines, ensuring the practical relevance of our study. The algorithm was executed on IBM's 133-qubit Heron quantum processor, demonstrating the feasibility of integrating quantum computing into industrial maintenance procedures. The presented results underscore the effectiveness of our quantum-based failure detection system, showcasing its capability to accurately identify anomalies in noisy time series data. This work not only highlights the potential of quantum computing in industrial diagnostics but also paves the way for more sophisticated quantum algorithms in the realm of predictive maintenance.

Bayesian optimisation is a sample efficient method for finding a global optimum of expensive black-box objective functions. Historic datasets from related problems can be exploited to help improve performance of Bayesian optimisation by adapting transfer learning methods to various components of the Bayesian optimisation pipeline. In this study we perform an empirical analysis of various ensemble-based transfer learning Bayesian optimisation methods and pipeline components. We expand on previous work in the literature by contributing some specific pipeline components, and three new real-time transfer learning Bayesian optimisation benchmarks. In particular we propose to use a weighting strategy for ensemble surrogate model predictions based on regularised regression with weights constrained to be positive, and a related component for handling the case when transfer learning is not improving Bayesian optimisation performance. We find that in general, two components that help improve transfer learning Bayesian optimisation performance are warm start initialisation and constraining weights used with ensemble surrogate model to be positive.

In this work we introduce three ideas that can further improve particle FRNN physics simulations running on RT Cores; i) a real-time update/rebuild ratio optimizer for the bounding volume hierarchy (BVH) structure, ii) a new RT core use, with two variants, that eliminates the need of a neighbor list and iii) a technique that enables RT cores for FRNN with periodic boundary conditions (BC). Experimental evaluation using the Lennard-Jones FRNN interaction model as a case study shows that the proposed update/rebuild ratio optimizer is capable of adapting to the different dynamics that emerge during a simulation, leading to a RT core pipeline up to $\sim 3.4\times$ faster than with other known approaches to manage the BVH. In terms of simulation step performance, the proposed variants can significantly improve the speedup and EE of the base RT core idea; from $\sim1.3\times$ at small radius to $\sim2.0\times$ for log normal radius distributions. Furthermore, the proposed variants manage to simulate cases that would otherwise not fit in memory because of the use of neighbor lists, such as clusters of particles with log normal radius distribution. The proposed RT Core technique to support periodic BC is indeed effective as it does not introduce any significant penalty in performance. In terms of scaling, the proposed methods scale both their performance and EE across GPU generations. Throughout the experimental evaluation, we also identify the simulation cases were regular GPU computation should still be preferred, contributing to the understanding of the strengths and limitations of RT cores.

Healthcare organizations are beginning to embed agentic AI into routine workflows, including clinical documentation support and early-warning monitoring. As these capabilities diffuse across departments and vendors, health systems face agent sprawl, causing duplicated agents, unclear accountability, inconsistent controls, and tool permissions that persist beyond the original use case. Existing AI governance frameworks emphasize lifecycle risk management but provide limited guidance for the day-to-day operations of agent fleets. We propose a Unified Agent Lifecycle Management (UALM) blueprint derived from a rapid, practice-oriented synthesis of governance standards, agent security literature, and healthcare compliance requirements. UALM maps recurring gaps onto five control-plane layers: (1) an identity and persona registry, (2) orchestration and cross-domain mediation, (3) PHI-bounded context and memory, (4) runtime policy enforcement with kill-switch triggers, and (5) lifecycle management and decommissioning linked to credential revocation and audit logging. A companion maturity model supports staged adoption. UALM offers healthcare CIOs, CISOs, and clinical leaders an implementable pattern for audit-ready oversight that preserves local innovation and enables safer scaling across clinical and administrative domains.

Whether Large Language Models (LLMs) truly possess human-like Theory of Mind (ToM) capabilities has garnered increasing attention. However, existing benchmarks remain largely restricted to narrow paradigms like false belief tasks, failing to capture the full spectrum of human cognitive mechanisms. We introduce CogToM, a comprehensive, theoretically grounded benchmark comprising over 8000 bilingual instances across 46 paradigms, validated by 49 human annotator.A systematic evaluation of 22 representative models, including frontier models like GPT-5.1 and Qwen3-Max, reveals significant performance heterogeneities and highlights persistent bottlenecks in specific dimensions. Further analysis based on human cognitive patterns suggests potential divergences between LLM and human cognitive structures. CogToM offers a robust instrument and perspective for investigating the evolving cognitive boundaries of LLMs.

PURPOSE OR GOAL: This study investigates how GenAI can be integrated with a criterion-referenced grading framework to improve the efficiency and quality of grading for mathematical assessments in engineering. It specifically explores the challenges demonstrators face with manual, model solution-based grading and how a GenAI-supported system can be designed to reliably identify student errors, provide high-quality feedback, and support human graders. The research also examines human graders' perceptions of the effectiveness of this GenAI-assisted approach. ACTUAL OR ANTICIPATED OUTCOMES: The study found that GenAI achieved an overall grading accuracy of 92.5%, comparable to two experienced human graders. The two researchers, who also served as subject demonstrators, perceived the GenAI as a helpful second reviewer that improved accuracy by catching small errors and provided more complete feedback than they could manually. A central outcome was the significant enhancement of formative feedback. However, they noted the GenAI tool is not yet reliable enough for autonomous use, especially with unconventional solutions. CONCLUSIONS/RECOMMENDATIONS/SUMMARY: This study demonstrates that GenAI, when paired with a structured, criterion-referenced framework using binary questions, can grade engineering mathematical assessments with an accuracy comparable to human experts. Its primary contribution is a novel methodological approach that embeds the generation of high-quality, scalable formative feedback directly into the assessment workflow. Future work should investigate student perceptions of GenAI grading and feedback.

Large language models (LLMs) can call tools effectively, yet they remain brittle in multi-turn execution: following a tool call error, smaller models often degenerate into repetitive invalid re-invocations, failing to interpret error feedback and self-correct. This brittleness hinders reliable real-world deployment, where the execution errors are inherently inevitable during tool interaction procedures. We identify a key limitation of current approaches: standard reinforcement learning (RL) treats errors as sparse negative rewards, providing no guidance on how to recover, while pre-collected synthetic error-correction datasets suffer from distribution mismatch with the model's on-policy error modes. To bridge this gap, we propose Fission-GRPO, a framework that converts execution errors into corrective supervision within the RL training loop. Our core mechanism fissions each failed trajectory into a new training instance by augmenting it with diagnostic feedback from a finetuned Error Simulator, then resampling recovery rollouts on-policy. This enables the model to learn from the precise errors it makes during exploration, rather than from static, pre-collected error cases. On the BFCL v4 Multi-Turn, Fission-GRPO improves the error recovery rate of Qwen3-8B by 5.7% absolute, crucially, yielding a 4% overall accuracy gain (42.75% to 46.75%) over GRPO and outperforming specialized tool-use agents.

In this report, we present the Qwen3-TTS series, a family of advanced multilingual, controllable, robust, and streaming text-to-speech models. Qwen3-TTS supports state-of-the-art 3-second voice cloning and description-based control, allowing both the creation of entirely novel voices and fine-grained manipulation over the output speech. Trained on over 5 million hours of speech data spanning 10 languages, Qwen3-TTS adopts a dual-track LM architecture for real-time synthesis, coupled with two speech tokenizers: 1) Qwen-TTS-Tokenizer-25Hz is a single-codebook codec emphasizing semantic content, which offers seamlessly integration with Qwen-Audio and enables streaming waveform reconstruction via a block-wise DiT. 2) Qwen-TTS-Tokenizer-12Hz achieves extreme bitrate reduction and ultra-low-latency streaming, enabling immediate first-packet emission ($97\,\mathrm{ms}$) through its 12.5 Hz, 16-layer multi-codebook design and a lightweight causal ConvNet. Extensive experiments indicate state-of-the-art performance across diverse objective and subjective benchmark (e.g., TTS multilingual test set, InstructTTSEval, and our long speech test set). To facilitate community research and development, we release both tokenizers and models under the Apache 2.0 license.

1-identification is a fundamental multi-armed bandit formulation on pure exploration. An agent aims to determine whether there exists a qualified arm whose mean reward is not less than a known threshold $μ_0$, or to output \textsf{None} if it believes such an arm does not exist. The agent needs to guarantee its output is correct with probability at least $1-δ$, while making expected total pulling times $\mathbb{E}τ$ as small as possible. We work on 1-identification with two main contributions. (1) We utilize an optimization formulation to derive a new lower bound of $\mathbb{E}τ$, when there is at least one qualified arm. (2) We design a new algorithm, deriving tight upper bounds whose gap to lower bounds are up to a polynomial of logarithm factor across all problem instance. Our result complements the analysis of $\mathbb{E}τ$ when there are multiple qualified arms, which is an open problem left by history literature.

Reinforcement Learning with Verifiable Rewards (RLVR) is a central paradigm for turning large language models (LLMs) into reliable problem solvers, especially in logic-heavy domains. Despite its empirical success, it remains unclear whether RLVR elicits novel capabilities or merely sharpens the distribution over existing knowledge. We study this by formalizing over-sharpening, a phenomenon where the policy collapses onto limited modes, suppressing valid alternatives. At a high level, we discover finite-batch updates intrinsically bias learning toward sampled modes, triggering a collapse that propagates globally via semantic coupling. To mitigate this, we propose inverse-success advantage calibration to prioritize difficult queries and distribution-level calibration to diversify sampling via a memory network. Empirical evaluations validate that our strategies can effectively improve generalization.

The rapid growth of live-streaming platforms such as Twitch has introduced complex challenges in moderating toxic behavior. Traditional moderation approaches, such as human annotation and keyword-based filtering, have demonstrated utility, but human moderators on Twitch constantly struggle to scale effectively in the fast-paced, high-volume, and context-rich chat environment of the platform while also facing harassment themselves. Recent advances in large language models (LLMs), such as DeepSeek-R1-Distill and Llama-3-8B-Instruct, offer new opportunities for toxicity detection, especially in understanding nuanced, multimodal communication involving emotes. In this work, we present an exploratory comparison of toxicity detection approaches tailored to Twitch. Our analysis reveals that incorporating emotes improves the detection of toxic behavior. To this end, we introduce ToxiTwitch, a hybrid model that combines LLM-generated embeddings of text and emotes with traditional machine learning classifiers, including Random Forest and SVM. In our case study, the proposed hybrid approach reaches up to 80 percent accuracy under channel-specific training (with 13 percent improvement over BERT and F1-score of 76 percent). This work is an exploratory study intended to surface challenges and limits of emote-aware toxicity detection on Twitch.

Current business environments require organizations to continuously reconfigure cross-functional processes, yet enterprise systems are still organized around siloed departments, rigid workflows, and hard-coded automation. Meanwhile large language models (LLMs) excel at interpreting natural language and unstructured data but lack deterministic, verifiable execution of complex business logic. To address this gap, here we introduce AUTOBUS, an Autonomous Business System that integrates LLM-based AI agents, predicate-logic programming, and business-semantics-centric enterprise data into a coherent neuro-symbolic AI architecture for orchestrating end-to-end business initiatives. AUTOBUS models an initiative as a network of tasks with explicit pre/post conditions, required data, evaluation rules, and API-level actions. Enterprise data is organized as a knowledge graph whose entities, relationships, and constraints are translated into logic facts and foundational rules, providing the semantic grounding for task reasoning. Core AI agents synthesize task instructions, enterprise semantics, and available tools into task-specific logic programs, which are executed by a logic engine that enforces constraints, coordinates auxiliary tools, and orchestrate execution of actions and outcomes. Humans define and maintain the semantics, policies and task instructions, curate tools, and supervise high-impact or ambiguous decisions, ensuring accountability and adaptability. We detail the AUTOBUS architecture, the anatomy of the AI agent generated logic programs, and the role of humans and auxiliary tools in the lifecycle of a business initiative.

Spiking Neural Networks (SNNs) are promising energy-efficient models and powerful framworks of modeling neuron dynamics. However, existing binary spiking neurons exhibit limited biological plausibilities and low information capacity. Recently developed ternary spiking neuron possesses higher consistency with biological principles (i.e. excitation-inhibition balance mechanism). Despite of this, the ternary spiking neuron suffers from defects including iterative information loss, temporal gradient vanishing and irregular distributions of membrane potentials. To address these issues, we propose Complemented Ternary Spiking Neuron (CTSN), a novel ternary spiking neuron model that incorporates an learnable complemental term to store information from historical inputs. CTSN effectively improves the deficiencies of ternary spiking neuron, while the embedded learnable factors enable CTSN to adaptively adjust neuron dynamics, providing strong neural heterogeneity. Furthermore, based on the temporal evolution features of ternary spiking neurons' membrane potential distributions, we propose the Temporal Membrane Potential Regularization (TMPR) training method. TMPR introduces time-varying regularization strategy utilizing membrane potentials, furhter enhancing the training process by creating extra backpropagation paths. We validate our methods through extensive experiments on various datasets, demonstrating remarkable performance advances.

This paper introduces a neural network-based nonlinear shrinkage estimator of covariance matrices for the purpose of minimum variance portfolio optimization. It is a hybrid approach that integrates statistical estimation with machine learning. Starting from the Ledoit-Wolf (LW) shrinkage estimator, we decompose the LW covariance matrix into its eigenvalues and eigenvectors, and apply a lightweight transformer-based neural network to learn a nonlinear eigenvalue shrinkage function. Trained with portfolio risk as the loss function, the resulting precision matrix (the inverse covariance matrix) estimator directly targets portfolio risk minimization. By conditioning on the sample-to-dimension ratio, the approach remains scalable across different sample sizes and asset universes. Empirical results on stock daily returns from Standard & Poor's 500 Index (S&P500) demonstrate that the proposed method consistently achieves lower out-of-sample realized risk than benchmark approaches. This highlights the promise of integrating structural statistical models with data-driven learning.

While modern Text-to-Speech (TTS) systems achieve high fidelity for read-style speech, they struggle to generate Autonomous Sensory Meridian Response (ASMR), a specialized, low-intensity speech style essential for relaxation. The inherent challenges include ASMR's subtle, often unvoiced characteristics and the demand for zero-shot speaker adaptation. In this paper, we introduce DeepASMR, the first framework designed for zero-shot ASMR generation. We demonstrate that a single short snippet of a speaker's ordinary, read-style speech is sufficient to synthesize high-fidelity ASMR in their voice, eliminating the need for whispered training data from the target speaker. Methodologically, we first identify that discrete speech tokens provide a soft factorization of ASMR style from speaker timbre. Leveraging this insight, we propose a two-stage pipeline incorporating a Large Language Model (LLM) for content-style encoding and a flow-matching acoustic decoder for timbre reconstruction. Furthermore, we contribute DeepASMR-DB, a comprehensive 670-hour English-Chinese multi-speaker ASMR speech corpus, and introduce a novel evaluation protocol integrating objective metrics, human listening tests, LLM-based scoring and unvoiced speech analysis. Extensive experiments confirm that DeepASMR achieves state-of-the-art naturalness and style fidelity in ASMR generation for anyone of any voice, while maintaining competitive performance on normal speech synthesis.

Large language models (LLMs) exhibit powerful capabilities but risk memorizing sensitive personally identifiable information (PII) from their training data, posing significant privacy concerns. While machine unlearning techniques aim to remove such data, they predominantly depend on access to the training data. This requirement is often impractical, as training data in real-world deployments is commonly proprietary or inaccessible. To address this limitation, we propose Data-Free Selective Unlearning (DFSU), a novel privacy-preserving framework that removes sensitive PII from an LLM without requiring its training data. Our approach first synthesizes pseudo-PII through language model inversion, then constructs token-level privacy masks for these synthetic samples, and finally performs token-level selective unlearning via a contrastive mask loss within a low-rank adaptation (LoRA) subspace. Extensive experiments on the AI4Privacy PII-Masking dataset using Pythia models demonstrate that our method effectively removes target PII while maintaining model utility.

Masked Diffusion Language Models (MDLMs) promise parallel token generation and arbitrary-order decoding, yet it remains unclear to what extent current models truly realize these capabilities. We characterize MDLM behavior along two dimensions -- parallelism strength and generation order -- using Average Finalization Parallelism (AFP) and Kendall's tau. We evaluate eight mainstream MDLMs (up to 100B parameters) on 58 benchmarks spanning knowledge, reasoning, and programming. The results show that MDLMs still lag behind comparably sized autoregressive models, mainly because parallel probabilistic modeling weakens inter-token dependencies. Meanwhile, MDLMs exhibit adaptive decoding behavior: their parallelism and generation order vary significantly with the task domain, the stage of reasoning, and whether the output is correct. On tasks that require "backward information" (e.g., Sudoku), MDLMs adopt a solution order that tends to fill easier Sudoku blanks first, highlighting their advantages. Finally, we provide theoretical motivation and design insights supporting a Generate-then-Edit paradigm, which mitigates dependency loss while retaining the efficiency of parallel decoding.

Managing perishable products with limited lifetimes is a fundamental challenge in inventory management, as poor ordering decisions can quickly lead to stockouts or excessive waste. We study a perishable inventory system with random lead times in which both the demand process and the lead time distribution are unknown. We consider a practical setting where orders are placed using limited historical data together with observed covariates and current system states. To improve learning efficiency under limited data, we adopt a marginal cost accounting scheme that assigns each order a single lifetime cost and yields a unified loss function for end-to-end learning. This enables training a deep learning-based policy that maps observed covariates and system states directly to order quantities. We develop two end-to-end variants: a purely black-box approach that outputs order quantities directly (E2E-BB), and a structure-guided approach that embeds the projected inventory level (PIL) policy, capturing inventory effects through explicit computation rather than additional learning (E2E-PIL). We further show that the objective induced by E2E-PIL is homogeneous of degree one, enabling a boosting technique from operational data analytics (ODA) that yields an enhanced policy (E2E-BPIL). Experiments on synthetic and real data establish a robust performance ordering: E2E-BB is dominated by E2E-PIL, which is further improved by E2E-BPIL. Using an excess-risk decomposition, we show that embedding heuristic policy structure reduces effective model complexity and improves learning efficiency with only a modest loss of flexibility. More broadly, our results suggest that deep learning-based decision tools are more effective and robust when guided by human knowledge, highlighting the value of integrating advanced analytics with inventory theory.

As large language models (LLMs) are increasingly deployed in real-world applications, safety guardrails are required to go beyond coarse-grained filtering and support fine-grained, interpretable, and adaptable risk assessment. However, existing solutions often rely on rapid classification schemes or post-hoc rules, resulting in limited transparency, inflexible policies, or prohibitive inference costs. To this end, we present YuFeng-XGuard, a reasoning-centric guardrail model family designed to perform multi-dimensional risk perception for LLM interactions. Instead of producing opaque binary judgments, YuFeng-XGuard generates structured risk predictions, including explicit risk categories and configurable confidence scores, accompanied by natural language explanations that expose the underlying reasoning process. This formulation enables safety decisions that are both actionable and interpretable. To balance decision latency and explanatory depth, we adopt a tiered inference paradigm that performs an initial risk decision based on the first decoded token, while preserving ondemand explanatory reasoning when required. In addition, we introduce a dynamic policy mechanism that decouples risk perception from policy enforcement, allowing safety policies to be adjusted without model retraining. Extensive experiments on a diverse set of public safety benchmarks demonstrate that YuFeng-XGuard achieves stateof-the-art performance while maintaining strong efficiency-efficacy trade-offs. We release YuFeng-XGuard as an open model family, including both a full-capacity variant and a lightweight version, to support a wide range of deployment scenarios.

Recent advancements in mobile and wireless networks are unlocking the full potential of robotic autonomy, enabling robots to take advantage of ultra-low latency, high data throughput, and ubiquitous connectivity. However, for robots to navigate and operate seamlessly, efficiently and reliably, they must have an accurate understanding of both their surrounding environment and the quality of radio signals. Achieving this in highly dynamic and ever-changing environments remains a challenging and largely unsolved problem. In this paper, we introduce MapViT, a two-stage Vision Transformer (ViT)-based framework inspired by the success of pre-train and fine-tune paradigm for Large Language Models (LLMs). MapViT is designed to predict both environmental changes and expected radio signal quality. We evaluate the framework using a set of representative Machine Learning (ML) models, analyzing their respective strengths and limitations across different scenarios. Experimental results demonstrate that the proposed two-stage pipeline enables real-time prediction, with the ViT-based implementation achieving a strong balance between accuracy and computational efficiency. This makes MapViT a promising solution for energy- and resource-constrained platforms such as mobile robots. Moreover, the geometry foundation model derived from the self-supervised pre-training stage improves data efficiency and transferability, enabling effective downstream predictions even with limited labeled data. Overall, this work lays the foundation for next-generation digital twin ecosystems, and it paves the way for a new class of ML foundation models driving multi-modal intelligence in future 6G-enabled systems.

Prompting is central to interaction with AI systems, yet many users struggle to explore alternative directions, articulate creative intent, or understand how variations in prompts shape model outputs. We introduce prompt recommender systems (PRS) as an interaction approach that supports exploration, suggesting contextually relevant follow-up prompts. We present PromptHelper, a PRS prototype integrated into an AI chatbot that surfaces semantically diverse prompt suggestions while users work on real writing tasks. We evaluate PromptHelper in a 2x2 fully within-subjects study (N=32) across creative and academic writing tasks. Results show that PromptHelper significantly increases users' perceived exploration and expressiveness without increasing cognitive workload. Qualitative findings illustrate how prompt recommendations help users branch into new directions, overcome uncertainty about what to ask next, and better articulate their intent. We discuss implications for designing AI interfaces that scaffold exploratory interaction while preserving user agency, and release open-source resources to support research on prompt recommendation.

This article critically investigates the limitations of the simulated annealing algorithm using probabilistic bits (pSA) in solving large-scale combinatorial optimization problems. The study begins with an in-depth analysis of the pSA process, focusing on the issues resulting from unexpected oscillations among p-bits. These oscillations hinder the energy reduction of the Ising model and thus obstruct the successful execution of pSA in complex tasks. Through detailed simulations, we unravel the root cause of this energy stagnation, identifying the feedback mechanism inherent to the pSA operation as the primary contributor to these disruptive oscillations. To address this challenge, we propose two novel algorithms, time average pSA (TApSA) and stalled pSA (SpSA). These algorithms are designed based on partial deactivation of p-bits and are thoroughly tested using Python simulations on maximum cut benchmarks that are typical combinatorial optimization problems. On the 16 benchmarks from 800 to 5,000 nodes, the proposed methods improve the normalized cut value from 0.8% to 98.4% on average in comparison with the conventional pSA.

Clinical empathy is essential for patient care, but physicians need continually balance emotional warmth with factual precision under the cognitive and emotional constraints of clinical practice. This study investigates how large language models (LLMs) can function as empathy editors, refining physicians' written responses to enhance empathetic tone while preserving underlying medical information. More importantly, we introduce novel quantitative metrics, an Empathy Ranking Score and a MedFactChecking Score to systematically assess both emotional and factual quality of the responses. Experimental results show that LLM edited responses significantly increase perceived empathy while preserving factual accuracy compared with fully LLM generated outputs. These findings suggest that using LLMs as editorial assistants, rather than autonomous generators, offers a safer, more effective pathway to empathetic and trustworthy AI-assisted healthcare communication.

Large Language Models (LLMs) are increasingly used to generate and edit scientific abstracts, yet their integration into academic writing raises questions about trust, quality, and disclosure. Despite growing adoption, little is known about how readers perceive LLM-generated summaries and how these perceptions influence evaluations of scientific work. This paper presents a mixed-methods survey experiment investigating whether readers with ML expertise can distinguish between human- and LLM-generated abstracts, how actual and perceived LLM involvement affects judgments of quality and trustworthiness, and what orientations readers adopt toward AI-assisted writing. Our findings show that participants struggle to reliably identify LLM-generated content, yet their beliefs about LLM involvement significantly shape their evaluations. Notably, abstracts edited by LLMs are rated more favorably than those written solely by humans or LLMs. We also identify three distinct reader orientations toward LLM-assisted writing, offering insights into evolving norms and informing policy around disclosure and acceptable use in scientific communication.

We present BanditLP, a scalable multi-stakeholder contextual bandit framework that unifies neural Thompson Sampling for learning objective-specific outcomes with a large-scale linear program for constrained action selection at serving time. The methodology is application-agnostic, compatible with arbitrary neural architectures, and deployable at web scale, with an LP solver capable of handling billions of variables. Experiments on public benchmarks and synthetic data show consistent gains over strong baselines. We apply this approach in LinkedIn's email marketing system and demonstrate business win, illustrating the value of integrated exploration and constrained optimization in production.

Personalized learning systems have emerged as a promising approach to enhance student outcomes by tailoring educational content, pacing, and feedback to individual needs. However, most existing systems remain fragmented, specializing in either knowledge tracing, diagnostic modeling, or resource recommendation, but rarely integrating these components into a cohesive adaptive cycle. In this paper, we propose ALIGNAgent (Adaptive Learner Intelligence for Gap Identification and Next-step guidance), a multi-agent educational framework designed to deliver personalized learning through integrated knowledge estimation, skill-gap identification, and targeted resource recommendation.ALIGNAgent begins by processing student quiz performance, gradebook data, and learner preferences to generate topic-level proficiency estimates using a Skill Gap Agent that employs concept-level diagnostic reasoning to identify specific misconceptions and knowledge deficiencies. After identifying skill gaps, the Recommender Agent retrieves preference-aware learning materials aligned with diagnosed deficiencies, implementing a continuous feedback loop where interventions occur before advancing to subsequent topics. Extensive empirical evaluation on authentic datasets from two undergraduate computer science courses demonstrates ALIGNAgent's effectiveness, with GPT-4o-based agents achieving precision of 0.87-0.90 and F1 scores of 0.84-0.87 in knowledge proficiency estimation validated against actual exam performance.

Existing cultural commonsense benchmarks treat nations as monolithic, assuming uniform practices within national boundaries. But does cultural commonsense hold uniformly within a nation, or does it vary at the sub-national level? We introduce Indica, the first benchmark designed to test LLMs' ability to address this question, focusing on India - a nation of 28 states, 8 union territories, and 22 official languages. We collect human-annotated answers from five Indian regions (North, South, East, West, and Central) across 515 questions spanning 8 domains of everyday life, yielding 1,630 region-specific question-answer pairs. Strikingly, only 39.4% of questions elicit agreement across all five regions, demonstrating that cultural commonsense in India is predominantly regional, not national. We evaluate eight state-of-the-art LLMs and find two critical gaps: models achieve only 13.4%-20.9% accuracy on region-specific questions, and they exhibit geographic bias, over-selecting Central and North India as the "default" (selected 30-40% more often than expected) while under-representing East and West. Beyond India, our methodology provides a generalizable framework for evaluating cultural commonsense in any culturally heterogeneous nation, from question design grounded in anthropological taxonomy, to regional data collection, to bias measurement.

Generalizing Multimodal Large Language Models (MLLMs) to novel video domains is essential for real-world deployment but remains challenging due to the scarcity of labeled data. While In-Context Learning (ICL) offers a training-free adaptation path, standard methods rely on large annotated pools, which are often impractical in specialized environments like industrial or surgical settings since they require the experts' annotations. To bridge this gap, we introduce VIOLA (Video In-cOntext Learning with minimal Annotation), a label-efficient framework that synergizes minimal expert supervision with abundant unlabeled data. First, to maximize the efficiency of a strict annotation budget, we propose density-uncertainty-weighted sampling. Unlike standard diversity or uncertainty strategies that risk selecting visual outliers, our method leverages density estimation to identify samples that are simultaneously diverse, representative, and informative. Second, to utilize the remaining unlabeled data without noise propagation, we construct a hybrid pool and introduce confidence-aware retrieval and confidence-aware prompting. These mechanisms explicitly model label reliability, retrieving demonstrations based on a composite score of similarity and confidence while enabling the MLLM to adaptively distinguish between verified ground truths and noisy pseudo-labels. Extensive experiments across nine diverse benchmarks using four MLLMs demonstrate that our framework significantly outperforms various baselines in low-resource settings, achieving robust adaptation with minimal annotation costs.

Real-world scientific applications frequently encounter incomplete observational data due to sensor limitations, geographic constraints, or measurement costs. Although neural operators significantly advanced PDE solving in terms of computational efficiency and accuracy, their underlying assumption of fully-observed spatial inputs severely restricts applicability in real-world applications. We introduce the first systematic framework for learning neural operators from partial observation. We identify and formalize two fundamental obstacles: (i) the supervision gap in unobserved regions that prevents effective learning of physical correlations, and (ii) the dynamic spatial mismatch between incomplete inputs and complete solution fields. Specifically, our proposed Latent Autoregressive Neural Operator~(\ours) introduces two novel components designed explicitly to address the core difficulties of partial observations: (i) a mask-to-predict training strategy that creates artificial supervision by strategically masking observed regions, and (ii) a Physics-Aware Latent Propagator that reconstructs solutions through boundary-first autoregressive generation in latent space. Additionally, we develop POBench-PDE, a dedicated and comprehensive benchmark designed specifically for evaluating neural operators under partial observation conditions across three PDE-governed tasks. \ours achieves state-of-the-art performance with 18--69$\%$ relative L2 error reduction across all benchmarks under patch-wise missingness with less than 50$\%$ missing rate, including real-world climate prediction. Our approach effectively addresses practical scenarios involving up to 75$\%$ missing rate, to some extent bridging the existing gap between idealized research settings and the complexities of real-world scientific computing.

A key task in ML is to optimize models at various stages, e.g. by choosing hyperparameters or picking a stopping point. A traditional ML approach is to use validation loss, i.e. to apply the training loss function on a validation set to guide these optimizations. However, ML for healthcare has a distinct goal from traditional ML: Models must perform well relative to specific clinical requirements, vs. relative to the loss function used for training. These clinical requirements can be captured more precisely by tailored metrics. Since many optimization tasks do not require the driving metric to be differentiable, they allow a wider range of options, including the use of metrics tailored to be clinically-relevant. In this paper we describe two controlled experiments which show how the use of clinically-tailored metrics provide superior model optimization compared to validation loss, in the sense of better performance on the clinical task. The use of clinically-relevant metrics for optimization entails some extra effort, to define the metrics and to code them into the pipeline. But it can yield models that better meet the central goal of ML for healthcare: strong performance in the clinic.

Continual Test-Time Adaptation (CTTA) seeks to update a pretrained model during deployment using only the incoming, unlabeled data stream. Although prior approaches such as Tent, EATA etc. provide meaningful improvements under short evolving shifts, they struggle when the test distribution changes rapidly or over extremely long horizons. This challenge is exemplified by the CCC benchmark, where models operate over streams of 7.5M samples with continually changing corruption types and severities. We propose RDumb++, a principled extension of RDumb that introduces two drift-detection mechanisms i.e entropy-based drift scoring and KL-divergence drift scoring, together with adaptive reset strategies. These mechanisms allow the model to detect when accumulated adaptation becomes harmful and to recover before prediction collapse occurs. Across CCC-medium with three speeds and three seeds (nine runs, each containing one million samples), RDumb++ consistently surpasses RDumb, yielding approx 3% absolute accuracy gains while maintaining stable adaptation throughout the entire stream. Ablation experiments on drift thresholds and reset strengths further show that drift-aware resetting is essential for preventing collapse and achieving reliable long-horizon CTTA.

Deep learning models, particularly Transformers, are often criticized as "black boxes" and lack interpretability. We propose Prism, a white-box attention-based architecture derived from the principles of Maximizing Coding Rate Reduction ($\text{MCR}^2$). By modeling the attention mechanism as a gradient ascent process on a distinct signal-noise manifold, we introduce two physical constraints: an overcomplete dictionary to expand the representational phase space, and an irrational frequency separation ($π$-RoPE) to enforce incoherence between signal and noise subspaces. We demonstrate that these geometric inductive biases can be viewed as a physical constraint and they are sufficient to induce unsupervised functional disentanglement alone. Using TinyStories as a controlled testbed for verifying spectral dynamics, we observe that Prism spontaneously specializes its attention heads into spectrally distinct regimes: low-frequency heads capturing long-range causal dependencies (signal) and high-frequency heads handling local syntactic constraints (noise). Our results suggest that interpretability and performance are not a trade-off, but can be unified through principled geometric construction.

Early detection of malignant skin lesions is critical for improving patient outcomes in aggressive, metastatic skin cancers. This study evaluates a comprehensive system for preliminary skin lesion assessment that combines the clinically established ABCD rule of dermoscopy (analyzing Asymmetry, Borders, Color, and Dermoscopic Structures) with machine learning classification. Using a 1,000-image subset of the HAM10000 dataset, the system implements an automated, rule-based pipeline to compute a Total Dermoscopy Score (TDS) for each lesion. This handcrafted approach is compared against various machine learning solutions, including traditional classifiers (Logistic Regression, Random Forest, and SVM) and deep learning models. While the rule-based system provides high clinical interpretability, results indicate a performance bottleneck when reducing complex morphology to five numerical features. Experimental findings show that transfer learning with EfficientNet-B0 failed significantly due to domain shift between natural and medical images. In contrast, a custom three-layer Convolutional Neural Network (CNN) trained from scratch achieved 78.5% accuracy and 86.5% recall on median-filtered images, representing a 19-point accuracy improvement over traditional methods. The results demonstrate that direct pixel-level learning captures diagnostic patterns beyond handcrafted features and that purpose-built lightweight architectures can outperform large pretrained models for small, domain-specific medical datasets.

Machine unlearning aims to remove specific knowledge (e.g., copyrighted or private data) from a trained model without full retraining. In practice, models are often quantized (e.g., 4-bit) for deployment, but we find that quantization can catastrophically restore forgotten information [1]. In this paper, we (1) analyze why low-bit quantization undermines unlearning, and (2) propose a quantization-aware unlearning method to mitigate this. We first compute weight-change statistics and bucket overlaps in quantization to show that typical unlearning updates are too small to cross quantization thresholds. Building on this insight, we introduce a logits space hinge loss: for each forget example, we force the output logits of the unlearned model to differ from the original model by at least a margin (half the quantization step). This ensures forgotten examples remain distinguishable even after quantization. We evaluate on language and classification tasks (including a Twitter misinformation dataset) and show our method preserves forgetting under 4-bit quantization, whereas existing methods almost entirely recover the forgotten knowledge.

A world model is an AI system that simulates how an environment evolves under actions, enabling planning through imagined futures rather than reactive perception. Current world models, however, suffer from visual conflation: the mistaken assumption that high-fidelity video generation implies an understanding of physical and causal dynamics. We show that while modern models excel at predicting pixels, they frequently violate invariant constraints, fail under intervention, and break down in safety-critical decision-making. This survey argues that visual realism is an unreliable proxy for world understanding. Instead, effective world models must encode causal structure, respect domain-specific constraints, and remain stable over long horizons. We propose a reframing of world models as actionable simulators rather than visual engines, emphasizing structured 4D interfaces, constraint-aware dynamics, and closed-loop evaluation. Using medical decision-making as an epistemic stress test, where trial-and-error is impossible and errors are irreversible, we demonstrate that a world model's value is determined not by how realistic its rollouts appear, but by its ability to support counterfactual reasoning, intervention planning, and robust long-horizon foresight.

Alzheimer's disease (AD), defined as an abnormal buildup of amyloid plaques and tau tangles in the brain can be diagnosed with high accuracy based on protein biomarkers via PET or CSF analysis. However, due to the invasive nature of biomarker collection, most AD diagnoses are made in memory clinics using cognitive tests and evaluation of hippocampal atrophy based on MRI. While clinical assessment and hippocampal volume show high diagnostic accuracy for amnestic or typical AD (tAD), a substantial subgroup of AD patients with atypical presentation (atAD) are routinely misdiagnosed. To improve diagnosis of atAD patients, we propose a machine learning approach to distinguish between atAD and non-AD cognitive impairment using clinical testing battery and MRI data collected as standard-of-care. We develop and evaluate our approach using 1410 subjects across four groups (273 tAD, 184 atAD, 235 non-AD, and 685 cognitively normal) collected from one private data set and two public data sets from the National Alzheimer's Coordinating Center (NACC) and the Alzheimer's Disease Neuroimaging Initiative (ADNI). We perform multiple atAD vs. non-AD classification experiments using clinical features and hippocampal volume as well as a comprehensive set of MRI features from across the brain. The best performance is achieved by incorporating additional important MRI features, which outperforms using hippocampal volume alone. Furthermore, we use the Boruta statistical approach to identify and visualize significant brain regions distinguishing between diagnostic groups. Our ML approach improves the percentage of correctly diagnosed atAD cases (the recall) from 52% to 69% for NACC and from 34% to 77% for ADNI, while achieving high precision. The proposed approach has important implications for improving diagnostic accuracy for non-amnestic atAD in clinical settings using only clinical testing battery and MRI.

Large Language Model (LLM)-based question-answering systems offer significant potential for automating customer support and internal knowledge access in small businesses, yet their practical deployment remains challenging due to infrastructure costs, engineering complexity, and security risks, particularly in retrieval-augmented generation (RAG)-based settings. This paper presents an industry case study of an open-source, multi-tenant platform that enables small businesses to deploy customised LLM-based support chatbots via a no-code workflow. The platform is built on distributed, lightweight k3s clusters spanning heterogeneous, low-cost machines and interconnected through an encrypted overlay network, enabling cost-efficient resource pooling while enforcing container-based isolation and per-tenant data access controls. In addition, the platform integrates practical, platform-level defences against prompt injection attacks in RAG-based chatbots, translating insights from recent prompt injection research into deployable security mechanisms without requiring model retraining or enterprise-scale infrastructure. We evaluate the proposed platform through a real-world e-commerce deployment, demonstrating that secure and efficient LLM-based chatbot services can be achieved under realistic cost, operational, and security constraints faced by small businesses.

Accurate prediction of traffic crash severity is critical for improving emergency response and public safety planning. Although recent large language models (LLMs) exhibit strong reasoning capabilities, their single-agent architectures often struggle with heterogeneous, domain-specific crash data and tend to generate biased or unstable predictions. To address these limitations, this paper proposes TransportAgents, a hybrid multi-agent framework that integrates category-specific LLM reasoning with a multilayer perceptron (MLP) integration module. Each specialized agent focuses on a particular subset of traffic information, such as demographics, environmental context, or incident details, to produce intermediate severity assessments that are subsequently fused into a unified prediction. Extensive experiments on two complementary U.S. datasets, the Consumer Product Safety Risk Management System (CPSRMS) and the National Electronic Injury Surveillance System (NEISS), demonstrate that TransportAgents consistently outperforms both traditional machine learning and advanced LLM-based baselines. Across three representative backbones, including closed-source models such as GPT-3.5 and GPT-4o, as well as open-source models such as LLaMA-3.3, the framework exhibits strong robustness, scalability, and cross-dataset generalizability. A supplementary distributional analysis further shows that TransportAgents produces more balanced and well-calibrated severity predictions than standard single-agent LLM approaches, highlighting its interpretability and reliability for safety-critical decision support applications.

We develop a two-stage retrieval system that combines multiple complementary retrieval methods with a learned reranker and LLM-based reranking, to address the TREC Tip-of-the-Tongue (ToT) task. In the first stage, we employ hybrid retrieval that merges LLM-based retrieval, sparse (BM25), and dense (BGE-M3) retrieval methods. We also introduce topic-aware multi-index dense retrieval that partitions the Wikipedia corpus into 24 topical domains. In the second stage, we evaluate both a trained LambdaMART reranker and LLM-based reranking. To support model training, we generate 5000 synthetic ToT queries using LLMs. Our best system achieves recall of 0.66 and NDCG@1000 of 0.41 on the test set by combining hybrid retrieval with Gemini-2.5-flash reranking, demonstrating the effectiveness of fusion retrieval.

The proliferation of XR devices has made egocentric hand pose estimation a vital task, yet this perspective is inherently challenged by frequent finger occlusions. To address this, we propose a novel approach that leverages the rich information in dorsal hand skin deformation, unlocked by recent advances in dense visual featurizers. We introduce a dual-stream delta encoder that learns pose by contrasting features from a dynamic hand with a baseline relaxed position. Our evaluation demonstrates that, using only cropped dorsal images, our method reduces the Mean Per Joint Angle Error (MPJAE) by 18% in self-occluded scenarios (fingers >=50% occluded) compared to state-of-the-art techniques that depend on the whole hand's geometry and large model backbones. Consequently, our method not only enhances the reliability of downstream tasks like index finger pinch and tap estimation in occluded scenarios but also unlocks new interaction paradigms, such as detecting isometric force for a surface "click" without visible movement while minimizing model size.

Hallucination in large language models (LLMs) remains an acute concern, contributing to the spread of misinformation and diminished public trust, particularly in high-risk domains. Among hallucination types, factuality is crucial, as it concerns a model's alignment with established world knowledge. Adversarial factuality, defined as the deliberate insertion of misinformation into prompts with varying levels of expressed confidence, tests a model's ability to detect and resist confidently framed falsehoods. Existing work lacks high-quality, domain-specific resources for assessing model robustness under such adversarial conditions, and no prior research has examined the impact of injected misinformation on long-form text factuality. To address this gap, we introduce AdversaRiskQA, the first verified and reliable benchmark systematically evaluating adversarial factuality across Health, Finance, and Law. The benchmark includes two difficulty levels to test LLMs' defensive capabilities across varying knowledge depths. We propose two automated methods for evaluating the adversarial attack success and long-form factuality. We evaluate six open- and closed-source LLMs from the Qwen, GPT-OSS, and GPT families, measuring misinformation detection rates. Long-form factuality is assessed on Qwen3 (30B) under both baseline and adversarial conditions. Results show that after excluding meaningless responses, Qwen3 (80B) achieves the highest average accuracy, while GPT-5 maintains consistently high accuracy. Performance scales non-linearly with model size, varies by domains, and gaps between difficulty levels narrow as models grow. Long-form evaluation reveals no significant correlation between injected misinformation and the model's factual output. AdversaRiskQA provides a valuable benchmark for pinpointing LLM weaknesses and developing more reliable models for high-stakes applications.

The use of Transfer Learning & Transformers has steadily improved accuracy and has significantly contributed in solving complex computation problems. However, this transformer led accuracy improvement in Applied AI Analytics specifically in sentiment analytics comes with the dark side. It is observed during experiments that a lot of these improvements in transformer led accuracy of one class of sentiment has been at the cost of polarization of another class of sentiment and the failing of neutrality. This lack of neutrality poses an acute problem in the Applied NLP space, which relies heavily on the computational outputs of sentiment analytics for reliable industry ready tasks.

Characters in novels have typically been modeled based on their presence in scenes in narrative, considering aspects like their actions, named mentions, and dialogue. This conception of character places significant emphasis on the main character who is present in the most scenes. In this work, we instead adopt a framing developed from a new literary theory proposing a six-component structural model of character. This model enables a comprehensive approach to character that accounts for the narrator-character distinction and includes a component neglected by prior methods, discussion by other characters. We compare general-purpose LLMs with task-specific transformers for operationalizing this model of character on major 19th-century British realist novels. Our methods yield both component-level and graph representations of character discussion. We then demonstrate that these representations allow us to approach literary questions at scale from a new computational lens. Specifically, we explore Woloch's classic "the one vs the many" theory of character centrality and the gendered dynamics of character discussion.

Drawing inspiration from recent findings including surprisingly decent performance of transformers without positional encoding (NoPE) in the domain of language models and how registers (additional throwaway tokens not tied to input) may improve the performance of large vision transformers (ViTs), we invent and test a variant of ViT called fractal ViT that breaks permutation invariance among the tokens by applying an attention mask between the regular tokens and ``summary tokens'' similar to registers, in isolation or in combination with various positional encodings. These models do not improve upon ViT with registers, highlighting the fact that these findings may be scale, domain, or application-specific.

Spatial transcriptomics enables spatial gene expression profiling, motivating computational models that capture spatially conditioned regulatory relationships. We introduce SAGE-FM, a lightweight spatial transcriptomics foundation model based on graph convolutional networks (GCNs) trained with a masked central spot prediction objective. Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that robustly recover masked genes, with 91% of masked genes showing significant correlations (p < 0.05). The embeddings generated by SAGE-FM outperform MOFA and existing spatial transcriptomics methods in unsupervised clustering and preservation of biological heterogeneity. SAGE-FM generalizes to downstream tasks, enabling 81% accuracy in pathologist-defined spot annotation in oropharyngeal squamous cell carcinoma and improving glioblastoma subtype prediction relative to MOFA. In silico perturbation experiments further demonstrate that the model captures directional ligand-receptor and upstream-downstream regulatory effects consistent with ground truth. These results demonstrate that simple, parameter-efficient GCNs can serve as biologically interpretable and spatially aware foundation models for large-scale spatial transcriptomics.

Volunteer-led lake monitoring yields irregular, seasonal time series with many gaps arising from ice cover, weather-related access constraints, and occasional human errors, complicating forecasting and early warning of harmful algal blooms. We study Secchi Disk Depth (SDD) forecasting on a 30-lake, data-rich subset drawn from three decades of in situ records collected across Maine lakes. Missingness is handled via Multiple Imputation by Chained Equations (MICE), and we evaluate performance with a normalized Mean Absolute Error (nMAE) metric for cross-lake comparability. Among six candidates, ridge regression provides the best mean test performance. Using ridge regression, we then quantify the minimal sample size, showing that under a backward, recent-history protocol, the model reaches within 5% of full-history accuracy with approximately 176 training samples per lake on average. We also identify a minimal feature set, where a compact four-feature subset matches the thirteen-feature baseline within the same 5% tolerance. Bringing these results together, we introduce a joint feasibility function that identifies the minimal training history and fewest predictors sufficient to achieve the target of staying within 5% of the complete-history, full-feature baseline. In our study, meeting the 5% accuracy target required about 64 recent samples and just one predictor per lake, highlighting the practicality of targeted monitoring. Hence, our joint feasibility strategy unifies recent-history length and feature choice under a fixed accuracy target, yielding a simple, efficient rule for setting sampling effort and measurement priorities for lake researchers.

In recent years, Rectified flow (RF) has gained considerable popularity largely due to its generation efficiency and state-of-the-art performance. In this paper, we investigate the degree to which RF automatically adapts to the intrinsic low dimensionality of the support of the target distribution to accelerate sampling. We show that, using a carefully designed choice of the time-discretization scheme and with sufficiently accurate drift estimates, the RF sampler enjoys an iteration complexity of order $O(k/\varepsilon)$ (up to log factors), where $\varepsilon$ is the precision in total variation distance and $k$ is the intrinsic dimension of the target distribution. In addition, we show that the denoising diffusion probabilistic model (DDPM) procedure is equivalent to a stochastic version of RF by establishing a novel connection between these processes and stochastic localization. Building on this connection, we further design a stochastic RF sampler that also adapts to the low-dimensionality of the target distribution under milder requirements on the accuracy of the drift estimates, and also with a specific time schedule. We illustrate with simulations on the synthetic data and text-to-image data experiments the improved performance of the proposed samplers implementing the newly designed time-discretization schedules.

Speculative Decoding (SD) accelerates autoregressive large language model (LLM) inference by decoupling generation and verification. While recent methods improve draft quality by tightly coupling the drafter with the target model, the verification mechanism itself remains largely unchanged, relying on strict token-level rejection sampling. In practice, modern LLMs frequently operate in low-margin regimes where the target model exhibits weak preference among top candidates. In such cases, rejecting plausible runner-up tokens yields negligible information gain while incurring substantial rollback cost, leading to a fundamental inefficiency in verification. We propose Margin-Aware Speculative Verification, a training-free and domain-agnostic verification strategy that adapts to the target model's local decisiveness. Our method conditions verification on decision stability measured directly from the target logits and relaxes rejection only when strict verification provides minimal benefit. Importantly, the approach modifies only the verification rule and is fully compatible with existing target-coupled speculative decoding frameworks. Extensive experiments across model scales ranging from 8B to 235B demonstrate that our method delivers consistent and significant inference speedups over state-of-the-art baselines while preserving generation quality across diverse benchmarks.

A common solution for mitigating outdated or incorrect information in Large Language Models (LLMs) is to provide updated facts in-context or through knowledge editing. However, these methods introduce knowledge conflicts when the knowledge update fails to overwrite the model's parametric knowledge, which propagate to faulty reasoning. Current benchmarks for this problem, however, largely focus only on single knowledge updates and fact recall without evaluating how these updates affect downstream reasoning. In this work, we introduce TRACK (Testing Reasoning Amid Conflicting Knowledge), a new benchmark for studying how LLMs propagate new knowledge through multi-step reasoning when it conflicts with the model's initial parametric knowledge. Spanning three reasoning-intensive scenarios (WIKI, CODE, and MATH), TRACK introduces multiple, realistic conflicts to mirror real-world complexity. Our results on TRACK reveal that providing updated facts to models for reasoning can worsen performance compared to providing no updated facts to a model, and that this performance degradation exacerbates as more updated facts are provided. We show this failure stems from both inability to faithfully integrate updated facts, but also flawed reasoning even when knowledge is integrated. TRACK provides a rigorous new benchmark to measure and guide future progress on propagating conflicting knowledge in multi-step reasoning.

Deep Learning (DL) libraries (e.g., PyTorch) are popular in AI development. These libraries are complex and contain bugs. Researchers have proposed various bug-finding techniques for such libraries. Yet, there is much room for improvement. A key challenge in testing DL libraries is the lack of API specifications. Prior testing approaches often inaccurately model the input specifications of DL APIs, resulting in missed valid inputs that could reveal bugs or false alarms due to invalid inputs. To address this challenge, we develop Centaur -- the first neurosymbolic technique to test DL library APIs using dynamically learned input constraints. Centaur leverages the key idea that formal API constraints can be learned from a small number of automatically generated seed inputs, and that the learned constraints can be solved using SMT solvers to generate valid and diverse test inputs. We develop a novel grammar that represents first-order logic formulae over API parameters and expresses tensor-related properties (e.g., shape, data types) as well as relational properties between parameters. We use the grammar to guide a Large Language Model (LLM) to enumerate syntactically correct candidate rules, validated using seed inputs. Further, we develop a custom refinement strategy to prune the set of learned rules to eliminate spurious or redundant rules. We use the learned constraints to systematically generate valid and diverse inputs by integrating SMT-based solving with randomized sampling. We evaluate Centaur for testing PyTorch and TensorFlow. Our results show that Centaur's constraints have a recall of 94.0% and a precision of 94.0% on average. In terms of coverage, Centaur covers 203, 150, and 9,608 more branches than TitanFuzz, ACETest and Pathfinder, respectively. Using Centaur, we also detect 26 new bugs in PyTorch and TensorFlow, 18 of which are confirmed.

Large Language Models (LLMs) exhibit significant social biases that can perpetuate harmful stereotypes and unfair outcomes. In this paper, we propose Multi-Persona Thinking (MPT), a novel inference-time framework that leverages dialectical reasoning from multiple perspectives to reduce bias. MPT guides models to adopt contrasting social identities (e.g., male and female) along with a neutral viewpoint, and then engages these personas iteratively to expose and correct biases. Through a dialectical reasoning process, the framework transforms the potential weakness of persona assignment into a strength for bias mitigation. We evaluate MPT on two widely used bias benchmarks across both open-source and closed-source models of varying scales. Our results demonstrate substantial improvements over existing prompting-based strategies: MPT achieves the lowest bias while maintaining core reasoning ability.

The rapid evolution of Retrieval-Augmented Generation (RAG) toward multimodal, high-stakes enterprise applications has outpaced the development of domain specific evaluation benchmarks. Existing datasets often rely on general-domain corpora or purely textual retrieval, failing to capture the complexity of specialized technical documents where information is inextricably multimodal and reasoning requires synthesizing disjoint evidence. We address this gap by introducing MiRAGE, a Multiagent framework for RAG systems Evaluation, that leverages a collaborative swarm of specialized agents to generate verified, domain-specific, multimodal, and multi-hop Question-Answer datasets. MiRAGE orchestrates a swarm of specialized agents: a recursive context optimization loop to aggregate scattered evidence, an adversarial verifier agent to guarantee factual grounding, and an agent to recognize the expert persona and the relevant domain to mimic expert cognitive workflows. Extensive empirical evaluation across four distinct domains (regulations, finance, quantitative biology, and journalism) demonstrates that MiRAGE generates datasets with significantly higher reasoning complexity (>2.3 average hops) and factual faithfulness. Our ablation studies point that MiRAGE can be powered by LLMs if textual descriptions of the images are available. Visual grounding still remains a frontier. By automating the creation of gold standard evaluation datasets that reflect the latent thematic structure of proprietary corpora, MiRAGE provides the necessary infrastructure to rigorously benchmark the next generation information retrieval systems.

Federal research funding shapes the direction, diversity, and impact of the US scientific enterprise. Large language models (LLMs) are rapidly diffusing into scientific practice, holding substantial promise while raising widespread concerns. Despite growing attention to AI use in scientific writing and evaluation, little is known about how the rise of LLMs is reshaping the public funding landscape. Here, we examine LLM involvement at key stages of the federal funding pipeline by combining two complementary data sources: confidential National Science Foundation (NSF) and National Institutes of Health (NIH) proposal submissions from two large US R1 universities, including funded, unfunded, and pending proposals, and the full population of publicly released NSF and NIH awards. We find that LLM use rises sharply beginning in 2023 and exhibits a bimodal distribution, indicating a clear split between minimal and substantive use. Across both private submissions and public awards, higher LLM involvement is consistently associated with lower semantic distinctiveness, positioning projects closer to recently funded work within the same agency. The consequences of this shift are agency-dependent. LLM use is positively associated with proposal success and higher subsequent publication output at NIH, whereas no comparable associations are observed at NSF. Notably, the productivity gains at NIH are concentrated in non-hit papers rather than the most highly cited work. Together, these findings provide large-scale evidence that the rise of LLMs is reshaping how scientific ideas are positioned, selected, and translated into publicly funded research, with implications for portfolio governance, research diversity, and the long-run impact of science.

Online encyclopedias are central to contemporary information infrastructures and have become focal points of debates over ideological bias. Wikipedia, in particular, has long been accused of left-leaning bias, while Grokipedia, an AI-generated encyclopedia launched by xAI, has been framed as a right-leaning alternative. This paper presents a comparative analysis of Wikipedia and Grokipedia on well-established politically contested topics. Specifically, we examine differences in semantic framing, political orientation, and content prioritization. We find that semantic similarity between the two platforms decays across article sections and diverges more strongly on controversial topics than on randomly sampled ones. Additionally, we show that both encyclopedias predominantly exhibit left-leaning framings, although Grokipedia exhibits a more bimodal distribution with increased prominence of right-leaning content. The experimental code is publicly available.

Standard autoregressive decoding in large language models (LLMs) is inherently short-sighted, often failing to find globally optimal reasoning paths due to its token-by-token generation process. While inference-time strategies like foresight sampling attempt to mitigate this by simulating future steps, they typically rely on ad-hoc heuristics for valuing paths and pruning the search space. This paper introduces Martingale Foresight Sampling (MFS), a principled framework that reformulates LLM decoding as a problem of identifying an optimal stochastic process. By modeling the quality of a reasoning path as a stochastic process, we leverage Martingale theory to design a theoretically-grounded algorithm. Our approach replaces heuristic mechanisms with principles from probability theory: step valuation is derived from the Doob Decomposition Theorem to measure a path's predictable advantage, path selection uses Optional Stopping Theory for principled pruning of suboptimal candidates, and an adaptive stopping rule based on the Martingale Convergence Theorem terminates exploration once a path's quality has provably converged. Experiments on six reasoning benchmarks demonstrate that MFS surpasses state-of-the-art methods in accuracy while significantly improving computational efficiency. Code will be released at https://github.com/miraclehetech/EACL2026-Martingale-Foresight-Sampling.

Emergency Department overcrowding is a critical issue that compromises patient safety and operational efficiency, necessitating accurate demand forecasting for effective resource allocation. This study evaluates and compares three distinct predictive models: Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX), EXtreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks for forecasting daily ED arrivals over a seven-day horizon. Utilizing data from an Australian tertiary referral hospital spanning January 2017 to December 2021, this research distinguishes itself by decomposing demand into eight specific ward categories and stratifying patients by clinical complexity. To address data distortions caused by the COVID-19 pandemic, the study employs the Prophet model to generate synthetic counterfactual values for the anomalous period. Experimental results demonstrate that all three proposed models consistently outperform a seasonal naive baseline. XGBoost demonstrated the highest accuracy for predicting total daily admissions with a Mean Absolute Error of 6.63, while the statistical SARIMAX model proved marginally superior for forecasting major complexity cases with an MAE of 3.77. The study concludes that while these techniques successfully reproduce regular day-to-day patterns, they share a common limitation in underestimating sudden, infrequent surges in patient volume.

The safe deployment of large language models (LLMs) in high-stakes fields like biomedicine, requires them to be able to reason about cause and effect. We investigate this ability by testing 13 open-source LLMs on a fundamental task: pairwise causal discovery (PCD) from text. Our benchmark, using 12 diverse datasets, evaluates two core skills: 1) \textbf{Causal Detection} (identifying if a text contains a causal link) and 2) \textbf{Causal Extraction} (pulling out the exact cause and effect phrases). We tested various prompting methods, from simple instructions (zero-shot) to more complex strategies like Chain-of-Thought (CoT) and Few-shot In-Context Learning (FICL). The results show major deficiencies in current models. The best model for detection, DeepSeek-R1-Distill-Llama-70B, only achieved a mean score of 49.57\% ($C_{detect}$), while the best for extraction, Qwen2.5-Coder-32B-Instruct, reached just 47.12\% ($C_{extract}$). Models performed best on simple, explicit, single-sentence relations. However, performance plummeted for more difficult (and realistic) cases, such as implicit relationships, links spanning multiple sentences, and texts containing multiple causal pairs. We provide a unified evaluation framework, built on a dataset validated with high inter-annotator agreement ($κ\ge 0.758$), and make all our data, code, and prompts publicly available to spur further research. \href{https://github.com/sydneyanuyah/CausalDiscovery}{Code available here: https://github.com/sydneyanuyah/CausalDiscovery}

This paper examines how to make large language models reliable for high-stakes legal work by reducing hallucinations. It distinguishes three AI paradigms: (1) standalone generative models ("creative oracle"), (2) basic retrieval-augmented systems ("expert archivist"), and (3) an advanced, end-to-end optimized RAG system ("rigorous archivist"). The authors introduce two reliability metrics -False Citation Rate (FCR) and Fabricated Fact Rate (FFR)- and evaluate 2,700 judicial-style answers from 12 LLMs across 75 legal tasks using expert, double-blind review. Results show that standalone models are unsuitable for professional use (FCR above 30%), while basic RAG greatly reduces errors but still leaves notable misgrounding. Advanced RAG, using techniques such as embedding fine-tuning, re-ranking, and self-correction, reduces fabrication to negligible levels (below 0.2%). The study concludes that trustworthy legal AI requires rigor-focused, retrieval-based architectures emphasizing verification and traceability, and provides an evaluation framework applicable to other high-risk domains.

Graph neural network (GNN) have demonstrated exceptional performance in solving critical problems across diverse domains yet remain susceptible to backdoor attacks. Existing studies on backdoor attack for graph classification are limited to single target attack using subgraph replacement based mechanism where the attacker implants only one trigger into the GNN model. In this paper, we introduce the first multi-targeted backdoor attack for graph classification task, where multiple triggers simultaneously redirect predictions to different target labels. Instead of subgraph replacement, we propose subgraph injection which preserves the structure of the original graphs while poisoning the clean graphs. Extensive experiments demonstrate the efficacy of our approach, where our attack achieves high attack success rates for all target labels with minimal impact on the clean accuracy. Experimental results on five dataset demonstrate the superior performance of our attack framework compared to the conventional subgraph replacement-based attack. Our analysis on four GNN models confirms the generalization capability of our attack which is effective regardless of the GNN model architectures and training parameters settings. We further investigate the impact of the attack design parameters including injection methods, number of connections, trigger sizes, trigger edge density and poisoning ratios. Additionally, our evaluation against state-of-the-art defenses (randomized smoothing and fine-pruning) demonstrates the robustness of our proposed multi-target attacks. This work highlights the GNN vulnerability against multi-targeted backdoor attack in graph classification task. Our source codes will be available at https://github.com/SiSL-URI/Multi-Targeted-Graph-Backdoor-Attack.

Training modern deep learning models is increasingly constrained by GPU memory and compute limits. While Randomized Numerical Linear Algebra (RandNLA) offers proven techniques to compress these models, the lack of a unified, production-grade library prevents widely adopting these methods. We present Panther, a PyTorch-compatible library that consolidates established RandNLA algorithms into a single high-performance framework. Panther engineers efficient, drop-in replacements for standard components including sketched linear layers, 2D convolution, multi-head attention, and randomized matrix decompositions (such as pivoted CholeskyQR). By implementing a custom C++/CUDA backend (pawX), Panther provides an optimized implementation that can run on both CPUs and GPUs. We demonstrate the effectiveness of RandNLA techniques and Panther's ease of adoption. By replacing standard PyTorch linear layers with Panther layers (requiring only a few lines of code) we achieve significant memory savings (up to 75%) on BERT while maintaining comparable loss. Source code is available (MIT License) at https://github.com/FahdSeddik/panther, along with demonstration video at https://youtu.be/7M3RQb4KWxs.

The prevalence and low cost of LLMs have led to a rise of synthetic content. From review sites to court documents, ``natural'' content has been contaminated by data points that appear similar to natural data, but are in fact LLM-generated. In this work we revisit fundamental learning theory questions in this, now ubiquitous, setting. We model this scenario as a sequence of learning tasks where the input is a mix of natural and synthetic data, and the learning algorithms are oblivious to the origin of any individual example. We study the possibilities and limitations of ERM in this setting. For the problem of estimating the mean of an arbitrary $d$-dimensional distribution, we find that while ERM converges to the true mean, it is outperformed by an algorithm that assigns non-uniform weights to examples from different generations of data. For the PAC learning setting, the disparity is even more stark. We find that ERM does not always converge to the true concept, echoing the model collapse literature. However, we show there are algorithms capable of learning the correct hypothesis for arbitrary VC classes and arbitrary amounts of contamination.

The integration of Large Language Models (LLMs) into the public health policy sector offers a transformative approach to navigating the vast repositories of regulatory guidance maintained by agencies such as the Centers for Disease Control and Prevention (CDC). However, the propensity for LLMs to generate hallucinations, defined as plausible but factually incorrect assertions, presents a critical barrier to the adoption of these technologies in high-stakes environments where information integrity is non-negotiable. This empirical evaluation explores the effectiveness of Retrieval-Augmented Generation (RAG) architectures in mitigating these risks by grounding generative outputs in authoritative document context. Specifically, this study compares a baseline Vanilla LLM against Basic RAG and Advanced RAG pipelines utilizing cross-encoder re-ranking. The experimental framework employs a Mistral-7B-Instruct-v0.2 model and an all-MiniLM-L6-v2 embedding model to process a corpus of official CDC policy analytical frameworks and guidance documents. The analysis measures the impact of two distinct chunking strategies, recursive character-based and token-based semantic splitting, on system accuracy, measured through faithfulness and relevance scores across a curated set of complex policy scenarios. Quantitative findings indicate that while Basic RAG architectures provide a substantial improvement in faithfulness (0.621) over Vanilla baselines (0.347), the Advanced RAG configuration achieves a superior faithfulness average of 0.797. These results demonstrate that two-stage retrieval mechanisms are essential for achieving the precision required for domain-specific policy question answering, though structural constraints in document segmentation remain a significant bottleneck for multi-step reasoning tasks.

Reflexive Thematic Analysis (RTA) is a critical method for generating deep interpretive insights. Yet its core tenets, including researcher reflexivity, tangible analytical evolution, and productive disagreement, are often poorly supported by software tools that prioritize speed and consensus over interpretive depth. To address this gap, we introduce Reflexis, a collaborative workspace that centers these practices. It supports reflexivity by integrating in-situ reflection prompts, makes code evolution transparent and tangible, and scaffolds collaborative interpretation by turning differences into productive, positionality-aware dialogue. Results from our paired-analyst study (N=12) indicate that Reflexis encouraged participants toward more granular reflection and reframed disagreements as productive conversations. The evaluation also surfaced key design tensions, including a desire for higher-level, networked memos and more user control over the timing of proactive alerts. Reflexis contributes a design framework for tools that prioritize rigor and transparency to support deep, collaborative interpretation in an age of automation.

We theoretically predict that Dzyaloshinskii-Moriya interaction can induce magnon equilibrium spin current in collinear antiferromagnets. Such a current, being a response to the effective magnon vector potential, can be considered as magnon analog of the superconducting supercurrent or the persistent current. Large amplitude of the predicted effect may compensate for the smallness of the Dzyaloshinskii-Moriya interaction, making the equilibrium spin currents to be experimentally observed. We suggest that external electric field can play the role of effective flux magnons interact with and propose an experiment based on the interference of magnons in the ring geometry as a verification of the concept.

Political moderation, a key attractor in democratic systems, proves highly fragile under realistic information conditions. We develop a stochastic model of opinion dynamics to analyze how noise and differential susceptibility reshape the political spectrum. Extending Marvel et al.'s deterministic framework, we incorporate stochastic media influence $ζ(t)$ and neuropolitically-grounded sensitivity differences ($σ_y > σ_x$). Analysis reveals the moderate population -- stable in deterministic models -- undergoes catastrophic collapse under stochastic forcing. This occurs through an effective deradicalization asymmetry ($u_{B}^{\text{eff}} = u + σ_y^2/2 > u_{A}^{\text{eff}}$) that drives conservatives to extinction, eliminating cross-cutting interactions that sustain moderates. The system exhibits a phase transition from multi-stable coexistence to liberal dominance, demonstrating how information flow architecture -- independent of content -- systematically dismantles the political center. Our findings reveal moderation as an emergent property highly vulnerable to stochastic perturbations in complex social systems.

Probing energy-dependent transport in quantum simulators requires preparing states with tunable energy and small energy variance. Existing approaches often study quench dynamics of simple initial states, such as computational basis states, which are far from energy eigenstates and therefore limit the achievable energy resolution. In this work, we propose using wavepackets to probe transport properties with improved energy resolution. To demonstrate the utility of this approach, we prepare and evolve wavepackets on Quantinuum's H2-2 quantum computer and identify an energy-dependent localization transition in the Anderson model on an 8x7 lattice--a finite-size mobility edge. We observe that a wavepacket initialized at low energy remains spatially localized under time evolution, while a high-energy wavepacket delocalizes, consistent with the presence of a mobility edge. Crucial to our experiments is an error mitigation strategy that infers the noiseless output bit string distribution using maximum-likelihood estimation. Compared to post-selection, this method removes systematic errors and reduces statistical uncertainty by up to a factor of 5. We extend our methods to the many-particle regime by developing a quantum algorithm for preparing quasiparticle wavepackets in a one-dimensional model of interacting fermions. This technique has modest quantum resource requirements, making wavepacket-based studies of transport in many-body systems a promising application for near-term quantum computers.

Understanding how to analyze highly entangled thermal eigenstates is a central challenge in the study of quantum many-body systems. In this Letter, we introduce a stabilizer-based approach to construct analytically tractable energy eigenstates of nonintegrable many-body Hamiltonians. Focusing on zero-energy eigenstates at infinite temperature, we prove a sharp no-go theorem: stabilizer eigenstates of two-body Hamiltonians cannot satisfy $k$-body microscopic thermal equilibrium for any $k\ge4$. We further show that this bound is tight by explicitly constructing two-body nonintegrable Hamiltonians whose stabilizer eigenstates reproduce thermal expectation values for all two-body and all three-body observables. Finally, we identify the structural origin of this limitation by characterizing the conditions under which a stabilizer state can appear as a zero-energy eigenstate of a Hamiltonian, thereby revealing a fundamental constraint imposed by the few-body nature of interactions.

Saturation of cumulative observables is widely observed in systems with propagating or spreading signals and is commonly modeled using system-specific mechanisms such as scattering statistics, coherence functions, or phenomenological decay laws. This work shows that such saturation follows directly from linear local relaxation alone. Any linear observable accumulated over the lifetime of a relaxing signal is bounded by a scale set by the relaxation time, independent of geometry, dimensionality, or microscopic dynamics. When relaxation is mapped to space through transport or spreading, this temporal bound yields a corresponding spatial saturation scale. A closed-form expression reveals a two-regime behavior: linear growth at short times followed by saturation beyond the relaxation time. The result provides a minimal and unified explanation for cumulative saturation across transport, diffusive, and stochastic systems.

In combinatorial optimization problems with degenerate ground states, fair sampling of degenerate solutions is essential. However, the quantum approximate optimization algorithm (QAOA) with a standard transverse-field mixer induces biases among degenerate states as circuit depth increases. Based on quantum-classical correspondence theory, we propose SBO-QAOA, which employs a temperature-dependent Hamiltonian encoding a Gibbs distribution as its ground state. Numerical simulations show that, unlike standard QAOA, SBO-QAOA yields ground-state probabilities converging to finite-temperature values with uniform distribution among degenerate states. These fairness and temperature-targeting properties are preserved even with only four variational parameters under a linear schedule.

Consistent with prior qualitative expectations for group IV-VI topological crystalline insulators, this work demonstrates, based on band structure and Chern number calculations, that Pb$_{1-x}$Sn$_x$Se/(PbSe)$_{1-y}$(EuS)$_y$ quantum wells constitute a promising and viable platform for realizing a variety of quantum anomalous Hall phases. The proposed basis transformation procedure for the multiband $\mathit{k} \cdot \mathit{p}$ Hamiltonian enables the treatment of wells grown along arbitrary crystallographic directions while explicitly accounting for the anisotropy of the material's isoenergetic surfaces. Numerical studies of $\langle 111\rangle$-, $\langle 110\rangle$- and $\langle 001\rangle$-oriented quantum wells predict attainable Chern numbers with magnitudes ranging from $1$ to $4$, depending on the quantum well width, Sn content, and relative orientation of the four projected $\mathrm{L}$ valleys with respect to the growth direction. The results further indicate that appropriate strain compensation is required to achieve high-quality quantization of the Hall conductance.

In this work, we combine coarse-grained Brownian dynamics simulations and mean-field theory to study supercoiling dynamics, as well as the steady-state profiles of twist and writhe, in an open DNA polymer where one of the free ends is subjected to a constant torque. Even though the other end is free, and hence can spin and release torsional stress, we observe that the entire chain transitions between a swollen and a plectonemic phase as the torque increases beyond a critical threshold. In the plectonemic phase, we observe a non-linear twist profile in the steady state, resulting from the mutual interconversion between the injected twist and geometrical writhe, which distributes inhomogeneously along the chain. We also show that the non-equilibrium dynamics of twist accumulation is diffusive, and that writhe diffusion is negligible in this geometry, as plectonemes remain localised near the end that is being rotated. We discuss the feasibility of testing our results with single-molecule experiments.

Hidden Markov models (HMMs) are ubiquitous in time-series modelling, with applications ranging from chemical reaction modelling to speech recognition. These HMMs are often large, with high-dimensional memories. A recently-proposed application of quantum technologies is to execute quantum analogues of HMMs. Such quantum HMMs (QHMMs) are strictly more expressive than their classical counterparts, enabling the construction of more parsimonious models of stochastic processes. However, state-of-the-art techniques for QHMM compression, based on tensor networks, are only applicable for a restricted subset of HMMs, where the transitions are deterministic. In this work we introduce a pipeline by which \emph{any} finite, ergodic HMM can be compressed in this manner, providing a route for effective quantum dimension reduction of general HMMs. We demonstrate the method on both a simple toy model, and on a speech-derived HMM trained from data, obtaining favourable memory--accuracy trade-offs compared to classical compression approaches.

We discuss the statistics of additive thermal (internal) noise in systems governed by the generalized Langevin equation with linear dissipation. To assess the equation's validity, it is common to assume that the system is ergodic and to verify that solutions approach correct equilibrium values at asymptotically long times. In this paper, we instead consider the consistency of the generalized Langevin equation with the Jarzynski equality at finite times and do not assume the system's ergodicity. Specifically, we consider a classical Brownian oscillator whose initial stiffness, or frequency, is perturbed by a rectangular pulse of duration $τ$. We find that the Jarzynski equality is satisfied unconditionally only up to the seventh order in $τ$; in higher orders, the Jarzynski equality holds if and only if the noise is Gaussian. These results imply that, unless it is exact, the Langevin equation can only be used to evaluate properties that are linear or quadratic in noise and its derivatives. Such properties are insensitive to the noise statistics, so the Langevin equation with linear dissipation and non-Gaussian noise (though not inconsistent by itself) is superfluous.

We investigate the splitting probability of a monitored continuous-time quantum walk with two targets and show that, in stark contrast to a classical random walk, it exhibits a nonanalytic, phase-transition-like behavior controlled by the sampling time at the targets. For large systems and sampling times smaller than a critical value $τ_c = 2π/ΔE$, where $ΔE$ is the energy bandwidth, the splitting probability is universal and equal to $1/2$, independent of the initial condition and the sampling time. Above the critical sampling, a nonuniversal regime emerges in which the splitting probability deviates from $1/2$ and develops a fluctuating pattern of pronounced peaks and dips dependent on both the sampling time and the initial condition. These results follow from a nontrivial mapping of the splitting problem onto a pair of single-target detection problems enabled by the superposition principle.

We introduce a novel operator to describe a random walk process on a simplicial complex. Walkers are allowed to wonder across simplices of various dimensions, bridging nodes to edges, and edges to triangles, via a nested organization that hierarchically extends to higher structures of arbitrary large, but finite, dimension. The asymptotic distribution of the walkers provides a natural ranking to gauge the relative importance of higher order simplices. Optimal search strategies in presence of stochastic teleportation are addressed and the peculiar interplay of noise with higher order structures unraveled.

We show that a propagating Dirac electron with intrinsic spin generically carries a real--space helical conserved current, even in the absence of orbital angular momentum. Using exact Dirac eigenstates in cylindrical confinement, we demonstrate that this helical structure possesses definite handedness, persists into evanescent regions, and is characterized by a geometric helix pitch independent of the longitudinal de~Broglie wavelength. This intrinsic helical geometry enables a local geometric coupling between a propagating electron and a chiral environment, yielding chirality--dependent spin selectivity through current geometry rather than through a spin--orbit coupling term.

The Fleming-Viot process describes a system of $N$ particles diffusing on a graph with an absorbing site. Whenever one of the particles is absorbed, it is replaced by a new particle at the position of one of the $N-1$ remaining particles. Here we consider the case where the particles lie on the semi-infinite line with a biased diffusion towards the origin which is the absorbing site. In the large $N$ limit, the evolution of the density becomes deterministic and has a number of characteristics similar to the Fisher-KPP equation: a one-parameter family of steady state solutions, dependence of the long time asymptotics on the initial conditions, Bramson logarithmic shift, etc. One noticeable difference, however, is that in the Fleming-Viot case, the solution can be computed explicitly for arbitrary initial conditions and at an arbitrary time. By modifying the diffusion rule near the origin, one can produce a transition in the flux of absorbed particles, very similar to the pushed-pulled transition in travelling waves. Lastly, using a cut-off approximation (which is known to be correct in the theory of travelling waves), we derive a number of predictions for the leading large $N$ correction of the flux of absorbed particles.

Graphene enables precise carrier-density control via gating, making it an ideal platform for studying electronic interactions. However, sample inhomogeneities often limit access to the low-density regimes where these interactions dominate. Enhancing carrier mobility is therefore crucial for exploring fundamental properties and developing device applications. Here, we demonstrate a significant reduction in external inhomogeneity using a double-layer graphene architecture separated by an ultra-thin hexagonal boron nitride layer. Mutual screening between the layers reduces scattering from random Coulomb potentials, resulting in a quantum mobility exceeding. Shubnikov de-Haas oscillations emerge at magnetic fields below 1 mT, while integer quantum Hall features are observed at 0.002T. Furthermore, we identify a fractional quantum Hall plateau at a filling factor of at 2T. These results demonstrate the platform's suitability for investigating strongly correlated electronic phases in graphene-based heterostructures.

Fingerprint labyrinthine patterns exhibit a level of structural complexity beyond simple stripe phases, combining local stripe order with a dense network of point-like defects. Unlike symmetry-breaking phases, where coarsening proceeds via diffusive defect annihilation, or conventional stripe phases, where curvature-driven motion of extended grain boundaries dominates, the coarsening of fingerprint labyrinths is governed primarily by localized junction and terminal defects. Using the Turing-Swift-Hohenberg equation, we study the nonequilibrium relaxation of fingerprint labyrinthine patterns following a quench. To go beyond conventional Fourier-based diagnostics, we employ a template-matching convolutional neural network (TM-CNN) to identify and track junctions and terminals directly in real space, enabling a quantitative characterization of defect statistics and spatial correlations. We show that, although these point-like defects drive coarsening, their motion is strongly constrained by the surrounding stripe geometry, leading to slow, nondiffusive dynamics that are qualitatively distinct from both conventional phase ordering and stripe coarsening. Together, these results establish defect-mediated dynamics as the central organizing principle of fingerprint labyrinthine coarsening and demonstrate the effectiveness of machine-learning-assisted approaches for complex pattern-forming systems.

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

Counterdiabatic driving realizes parameter displacement of an energy eigenstate of a given parametrized Hamiltonian using the adiabatic gauge potential. In this paper, we propose a universal method of digitized counterdiabatic driving, constructing the adiabatic gauge potential in a digital way with the idea of universal counterdiabatic driving. This method has three advantages over existing universal counterdiabatic driving and/or digitized counterdiabatic driving: it does not introduce any many-body and/or nonlocal interactions to an original target Hamiltonian; it can incorporate infinite nested commutators, which constitute the adiabatic gauge potential; and it gives explicit expression of rotation angles for digital implementation. We show the consistency of our method to the exact theory in an analytical way and the effectiveness of our method with the aid of numerical simulations.

For a few decades, Fluorescent Molecular Rotors have been commonly employed as local probes of microviscosity in complex materials. However, without proper calibration, relating microviscosity to a physical parameter is unclear, which strongly limits their quantitative use in biological media for instance. In this study, the response of a molecular rotor in binary and ternary macromolecular aqueous solutions of polyethylene glycol (PEG) of different molecular weights is investigated in order to better rationalize the sensitivity of rotors to their cybotactic environment. More precisely, for the investigated composition range of ternary mixtures, it is shown that a linear mixing rule applies for fluorescence lifetime with the proportion of the two PEG, and with an increasing ratio of heavy PEG leading to larger lifetimes. These results allow to test more precisely the free volume theory, which has been proposed in the context of probing glass transition. Analysis show that while this theory semi-quantitatively captures the observation, its precise use raises some questions.

Partitioning large networks into stable clusters of synchronized nodes is a challenging task. Recent approaches based on spectral analysis can provide exact results on specific dynamics but remain unfeasible for very large networks. Moreover, within a stochastic framework, it is unclear which dynamics should be chosen to study synchronization. Here we propose an unbiased and scalable method based on the message-passing algorithm. By exploiting the collective behavior emerging across critical points of an effective Ising-like model, we identify dynamically coherent clusters of synchronized nodes and illustrate the approach on some large real-world networks. We find that, unlike continuous-time dynamics, abrupt desyncrhronization occurs even in simple graphs, without the need to invoke higher order interactions. However, when noise is included, the transition to synchronization becomes smoother and proceeds through the formation of plateaus, albeit at the cost of requiring larger coupling strengths.

The maximum work extractable from a quantum system is achieved when the system is driven adiabatically. Frictional work then quantifies the difference in work output between adiabatic and non-adiabatic driving. Here we show that frictional work in a non-integrable spin chain is well-described by the diagonal entropy production associated with the build up of quantum coherence. The relationship is characterized by an effective temperature of the final adiabatic state and holds for slow to moderate driving protocols. For fast protocols, the frictional work is instead described by the quantum relative entropy between the final non-adiabatic and adiabatic states. We compare our results to those obtained from an integrable spin chain, in which case the adiabatic state is no longer described by a single temperature. In this case, the frictional work is described by a sum of terms for each independent subspace of the spin chain, which are at different effective temperatures. We show how integrability breaking can enhance work extraction in the adiabatic limit, but degrade work extraction in sufficiently non-adiabatic regimes.

We study computationally the creep and yielding of athermal gels and fibre network materials under a constant imposed shear stress, within a minimal model of interconnected filaments with central forces in $d=2$ spatial dimensions. Each filament is assumed Hookean initially, then breaks irreversibly above a threshold strain. At early times after the imposition of a small stress, we find purely viscoelastic creep response associated with non-affine deformations within the material, with solid terminal behaviour for a network coordination $Z>2d=4$ and initially floppy response for $Z<4$. For a marginally connected network, $Z=4$, we find sustained power law creep with a strain rate $\dotγ\sim t^{-1/2}$ and strain $γ\sim t^{1/2}$ as a function of time $t$ after the imposition of the stress. This viscoelastic regime gives way at later times to irreversible elastoplastic creep arising from filament breakage, broadening the range of values of $Z$ and time over which power law creep occurs, compared to a network with filament breakage disallowed. This accumulating damage can weaken the network to such an extent that catastrophic material failure then occurs after a long delay, which we characterise. Finally, we consider the implications of viscoelastic versus elastoplastic deformation for the extent to which a material will recover its original shape if the load is removed after some interval of creep.

Solid-state quantum emitters (QEs) in two-dimensional semiconductors offer compact, chip-compatible sources for quantum photonics. In transition-metal dichalcogenides (TMDs), nanopillars are widely used to induce localized emission, yet the underlying confinement mechanism and the relative roles of strain versus dielectric environment remain unclear. The general problem addressed here is whether strain alone explains quantum emitter formation and placement in MoSe$_2$, or whether dielectric contrast at suspended-supported interfaces is also required. Here, we combine hyperspectral superlocalization of photoluminescence with co-registered AFM topography and phase to map the positions of localized states (LS) in MoSe$_2$ suspended on GaN pillars and correlate them with bending strain and the local dielectric context. Contrary to the common assumption of purely strain-driven activation, LS frequently occur at suspended--supported interfaces around the pillar apex and span a broad strain range without a clear threshold, while being scarce along high-strain ripples. Our data indicate that deterministic emitter positioning in Mo-based TMDs benefits from co-engineering both strain gradients and nanoscale dielectric heterogeneity, rather than strain alone. More broadly, this combined optical-mechanical characterization approach provides a general framework for mapping structure-property relationships in 2D quantum materials at the single-emitter level.

We study a quasi-one-dimensional fluid of hard dumbbells with continuous orientational degrees of freedom using an exact transfer-matrix formulation. The model allows for a complete analytical characterization of thermodynamic properties, orientational ordering, and correlation functions in terms of the spectral properties of an integral operator. We derive exact expressions for the equation of state, the orientational distribution function, and both partial and total radial distribution functions. Their asymptotic behavior is governed by the complex poles of the Laplace-transformed correlation functions, which determine the positional and orientational correlation lengths. As density increases, the system exhibits a continuous crossover from a weakly ordered regime with a unimodal orientational distribution to a strongly constrained regime characterized by bimodal orientational ordering. This crossover is accompanied by a nonmonotonic behavior of the pressure relative to the Tonks gas and by a qualitative change in the decay of correlation functions from oscillatory to monotonic. In the high-pressure limit, we show that orientational and positional fluctuations contribute equally to the pressure, leading to a universal ratio of twice the Tonks pressure. The theoretical predictions are supported by numerical solutions of the discretized transfer operator and by scaling arguments that elucidate the high-pressure behavior of ordering and correlation lengths.

Mobility edges commonly arise in one-dimensional quasiperiodic systems once exact self-duality is broken, yet their origin is typically understood only at the level of individual Hamiltonians. Here we show that mobility edge positions are not independent spectral features of individual Hamiltonians, but are structurally constrained across quasiperiodic Hamiltonians related by an isospectral duality. Using a bichromatic Aubry--André model as a minimal setting, we demonstrate that this constraint is encoded in an exact identity for Lyapunov exponents derived from the Thouless formula. As a consequence, the mobility edge positions are restricted to a reduced set of energies. In the self-dual limit, these mobility edge positions coincide at a single localization--delocalization transition. This structural constraint enforces a linear critical scaling of the physical Lyapunov spectrum near the self-dual point. Numerical results confirm a critical exponent consistent with the standard Aubry--André value of $ν= 1$, while simultaneously revealing a novel, non-universal energy-dependent prefactor.

The fluctuations are termed mesoscopic, when their typical size is essentially larger then the average distance between the nearest neighbors, while being much smaller than the overall system size. Since the features of mesoscopic fluctuations are essentially different from those of the surrounding matter, they can be interpreted as fluctuations of one phase occurring inside another host phase. In condensed matter, these fluctuations are of nanosize. They can occur in many-body systems of different nature, for instance, they are typical for condensed matter, can appear in systems of trapped atoms, and also arise in biological and social systems. A survey of the experimental evidence for the occurrence of mesoscopic fluctuations in different materials and systems is given. The main attention is paid to a general theoretical approach for describing them. Applications of the approach are also discussed.

We present a geometric framework for the inverse design of smart woven fabrics composed of non-uniformly shrinking threads. A sufficiently tight weaving structure imposes strong local criteria on the material deformation and reduces the local geometry to a single scalar degree of freedom. Control over this degree of freedom can be achieved through simple calibration for each specific material system, via either mechanical experiments or numerical simulations. This reduction allows us to inverse-design a woven smart fabric, that conforms to an arbitrary target geometry when actuated, by solving a nonlinear hyperbolic partial differential equation. We validate this approach by deriving the thread-level actuation required for specific target geometries. We present both exact analytic solutions for symmetric shapes and a numerical optimization method for arbitrary freeform surfaces. These results confirm the practicality of our framework in achieving programmable, complex three-dimensional shaping.

Chemically fueled supramolecular systems can exhibit complex, time-dependent behaviors reminiscent of living matter when maintained far from equilibrium by continuous energy or fuel consumption. Here, we introduce a minimal reaction-diffusion model that captures the essential dynamics of a cooperative supramolecular polymerization network driven by monomer activation and deactivation. We show that a balance between autocatalytic growth and inhibitory decay sustains a nonequilibrium steady state in the model that undergoes a Hopf bifurcation, giving rise to autonomous oscillations. When spatial transport is introduced through diffusion, the system displays rich spatiotemporal phenomena, such as traveling wavefronts and transient polygonal patterns. Our results demonstrate that the interplay between reaction kinetics and diffusion can spontaneously generate self-organized, life-like dynamics in synthetic supramolecular polymer systems. This theoretical framework not only bridges molecular self-assembly and active matter dynamics but also provides design principles for creating adaptive, oscillatory, and self-patterning materials powered by chemical fuels.

We report an experimental study of induced superconductivity in Ge hut nanowire Josephson junctions. The Ge hut nanowires are grown on prepatterned SiGe ridges via molecular beam epitaxy (MBE) and Josephson junction devices are fabricated by contacting the nanowires with Al electrodes. Low-temperature current-bias transport measurements of the Josephson junctions are performed and the measurements show that the devices exhibit gate-tunable supercurrent and excess current. The analysis of excess current indicates that the transparency of the Ge hut nanowire Josephson junctions is as high as 85%. Voltage-bias spectroscopy measurements of the devices show multiple Andreev reflections up to the fourth order. With magnetic field and temperature-dependent measurements of the multiple Andreev reflections, the critical field and the critical temperature of the induced superconductivity in the Josephson junctions are extracted to be ~0.12 T and ~1.4 K. The success in introducing superconductivity into Ge hut nanowires will stimulate their applications in building advanced quantum processors.

The valley degree of freedom in atomically thin transition-metal dichalcogenides provides a natural binary index for information processing. Exciton-polaritons formed under strong light-matter coupling offer a promising route to overcome the limited lifetime and transport of bare valley excitons. Here we report an anomalous optical valley Hall effect in a monolayer WS2 exciton-polariton system. Using polarization- and time-resolved real-space imaging, we directly visualize a symmetry-breaking spatial separation of polaritons from opposite valleys under linearly polarized excitation, accompanied by an ultrafast Hall drift velocity on the order of 10^5 m/s. This behaviour cannot be accounted for by conventional cavity-induced mechanisms and instead points to a strain-induced synthetic pseudomagnetic field acting on the excitonic component of polaritons. Our results establish exciton-polaritons as a high-speed and optically accessible platform for valley transport, opening pathways towards tunable valleytronic and topological photonic devices.

Quantum interference effects such as weak localization (WL) and universal conductance fluctuations (UCF) normally yield consistent electronic phase-coherence lengths in homogeneous conductors. Here we show that in individual carbon nanotube bundles exfoliated from highly conductive solution-spun fibers, different probes, including the field scales and magnitudes of WL and UCF and nonlocal magnetoconductance, lead to strikingly disparate estimates of coherence lengths. WL magnetoconductance measured in a perpendicular magnetic field yields a phase-coherence length of approximately 50 nm. In contrast, UCF amplitudes are comparable to e squared over h even for an 8 micrometer long segment, and nonlocal magnetoconductance persists across a 4 micrometer separation of electrodes, revealing phase-coherent transport over micrometer length scales within a single bundle. The coexistence of short- and long-range coherence implies that locally diffusive electrons remain partially phase-correlated among nanotubes within the same bundle. These findings challenge the conventional single-scale picture of mesoscopic coherence and establish carbon nanotube bundles as a model platform for emergent, network-level quantum transport.

Transitions from bound states in the continuum (BICs) to quasi-BICs (qBICs) are typically realized by introducing in-plane asymmetry, including permittivity asymmetry (ε-qBICs) and geometry asymmetry (g-qBICs). Here, we demonstrate that when the in-plane symmetry is rigorously kept, the transition can also be occurred, provided the out-of-plane asymmetry is designed, which is called σh -qBICs in this work. When the {σh symmetry is gradually broken, the system undergoes a topological phase transition characterized by a Zak phase inversion, leading to a band inversion between quadrupole and dipole modes. This process not only enables controlled radiation coupling of BICs but also introduces a defect-immune qBIC regime. Our findings establish a general mechanism for engineering high-Q resonances and topologically robust plasmonic cavities.

We show that quantum information scrambling can enable a generic SWAP gate between collective degrees of freedom in systems without universal local control. Our protocol combines the Hayden-Preskill recovery scheme, associated with the black hole information paradox, with quantum teleportation and runs them in parallel and in opposite directions, enabling bidirectional exchange of quantum states through global interactions alone. This approach cleanly distinguishes the roles of information spreading, entanglement, and chaos for enabling both coherent state transfer and recovery. We propose an experimental realization using the Dicke model, which can be realized in cavity-QED and trapped-ion platforms, highlighting the utility of holography in designing practical quantum gates.

Biological membranes are dynamic surfaces whose shape and function are critically influenced by protein inclusions (PIs). While membrane deformations induced by PIs have been extensively studied in the small-deformation regime, a variety of processes involves strong membrane deformations. We investigate the interaction between lipid membranes and PIs in the large deformation (LD) regime, with the finite-element method. We develop an approximate analytical solution that captures key features of the LD regime. We show that the force exerted by the membrane on a PI displays a non-monotonic behavior with respect to the PI vertical displacement. The qualitative features of this force appear to be independent of the protein geometry. For two interacting PIs, the membrane-mediated potential exhibits sub-power-law decay with inter-protein distance, reflecting the complex nature of the elastic medium. The interaction potential shows that conical PIs with identical and opposite orientations repel and attract, respectively, confirming the analogy between PI orientation and electric charge, in the LD regime. In the presence of membrane flows, we identify a characteristic velocity that separates two regimes in which bending rigidity and viscous effects dominate, respectively, implying the onset of flow-induced deformations above such velocity threshold. Overall, our results provide quantitative predictions for membrane-protein systems in biologically relevant scenarios involving LDs, with implications for protein sorting, clustering, and membrane trafficking.

We investigate the dissipative dynamics of quantum population and coherence among different network topologies of a quantum network using a quantum spin model coupled to a thermal bosonic reservoir. Our study proceeds in two parts. First, we analyze a small network of Ising spins embedded in a large dissipative bath, modeled via the Lindblad master equation, where temperature arises naturally from system-bath coupling. This approach reveals how network topology shapes quantum dissipative dynamics, providing a basis for controlling quantum coherence through tailored network structures. Second, we propose a mean-field approach that extends the network to larger scales and captures dissipative dynamics in large-scale networks, connecting network topology to quantum coherence in complex systems and revealing the sensitivity of quantum coherence to network structure. Our results highlight how dissipative quantum dynamics depend on network topology, providing insight into the coherent dynamics of entangled states in networks. These results may be extended to dynamics in complex systems such as opinion propagation in social models, epidemiology, and various condensed-phase and biological systems.

We report the fabrication and operation of a source-drain-gate three-terminal field-effect electronic device with an electron mobility exceeding $40\times 10^6$ cm$^2$ / (Vs). Several devices were fabricated, with the highest achieved electron mobility obtained using a symmetrically-doped GaAs/AlGaAs quantum well forming a two-dimensional electron gas (2DEG) with a density of $1.47(1) \times 10^{11}$ cm$^{-2}$ and a pristine, pre-fabrication electron mobility of $44(2) \times 10^6$ cm$^2$/(\text{Vs}). To circumvent the well-known degradation of electron mobility during fabrication, devices were fabricated using a flip-chip technique where all lithographic processing steps were performed on a separate sapphire substrate. This method demonstrates the successful operation of various gate assembly designs on distinct 2DEGs without observable mobility degradation. This advance doubles the previous record for field-effect electronic device mobility and enables access to new regimes of quantum transport and applications that were previously unfathomable due to mobility limitations.

We introduce a family of generalized Rokhsar-Kivelson (RK) Hamiltonians, which are reverse-engineered to have an arbitrary edge-weighted superposition of dimer coverings as their exact ground state at the RK point. We then focus on a quantum dimer model on the triangular lattice, with doubly-periodic edge weights. For simplicity we consider a $2\times1$ periodic model in which all weights are set to one except for a tunable horizontal edge weight labeled $α$. We analytically show that the model exhibits a continuous quantum phase transition at $α=3$, changing from a topological $\mathbb{Z}_2$ quantum spin liquid ($α<3$) to a columnar ordered state ($α>3$). The dimer-dimer correlator decays exponentially on both sides of the transition with the correlation length $ξ\propto1/|α-3|$ and as a power-law at criticality. The vison correlator exhibits an exponential decay in the spin liquid phase, but becomes a constant in the ordered phase. We explain the constant vison correlator in terms of loops statistics of the double-dimer model. Using finite-size scaling of the vison correlator, we extract critical exponents consistent with the 2D Ising universality class.

A central open question in chiral-induced spin selectivity (CISS) is whether weak electron-phonon coupling in a helical molecular junction can generate a sizable spin polarization in two-terminal transport without invoking additional strong symmetry-breaking ingredients. We address this question by implementing a self-consistent nonequilibrium Green's function (NEGF) calculation for a helical tight-binding model with spin-orbit coupling and electron-phonon interactions. The electron-phonon self-energies are evaluated self-consistently, and the transport signal is extracted using the standard magnetization-reversal protocol with a spin-polarized analyzer lead. We benchmark a fully self-consistent NEGF within the self-consistent Born approximation (SCBA) treatment for both global and local electron-phonon couplings against commonly used approximations, including diagonal self-energy schemes. We quantify how the resulting transport regime and spin polarization depend on phonon frequency, coupling strength, bias, temperature, and system size. In contrast to large polarizations and anomalous size trends reported under approximate treatments, the fully self-consistent calculation yields negligible spin polarization, additionally the electron-phonon coupling mainly renormalizes the spectrum, and transport remains quasi-ballistic across the explored parameter range.

We demonstrate widefield magnetic imaging of current flow in hydrogen terminated diamond field effect transistors (FETs) through in-substrate nitrogen vacancy (NV) centers. Hydrogen termination of the diamond surface induces a two dimensional hole gas (2DHG), while an ensemble of near surface NV centers located $ \sim 1~μm$ below the surface enables noninvasive magnetic imaging of current flow with micrometer scale spatial resolution. The FETs were electrically characterized over a range of drain source biases $V_{ds}= 0$ to $-15V$ and gate voltages,$V_{gs}= +3$ to $-9V$ followed by in situ widefield NV magnetometry during device operation. Magnetic field maps and reconstructed current density distributions directly visualize current injection at the source drain contacts and transport beneath the hBN gated channel. Magnetic field maps reveal current density variations in the channel region owing to non-uniformities or defects in the gate dielectric. In addition, we observe a pronounced enhancement of the drain current ($\sim 600-900 μA$) and a shift in the apparent threshold voltage during laser illumination, reflecting photo induced changes in channel electrostatics. By correlating gate dependent magnetic images with simultaneous electrical measurements, we directly link spatial current distributions to FET transfer characteristics, providing new insight into buried interface transport and non-uniform gating effects in the transistor channel. As the methodology is compatible with top gated FETs, it can be used to map channel current distributions with micrometer resolution in emerging channel materials, such as 2D materials and wide bandgap channels, and establish widefield NV magnetometry as a powerful platform for probing charge transport in transistors and Van der Waals dielectric heterostructures.

A result by Gioia and Wang [Phys Rev X 12, 031007 (2022)] showed that translationally symmetric states having nonzero momentum are necessarily long range entangled (LRE). Here, we consider the question: can a notion of momentum for non-translation symmetric states directly encode the nature of their entanglement, as it does for translation symmetric states? We show the answer is affirmative for 1D systems, while higher dimensional extensions and topologically ordered systems require further work. While Gioia and Wang's result applies to states connected via finite depth quantum circuits to a translation symmetric state, it is often impractical to find such a circuit to determine the nature of the entanglement of states that break translation symmetry. Here, instead of translation eigenstates, we focus on the many-body momentum distribution and the expectation value of the translation operator in many-body states of systems having broken translation symmetry. We show that in the continuum limit the magnitude of the expectation value of the translation operator $|<T>|$ necessarily goes to $1$ for delocalized states, a proxy for LRE states in 1D systems. This result can be seen as a momentum-space version of Resta's formula for the localization length. We investigate how accurate our results are in different lattice models with and without well-defined continuum limits. To that end, we introduce two models: a deterministic version of the random dimer model, illustrating the role of the thermodynamic and continuum limits for our result at a lattice level, and a simplified version of the Aubry-Andre model, with commensurate hopping for both momentum and position space. Finally, we use the random dimer model as a test case for the accuracy of $|<T>|$ as a localization (and thus entanglement) probe for 1D periodic lattice models without a well-defined continuum limit.

Physical neural networks typically train linear synaptic weights while treating device nonlinearities as fixed. We show the opposite - by training the synaptic nonlinearity itself, as in Kolmogorov-Arnold Network (KAN) architectures, we yield markedly higher task performance per physical resource and improved performance-parameter scaling than conventional linear weight-based networks, demonstrating ability of KAN topologies to exploit reconfigurable nonlinear physical dynamics. We experimentally realise physical KANs in silicon-on-insulator devices we term 'Synaptic Nonlinear Elements' (SYNEs), operating at room temperature, 0.1-1 microampere currents, and 2 MHz speeds with no observed degradation over 10^13 measurements and months-long timescales. We demonstrate nonlinear function regression, classification, and prediction of Li-Ion battery dynamics from noisy real-world multi-sensor data. Physical KANs outperform equivalently-parameterised software multilayer perceptron networks across all tasks, with up to two orders of magnitude fewer parameters, and two orders of magnitude fewer devices than linear weight based physical networks. These results establish learned physical nonlinearity as a hardware-native computational primitive for compact and efficient learning systems, and SYNE devices as effective substrates for heterogenous nonlinear computing.

We theoretically study the critical speed for superfluid flow of a two-dimensional (2D) binary superfluid of light past a polarization-sensitive optical obstacle. This speed corresponds to the maximum mean flow velocity below which dissipation is absent. In the weak-obstacle regime, linear-response theory shows that the critical speed is set by Landau's criterion applied to the density and spin Bogoliubov modes, whose relative ordering can be inverted due to saturation of the optical nonlinearity. For obstacles of arbitrary strength and large spatial extent, we determine the critical speed from the conditions for strong ellipticity of the stationary hydrodynamic equations within the hydraulic and incompressible approximations. Numerical simulations in this regime reveal that the breakdown of superfluidity is initiated by the nucleation of vortex-antivortex pairs for an impenetrable obstacle, and of Jones-Roberts soliton-type structures for a penetrable obstacle. Beyond superfluids of light, our results provide a general framework for the critical speed of 2D binary nonlinear Schrödinger superflows, including Bose-Bose quantum mixtures.

We investigate conservation laws of the two-dimensional Zakharov-Kuznetsov (ZK) equation, a natural higher-dimensional and non-integrable extension of the Korteweg--de Vries equation. The ZK equation admits three scalar conserved quantities -- mass, momentum, and energy -- represented as $I_1$, $I_2$, and $I_3$, as well as a vector-valued quantity $\bm{I}_4$. In high-accuracy numerical simulations on a finite double-periodic domain, most of these quantities are well preserved, while a systematic temporal drift is observed only in the $x$-component $I_{4x}$. We show that the nontrivial evolution of $I_{4x}$ originates from an explicit boundary-flux contribution, which is induced by fluctuations of the solution and its spatial derivatives at the domain boundaries. We successfully identify the source of the inaccuracy in the numerical solutions. Motivated by this analysis, we define a modified center-of-mass quantity $I_{4x}^{\mathrm{mod}}$ and demonstrate its conservation numerically for single-pulse configurations. The modified quantity thus provides a consistent conservation law for the ZK equation and yields an appropriate description of center-of-mass motion in finite-domain numerical simulations.

We present dla-ideal-solver, a high-performance framework for simulating two-dimensional Diffusion-Limited Aggregation (DLA) using Numba-accelerated Python. By leveraging just-in-time (JIT) compilation, we achieve computational throughput comparable to legacy static implementations while retaining high-level flexibility. We investigate the Laplacian growth instability across varying injection geometries and walker concentrations. Our analysis confirms the robustness of the standard fractal dimension $D_f \approx 1.71$ for dilute regimes, consistent with the Witten-Sander universality class. However, we report a distinct crossover to Eden-like compact growth ($D_f \approx 1.87$) in high-density environments, attributed to the saturation of the screening length. Beyond standard mass-radius scaling, we employ generalized Rényi dimensions and lacunarity metrics to quantify the monofractal character and spatial heterogeneity of the aggregates. This work establishes a reproducible, open-source testbed for exploring phase transitions in non-equilibrium statistical mechanics.

Cooperative effects in oscillator networks are often associated with enhanced stability of phase-locked solutions, which increases with system size. We show that the stabilization of persistent currents in annular atomic superfluids with periodic barriers is a concrete manifestation of this phenomenon. Under the simplifying assumption of continuity of atomic flow across identical barriers, the system reduces to a ring of locally coupled Kuramoto-like oscillators. We analytically derive the stability diagram of phase-locked configurations and quantify their robustness to noise and small random initial imperfections, finding excellent agreement with experimental observations. These results are inherent to the ring topology and independent of the specific physical platform.

Transcription factors (TFs) regulate gene expression by binding to specific genomic loci determined by DNA sequence. Their sequence specificity is commonly summarized by a consensus binding motif. However, eukaryotic genomes contain billions of low-affinity DNA sequences to which TFs associate with a sequence-dependent binding energy. We currently lack insight into how the genomic sequence defines this spectrum of binding energies and the resulting pattern of TF localization. Here, we set out to obtain a quantitative understanding of sequence-dependent TF binding to both motif and non-motif sequences. We achieve this by first pursuing accurate measurements of physical binding energies of the human TF Klf4 to a library of short DNA sequences in a fluorescence-anisotropy-based bulk competitive binding assay. Second, we show that the highly non-linear sequence dependence of Klf4 binding energies can be captured by combining a linear model of binding energies with an Ising model of the coupled recognition of nucleotides by a TF. We find that this statistical mechanics model parametrized by our in vitro measurements captures Klf4 binding patterns on individual long DNA molecules stretched in the optical tweezer, and is predictive for Klf4 occupancy across the entire human genome without additional fit parameters.

We present the development and validation of a novel multimodal optical imaging platform that integrates hyperspectral imaging (HSI) and laser speckle contrast imaging (LSCI) to enable real-time, non-invasive mapping of tissue oxygenation, perfusion and metabolism, via blood flowmetry and targeting of oxy- (HbO2), deoxyhemoglobin (HHb), as well as oxidized cytochrome-c-oxidase (oxCCO). The system architecture features a single high-speed camera and dual optical path, with synchronized alternating illumination: a filtered, supercontinuum laser for HSI and a He-Ne laser for LSCI. The system performances were evaluated through in vivo experiments on rat spinal cord under normoxic and hypoxic conditions, revealing coherent physiological changes in hemodynamics, metabolism and relative blood flow index (rBFI). These results demonstrate the potential of the platform for functional tissue imaging and quantitative dynamic monitoring of both oxygen delivery and consumption.

This paper presents a numerical study of the dynamics of DNA double helix breakage under the influence of external forces using the Peyrard-Bishop-Dauxois (PBD) model. The PBD model represents DNA as a chain of nonlinearly coupled oscillators, which makes it possible to analyze the processes of melting and mechanical denaturation. The main focus is on cases where an external force is applied to the terminal or central site of a DNA molecule, simulating stretching at a constant rate. The critical force required to break hydrogen bonds, which depends on the point of application of the force, is calculated. It is found that the denaturation process occurs stepwise, with characteristic peaks in the force-time graphs. The phenomenon of hysteresis under periodic exposure to external forces is also studied, which is important for understanding energy losses and heating of the system.

Reliable identification of microplastic fibers is crucial for environmental monitoring but remains analytically challenging. We report an explainable deep-learning framework for classifying microplastic and natural microfibers using polarization-resolved digital holographic microscopy. From multiplexed holograms, the complex Jones matrix of each fiber was reconstructed to extract polarization eigen-parameters describing optical anisotropy. Statistical descriptors of nine polarization characteristics formed a 72-dimensional feature vector for a total of 296 fibers spanning six material classes, including polyamide 6, polyethylene terephthalate, polyamide 6.6, polypropylene, cotton and wool. The designed fully connected deep neural network achieved an accuracy of 96.7 % on the validation data, surpassing that of common machine-learning classifiers. Explainable artificial intelligence analysis with Shapley additive explanations identified eigenvalue-ratio quantities as dominant predictors, revealing the physical basis for classification. An additional reduced-feature model with the preserved architecture exploiting only these most significant eigenvalue-based characteristics retained high accuracy (93.3 %), thereby confirming their dominant role while still outperforming common machine-learning classifiers. These results establish polarization-based features as distinctive optical fingerprints and demonstrate the first explainable deep-learning approach for automated microplastic fiber identification.

In the study of time-dependent (i.e., temporal) networks, researchers often examine the evolution of communities, which are sets of densely connected sets of nodes that are connected sparsely to other nodes. An increasingly prominent approach to studying community structure in temporal networks is statistical inference. In the present paper, we study the performance of a class of statistical-inference methods for community detection in temporal networks. We represent temporal networks as multilayer networks, with each layer encoding a time step, and we illustrate that statistical-inference models that generate community assignments via either a uniform distribution on community assignments or discrete-time Markov processes are biased against generating communities with large or small numbers of nodes. In particular, we demonstrate that statistical-inference methods that use such generative models tend to poorly identify community structure in networks with large or small communities. To rectify this issue, we introduce a novel statistical model that generates the community assignments of the nodes in given layer (i.e., at a given time) using all of the community assignments in the previous layer. We prove results that guarantee that our approach greatly mitigates the bias against large and small communities, so using our generative model is beneficial for studying community structure in networks with large or small communities. Our code is available at https://github.com/tfaust0196/TemporalCommunityComparison.

The conformal integration of nitrogen-vacancy (NV) center nanodiamond arrays onto soft, hydrated, curvilinear biological interfaces remain a fundamental challenge for in vivo quantum sensing and imaging. Conventional transfer techniques often fail due to reliance on high temperature, corrosive chemicals, or mechanical peeling, leading to pattern damage, low fidelity, or poor biocompatibility. Here, we report a transfer strategy utilizing polyvinyl alcohol (PVA) carrier soluble tape, enabling rapid, residue-free, high-fidelity transfer of nanodiamond patterns onto diverse biointerfaces. The success of this method is rooted in a unique "hydrate-soften-expand-self-peel" mechanism of the soluble tape with PVA backing. In situ mechanical tracking reveals non-uniform PVA swelling upon hydration generates transient local normal and shear stresses at the interface. These stresses delaminate the tape within 3 minutes at room temperature while promoting adhesion of the nanodiamond array to the substrate. In contrast, conventional water-soluble tapes with composite structures undergo passive dissolution and collapse, causing residue contamination and reduced efficiency. Leveraging this mechanism, we achieve conformal patterning on ultra-soft hydrogels (~0.6 kPa) and highly curved bio-surfaces (hair, 100 μm^-1). Additionally, we demonstrate a dual-identity verification system integrating data storage and physical unclonable functions on a hydrogel contact lens. This work provides a versatile tool for bio-interface engineering and a general framework for gentle, efficient transfer of functional nanomaterials.

The impact of rising consumption on wealth inequality remains an open question. Here we revisit and extend the Social Architecture of Capitalism agent-based model proposed by Ian Wright, which reproduces stylized facts of wealth and income distributions. In a previous study, we demonstrated that the macroscopic behavior of the model is predominantly governed by a single dimensionless parameter, the ratio between average wealth per capita and mean salary, denoted by R. The shape of the wealth distribution, the emergence of a two-class structure, and the level of inequality -- summarized by the Gini index -- were found to depend mainly on R, with inequality increasing as R increases. In the present work, we examine the robustness of this result by relaxing some simplifying assumptions of the model. We first allow transactions such as purchases, salary payments, and revenue collections to occur with different frequencies, reflecting the heterogeneous temporal dynamics of real economies. We then impose limits on the maximum fractions of wealth that agents can spend or collect at each step, constraining the amplitude of individual transactions. We find that the dependence of the inequality on R remains qualitatively robust, although the detailed distribution patterns are affected by relative frequencies and transaction limits. Finally, we analyze a further variant of the model with adaptive wages emerging endogenously from the dynamics, showing that self-organized labor-market feedback can either stabilize or amplify inequality depending on macroeconomic conditions.

The global demand for green hydrogen and its derivatives is growing rapidly as a cornerstone for decarbonizing hard-to-abate sectors. Morocco, endowed with abundant solar and wind resources, ambitions to capture up to 4% of the global PtX market by 2030, positioning itself as a strategic partner for Europe's energy transition. Yet, uncertainty persists regarding European demand trajectories, infrastructure readiness, and investment risks. This study evaluates Morocco's hydrogen transition through 2035 using a sector-coupled capacity expansion model. We compare industry reallocation and hydrogen export-oriented scenarios, assessing their impacts under interannual weather variability and financial sensitivities. Both scenarios require a tripling of current renewable and electrolyzer capacities, with hydrogen demand reaching approximately up to 38 TWh by 2035. Lower financing costs (WACC) have a greater effect on system costs and competitiveness than stricter CO2 constraints or weather variability. The trade- off between domestic energy security and export competitiveness is pronounced, but both pathways are technically feasible and aligned with Morocco's strategic energy goals. These findings provide evidence-based guidance for policymakers to balance Morocco's domestic and export ambitions in the evolving hydrogen market.

This work primarily focuses on the study of three gradient reconstruction techniques applied to the calculation of viscous terms in a cell-centered, finite volume formulation for general unstructured grids. The work also addresses different ways of formulating the limiter functions necessary to maintain stability in the presence of flow discontinuities. The flows of interest are simulated using the compressible Reynolds-averaged Navier-Stokes equations, and the negative Spalart-Allmaras model is used for turbulence closure. Definition of interface inviscid terms uses the Roe approximate Riemann solver, whereas the interface viscous terms are calculated with a standard centered scheme together with appropriate definitions of the interface gradients. Steady state solutions are obtained using an implicit time-integration method, together with a novel convergence acceleration technique. This new approach defines a set of three simple rules for controlling the global CFL number based on the residue evolution. The work considers three test cases, namely, the subsonic bump-in-channel flow, the subsonic NASA high-lift Common Research Model multielement airfoil and the transonic ONERA M6 wing. Present results are compared to experimental and numerical data available in the literature. Severe numerical instabilities are observed when the simplest gradient reconstruction technique is used, while more sophisticated formulations are able to provide excellent agreement with the existing literature. Current results are demonstrated to be highly insensitive to modifications made to the numerical flux entropy fix terms. Integrated aerodynamic forces are shown to be mildly dependent on the limiter formulation used, even in the absence of shock waves. The proposed convergence acceleration procedure manages to quickly drive the residue terms to machine zero, provided no major instabilities are present.

The effects induced by numerical schemes and mesh geometry on the solution of two-dimensional supersonic inviscid flows are investigated in the context of the compressible Euler equations. Five different finite-difference schemes are considered: the Beam and Warming implicit approximate factorization algorithm, the original Steger and Warming flux vector splitting algorithm, the van Leer approach on performing the flux vector splitting and two different novel finite-difference interpretations of the Liou AUSM+ scheme. Special focus is given to the shock wave resolution capabilities of each scheme for the solution of an external supersonic inviscid flows around a blunt body. Significant changes in the shock structure are observed, mainly due to special properties of the scheme in use and the influence of the domain transformation procedure. Perturbations in the supersonic flow upstream of the shock are also seen in the solution, which is a non-physical behavior. Freestream subtraction, flux limiting and the explicit addition of artificial dissipation are employed in order to circumvent these problems. One of the AUSM+ formulations presented here is seen to be particularly more robust in avoiding the appearance of some of these numerically-induced disturbances and non-physical characteristics in the solution. Good agreement is achieved with both numerical and experimental results available in the literature.}

This study investigates the dynamics of red blood cells (RBCs) under the influence of evanescent waves generated by total internal reflection (TIR). Using a 1064 nm laser system and a dual-chamber prism setup, we quantified the mobility of erythrocytes in different glucose environments. Our methodology integrates automated tracking via TrackMate\c{opyright} to analyze over 60 trajectory sets. The results reveal a significant decrease in mean velocity, from 11.8 μm/s in 5 mM glucose to 8.8 μm/s in 50 mM glucose (p = 0.019). These findings suggest that evanescent waves can serve as a non-invasive tool to probe the mechanical properties of cell membranes influenced by biochemical changes.

Granular materials subjected to impact loading exhibit highly heterogeneous spatiotemporal dynamics governed by wave propagation, pore collapse, and grain-scale rearrangements. Mesoscale hydrocodes resolve these processes but are computationally expensive, limiting their use in parametric studies and uncertainty quantification. In this work, we develop a convolutional Long Short-Term Memory (ConvLSTM) neural network as a spatiotemporal surrogate for mesoscale simulations of weak shock propagation in granular media. Using two-dimensional hydrocode simulations as training data, we first consider a simplified "billiard break" problem in which a cue ball impacts a cluster of nine circular balls, all deformable. Sequences of pressure-field images serve as input-output pairs for a sequence-to-sequence ConvLSTM, which is trained to predict future frames from a short history. We compare several architectures and show that a relatively compact encoder-decoder ConvLSTM accurately reproduces the propagation of the pressure wave and the resulting particle motion for an unseen combination of cue-ball position and impact velocity. As a proof-of-concept extension, we apply the same ConvLSTM framework to previously published mesoscale simulations of weak shock compaction in a granular ensemble. When evaluated at piston impact speeds that were completely withheld from training, the surrogate captures the position and shape of the compaction front and its dependence on impact speed, while smoothing fine pore-scale details in the highly compacted region as expected. These results demonstrate that ConvLSTM models can serve as satisfactory surrogates for spatiotemporal mesoscale simulations of granular wave propagation, enabling accelerated exploration of parameter space and laying the groundwork for physics-informed, mesoscale simulations of granular materials under shock loading.

Foundational machine-learned interatomic potentials have emerged as powerful tools for atomistic simulations, promising near first-principles accuracy across diverse chemical spaces at a fraction of the cost of quantum-mechanical calculations. However, the most accurate equivariant architectures rely on Clebsch-Gordan tensor products whose computational cost scales steeply with angular resolution, creating a trade-off between model expressiveness and inference speed that ultimately limits practical applications. Here we introduce Geodite, an equivariant message-passing architecture that replaces tensor products while incorporating physical priors to ensure smooth, well-behaved potential energy surfaces. Trained on the Materials Project trajectories dataset of inorganic crystals, Geodite-MP achieves accuracy competitive with leading methods on benchmarks for materials stability prediction, thermal conductivity, phonon-derived properties, and nanosecond-scale molecular dynamics, while running $3\text{--}5\times$ faster than models performing similarly. By combining predictive accuracy, computational efficiency, and physicality, Geodite enables faster large-scale atomistic simulations and high-throughput screening that would otherwise be computationally prohibitive.

The design of spacecraft thermal protection systems (TPS) requires accurate knowledge of thermal transport properties across wide ranges of temperature and pressure. For fibrous insulation, conventional measurement techniques in laboratory settings are typically limited to temperatures much lower than what is reached in atmosphere entry scenarios. Moreover, it is often the case that only temperature measurements are available, meaning that the thermal conductivity of the insulation must be indirectly inferred as an inverse problem. We propose a Bayesian framework using information field theory (IFT) to reconstruct the thermal conductivity of high-temperature fibrous insulation from sparse experimental data. Under IFT, the conductivity is represented as a Gaussian process, and the physics is enforced via a physics-informed prior over the temperature derived from the heat equation. Bayes's rule produces an infinite-dimensional posterior distribution that quantifies uncertainty about the conductivity which can be evaluated in extrapolation regimes. We apply the method to Opacified Fibrous Insulation with both synthetic and experimental data to reconstruct the thermal conductivity beyond the experimental regime. The inferred conductivities are validated against reference data and then propagated into high-fidelity digital twins of flexible TPS performance under Mars and Earth entry trajectories. The results show that IFT yields accurate predictions with quantified uncertainty, enabling robust TPS sizing in regimes inaccessible to direct measurement.

Graphynes offer a chemically heterogeneous $sp/sp^{2}$ carbon framework with distinct electronic regimes and site-selective reactivity. Here, Density Functional Theory and Reactive Molecular Dynamics Simulations are combined to evaluate pristine, B-doped, N-doped, and B-N co-doped $γ$-graphyne and 6,6,12-graphyne (meta/ortho/para). $γ$-graphyne is a semiconductor, while 6,6,12-graphyne exhibits an anisotropic Dirac-like semi-metallic dispersion. B/N substitution reconstructs near-$E_F$ states via dopant $π$ hybridization, and B-N pairing stabilizes defects through donor-acceptor compensation, with the ortho substitutions being the most favorable. Hydrogen adsorption remains weak on pristine lattices but becomes locally optimized upon doping, with near thermo-neutral $ΔG_{\mathrm{ads}}$ 'hot spots' predominantly on $sp$-proximate carbon sites adjacent to the dopants. Reactive MD at 300 K further reveals an activity stability trade-off: B-N ortho in $γ$-graphyne sustains controlled hydrogen uptake without catastrophic bond scission, whereas B-N meta/para degrade, and 6,6,12-graphyne is generally more susceptible to over-hydrogenation. These results identify the B-N geometry as a key design variable for graphyne-based HER catalysts, which require both a favorable $ΔG_{\mathrm{ads}}$ and finite-temperature hydrogenation stability.

Building upon previous research, this paper further explores the topological foundations for maintaining universal rules within ultra-large-scale societies. It finds that in small-scale societies, absolute egalitarianism and the rule of law can be compatible through peer monitoring within a fully connected network. However, in ultra-large-scale societies, to maintain high-dimensional rules capable of protecting innovation and property rights, a complex hierarchical structure including "high-fragility" nodes must be constructed. Through quantitative analysis of power structures, this paper proves that a flattened, two-tier structure inevitably leads to the degradation of the rule of law. Only a social topology with sufficient hierarchical depth can escape the deathly trap of the Leviathan while expanding in scale, thereby sustaining the dynamic evolution of civilization.

Scientific attention is unevenly distributed, creating inequities in recognition and distorting access to opportunities. Using citations as a proxy, we quantify disparities in attention by gender and institutional prestige. We find that women receive systematically fewer citations than men, and that attention is increasingly concentrated among authors from elite institutions -- patterns not fully explained by underrepresentation alone. To explain these dynamics, we introduce a model of citation network growth that incorporates homophily (tendency to cite similar authors), preferential attachment (favoring highly cited authors) and group size (underrepresentation). The model shows that disparities arise not only from group size imbalances but also from cumulative advantage amplifying biased citation preferences. Importantly, increasing representation alone is often insufficient to reduce disparities. Effective strategies should also include reducing homophily, amplifying the visibility of underrepresented groups, and supporting equitable integration of newcomers. Our findings highlight the challenges of mitigating inequities in asymmetric networks like citations, where recognition flows in one direction. By making visible the mechanisms through which attention is distributed, we contribute to efforts toward a more responsible web of science that is fairer, more transparent, and more inclusive, and that better sustains innovation and knowledge production.

Accurate determination of transition states is central to an understanding of reaction kinetics. Double-endpoint methods where both initial and final states are specified, such as the climbing image nudged elastic band (CI-NEB), identify the minimum energy path between the two and thereby the saddle point on the energy surface that is relevant for the given transition, thus providing an estimate of the transition state within the harmonic approximation of transition state theory. Such calculations can, however, incur high computational costs and may suffer stagnation on exceptionally flat or rough energy surfaces. Conversely, methods that only require specification of an initial set of atomic coordinates, such as the minimum mode following (MMF) method, offer efficiency but can converge on saddle points that are not relevant for transition of interest. Here, we present an adaptive hybrid algorithm that integrates the CI-NEB with the MMF method so as to get faster convergence to the relevant saddle point. The method is benchmarked for the Baker-Chan (BC) saddle point test set using the PET-MAD machine-learned potential as well as 59 transitions of a heptamer island on Pt(111) from the OptBench benchmark set. A Bayesian analysis of the performance shows a median reduction in energy and force calculations of 46% [95% CrI: -55%, -37%] relative to CI-NEB for the BC set, while a 28% reduction is found for the transitions of the heptamer island. These results establish this hybrid method as a highly effective tool for high-throughput automated chemical discovery of atomic rearrangements.

The last decade has seen significant advances in computer-aided diagnostics for cytological screening, mainly through the improvement and integration of scanning techniques such as whole slide imaging (WSI) and the combination with deep learning. Simultaneously, new imaging techniques such as quantitative phase imaging (QPI) are being developed to capture richer cell information with less sample preparation. So far, the two worlds of WSI and QPI have not been combined. In this work, we present a reconstruction algorithm which makes whole slide imaging of cervical smears possible by using a self-referencing three-wave digital holographic microscope. Since a WSI is constructed by combining multiple patches, the algorithm is adaptive and can be used on partial holograms and patched holograms. We present the algorithm for a single shot hologram, the adaptations to make it flexible to various inputs and show that the algorithm performs well for the tested epithelial cells. This is a preprint of our paper, which has been accepted for publication in 2026 IEEE International Symposium on Biomedical Imaging (ISBI).

Volitional memory control supports adaptive cognition by enabling intentional Recall of goal-relevant information and Suppression of unwanted memories. While neural mechanisms underlying Recall and Suppression have been studied largely in isolation, less is known about the large-scale brain networks supporting these processes under competing cognitive demands, particularly as a function of subclinical anxiety. Here, we examined control of emotionally valenced memories during directed Recall and Suppression while 47 participants concurrently performed an independent visual working memory task. Cognitive control efficiency was quantified using the Balanced Integration Score (BIS), and seed-to-voxel resting-state functional connectivity (rsFC) was used to characterize intrinsic network organization. Dissociable rsFC profiles were associated with memory control efficiency across emotional valences and were selectively moderated by anxiety. More efficient Suppression of positive memories was linked to reduced connectivity between the anterior cingulate cortex and posterior perceptual-midline regions, as well as diminished hippocampal-frontal pole coupling. In contrast, efficient Suppression of negative memories was associated with increased connectivity between posterior parietal and lateral occipital regions. Anxiety moderated relationships between cognitive efficiency and prefrontal connectivity during Suppression of positive memories and Recall of positive and neutral memories. Direct comparisons further revealed stronger hippocampal-thalamic rsFC during Suppression relative to Recall of positive memories. Together, these findings delineate the functional brain architecture supporting volitional control of emotional memories under cognitive load and demonstrate that anxiety severity selectively shapes these network-level mechanisms across the anxiety continuum.

We study the ChorHendySnir2006 evolutionary model, which consists of a rooted phylogenetic tree with three leaves, subject to the Jukes--Cantor (JC69) molecular evolutionary model and molecular clock. We show that the likelihood function associated with this model has a unique maximum which depends analytically of the parameters (as it was conjectured in ChorHendySnir2006), assuming that these parameters verify some very precise inequalities; some of which arise naturally from the model. With a typical argument of differential topology we reduce the proof to answer a question of algebra, very simple, although computationally involved, that we solve using some Maple libraries. We are very indebted to Marta Casanellas, who presented the problem to us and gave us the first insights on it.

Strong, correlated noise in recurrent neural circuits often passes through nonlinear transfer functions, complicating dynamical mean-field analyses of complex phenomena such as transients and bifurcations. We introduce a method that replaces nonlinear functions of Ornstein-Uhlenbeck (OU) noise with a Gaussian-equivalent process matched in mean and covariance, and combine this with a lognormal moment closure for expansive nonlinearities to derive a closed dynamical mean-field theory for recurrent neuronal networks. The resulting theory captures order-one transients, fixed points, and noise-induced shifts of bifurcation structure, and outperforms standard linearization-based approximations in the strong-fluctuation regime. More broadly, the approach applies whenever dynamics depend smoothly on OU processes via nonlinear transformations, offering a tractable route to noise-dependent phase diagrams in computational neuroscience models.

We consider a SIRD epidemic model for a population composed of two groups of individuals with asymmetric interactions, where the force of infection depends on the active (alive) population in each group, rather than on the total population, as in the classical formulation. We prove that the final state for susceptible individuals is always positive and characterize it as the unique fixed point of a map. We also relate the final size to the basic reproduction number and show that the final number of susceptibles decreases when transmission rates increase. Numerical simulations compare the active-population and classical two-group SIRD models, showing differences in final size and the occurrence of multiple epidemic waves. The convergence of the fixed point approach is also illustrated.

Collective movement is observed widely in nature, where individuals interact locally to produce globally ordered, coherent motion. In typical models of collective motion, each individual takes the average direction of multiple neighbors, resulting in ordered movement. In small flocks, noise induced order can also emerge with individuals copying only a randomly chosen single neighbor at a time. We propose a new model of collective movement, inspired by how real animals move, where individuals can move in two directions or remain stationary. We demonstrate that when individuals interact with a single neighbor through a novel form of halting interaction -- where an individual may stop upon encountering an oppositely moving neighbor rather than instantly aligning -- persistent collective order can emerge even in large populations. This represents a fundamentally different mechanism from conventional averaging-based or noise-induced ordering. Using deterministic and stochastic mean-field approximations, we characterize the conditions under which such ``flocking by stopping'' behavior can occur, and confirm the mean-field predictions using individual-based simulations. Our results highlight how incorporating a stopped state and halting interactions can generate new routes to order in collective movement.

Precision psychiatry aspires to elucidate brain-based biomarkers of psychopathology to bolster disease risk assessment and treatment development. To this end, functional magnetic resonance imaging (fMRI) has helped triangulate brain circuits whose functional features are correlated with or even predictive of forms of psychopathology. Yet, fMRI biomarkers to date remain largely descriptive identifiers of where, rather than how, neurobiology is aberrant, limiting their utility for guiding treatment. We present a method for decomposing fMRI-based functional connectivity (FC) into constituent biomechanisms - output drive, input responsivity, modulator gating - with clearer alignment to differentiable therapeutic interventions. Neurocircuit mechanism decomposition (NMD) integrates (i) a graph-constrained, lag-aware transformer to estimate directed, pathway-specific routing distributions and drive signals, with (ii) a measurement-aware state-space model (SSM) that models hemodynamic convolution and recovers intrinsic latent dynamics. This dual-head architecture yields interpretable circuit parameters that may provide a more direct bridge from fMRI to treatment strategy selection. We instantiate the model in an anatomically and electrophysiologically well-defined circuit: the cortico-basal ganglia-thalamo-cortical loop.

Perturbation screens hold the potential to systematically map regulatory processes at single-cell resolution, yet modeling and predicting transcriptome-wide responses to perturbations remains a major computational challenge. Existing methods often underperform simple baselines, fail to disentangle measurement noise from biological signal, and provide limited insight into the causal structure governing cellular responses. Here, we present the latent causal diffusion (LCD), a generative model that frames single-cell gene expression as a stationary diffusion process observed under measurement noise. LCD outperforms established approaches in predicting the distributional shifts of unseen perturbation combinations in single-cell RNA-sequencing screens while simultaneously learning a mechanistic dynamical system of gene regulation. To interpret these learned dynamics, we develop an approach we call causal linearization via perturbation responses (CLIPR), which yields an approximation of the direct causal effects between all genes modeled by the diffusion. CLIPR provably identifies causal effects under a linear drift assumption and recovers causal structure in both simulated systems and a genome-wide perturbation screen, where it clusters genes into coherent functional modules and resolves causal relationships that standard differential expression analysis cannot. The LCD-CLIPR framework bridges generative modeling with causal inference to predict unseen perturbation effects and map the underlying regulatory mechanisms of the transcriptome.

Continuous attractor networks (CANs) are a well-established class of models for representing low-dimensional continuous variables such as head direction, spatial position, and phase. In canonical spatial domains, transitions along the attractor manifold are driven by continuous displacement signals, such as angular velocity-provided by sensorimotor systems external to the CAN itself. When such signals are not explicitly provided as dedicated displacement inputs, it remains unclear whether attractor-based circuits can reliably acquire recurrent dynamics that support stable state transitions, or whether alternative predictive strategies dominate. In this work, we present an experimental framework for training CANs to perform successor-like transitions between stable attractor states in the absence of externally provided displacement signals. We compare two recurrent topologies, a circular ring and a folded snake manifold, and systematically vary the temporal regime under which stability is evaluated. We find that, under short evaluation windows, networks consistently converge to impulse-driven associative solutions that achieve high apparent accuracy yet lack persistent attractor dynamics. Only when stability is explicitly enforced over extended free-run periods do genuine attractor-based transition dynamics emerge. This suggests that shortcut solutions are the default outcome of local learning in recurrent networks, while attractor dynamics represent a constrained regime rather than a generic result. Furthermore, we demonstrate that topology strictly limits the capacity for learned transitions. While the continuous ring topology achieves perfect stability over long horizons, the folded snake topology hits a geometric limit characterized by failure at manifold discontinuities, which neither curriculum learning nor basal ganglia-inspired gating can fully overcome.

Large Language Models (LLMs) possess strong representation and reasoning capabilities, but their application to structure-based drug design (SBDD) is limited by insufficient understanding of protein structures and unpredictable molecular generation. To address these challenges, we propose Exploration-Augmented Latent Inference for LLMs (ELILLM), a framework that reinterprets the LLM generation process as an encoding, latent space exploration, and decoding workflow. ELILLM explicitly explores portions of the design problem beyond the model's current knowledge while using a decoding module to handle familiar regions, generating chemically valid and synthetically reasonable molecules. In our implementation, Bayesian optimization guides the systematic exploration of latent embeddings, and a position-aware surrogate model efficiently predicts binding affinity distributions to inform the search. Knowledge-guided decoding further reduces randomness and effectively imposes chemical validity constraints. We demonstrate ELILLM on the CrossDocked2020 benchmark, showing strong controlled exploration and high binding affinity scores compared with seven baseline methods. These results demonstrate that ELILLM can effectively enhance LLMs capabilities for SBDD.

Motivation: Opioid use disorder (OUD) often arises after prescription opioid exposure and follows transitions among onset, remission, and relapse. Linked EHR-survey resources such as the All of Us Research Program enable stage-specific risk modeling and connection to intervention options. Results: We built a multi-stage framework to model time-to-onset, time-to-remission, and time-to-relapse after remission using All of Us EHR and survey data. For each participant we derived longitudinal predictors from clinical conditions and survey concepts, including recent (1/3/12-month) event counts, cumulative exposures, and time since last event. We fit regularized Cox models for each transition and aggregated selection frequencies and hazard ratios to identify a compact set of high-confidence predictors. Pain, mental health, and polysubstance use contributed across stages: chronic pain syndromes, tobacco/nicotine dependence, anxiety and depressive disorders, and cannabis dependence prominently predicted onset and relapse, whereas tobacco dependence during remission and other remission-coded conditions were strongly associated with transition to remission. To support therapeutic prioritization, we constructed a therapy knowledge graph integrating genetic targets, biological pathways, and published evidence to map identified risk factors to candidate treatments in recent OUD studies and clinical guidelines.

Background: Conventional electrocardiogram (ECG) analysis faces a persistent dichotomy: expert-driven features ensure interpretability but lack sensitivity to latent patterns, while deep learning offers high accuracy but functions as a black box with high data dependency. We introduce ECGomics, a systematic paradigm and open-source platform for the multidimensional deconstruction of cardiac signals into digital biomarker. Methods: Inspired by the taxonomic rigor of genomics, ECGomics deconstructs cardiac activity across four dimensions: Structural, Intensity, Functional, and Comparative. This taxonomy synergizes expert-defined morphological rules with data-driven latent representations, effectively bridging the gap between handcrafted features and deep learning embeddings. Results: We operationalized this framework into a scalable ecosystem consisting of a web-based research platform and a mobile-integrated solution (https://github.com/PKUDigitalHealth/ECGomics). The web platform facilitates high-throughput analysis via precision parameter configuration, high-fidelity data ingestion, and 12-lead visualization, allowing for the systematic extraction of biomarkers across the four ECGomics dimensions. Complementarily, the mobile interface, integrated with portable sensors and a cloud-based engine, enables real-time signal acquisition and near-instantaneous delivery of structured diagnostic reports. This dual-interface architecture successfully transitions ECGomics from theoretical discovery to decentralized, real-world health management, ensuring professional-grade monitoring in diverse clinical and home-based settings. Conclusion: ECGomics harmonizes diagnostic precision, interpretability, and data efficiency. By providing a deployable software ecosystem, this paradigm establishes a robust foundation for digital biomarker discovery and personalized cardiovascular medicine.

Bell tests based on homodyne detection are strongly constrained in continuous-variable systems. Can Gottesman-Kitaev-Preskill (GKP) encoding turn homodyne detection into a practical tool for revealing Bell nonlocality? We consider a physically motivated model in which each party performs homodyne detection and digitizes the continuous outcome via a fixed periodic binning, corresponding to logical Pauli measurements. Within this framework, we derive a bipartite no-go: CHSH cannot be violated for Bell-pair states. Moving beyond two parties, we show that finitely squeezed GKP-encoded GHZ and W states nevertheless exhibit strong multipartite nonlocality, violating multipartite Bell inequalities with homodyne-only readout. We quantify the required squeezing thresholds and robustness to loss, providing a route toward homodyne-based Bell tests in continuous-variable systems.

At the heart of quark confinement and hadronization, the physics of flux strings has recently become a focal point in the field of quantum simulation of high-energy physics (HEP). Despite considerable progress, a detailed understanding of the behavior of flux strings in quantum simulation-relevant lattice formulations of gauge theories has remained limited to the lowest truncations of the gauge field, which are severely limited in their ability to draw conclusions about the quantum field theory limit. Here, we employ tensor network simulations to investigate the behavior of flux strings in a quantum link formulation of $2+1$D quantum electrodynamics (QED) with a spin-$1$ representation of the gauge field. We first map out the ground-state phase diagram of this model in the presence of two spatially separated static charges, revealing distinct microscopic processes responsible for string breaking, including a two-stage breaking mechanism not possible in the spin-$\frac{1}{2}$ formulation. Starting in different initial product state string configurations, we then explore far-from-equilibrium quench dynamics across various parameter regimes, demonstrating genuine $2+1$D real-time string breaking and glueball-like bound state formation, with the latter not possible in the spin-$\frac{1}{2}$ formulation. In and out of equilibrium, we consider different values and placements of the static charges. Finally, we provide efficient qudit circuits for a quantum simulation experiment in which our results can be observed in state-of-the-art ion-trap setups. Our findings lay the groundwork for quantum simulations of flux strings towards the quantum field theory limit.

We present a method for associating labeled directed graphs to finite-dimensional Lie algebras, thereby enabling rapid identification of key structural algebraic features. To formalize this approach, we introduce the concept of graph-admissible Lie algebras and analyze properties of valid graphs given the antisymmetry property of the Lie bracket as well as the Jacobi identity. Based on these foundations, we develop graph-theoretic criteria for solvability, nilpotency, presence of ideals, simplicity, semisimplicity, and reductiveness of an algebra. Practical algorithms are provided for constructing such graphs and those associated with the lower central series and derived series via an iterative pruning procedure. This visual framework allows for an intuitive understanding of Lie algebraic structures that goes beyond purely visual advantages, since it enables a simpler and swifter grasping of the algebras of interest beyond computational-heavy approaches. Examples, which include the Schrödinger and Lorentz algebra, illustrate the applicability of these tools to physically relevant cases. We further explore applications in physics, where the method facilitates computation of similtude relations essential for determining quantum mechanical time evolution via the Lie algebraic factorization method. Extensions to graded Lie algebras and related conjectures are discussed. Our approach bridges algebraic and combinatorial perspectives, offering both theoretical insights and computational tools into this area of mathematical physics.

Many physical phenomena, including thermalization in open quantum systems and quantum Gibbs sampling, are modeled by Lindbladians approximating a system weakly coupled to a bath. Understanding the convergence speed of these Lindbladians to their steady states is crucial for bounding algorithmic runtimes and thermalization timescales. We study two such families of processes: one characterizing a repeated-interaction Gibbs sampling algorithm, and another modeling open many-body quantum thermalization. We prove that both converge in polynomial time for several non-commuting systems, including high-temperature local lattices, weakly interacting fermions, and 1D spin chains. These results demonstrate that simple dissipative quantum algorithms can prepare complex Gibbs states and that Lindblad dynamics accurately capture thermal relaxation. Our proofs rely on a novel technical result that extrapolates spectral gap lower bounds from quasi-local Lindbladians to the non-local generators governing these dynamics.

We introduce a class of pseudo-bosonic Klein-Gordon fields in 1+1 dimensions and we discuss some of their properties. This work originates from non Hermitian quantum mechanics and deformed canonical commutation relations. We show that, within this class of fields, there exist a specific subclass with the interesting feature of having finite equal space-time two-points function, contrarily to what happens for {\em standard} Klein-Gordon fields. This, in our opinion, is a relevant aspect of our proposal which is a good motivation to undertake a deeper analysis of this (and related) quantum fields.

Real-time decoding of quantum error correction (QEC) is essential for enabling fault-tolerant quantum computation. A practical decoder must operate with high accuracy at low latency, while remaining robust to spatial and temporal variations in hardware noise. We introduce a hardware-conditioned neural decoder framework designed to exploit the natural separation of timescales in superconducting processors, where calibration drifts occur over hours while error correction requires microsecond-scale responses. By processing calibration data through a graph-based encoder and conditioning a lightweight convolutional backbone via feature-wise linear modulation (FiLM), we decouple the heavy processing of device statistics from the low-latency syndrome decoding. We evaluate this approach using the 1D repetition code as a testbed on IBM Fez, Kingston, and Pittsburgh processors, collecting over 2.7 million experimental shots spanning distances up to d = 11. We demonstrate that a single trained model generalizes to unseen qubit chains and new calibration data acquired days later without retraining. On these unseen experiments, the FiLM-conditioned decoder achieves up to an 11.1x reduction in logical error rate relative to modified minimum-weight perfect matching. We observe that by employing a network architecture that exploits the highly asynchronous nature of system calibration and decoding, hardware-conditioned neural decoding demonstrates promising, adaptive performance with negligible latency overhead relative to unconditioned baselines.

McDonald and Clerk [Phys.\ Rev.\ Research 5, 033107 (2023)] showed that for linear open quantum systems the Liouvillian spectrum is independent of the noise strength. We first make this noise-independence principle precise in continuous time for multimode bosonic Gaussian Markov semigroups: for Hurwitz drift, a time-independent Gaussian similarity fixed by the Lyapunov equation gauges away diffusion for all times, so eigenvalues and non-diagonalizability are controlled entirely by the drift, while diffusion determines steady states and the structure of eigenoperators. We then extend the same separation to discrete time for general stable multimode bosonic Gaussian channels: for any stable Gaussian channel, we construct an explicit Gaussian similarity transformation that gauges away diffusion at the level of the channel parametrization. We illustrate the method with a single-mode squeezed-reservoir Lindbladian and with a non-Markovian family of single-mode Gaussian channels, where the exceptional-point manifolds and the associated gauging covariances can be obtained analytically.

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

While quantum metrology enables measurement precision beyond classical limits, its performance is often susceptible to experimental imperfections. Most prior studies have focused on imperfections in quantum states and operations. Here, we investigate the effect of coarse graining in quantum measurement through both theoretical analysis and experimental demonstration. Using an interferometer with a squeezed vacuum and a laser input, we analyze how coarse graining in homodyne detection affects the precision of phase estimation. We evaluate the Fisher information under various coarse-graining conditions and determine, in each case, an optimal estimation strategy that saturates the Cramér-Rao bound. Remarkably, even extremely coarse-grained measurement -- with only two bins -- enables phase estimation beyond the standard quantum limit and even achieves a precision that follows the Heisenberg scaling. We experimentally demonstrate quantum-enhanced phase estimation under coarse-grained homodyne detection. To determine an optimal estimation strategy, we employ the method of moments and present calibration procedures that enable its application to general experimental settings. Using only two bins, we observe a quantum enhancement of 1.2 dB compared to the classical method using the ideal measurement, improving towards 3.8 dB as the bin number increases. These results highlight a practical pathway to achieving quantum enhancement under the presence of severe experimental imperfections.

Due to the tidal forces that a black hole can produce, certain types of compact objects may undergo disruption as they approach the black hole. This disruption point is known as the Roche limit (or Roche radius). In this work, we studied the tidal forces arising from the presence of the Simpson--Visser black bounce. We analyzed the tidal forces both for a static observer and for a radially infalling observer and showed that differences arise depending on the choice of observer. We used the tidal forces together with the stellar binding forces to determine the Roche radius for neutron stars, white dwarfs, and Sun-like stars, and to investigate how the Simpson--Visser regularization affects the tidal disruption of these astrophysical objects. We also examined whether, for astrophysical black holes such as M87* and Sgr~A*, these stellar disruption processes occur inside or outside the event horizon, and thus whether they are observable.

A relevant signal in the quantum domain may manifest as a displacement or a phase shift operator in the bosonic phase space. For a real parameter $β$ embedded in such a displacement operator, the task of determining if $β\in [β_{-th}, β_{+th}]$ for real asymmetric thresholds $(β_{-th} \ne -β_{+th})$ is a binary decision problem. We propose a framework based on generalized quantum signal processing interferometry (GQSPI) on hybrid qubit-bosonic oscillator systems that addresses this parameter detection problem by recasting the practical task of active binary hypothesis testing on quantum systems to that of a polynomial approximation. We achieve a small decision error probability $p_{err}$ on the order of $O(\frac{1}{d}\log{(d)})$, with $d$ as the circuit depth. We analyze the protocol when (i) $β$ is a deterministic parameter, and (ii) when $β$ is drawn randomly from a known prior distribution. The performance of the sensing protocol under dephasing noise is also shown to be robust. We further extend our protocol from two thresholds to more general multi-threshold cases as well. Overall, the proposed framework enables decision-making over arbitrary thresholds for any general displacement signal in a single or a few shots.

Fast radio bursts (FRBs) are commonly divided into repeating and apparently non-repeating sources, but whether these represent distinct physical populations remains uncertain. In this work, we apply an unsupervised machine learning methods combining Uniform Manifold Approximation and Projection (UMAP) with density-based clustering to analyze CHIME/FRB Catalog 2. We find that FRBs remain primarily separated into two clusters in the multi-dimensional parameter space, with a recall of 0.94 for known repeaters, indicating strong robustness. Consistent with Catalog 1 analyses, we confirm that the spectral morphology parameter, specifically spectral running remains the key discriminator between the two populations, indicating that narrowband emission is an intrinsic and persistent property of repeating FRBs. With the enlarged Catalog 2 sample, we further identify a stable subclass of atypical repeaters (about $6\%$ of repeating bursts) that are broadband, shorter in duration, and more luminous, resembling non-repeating bursts. The Nonrepeater-like cluster also shows higher inferred energies and dispersion measures, consistent with a scenario in which apparently non-repeating FRBs may result from observational incompleteness, with low-energy repeating bursts remaining undetected. Our results provide new statistical evidence for a physical connection between repeating and non-repeating FRBs.

In this work, we show that if the bumblebee field in the Einstein-bumblebee theory is given by its vacuum expectation value ($B_μ=b_μ$) and it is not dynamical ($\partial_μB_ν-\partial_νB_μ=0$), then these conditions uniquely provide a generating technique, allowing us to construct exact solutions to bumblebee gravity from the vacuum solutions by adding a term $\sim b_μb_ν$ to the metric tensor. Also, we show that the bumblebee field within this technique is proportional to the tangential vector of the (timelike or spacelike) geodesic curve in the background vacuum spacetime, and can be easily found knowing the solution to the Hamilton-Jacobi equation. Moreover, we prove that this technique can be extended to the case of any non-zero cosmological constant and the presence of the electromagnetic field. We apply this generating technique and obtain the bumblebee extension of the Kerr-Newman-Taub-NUT-(anti-)de Sitter spacetime. We show that this extension is not unique, as it depends on the exact geodesic curve one chooses to associate a bumblebee field with. Then, by considering various special cases of this generic solution, we demonstrate that the condition of the global reality of the bumblebee field limits the set of geodesics with which we can associate it.

Quantum metrology typically demands the preparation of exotic quantum probe states, such as entangled or squeezed states, to surpass classical limits. However, the need for carefully calibrated system parameters and finely optimized quantum controls imposes limitations on scalability and robustness. Here, we circumvent these limitations by introducing an echoed random process that achieves sensitivity approaching the Heisenberg limit while remaining blind to the random probe state. We demonstrate that by simply driving a Kerr nonlinear mode with random pulses, the emergence of sub-Planck phase-space structures grants high sensitivity, eliminating the need for complex quantum control. The protocol is statistically robust, yielding high performance across broad driving parameter ranges while exhibiting resilience to control fluctuations and photon loss. Broadly applicable to both bosonic and qubit platforms, our work reveals a practical, hardware-efficient, scalable, and optimization-free route to quantum-enhanced metrology in high-dimensional Hilbert spaces.

Nonlinear effects play a fundamental role in the late-time ringdown of black holes, with direct implications for gravitational-wave observations. For massive fields, these dynamics become richer, yet their nonlinear signatures remain poorly understood. Here, we systematically study nonlinear tails of massive scalar perturbations, from a toy model with ingoing and outgoing sources to a self-interacting scalar model, revealing nonlinear tails and contrasting the results with their linear counterparts. We find that the nonlinear tails of massive scalar fields, opposite to massless ones, decay as the same rate as linear tails in the intermediate time, independent of source parameters or initial conditions. Nevertheless, quadratic quasinormal modes could serve as a probe to the nonlinear effects of massive fields.

We implement and benchmark on IBM Quantum hardware the circuit family proposed by Violaris for estimating operational inter-branch communication witnesses, defined as correlations in classical measurement records produced by compiled Wigner's-friend-style circuits. We realize a five-qubit instance of the protocol as an inter-register message-transfer pattern within a single circuit, rather than physical signaling, and evaluate its behavior under realistic device noise and compilation constraints. The circuit encodes branch-conditioned evolution of an observer subsystem whose dynamics depend on a control qubit, followed by a controlled transfer operation that probes correlations between conditional measurement contexts. Executing on the ibm_fez backend with 20000 shots, we observe population-based visibility of 0.877, coherence witnesses of 0.840 and -0.811 along orthogonal axes, and a phase-sensitive magnitude of approximately 1.17. While the visibility metric is insensitive to some classes of dephasing, the coherence witnesses provide complementary sensitivity to off-diagonal noise. This work does not test or discriminate among interpretations of quantum mechanics. Instead, it provides a reproducible operational constraint pipeline for evaluating detectability of non-ideal channels relative to calibrated device noise.

We propose a new approach to engineer the quantum Mpemba effect (QME) -- wherein an initial state farther from system relaxes faster than a close one -- by the Liouvillian skin effect (LSE) in open quantum systems. Moreover, the LSE serves as an ideal platform for realizing the QME and the spatial profile of the LSE provides a straightforward pathway for the initial state preparation, thereby enabling readily accessible experimental preparation. Focusing on the quadratic Lindbladians, we consider two concrete cases to design the initial states, thereby realizing the QME. Interestingly, we uncover a new kind of QME (QME-III) that is distinct from the two typical scenarios, manifested as two reversals in the Hilbert-Schmidt distance at two different times. In particular, the LSE provides a physically more intuitive understanding of the QME.

We study a general bipartite quantum system consisting of a spin interacting with a bosonic field, with the initial state prepared as the product of a spin coherent state and a canonical coherent state. Our goal is to develop a semiclassical framework to describe the entanglement dynamics between these two subsystems. Using appropriate approximations, we derive a semiclassical expression for the entanglement entropy that depends exclusively on the trajectories of the underlying classical description. By analytically extending the classical phase space into the complex domain, we identify additional complex trajectories that significantly improve the accuracy of the semiclassical description. The inclusion of these complex trajectories allows us to capture the entanglement dynamics with remarkable precision, even well beyond the Ehrenfest time. The approach is illustrated with a representative example, where the role of real and complex trajectories in reproducing the quantum entanglement entropy is explicitly demonstrated.

We develop the covariant phase space formulation of Weyl-transverse gravity (WTG) in the presence of general timelike and spacelike boundaries. WTG is classically equivalent to General Relativity (GR) but possesses a reduced gauge symmetry consisting of Weyl transformations and transverse diffeomorphisms, together with a fixed background volume form. This structure modifies the variational principle and the definition of conserved quantities relative to GR. We derive the symplectic potential, presymplectic current, and Hamiltonian generators associated with transverse diffeomorphisms, and we identify a set of boundary conditions under which the WTG action is differentiable. These include Dirichlet and Neumann conditions for both the auxiliary Weyl-invariant metric and the dynamical metric, as well as a natural implementation of York boundary conditions, for which WTG exhibits a particularly transparent geometric formulation. We obtain the Noether current and surface charge, clarify the role of the Lagrangian ambiguity related to the cosmological constant, and evaluate the Hamiltonian identity on spacetimes containing a bifurcate Killing horizon. The resulting first-law relation shows that variations of the cosmological constant can contribute nontrivially unless additional physical restrictions are imposed.

Superconducting nanowire single-photon detectors (SNSPDs) have emerged as the highest performing photon-counting detectors, making them a critical technology in quantum photonics and photon-starved optical sensing. However, the performance of SNSPDs is limited not by the intrinsic properties of the superconducting film, but by edge-induced current crowding. Despite extensive materials optimization and increasingly demanding fabrication strategies aimed at mitigating this edge-limited behavior, the device edges continue to limit the maximum device operating current, thereby degrading key performance metrics. Here, we demonstrate for the first time in situ tuning of a detector from an edge-limited to a bulk-limited regime, allowing the device to reach its intrinsic performance limit. Our approach is based on current-biased superconducting "rails" placed on either side of the detector to suppress current crowding at the edges. We show that activation of the rails reduces the dark count rate by nine orders of magnitude and extends the photon detection plateau at 1550 nm by more than 40%. These results are demonstrated on detectors up to 0.1 mm wide, establishing an entirely new class of ultra-wide strip detectors that we call superconducting strip photon detectors (SSPD). Moreover, the ability to suppress edge current crowding using the rails provides a pathway toward SSPDs with strip widths extending into the mm-scale. Such devices will enable large-area, high efficiency SSPD arrays with infrared sensitivity and open new opportunities in applications ranging from biomedical imaging to deep space optical communication.

We present a non-perturbative, first-principles derivation of renormalization relations for waveguide-QED models, explicitly accounting for the infrared (IR) and ultraviolet (UV) cutoffs that are necessarily introduced in numerical simulations. By formulating the atomic dynamics in the time domain, we obtain explicit expressions linking the bare model parameters to the physically observable atomic frequency and decay rate, and verify their consistency with scattering theory. We further connect these results to standard Feynman diagrams, providing a transparent physical interpretation and ensuring the generality of the approach. Finally, we show how these renormalization relations can be used to parameterize simulations with a minimal frequency bandwidth, simultaneously preserving physical accuracy and reducing computational cost, thereby paving the way for efficient and reliable multi-photon light-matter simulations.

Quantum circuit optimization - the process of transforming a quantum circuit into an equivalent one with reduced time and space requirements - is crucial for maximizing the utility of current and near-future quantum devices. While most automated optimization techniques focus on transforming circuits into equivalent ones that implement the same unitary, we show that substantial new opportunities for resource reduction can be achieved by (1) allowing approximate local transformations and (2) employing mixed quantum channels to approximate pure circuits. Our novel automated protocol for approximate circuit rewriting is a refined evolution of automated optimization techniques based on the ZX-calculus, where we add a greedy strategy that selectively replaces ZX-diagrams with small phase angles with stochastic mixtures of the identity and carefully chosen over-rotations, which are designed to reduce the overall gate count in expectation while staying within a strict error budget. This approach yields modest two-qubit gate count reduction in random quantum circuits, and achieves a substantial reduction in structured circuits such as the quantum Fourier transform. Fundamentally, our protocol converts experimental noise due to gate applications into deliberately engineered random noise, outperforming many other approximation methods on average. These results highlight the potential of mixed-channel approximations to enhance future quantum circuit performance, suggesting new directions for resource-aware automated quantum compilation beyond pure unitary channels.

Cosmological first-order phase transitions are a well-motivated source of stochastic gravitational waves (GWs), but most predictions are made based on the highly idealized model of perfectly spherical vacuum bubbles, neglecting thermal fluctuations. In this work we use $(3+1)$-dimensional lattice simulations of a scalar model with thermal initial conditions to quantify how thermal fluctuations distort bubble profiles and modify the resulting GW spectrum. We find that thermal fluctuations can strongly break spherical symmetry at early times, allowing even an isolated bubble to emit GWs. In multi-bubble simulations, thermal fluctuations systematically reshape the spectrum, suppressing the infrared part while enhancing and broadening the high-$k$ tail. We further provide an analytical estimate for the ultraviolet regime of the GW spectrum, which is in good agreement with our lattice results and suggests that this regime is dominated by thermal fluctuations. These effects could leave observable imprints in future GW searches.

In this work the machinery of q-deformed algebras are used to enhance cryptographic security during teleportation. We use q-deformed harmonic oscillator states to develop a novel method of teleportation. The deformed states can be expressed in terms of standard oscillator states and the expressions contain certain arbitrary functions of $q$. It is the presence of these arbitrary functions that allows an enhancement of cryptographic security. The specifics are : (a) q-deformed Bell-like states are constructed which reduce to the usual Bell states when the deformation parameter $q\rightarrow 1$. These deformed states form an orthonormal basis for q-deformed entangled bipartite states when certain arbitrary functions of $q$ satisfy a constraint. (b) We discuss the generalisation of the usual teleportation protocol with non-maximally entangled states. This generalisation is then employed to construct two new protocols using q-deformed non-maximally entangled states. These states have additional parameters and these have to be shared for decryption after teleportation. Consequently, the cryptographic security is improved.

We consider discrete spacetime models known as quantum walks, which can be used to simulate Dirac particles. In particular we look at fermion doubling in these models, in which high momentum states yield additional low energy solutions which behave like Dirac particles. The presence of doublers carries over to the `second quantised' version of the walks represented by quantum cellular automata, which may lead to spurious solutions when introducing interactions. Moreover, we also consider pseudo-doublers, which have high energy but behave like low energy Dirac particles, and cause potential problems regarding the stability of the vacuum. To address these issues, we propose a family of quantum walks, that are free of these doublers and pseudo-doublers, but still simulate the Dirac equation in the continuum limit. However, there remain a small number of additional low energy solutions which do not directly correspond to Dirac particles. While the conventional Dirac walk always has a zero probability for the walker staying at the same point, we obtain the family of walks by allowing this probability to be non-zero.

We study a spacetime obtained from the semi-classical backreaction computed via the Thermofield dynamics approach in the Poincare patch of de Sitter spacetime. The resulting bulk equation takes the Whittaker form and we examine two distinct applications. At leading order, the co-moving curvature perturbations are shown to match a constant-roll model in the frozen attractor regime, corresponding to a UV enhancement of the spectrum with $n_S \sim 2$. In the holographic context, we compute the CFT two-point function at the future boundary, and away from it we construct the flow-equation of the dual QFT that matches beta-function of the Sp(N ) model in three dimensions.

Simulation of quantum field theories and fundamental interactions are one of the most challenging tasks in modern particle physics. Classical computers generally fail to reproduce accurate results when it comes to strongly coupled theories such as QCD. Recent developments in quantum technologies open up the possibility of simulating such physical regimes by using quantum computers. In this paper, we study the quantum resource related to the simulability of a quantum theory, i.e. non-stabilizerness for Lattice Gauge Theory (LGT) with discrete symmetry gauge groups. We show that enforcing gauge constraints for $\mathbb{Z}_l$ LGTs has no cost in terms of this resource and discuss the relation between non-abelianity of the gauge group with the average non-stabilizerness of the gauge invariant Hilbert space.

We describe a categorical model of MALL (Multiplicative Additive Linear Logic) inspired by the Heisenberg-Schrödinger duality of finite-dimensional quantum theory. Proofs of formulas with positive logical polarity correspond to CPTP (completely positive trace-preserving) maps in our model, i.e. the quantum operations in the Schrödinger picture, whereas proofs of formulas with negative logical polarity correspond to CPU (completely positive unital) maps, i.e. the quantum operations in the Heisenberg picture. The mathematical development is based on noncommutative geometry and finite-dimensional von Neumann (co)algebras, which can be defined as special kinds of (co)monoid objects internal to the category of finite-dimensional operator spaces.

The Galactic Center black hole environment gives us new opportunity to test deviation from General Relativity and black hole physics. In this work we analytically generate the shape of the Galactic Center black hole by using a recently developed exact stationary, axisymmetric and vacuum solution of $f(R)$ gravity theory. By using scalaron mass as a free parameter we find that the shadow shape along with displacement and asymmetry is sensitive to the scalaron mass, even after keeping the black hole spin low. We recognize scalaron mass which is compatible with Kerr like quadrupole moment and hence black hole "no-hair" theorem. The same mass scale is found to reproduce the PPN parameter ($γ$) constrained in the weak field limit of the solar system. Gravitational identifiers, the Kretschmann scalar ($κ$) and gravitational potential ($φ$) have been used to infer scalaron masses in the regime of S-stars which are found to be consistent with the limits obtained using shadow scales. We ensure that $f(R)$ gravity scalaron has an appropriate general relativistic limit in the horizon scale of the black hole. We also identify the possibility of scale invariance of the General Relativistic limit.

Simulating noiseless quantum dynamics classically faces a fundamental dilemma: tensor-network methods become inefficient as entanglement saturates, while Pauli-truncation approaches typically rely on noise or randomness. To close the gap, we propose the Low-weight Pauli Dynamics (LPD) algorithm that efficiently approximates local observables for short-time dynamics in the absence of noise. We prove that the truncation error admits an average-case bound without assuming randomness, provided that the state is sufficiently entangled. Counterintuitively, entanglement--usually an obstacle for classical simulation--alleviates classical simulation error. We further show that such entangled states can be generated either by tensor-network classical simulation or near-term quantum devices. Our results establish a rigorous synergy between existing classical simulation methods and provide a complementary route to quantum simulation that reduces circuit depth for long-time dynamics, thereby extending the accessible regime of quantum dynamics.

Quantum approximate optimization algorithm (QAOA) have promising applications in combinatorial optimization problems (COPs). We investigated the MaxCut problem in three different families of graphs using QAOA ansats with parameter transfer initialization followed by targeted single layer optimization. For 3 regular (3R), Erdos Renyi (ER), and Barabasi Albert (BA) graphs, the parameter transfer approach achieved mean approximation ratios of 0.9443 for targeted-single layer optimization as compared to 0.9551 of full optimization. It represents 98.88 percent optimal performance, with 8.06 times computational speedup in unweighted graphs. But, in weighted graph families, optimal performance is relatively low (less than 90 percent) for higher nodes graph, suggesting parameter transfer followed by targeted-single-layer optimization is not ideal for weighted graph families, however, we find that for some weighted families (weighted 3-regular) this approach works perfectly. In 8.92 percent test cases, targeted single layer optimization outperformed the full optimization, indicating that complex parameter landscape can trap full optimization in sub-optimal local minima. To mitigate this inconsistency, ridge (L2) regularization is used to smoothen the solution landscape, which helps the optimizer to find better optimum parameters during full optimization and reduces these inconsistent test cases from 8.92 percent to 3.81 percent. This work demonstrates that efficient parameter initialization and targeted-single-layer optimization can improve the efficiency of QAOA with minimal performance degradation.

In conventional lattice models, the dispersion relation $ω(k)$ is assumed to be a smooth function. We prove that this smoothness implies the splitting of an initially localized excitation into counter-propagating wave packets. Consequently, unsplit spreading can occur only when $ω(k)$ develops singular features, precisely what long-range interactions enable. Remarkably, this phenomenon was clearly visible in published quantum simulation experiments as early as 2014, yet it has remained unrecognized or discussed as a distinct physical effect. We show that unsplit spreading emerges in realistic open quantum systems, such as 1D and 2D subwavelength atomic arrays, where the long-lived subradiant states host effective dispersion with the required singularities. Our work establishes unsplit spreading as an experimentally accessible, smoking-gun signature of singular band structure induced by long-range physics.

We generalize the generalized-squeezing problem to include fractional values of the squeezing order $n$. This approach allows us to determine the locations of critical points at which qualitative changes in behaviour occur and accurately predict the behaviour at these critical points, which are challenging for conventional computational methods. Based on our numerical calculations, we identify with a high degree of confidence the point at which the spectrum turns from continuous to discrete and the point at which oscillations turn from having asymptotically infinite amplitudes to finite amplitudes. Furthermore, we numerically investigate the behaviour in the large $n$ regime and provide an intuitive explanation that coincides with the numerical results.

We develop a workflow within the ONIOM framework and demonstrate it on the hybrid computing system consisting of the supercomputer Fugaku and the Quantinuum Reimei trapped-ion quantum computer. This hybrid platform extends the layered approach for biomolecular chemical reactions to accurately treat the active site, such as a protein, and the large and often weakly correlated molecular environment. Our result marks a significant milestone in enabling scalable and accurate simulation of complex biomolecular reactions

We observe a metrological advantage in phase-space sensitivity for photon-added cat and kitten states over their original forms, due to phase-space broadening from increased amplitude via photon addition, albeit with higher energy cost. Using accessible non-classical resources, weak squeezing and displacement, we construct a squeezed state and two superposed states: the squeezed cat state and the symmetrically squeezed state. Their photon-added variants are compared with parity-matched cat and KSs using quantum Fisher information and fidelity. The QFI isocontours reveal regimes where KS exhibit high fidelity and large amplitude, enabling their preparation via Gaussian operations and photon addition. Similar regimes are identified for cat states enhanced by squeezing and photon addition, demonstrating improved metrological performance. Moreover, increased amplitude and thus larger phase-space area reduces the size of interferometric fringes, enhancing the effectiveness of quantum error correction in cat codes.

A couple of milliseconds after the merger of a binary system of neutron stars can play a fundamental role in amplifying the comparatively low initial magnetic fields into magnetar strengths. The basic mechanism responsible for this amplification is the Kelvin-Helmholtz instability (KHI) and we here report the first systematic study of the impact of rapid rotation on the KHI-amplification process exploiting general-relativistic magnetohydrodynamic simulations at very high-resolutions of $35\,{\rm m}$. Concentrating on four different spinning configurations, we find that aligned, anti-aligned, and mixed (aligned/anti-aligned) spin configurations lead to markedly different growth rates of the electromagnetic (EM) energy, field topologies, and vortex properties when compared to the irrotational case. These differences arise from intrinsic variations in the system dynamics, such as tidal deformation, collision strength, and contact surface area, with the anti-aligned configuration producing the largest vorticity and growth in EM energy. Importantly, while different spin configurations lead to significantly different initial growth rates of the poloidal/toroidal components, all systems converge to a specific topological partition. Our simulations are confined to a short window in time, but the different EM energies produced as a result of spin will imprint the EM emission at merger and provide information on the spinning state at merger.

We propose a phase-difference estimation algorithm based on the tensor-network circuit compression, leveraging time-evolution data to pursue scalability and higher accuracy on a quantum phase estimation (QPE)-type algorithm. Using tensor networks, we construct circuits composed solely of nearest-neighbor gates and extract time-evolution data by four-type circuit measurements. In addition, to enhance the accuracy of time-evolution and state-preparation circuits, we propose techniques based on algorithmic error mitigation and on iterative circuit optimization combined with merging into matrix product states, respectively. Verifications using a noiseless simulator for the 8-qubit one-dimensional Hubbard model using an ancilla qubit show that the proposed algorithm achieves accuracies with 0.4--4.7\% error from a true energy gap on an appropriate time-step size, and that accuracy improvements due to the algorithmic error mitigation are observed. We also confirm the enhancement of the overlap with matrix product states through iterative optimization. Finally, the proposed algorithm is demonstrated on IBM Heron devices with Q-CTRL error suppression for 8-, 36-, and 52-qubit models using more than 5,000 2-qubit gates. These largest-scale demonstrations for the QPE-type algorithm represent significant progress not only toward practical applications of near-term quantum computing but also toward preparation for the era of error-corrected quantum devices.

The formation of a cusp on a black hole shadow is a striking signature of physics beyond the Kerr paradigm. We demonstrate that this morphological change fundamentally alters the shadow's topology with the topological charge flipping from 1 to -1. To analyze this topological transition, we introduce a gravitational equal-area law, analogous to Maxwell's construction in thermodynamics, and identify a critical point for cusp formation. Near this point, we uncover universal behavior characterized by a critical exponent 1/2, which places this gravitational lensing system within the mean-field universality class. These results establish a new framework for testing fundamental physics of black hole shadows, reframing the search for deviations from general relativity as a targeted hunt for a distinct topological and critical phenomenon.

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

Atom interferometers require both high efficiency and robust performance in their mirror pulses under experimental inhomogeneities. In this work, we demonstrated that quantum optimal control designed mirror pulse significantly enhance interferometer performance by using novel adaptive sliced structure. Using gradient ascent pulse engineering (GRAPE), optimized mirror pulse for a Mach-Zehnder light-pulse atom interferometer was designed by discretizing the control into non-uniform phase slices. This design broadened the tolerence to experimentally relevant variations in detuning $[-Ω_0,Ω_0]$ and Rabi frequency $[0.1\timesΩ_0,1.9\timesΩ_0]$ ($Ω_0=2π\times25$ kHz), while maintaining high transfer efficiency even when the response-time delays up to 1.6 $\rm{μs}$. The optimized pulse was found to be robust to coupling inhomogeneity and velocity spread, offering a significant improvement in robustness over conventional pulse. The adaptive pulse slicing method provides a minimalist strategy that reduces experimental complexity while enhancing robustness and scalability, offering an innovative scheme for quantum optimal control in high precision atom interferometry.

Squeezed states of light enable enhanced measurement precision by reducing noise below the standard quantum limit. A key application of squeezed light is nonlinear microscopy, where state-of-the-art performance is limited by photodamage and quantum-limited noise. Such microscopes require bright, pulsed light for optimal operation, yet generating and detecting bright pulsed squeezing at high levels remains challenging. In this work, we present an efficient technique to generate high levels of bright picosecond pulsed squeezed light using a $χ^2$ optical parametric amplification process in a waveguide. We measure $-3.2~\mathrm{dB}$ of bright squeezing with optical power compatible with nonlinear microscopy, as well as $-3.6~\mathrm{dB}$ of vacuum squeezing. Corrected for losses, these squeezing levels correspond to $-15.4^{+2.7}_{-8.7}~\mathrm{dB}$ of squeezing generated in the waveguide. The measured level of bright amplitude pulsed squeezing is to our knowledge the highest reported to date, and will contribute to the broader adoption of quantum-enhanced nonlinear microscopy in biological studies.

Color center spins in 4H-SiC offer a rare combination of wafer-scale materials maturity with long spin coherence and chip-level photonics, making them promising building blocks for scalable quantum technologies. In particular, the silicon vacancy hosts an S=3/2 ground state, a native qudit that enables compact encodings and subspace-selective control, but also introduces spectator transitions: short, detuned pulses can coherently drive non-addressed level pairs and create crosstalk. Here we use broadband Ramsey interferometry to reveal and quantify such spectator-transition crosstalk. Experimentally, the Ramsey Fourier spectra display multiple lines beyond the addressed single-quantum transition. Analytically, we map each line to a pairwise energy difference between qudit levels of the rotating-frame Hamiltonian and assign its weight via compact amplitudes set by the prepared state and the microwave pulse parameters, predicting a deterministic six-branch structure. Numerical time-domain propagation with the experimental sampling reproduces the detuning map, and the measured peak positions coincide with the analytic branch lines without frequency fitting. Together these results provide a practical, spectator-aware framework for multilevel control in the silicon vacancy qudit. The approach offers clear guidance to suppress crosstalk or, conversely, to exploit spectator lines, for example as additional constraints for in situ pulse calibration and for phase-sensitive quantum state and process estimation.

Bouncing cosmologies offer an alternative to inflation by resolving the initial singularity through a contracting phase followed by a bounce into expansion. In many such models, the contracting phase is dominated by a single matter component, typically pressureless dust, which leads to an almost scale-invariant spectrum of scalar cosmological perturbations with a slight blue tilt, so that generating the observed red-tilted spectrum within this framework was challenging. In this work, we consider a more realistic scenario in which the contracting phase includes both matter and radiation, as required on physical grounds. We show that the presence of radiation can naturally induce a red tilt in the spectrum of curvature perturbations seeded by quantum vacuum fluctuations in the remote past of the contraction. Since the perturbations of the two fluids are coupled via gravity, vacuum initial conditions must be carefully defined. We demonstrate that, without fine-tuning, the resulting entropy perturbations are subdominant with respect to curvature perturbations. This suggests that a minimal two-component bounce model, involving only ordinary matter and radiation, can connect to the standard expanding cosmology with observationally viable initial conditions.

The Fokker-Planck equation models rare events across sciences, but its high-dimensional nature challenges classical computers. Quantum algorithms for such non-unitary dynamics often suffer from exponential {decay in} success probability. We introduce a quantum algorithm that overcomes this for computing reaction rates. Using a sum-of-squares representation, we develop a Gaussian linear combination of Hamiltonian simulations (Gaussian-LCHS) to represent the non-unitary propagator with $O\left(\sqrt{t\|H\|\log(1/ε)}\right)$ queries to its block encoding. Crucially, we pair this with {a} novel technique to directly estimate matrix elements without exponential decay. For $η$ pairwise interacting particles discretized with $N$ plane waves per degree of freedom, we estimate reactive flux to error $ε$ using $\widetilde{O}\left((η^{5/2}\sqrt{tβ}α_V + η^{3/2}\sqrt{t/β}N)/ε\right)$ quantum gates, where $α_V = \max_{r}|V'(r)/r|$. For non-convex potentials, the {sharpest classical} worst-case analytical bounds to simulate the related overdamped Langevin {equation} scale as $O(te^{Ω(η)}/ε^4)$. This {implies} an exponential separation in particle number $η$, a quartic speedup in $ε$, and quadratic speedup in $t$. While specialized classical heuristics may outperform these bounds in practice, this demonstrates a rigorous route toward quantum advantage for high-dimensional dissipative dynamics.

This whitepaper presents NWQWorkflow, an end-to-end workflow for quantum application development, compilation, error correction, benchmarking, numerical simulation, control, and execution on a prototype superconducting testbed. NWQWorkflow integrates NWQStudio (programming GUI environment), NWQASM (intermediate representation), QASMTrans (compiler), NWQEC (quantum error correction), QASMBench (benchmarking and characterization), NWQSim (HPC simulation), NWQLib (algorithm library), NWQData (data sets), NWQControl (quantum control), and NWQSC (superconducting testbed). The system enables closed-loop software-hardware co-design and reflects the past eight years of quantum computing research the author has led at PNNL (2018-2026). By releasing most software components as open source or planning their open-source availability, we aim to cultivate a collaborative quantum information science (QIS) ecosystem and support the transition toward a scalable quantum supercomputing era.

The quantum hashing bound guarantees that rates up to $1-H(p_I, p_X, p_Y, p_Z)$ are achievable for memoryless Pauli channels, but it is not generally tight. A known way to improve achievable rates for certain asymmetric Pauli channels is to apply a small inner stabilizer code to a few channel uses, decode, and treat the resulting logical noise as an induced Pauli channel; reapplying the hashing argument to this induced channel can beat the baseline hashing bound. We generalize this induced-channel viewpoint to arbitrary stabilizer codes used purely as channel transforms. Given any $ [\![ n, k ]\!] $ stabilizer generator set, we construct a full symplectic tableau, compute the induced joint distribution of logical Pauli errors and syndromes under the physical Pauli channel, and obtain an achievable rate via a hashing bound with decoder side information. We perform a structured search over small transforms and report instances that improve the baseline hashing bound for a family of Pauli channels with skewed and independent errors studied in prior work.

We construct four-dimensional de Sitter space as an excited state, rather than as a vacuum configuration, in type IIB, heterotic SO(32), and heterotic E_8 \times E_8 string theories. This framework provides a mechanism to evade vacuum-based no-go theorems for de Sitter solutions in string theory. Starting from a generic M-theory configuration, we obtain de Sitter isometry in the dual string theories through appropriate dynamical duality sequences in the late-time limit. The excited state, identified as a Glauber-Sudarshan state, is constructed as the expectation value of the metric operator in M-theory using path-integral techniques. We further analyze the conditions required for the existence of a well-defined effective field theory description and show that these conditions are equivalent to the Null Energy Condition for a (3+1)-dimensional FLRW cosmology. Finally, we investigate constraints arising from axionic cosmology and demonstrate how the time-dependent solutions are modified when experimental bounds on the axionic coupling constant are taken into account. This article serves as a computational companion to sections 3 and 4 of the paper https://doi.org/10.48550/arXiv.2511.03798.

Quantum low-density parity-check (qLDPC) codes can be implemented by measuring only low-weight checks, making them compatible with noisy quantum hardware and central to the quest to build noise-resilient quantum computers. A fundamental open question is how constraints on check weight limit the achievable parameters of qLDPC codes. Here, we study stabilizer and subsystem codes with constrained check weight, combining analytical arguments with numerical optimization to establish strong upper bounds on their parameters. We show that stabilizer codes with checks of weight at most three cannot have nontrivial distance. We also prove tight tradeoffs between rate and distance for broad families of CSS stabilizer and subsystem codes with checks of weight at most four and two, respectively. Notably, our bounds are applicable to general qLDPC codes, as they rely only on check-weight constraints without assuming geometric locality or special graph connectivity. In the finite-size regime, we derive numerical upper bounds using linear programming techniques and identify explicit code constructions that approach these limits, delineating the landscape of practically relevant qLDPC codes with tens or hundreds of physical qubits.

Inspired by regular black holes (RBHs) sourced by dark matter halos, we generalize the anisotropic energy-momentum tensor by relaxing the $P_r = -ρ$ condition between radial pressure and density. We demonstrate that while RBHs are a unique special case, a broader class of relations yields horizonless compact stars. Under specific parameter limits, these objects approach a ``frozen state," mimicking black hole features without an event horizon. These compact star solutions could satisfy weak energy conditions and provide a robust mechanism for dark matter-sourced black hole mimickers.

Certain families of quantum mechanical models can be described and solved efficiently on a classical computer, including qubit or qudit Clifford circuits and stabilizer codes, free-boson or free-fermion models, and certain rotor and GKP codes. We show that all of these families can be described as instances of the same algebraic structure, namely quadratic functions over abelian groups, or more generally over (super) Hopf algebras. Different kinds of degrees of freedom correspond to different "elementary" abelian groups or Hopf algebras: $\mathbb{Z}_2$ for qubits, $\mathbb{Z}_d$ for qudits, $\mathbb{R}$ for continuous variables, both $\mathbb{Z}$ and $\mathbb{R}/\mathbb{Z}$ for rotors, and a super Hopf algebra $\mathcal F$ for fermionic modes. Objects such as states, operators, superoperators, or projection-operator valued measures, etc, are tensors. For the solvable models above, these tensors are quadratic tensors based on quadratic functions. Quadratic tensors with $n$ degrees of freedom are fully specified by only $O(n^2)$ coefficients. Tensor networks of quadratic tensors can be contracted efficiently on the level of these coefficients, using an operation reminiscent of the Schur complement. Our formalism naturally includes models with mixed degrees of freedom, such as qudits of different dimensions. We also use quadratic functions to define generalized stabilizer codes and Clifford gates for arbitrary abelian groups. Finally, we give a generalization from quadratic (or 2nd order) to $i$th order tensors, which are specified by $O(n^i)$ coefficients but cannot be contracted efficiently in general.

Quantum channel capacities are fundamental to quantum information theory. Their definition, however, does not limit the computational resources of sender and receiver. In this work, we initiate the study of computational quantum capacities. These quantify how much information can be reliably transmitted when imposing the natural requirement that en- and decoding have to be computationally efficient. We focus on the computational two-way quantum capacity and showcase that it is closely related to the computational distillable entanglement of the Choi state of the channel. This connection allows us to show a stark computational capacity separation. Under standard cryptographic assumptions, there exists a quantum channel of polynomial complexity whose computational two-way quantum capacity vanishes while its unbounded counterpart is nearly maximal. More so, we show that there exists a sharp transition in computational quantum capacity from nearly maximal to zero when the channel complexity leaves the polynomial realm. Our results demonstrate that the natural requirement of computational efficiency can radically alter the limits of quantum communication.

Violations of classicality can be probed through measurements performed on a system at different times, as proposed by Leggett and Garg. Specifically, violations of Leggett-Garg inequalities suggest the presence of quantum effects in macroscopic systems. Long-baseline neutrino experiments provide some of the longest available propagation distances over which such tests can be performed. Previous studies of Leggett-Garg tests in the neutrino sector have largely focused on showing that the oscillation probabilities can violate classical bounds for certain parameter choices. In this work, we develop a more complete and data-driven framework that treats both the distributions representing the classical and quantum behavior, as well as the experimental uncertainties. We consider MINOS, T2K, NOvA, as well as the upcoming DUNE, and present the respective statistical significance for distinguishing quantum behavior from classical scenarios at these long-baseline neutrino experiments. Among them, we find that T2K yields the most significant violation of classicality, at the level of $\sim 14 σ$, with NOvA and projections for DUNE also resulting in a significance of more than $5σ$.

LISA, the Laser Interferometer Space Antenna, due to launch mid-2035, is a large class space mission by the European Space Agency (ESA). In partnership with NASA and ESA-member states, ESA is on track to launch what is expected to be the first space-based gravitational wave detector. By hosting detectors in space, one gains access to a lower frequency band of gravitational wave sources and with them, a plethora of new science. To maximise this scientific gain, ESA and NASA selected 20 scientists for the LISA Science Team, to carry out and/or lead necessary actions on the run up to LISA launch. We give a short overview and update of the LISA mission, some of its science objectives and related waveforms, as well as the work of the LISA Science Team as of December 2025.

Quantum error correction is indispensable to achieving reliable quantum computation. When quantum information is encoded redundantly, a larger Hilbert space is constructed using multiple physical qubits, and the computation is performed within a designated subspace. When applying deep learning to the decoding of quantum error-correcting codes, a key challenge arises from the non-uniqueness between the syndrome measurements provided to the decoder and the corresponding error patterns that constitute the ground-truth labels. Building upon prior work that addressed this issue for the toric code by re-optimizing the decoder with respect to the symmetry inherent in the parity-check structure, we generalize this approach to arbitrary stabilizer codes. In our experiments, we employed multilayer perceptrons to approximate continuous functions that complement the syndrome measurements of the Color code and the Golay code. Using these models, we performed decoder re-optimization for each code. For the Color code, we achieved an improvement of approximately 0.8% in decoding accuracy at a physical error rate of 5%, while for the Golay code the accuracy increased by about 0.1%. Furthermore, from the evaluation of the geometric and algebraic structures in the continuous function approximation for each code, we showed that the design of generalized continuous functions is advantageous for learning the geometric structure inherent in the code. Our results also indicate that approximations that faithfully reproduce the code structure can have a significant impact on the effectiveness of reoptimization. This study demonstrates that the re-optimization technique previously shown to be effective for the Toric code can be generalized to address the challenge of label degeneracy that arises when applying deep learning to the decoding of stabilizer codes.

A recent result of Jaracz has established nonexistence of global solutions to the coupled generalized Jang equation and zero divergence system which satisfy the asymptotic conditions needed to prove the Penrose conjecture by identifying a breakdown mechanism for the Jang slope at finite radius. In this work, we investigate whether a similar obstruction arises when the generalized Jang equation is instead coupled to the conformal flow of metrics. Restricting to spherical symmetry and time-symmetric initial data, we formulate a numerically tractable version of the Jang/conformal flow system. Our numerical results show no evidence of a finite radius breakdown analogous to that observed by Jaracz. Instead, the Jang slope remains regular and approaches its limiting value asymptotically. This behavior persists under controlled perturbations of the warping factor, indicating robustness of the observed phenomenon. These findings suggest that coupling to conformal flow of metrics alters the obstruction mechanism present in the Jang/zero divergence system, and hence that this system may still be viable for proving the Penrose conjecture.

We experimentally demonstrate the simultaneous estimation of the three parameters characterizing a pair of incoherent optical sources in the sub-Rayleigh regime, enabling super-resolved scene characterization. Using spatial-mode demultiplexing (SPADE) with two demultiplexers--one deliberately shifted--we determine separations well below the diffraction limit and achieve sensitive joint estimation of separation, centroid, and relative brightness over a broad range of scene configurations in a single experimental setting. We benchmark our performance using Fisher-information-based Cramér-Rao bounds, and discuss the corresponding quantum limits. We investigate two complementary scenarios: a realistic case with slightly non-identical sources, and an idealized case of indistinguishable sources.

We propose a measurement-based quantum metrology protocol in a composite model, where the probe system (a spin ensemble) is coupled to an ancillary two-level system (qubit) with a general Heisenberg XXZ interaction. With an optimized and weak probe-ancilla coupling strength and a proper duration of joint evolution, the two parallel evolution paths of the probe system induced by the unconditional measurement on qubit can transform an eigenstate of the collective angular momentum operator of spin ensemble to be a two-component state with a large distance in eigenspace. The quantum Fisher information about the phase encoded in the probe system of polarized states or their superposition, that could be relaxed to mixed states, can therefore manifest an exact or asymptotic quadratic scaling with respect to the probe size (spin number) $N$. The quadratic scaling behavior is found to be insensitive to the imperfect encoding operator and coupling strength. By virtue of the parity detection on the ancillary qubit or the probe system, the phase sensitivity can approach the Heisenberg limit. We suggest that the unconditional measurement on qubit could become an efficient resource to replace Greenberger-Horne-Zeilinger-like states and squeezing Hamiltonian for exceeding the standard quantum limit in metrology precision.

We study skein modules of 3-manifolds by embedding them into the Hilbert spaces of 4d ${\cal N}=4$ super-Yang-Mills theories. When the 3-manifold has reduced holonomy, we present an algorithm to determine the dimension and the list of generators of the skein module with a general gauge group. The analysis uses a deformation preserving ${\cal N}=1$ supersymmetry to express the dimension as a sum over nilpotent orbits. We find that the dimensions often differ between Langlands-dual pairs beyond the A-series, for which we provide a physical explanation involving chiral symmetry breaking and 't Hooft operators. We also relate our results to the structure of $\mathbb{C}^*$-fixed loci in the moduli space of Higgs bundles. This approach helps to clarify the relation between the gauge-theoretic framework of Kapustin and Witten with other versions of the geometric Langlands program, explains why the dimensions of skein modules do not exhibit a TQFT-like behavior, and provides a physical interpretation of the skein-valued curve counting of Ekholm and Shende.

We present cosmological parameters measurements from the full combination of DESI DR1 galaxy clustering data described with large-scale structure effective field theory. By incorporating additional datasets (photometric galaxies and CMB lensing cross-correlations) and extending the bispectrum likelihood to smaller scales using a consistent one-loop theory computation, we achieve substantial gains in constraining power relative to previous analyses. Combining with the latest DESI baryon acoustic oscillation data and using cosmic microwave background (CMB) priors on the power spectrum tilt and baryon density, we obtain tight constraints on the $Λ$CDM model, finding the Hubble constant $H_0=69.08\pm 0.37~\mathrm{km}\,\mathrm{s}^{-1}\mathrm{Mpc}^{-1}$, the matter density fraction $Ω_m=0.2973\pm 0.0050$, and the mass fluctuation amplitude $σ_8 = 0.815\pm 0.016$ (or the lensing parameter $S_8\equivσ_8\sqrt{Ω_m/0.3}=0.811\pm 0.016$), corresponding to $0.6\%$, $1.7\%$, and $2\%$ precision respectively. Adding the Pantheon+ supernova sample (SNe), we find a preference of $2.6σ$ for the $w_0w_a$ dynamical dark energy model from low-redshift data alone, which increases to $2.8σ$ when exchanging the SNe with Planck CMB data. Combining full-shape data with BAO, CMB, and SNe likelihoods, we improve the dark energy figure-of-merit by $18\%$ and bound the sum of the neutrino masses to $M_ν<0.057$ eV in $Λ$CDM and $M_ν<0.095$ eV in the $w_0w_a$ dynamical dark energy model (both at 95\% CL). This represents an improvement of $25\%$ over the background expansion constraints and the strongest bound on neutrino masses in $w_0w_a$CDM to date. Our results suggest that the preference for the normal ordering of neutrino mass states holds regardless of the cosmological background model, and is robust in light of tensions between cosmological datasets.

We compute one-loop QCD corrections to the triple differential width of the inclusive decay $\bar{B} \to X_c \ell \barν_\ell$ including contributions from all relevant dimension-six operators in the Weak Effective Theory (WET), at leading power in heavy quark expansion. Furthermore we derive for the first time up to order $\mathcal{O}(α_s)$ analytic expressions for the first three moments of the distribution in the lepton energy, hadronic invariant mass and dilepton invariant mass, in the presence of beyond the Standard Model contributions from the WET.

We propose a bootstrap approximation method for the Hermitian one-matrix model that does not rely on positivity constraints. The theoretical foundation of this method is that the one-matrix model admits an eigenvalue distribution $ρ(λ)$, and that the moments $w_n$ generated from it satisfy the loop equations. Our framework is designed to numerically determine a self-consistent pair of $ρ(λ)$ and $w_n$ that simultaneously satisfies these two requirements. In the concrete implementation, we employ a least-squares method, for which no sign problem arises in principle, and therefore the method can be formally applied also to Minkowski-type models. Actual numerical calculations show that this bootstrap approximation reproduces, with very high accuracy, the exact solutions for Euclidean-type models and the perturbative results for Minkowski-type models.

We investigate the combined effects of carrier envelope phase and laser pulse shape on electron-positron pair production in the presence of an external time-dependent asymmetric electric field by solving the quantum Vlasov equation. We analyze how the pulse asymmetry, the envelope type (Gaussian, Lorentzian and Sauter), and the carrier envelope phase jointly influence the momentum distribution and the number density of created pairs. Our results show that pair production exhibits extreme sensitivity to both the degree of temporal asymmetry and the steepness of the envelope on either side of the pulse. These effects are qualitatively explained through a turning-point analysis, which, for the first time, is carried out for a non-analytic electric field using a regularization scheme. We observed that multiphoton pair production dominates the Schwinger mechanism in the case of a long falling-pulse asymmetry. For a short falling pulse with a flat-topped profile, pair production is further facilitated. We demonstrate that the number density can be enhanced by two to three orders of magnitude by choosing certain field parameters.

We bootstrap tree-level supergravity four-point correlators on AdS$_5\times$S$^5$ with one external half-BPS double-particle operator and three half-BPS single-particle operators. Our only input is the consistency of the operator product expansion of $SU(N)$ ${\cal N} = 4$ super Yang-Mills theory at large $N$ and large 't Hooft coupling. Even though the leading order OPE does not close on double-particle operators, but involves triple-particle operators, the CFT data of the double-particle operators, both long and protected, is sufficient to uniquely fix the correlators. We then verify that our results for the four-point correlators with one double-particle and three single-particle operators are reproduced by the appropriate double-particle limit of the five-point tree-level correlators of single-particle operators, with arbitrary Kaluza-Klein levels, recently conjectured in arXiv:2507.14124. Our study thus provides further evidence for the latter result.

We study $b \to s \ell_1^+ \ell_2^-$ transitions, both for the lepton flavour conserving $\ell_1=\ell_2$ and violating case $\ell_1 \neq \ell_2$, in a minimal extension of the Standard Model proposed in [1]. In this framework, the Standard Model (SM) gauge group is enlarged by a new $U(1)^\prime$ component. The fermion $U(1)^\prime$ charges are assigned in a generation-dependent way, and involve three rational parameters $ε_{1,2,3}$ summing to zero by the condition of cancellation of the gauge anomalies. Each $ε_i$ is common to all fermions in a generation, which produces correlations among quark and lepton observables. The new neutral gauge boson $Z^\prime$ has flavour violating couplings to quarks and leptons. For SM allowed processes, small deviations with respect to the SM predictions are found: this is a consequence of a feature of the model where quark and lepton sectors preclude each other large deviations from SM. Lepton flavour violating processes are allowed at tree-level. The experimental upper bounds for the rates of the processes $τ^- \to μ^- μ^+ μ^-$, $μ^- \to e^- γ$, $μ^- \to e^- e^+ e^-$ and the $ μ^- \to e^-$ conversion in nuclei play a hierarchical role in constraining the branching fractions of lepton flavour violating $B$ and $B_s$ decays.

The SABRE North experiment is developing ultra-high radiopurity NaI(Tl) detectors to investigate dark matter. To achieve this, SABRE North utilizes the technique called zone refining for NaI powder purification. This work details the mathematical model developed to describe the purification process. By comparing this model to the results of the commissioning and production runs conducted prior to crystal growth, the distribution coefficients were determined for various impurities, contained in the powder at the parts-per-billion (ppb) level. Furthermore, the synthesis of data from both zone refining and normal freezing is discussed. These findings can be used to predict the SABRE North detectors background level in the energy region-of-interest for dark matter search and to optimize the production of ultra-high purity crystals through multiple purification strategies.

Compact astrophysical systems such as neutron stars and black holes provide powerful laboratories for testing feebly coupled dark photons (DPs). We investigate light DPs kinetically mixed with the visible photon that need not be the dark matter, focusing on resonant photon-DP oscillations in magnetized, modeled plasma environments. We show that realistic non-monotonic plasma density profiles generically enhance resonant conversion relative to monotonic models, leading to substantially stronger constraints on the photon-DP kinetic mixing parameter ($ε$). Using spectral data from the supermassive black hole (SMBH) M87*, extending to the LOFAR band, we derive a bound $ε\simeq 7\times10^{-6}$ at the DP mass $m_{A'} \simeq 5\times10^{-7}\,\mathrm{eV}$ for oscillation distance $3r_{\rm ph}$, where $r_{\rm ph}$ denotes the photon sphere radius. From the Crab pulsar-wind Nebula, we obtain an even stronger constraint, $ε\simeq 8\times10^{-7}$ at $m_{A'} \simeq 4\times10^{-9}\,\mathrm{eV}$ for oscillation baselines of order $10^{3}\,\mathrm{km}$, surpassing existing astrophysical limits in realistic plasma backgrounds. While laboratory and cosmological bounds remain slightly stronger at comparable masses, observation of compact objects with larger surface magnetic fields and measurements of photon spectra at lower frequencies would enhance the limits on the photon-DP coupling by orders of magnitude.

The extension of the Higgs sector with an additional complex triplet field is often considered for generating the neutrino mass by the Type-II seesaw mechanism. Such an extension generally predicts $ρ\neq1$, where $ρ$ is the electroweak rho parameter at the tree level, so that the renormalization of the electroweak parameters is different from models like the standard model (SM) and two Higgs doublet models. In this paper, we present a full set of radiative corrections to decays of the 125 GeV Higgs boson ($h$) in this model. One-loop contributions of the extra Higgs bosons as well as SM fermions and gauge bosons to the decay rates of $h$ are calculated in the on-shell scheme. Gauge dependence appearing in the counter terms of mixing angles is eliminated by the pinch technique. Higher-order QCD corrections are also implemented. We find that the decay rates can significantly deviate from the predictions in the SM and other extensions such as the two Higgs doublet models and the singlet model. For example, the decay rates of $h\to WW^\ast$ and $h\to ZZ^\ast$ can be a few percent larger than the SM value under current experimental and theoretical constraints. In this case, deviations in $h\to γγ$ and Higgs self-coupling can reach about $-20\%$ and $100\%$, respectively. The pattern of the deviations is different from the other extended Higgs models. These characteristic predictions are expected to be detected at the High-Luminosity LHC or future Higgs factories.

We define Deligne-Beilinson (DB) cohomology on a cubic lattice and use it to formulate and analyze lattice $U(1)$ Chern-Simons theory at even levels. The continuum DB cohomology provides a refined mathematical framework for continuum $U(1)$ connections constructed in a patchwise manner. The lattice DB cohomology we construct retains many essential properties of the continuum DB cohomology and naturally incorporates a notion of self-linking number. The lattice $U(1)$ Chern-Simons action formulated using the lattice DB cohomology is expressed as a simple quadratic form via the star product, which naturally exhibits level quantization. Framed Wilson lines respecting staggered symmetry are defined in a gauge-invariant manner, and their expectation values are shown to be given by the self-linking number, as follows from completing the square. Using the lattice Hodge decomposition, we explicitly characterize the DB cohomology on a three-dimensional cubic toroidal lattice and present a gauge-fixed, rigorous path integral for the lattice Chern-Simons theory. To regulate divergences in the lattice Chern-Simons path integral arising from staggered symmetry, we introduce a small Maxwell term. The resulting error is controlled by the linear order in the small Maxwell coupling.

In the new era of HL-LHC experiments, fast-timing detectors are emerging as critical tools for background rejection. Typical requirements include a temporal hit resolution of about 50 ps, a spatial resolution of around 12 $μ$m, and radiation hardness up to $10^{17}$n$_\text{eq}$/cm$^2$. To address these challenges, the development of non-standard sensor designs and advanced fast readout electronics is required. The OPTIMA multichannel board addresses the need for testing small sensor demonstrators when they cannot yet be bonded to dedicated readout ASICs. It provides fast readout of up to 16 channels and can be integrated into various test setups, including test beam environments. This contribution presents the design of the OPTIMA board, its integration in test beams, and the first experimental results.

We explicitly demonstrate the importance of a new thermal resonant channel in the context of low-scale leptogenesis, which goes beyond the well-known mixing and oscillation of massive singlet neutrinos. This new channel is always present when considering the thermally-induced Higgs decay to leptons and relativistic sterile neutrinos, and can become dominant thanks to thermally-generated lepton-doublet flavour oscillations. This mechanism can yield the observed baryon asymmetry in our universe, even if there is no resonant enhancement from quasi-degenerate sterile neutrinos. The required active-to-sterile neutrino mixing differs from the other two low-scale leptogenesis channels and can be probed in fixed-target and long-lived particle experiments, and by displaced vertex searches at high-energy colliders.

We study parity-violating birefringence induced by an asymmetric neutrino background in plasmas associated with active galactic nuclei (AGN). We derive a directionality factor arising from the relative bulk motion between the neutrino medium and plasma, and show that it can produce an anomalous frequency dependence of the polarization-rotation angle, distinct from the $ω^{-2}$ scaling of Faraday rotation. This anomalous scaling can occur either at the resonance plasma frequency condition $ω\simeq ω_p$, or when $E_ν^{0}\simeq m_νω/ω_p$ lies within the range of the neutrino energy spectrum. We estimate the effect for three scenarios: jets propagating through the cosmic neutrino background (C$ν$B), jets with an internal flux of high-energy neutrinos, and accretion-disk plasma permeated by the C$ν$B. Of the three scenarios, the latter gives the largest rotation angle $φ_{\rm d} \sim 10^{-35}\,\mathrm{rad}$, at X-ray frequencies. Although the predicted rotation angles are below current polarimetric sensitivity, the identified spectral signatures provide a theoretical framework for probing neutrino asymmetries and AGN plasma properties independent of magnetic field models.

Threshold corrections provide a universal link between experimentally measured broken-phase parameters and $\overline{\hbox{MS}}$ renormalised symmetric-phase parameters used in SMEFT renormalisation-group (RG) analyses and matching to new-physics models. In this work we compute in analytic form the complete set of one-loop threshold corrections in dimension-six SMEFT, using two electroweak input schemes and including tadpole effects in the Fleischer-Jegerlehner (FJ) scheme. As a by-product of the analysis, we obtain the full one-loop running of the strong coupling in SMEFT, including electroweak corrections, and the associated decoupling constants when the top quark and heavy electroweak bosons are integrated out. Although generally moderate, the loop-level dimension-six corrections can be enhanced through tadpole effects and top-quark loops, and can shift symmetric phase parameters at the 5% level under reasonable assumptions. Our results constitute a process-independent ingredient which can be readily implemented in public RG-running and matching codes in SMEFT.

The dark matter and existence of new particles are now a possible explanation of several physics phenomena which evade the predictions of the Standard Model. In this context Axion-Like-Particles (ALP) with masses in the MeV to GeV range with additional Peccei-Quinn breaking contribution, and which are coupled to the Standard Model have been postulated. To search for the existence of such new particles, we have launched dedicated analysis of a high statistics data sample collected by High-Acceptance Di-Electron Spectrometer (HADES) operating at GSI in Darmstadt. In particular, we study $η$ meson decays into $π^{+}π^{-}e^{+}e^{-}$, where hypothesized isoscalar gauge boson $a$ could be produced in the intermediate state $η\toπ^{+}π^{-}a$ decaying predominantly to $e^{+}e^{-}$. In this report we describe the analysis strategy we applied to search for a resonant peak in the dilepton invariant mass spectrum $η\toπ^{+}π^{-}a\to π^{+}π^{-}e^{+}e^{-}$ and present the method for event selection and particle identification.

We search for the reaction channel $e^+ e^- \to ηη\,J/ψ$ in a data sample with center-of-mass energies from 4.226 to 4.950~GeV, which was collected by the BESIII detector operating at the Beijing Electron Positron Collider (BEPCII). The data analysis is performed with two different methods, exclusively and semi-inclusively, which enabling a comparison and combination of the results. Only in a few cases a statistical significance of the cross sections with more than $3σ$ is observed with one of the methods. Only at 4.750~GeV the significance of the cross section measurement is 8.9$σ$ (observation) with the exclusive analysis and 3.4$σ$ (evidence) with the semi-inclusive analysis. Therefore, the corresponding upper limits of the cross section at the 90% confidence level are determined. The energy dependent results show clear deviations from the the line shape expected from three-body phase space alone. Since the statistical significance for almost all center-of-mass energies is low, the upper limits for the reaction channel $e^+ e^- \to ηη\,J/ψ$ also serve as limits for the existence of a possible isospin partner to the charmonium-like isospin triplet $Z_{\rm c}(3900)$ which decays to $J/ψη$.

The internal structure of the Standard Model implies a natural $\mathbb{Z}_4 \times \mathbb{Z}_3$ discrete gauge symmetry. Cancellation of the corresponding Dai--Freed anomalies requires the introduction of three right-handed neutrinos and three additional Majorana fermions $χ_i$. This gauge symmetry forbids the decay of the lightest fermion $χ_1$ into Standard Model particles, rendering it automatically stable and providing a dark matter candidate without introducing an ad hoc stabilizing symmetry and domain-wall problem. The mass of $χ_1$ is generated by the vacuum expectation value of a singlet scalar near the electroweak scale, naturally realizing a weakly interacting massive particle (WIMP) freeze-out scenario. Dark matter annihilation proceeds through scalar mediation, allowing the observed relic abundance to be reproduced while remaining consistent with current direct-detection constraints. It naturally realizes the secluded dark matter scenario and can be further tested in the next generation of experiments.

We couple the two-flavor Nambu--Jona-Lasinio model to a gluonic background corresponding to the gauge-field expectation value in the center-symmetric Landau gauge. Low-energy features in this gauge are captured by a center-symmetric extension of the Curci-Ferrari model and provide a good grasp on key aspects of the confinement/deconfinement transition. Within this framework, we can investigate the interplay between the chiral and deconfinement transitions. Compared to other approaches based on multi-parameter Ansätze of the Polyakov loop potential fixed from comparison to finite-temperature lattice data, the modeling of the glue sector in the present set-up depends on only one phenomenological parameter that can be fixed by comparison to lattice data in the vacuum. We detail the structure of the phase diagram, with special emphasis on the finite density axis, and compute thermodynamical observables relevant for applications. We also highlight the properties of the recently introduced net quark number response of the medium as a sensible probe of the phases of QCD, in particular as a tool to disambiguate the nature of certain regions of the phase diagram where the use of the Polyakov loops could lead to misinterpretations. Finally, we critically assess the sensitivity of our results to the various parameters, both in the glue sector and in the chiral sector.

The ESSnuSB is a design study for a long-baseline neutrino experiment to precisely measure the CP violation in the leptonic sector, at the second neutrino oscillation maximum, using a beam driven by the uniquely powerful ESS linear accelerator. The ESSnuSB-plus (ESSnuSB+) design study program, which is an extension phase of the ESSnuSB project, aims in designing two new facilities, a Low Energy nuSTORM and a Low Energy Monitored Neutrino Beam and use them to precisely measure the (anti)neutrino-nucleus cross-section in the energy range of 200-600 MeV, where experimental data are lacking or imprecise. In addition, new target stations and a new water-Cherenkov near-near detector will be designed to measure cross sections and serve to explore sterile neutrino physics. An overall status of the project will be presented together with the ESSnuSB+ additions.

Advances in accelerator technology have led to significant improvements in the quality of muon beams over the past decades. Investigations of the muon and muonium enable precise measurements of fundamental constants, as well as searching for new physics beyond the Standard Model. Furthermore, by utilizing muon beams with high intensity and polarization, studies of the dynamics of the muon and muonium within atom level can offer valuable insights into material science. This review presents recent progress and prospects at the frontiers of muon beams and high-precision muonium physics. It also provides an overview of novel methods and detection techniques to achieve high sensitivities in different areas, including particle physics, nuclear physics, materials science and beyond.

We investigate the effect of decay channels on the near-threshold mass scaling by employing potential models. By varying the attractive strength of a square-well potential, we examine the pole trajectory associated with the transition of an $s$-wave bound state into a resonance state, incorporating decay-channel effects through both a single-channel complex potential model and a coupled-channel real potential model. As a result, we show that the pole of a quasibound state below the threshold is not continuously connected to that of a resonance state above the threshold. Furthermore, by comparing the results obtained from the single-channel and coupled-channel models, we clarify the correspondence between the pole trajectories in the two approaches.

The development of the Forward Spectator Detector (FSD) for the CBM experiment represents a crucial step toward successful realization of the CBM physics program - understanding of highly compressed nuclear matter at the forthcoming FAIR facility. Designed for detecting collision participants at high collision rates at the SIS-100 accelerator, the FSD employs scintillator-based detector technology to accurately reconstruct the reaction plane and to determine the centrality of the collision. Overview of the technical design and performance studies conducted for the FSD is provided.

We propose two-loop rainbow type of the neutrino mass model via ${Z}_2$ gauging of ${Z}_6$ non-invertible symmetry in which we introduce three families of isospin doublet vector-like fermions, heavy right-handed neutrinos and isospin doublet and singlet bosons. All new fields, which have nonzero charges under the non-invertible symmetry, can be dark matter candidates, since the non-invertible symmetry possesses a remnant ${Z}_2$ symmetry that plays a role in assuring the stability of our dark matter candidate. Even though the non-invertible symmetry is dynamically broken at one-loop level, its violation does not affect our scenario. In this paper, we especially consider the lightest mode of the neutral components in the doublet vector-like fermions as our main dark matter candidate. The dark matter is potentially degenerated to the other two families of neutral fermions, since the mass difference is induced at one-loop level. Thus, we consider our dark matter candidate in rather simpler co-annihilation system among their particles. Considering all the constraints of neutrino oscillation data, lepton flavor violations, muon $g-2$ and the relic density of dark matter, we perform the numerical analysis and show some allowed regions for these phenomenology. Due to our dark matter nature, the sum of neutrino masses in case of normal hierarchy is larger than that in case of inverted hierarchy, which is opposite situation compared with typical active neutrino models.

We present a comprehensive theoretical framework for probing quantum entanglement in the decay angular distributions of a spin-$S$ particle-antiparticle pair $A\bar{A}$, where each particle decays sequentially into a two-body final state, $A\to B+C$ and $\bar{A}\to\bar{B}+\bar{C}$, with $B(\bar{B})$ carrying spin $b$ and $C(\bar{C})$ being spinless. Starting from the most general polarized initial state, we derive the fully differential angular distribution $\mathcal{W}(θ_1,θ_2,φ_1,φ_2)$ and identify observables $\langle\cos(2S(φ_1\mpφ_2))\rangle$ whose expectation values directly depend on the entanglement-sensitive coefficients $\text{Re}\left(α_{-S,\mp S}α_{S,\pm S}^*\right)$ of the initial state. The proportionality factor $\mathcal{C}(S,b)$ in these relations is computed explicitly. For bosonic decays ($b=0,1,2,\ldots$), $\mathcal{C}(S,b)$ is universal and independent of decay dynamics; in particular, $\mathcal{C}(S,0)=1/2$ for any $S$, and $\mathcal{C}(1,1)=1/8$, matching known results for $W^+W^-$ decays. For fermionic decays ($b=\frac{1}{2},\frac{3}{2},\frac{5}{2}\ldots$), $\mathcal{C}(S,b)$ depends explicitly on the spin analysis powers $α_{A/\bar{A}}$, making entanglement extraction more decay-dependent. We further demonstrate, within the context of $e^+e^-\toγ^*\to A\bar{A}$ production, how $α_{A/\bar{A}}$ can be determined experimentally using specific angular observables restricted to the beam-axis region. Our results highlight the special role of bosonic decays in providing clean, model-independent tests of quantum entanglement at colliders, while outlining a pathway for entanglement measurement in fermionic cases through supplementary polarization information.

We analyze spin entanglement in $Λ\barΛ$ pairs produced in the decays of scalar ($h$) and pseudoscalar ($a$) particles, contrasting predictions from quantum field theory (QFT) with those of local hidden-variable theories (LHVTs). Using the self-analyzing weak decays $Λ\to pπ^-$ and $\barΛ \to \bar{p}π^+$, we derive the joint angular distributions within QFT. Our key findings are: For scalar production $h \to Λ\barΛ$, no LHVT respecting locality and angular-momentum conservation can reproduce the QFT distribution. For pseudoscalar production $a \to Λ\barΛ$, a CPT-symmetric LHVT is excluded by positivity constraints given the measured analyzing powers; however, if CPT symmetry is relaxed, an explicit LHVT construction, with uniform hidden-variable measure and response functions satisfying $b_1 c_1 = 3α_Λα_{\barΛ}$, can match the QFT result. These distinct signatures provide clear, experimentally testable criteria to discriminate between QFT and LHVT in $Λ\barΛ$ systems across different production mechanisms.

We study holographic defect conformal field theories which are dual to probe branes with bottom-up methods. First we determine the embedding of codimension-1 probe branes in AdS space. Then we compute defect one and two-point correlation functions of heavy scalar operators at strong coupling. In particular, we use geodesic approximations to compute scalar two-point functions across the defect, reflected two-point functions, and ambient channel two-point functions. In appropriate limits, our results agree with the boundary operator expansion (BOE) and the operator product expansion (OPE).

Extending the scalar sector is one of the standard approaches to exploring scenarios beyond the Standard Model. In this work, we examine the collider phenomenology of the Three Higgs Doublet Model (3HDM) in the Type-Z or the democratic Yukawa interaction setup at the LHC. The scalar spectrum of the 3HDM includes three CP-even scalars, two CP-odd scalars, and four charged Higgs bosons. Focusing on cascade decay topologies, we investigate the collider signatures of the neutral scalars through the process $pp \rightarrow SV$, where $S$ is a neutral scalar and $V$ is a vector boson. We perform a cross-section analysis across multiple benchmark points that satisfy both theoretical and experimental constraints, considering two mass hierarchy scenarios: (i) Regular Hierarchy, where the SM-like Higgs is the lightest CP-even scalar, and (ii) Medial Hierarchy, featuring one Higgs boson lighter than the SM Higgs and one heavier. For both scenarios, we study the specific process $pp \rightarrow A \rightarrow HZ \rightarrow b \bar{b} l^+l^-$, performing a cut and count analysis at $\sqrt{s}=14$ TeV. Our results demonstrate that while the Medial Hierarchy scenario allows discovery-level sensitivity for both the CP-even and CP-odd scalars, achieving the same sensitivity in the Regular Hierarchy setup necessitates substantially higher luminosity.

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

We present a study of the inclusive charmless semileptonic $b$ decay, $b\to X_u\ell\barν$ at next-to-next-to-leading order (NNLO) in perturbative QCD, with the primary aim of extracting the hadronic structure functions $W_i$ at NNLO. The analysis is based on a numerical calculation of the relevant kinematic distributions using a phase-space slicing method to handle infrared-sensitive contributions from real gluon emissions. We use known results from Heavy Quark Effective Theory and Soft-Collinear Effective Theory to extract the singular terms and construct a model for the regular contributions to the structure functions at NNLO, then perform a fit to the numerical results. We use our approximate structure functions to compute various kinematic distributions and moments: the comparison with existing analytic and numerical results shows very good agreement, which is further improved including the available analytic results in the fit.

We consider mixed four-point correlators of 1/2-BPS operators $\mathcal{O}_{k_i}$ in the maximally supersymmetric CFTs, i.e. the 3d $\mathcal{N}=8$, 4d $\mathcal{N}=4$, and 6d $\mathcal{N}=(2,0)$ theories. In \cite{Dolan:2004mu}, Dolan, Gallot, and Sokatchev demonstrated that four-point correlators of identical $\mathcal{O}_{k_i}$ in these SCFTs can be expressed in terms of unconstrained ``reduced correlators" $\mathcal{T}^{\{k_i\}}_{I,J}(U,V)$, $h^{\{k_i\}}_I(z)$ acted on by a $2(\varepsilon-1)^\text{nd}$ order differential operator $Δ_\varepsilon$, which is non-local in odd dimensions $d=2(\varepsilon+1)$. We generalize this construction to mixed correlators $\langle \mathcal{O}_{k_1}\mathcal{O}_{k_2}\mathcal{O}_{k_3}\mathcal{O}_{k_1+k_2+k_3-2\mathcal{E}}\rangle$ up to extremality $\mathcal{E}=2$. To construct superconformal blocks, we generalize the R-symmetry channel equations and use Jack polynomial expansions to recursively generate the full spectrum of conformal blocks in a superblock from a single channel. We observe that this channel equation can be inverted to expand $\mathcal{T}^{\{k_i\}}_{I,J}$, $h^{\{k_i\}}_I$ in ``reduced" blocks with shifted kinematics $(\tildeΔ_{12},\tildeΔ_{34})=\left(Δ_{12},Δ_{34}-2(\varepsilon-1)\right)$. These reduced blocks reproduce what is known in 4d, generalize the known $\langle \mathcal{O}_{2}\mathcal{O}_{2}\mathcal{O}_{k}\mathcal{O}_{k}\rangle$ case in 6d, and offer a novel result in 3d.

We discuss the marginal $\sqrt{T\overline{T}}$ deformation of massless scalar field theories in two dimensions from a dynamical perspective. The operator flow equations for such deformations induce a particular Legendre Transformation between flowed Lagrangians and flowed Hamiltonians. The marginal deformation does not change the conformal symmetries of the theory, until some special points in the moduli space are reached, and the relativistic conformal algebra smoothly changes to the Carrollian conformal (equivalently BMS) one. We investigate this change of symmetry from both configuration space and phase space point of view, while keeping the notion of Legendre Transformation unchanged during the flow. By expanding the actions, in the extreme limits of the flow parameter, we recover the usual ``Electric'' Carroll theory and further uncover a novel ``Magnetic'' counterpart. We discuss the intriguing geometric understanding of such dynamical maps for the deformed theories, and also provide a concrete example for the same from a deformed string theory in flat space.

A dense neutrino plasma can exhibit collective flavor evolution caused by neutrino--neutrino refraction. Recently, a new class of exact nonlinear inhomogeneous solutions was discovered: single-wave (SW) solutions of the fast flavor system. The key property is that the flavor occupation numbers remain homogeneous, whereas the field of flavor coherence varies spatially with a single wave vector. The equations of motion for this structure resemble those of a collection of classical spins, in analogy with the homogeneous slow and fast flavor cases. In contrast, the SW system is not integrable (it does not possess Gaudin invariants) so that, while two-beam pendulum solutions are inevitable, they do not extend to a multi-angle system. We develop a taxonomy of all known nonlinear collective flavor solutions, explaining the overlap between categories and their differences.

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

We consider Loryons, particles beyond the Standard Model that receive a significant fraction of their masses from electroweak symmetry breaking, in the context of a two Higgs doublet model. Using scalar Loryons in the the $[1,1]$, $[1,3]$ (as well as the equivalent $[3,1]$) and the $[2,2]$ representations of the custodial $SU(2)_L \times SU(2)_R$ global symmetry as benchmarks, we study the constraints on the Loryon parameter space, focusing on unitarity, Higgs decay observables, and the absence of Loryon vacuum expectation values. We find that while neutral singlet Loryons remain viable for masses up to 700 GeV, representations containing charged scalars are severely constrained by LHC data, particularly as the fraction of mass generated by symmetry breaking increases.

We present the first dedicated analysis of cosmic-ray boosted dark matter (CRDM) in paleo detectors. Owing to their large kinetic energies, CRDM particles generate nuclear-recoil tracks that extend to substantially larger lengths than those produced by dominant backgrounds from neutrinos and intrinsic radioactivity. Combined with the ultra-large effective geological exposure of $\mathcal{O}(10^{5})~\mathrm{t\,yr}$, paleo detectors provide a uniquely sensitive probe of sub-GeV DM. Considering both constant and vector-mediator interactions, we find that paleo detectors improve the sensitivity to the DM--proton scattering cross section by one to two orders of magnitude compared with the latest XENONnT limits.

The binary black hole signal GW250114, the clearest gravitational wave detected to date, offers a unique opportunity to test general relativity in the relativistic strong-gravity regime. How well does GW250114 agree with Einstein's predictions in the plunge-merger-ringdown stage? To address this point, we constrain deviations from general relativity across the plunge-merger-ringdown stage of spin-precessing binaries with a parametrized waveform model within the effective-one-body formalism. We find that deviations from the peak gravitational-wave amplitude and instantaneous frequency of the $(\ell, |m|)=(2,2)$ mode are constrained to about $10\%$ and $4\%$, respectively, at $90\%$ credible level. These constraints are, respectively, two and four times more stringent than those obtained by analyzing GW150914. We also constrain, for the first time, the instantaneous frequency of the $(\ell, |m|)=(4,4)$ mode at merger to about $6\%$, and the time at which the gravitational-wave amplitude peaks to about $5~\mathrm{ms}$. These results are the most precise tests of general relativity in the nonlinear regime to date, and can be employed to constrain extensions of Einsten's theory.

We investigate a minimal extension of the Standard Model with a real singlet scalar and a singlet Dirac fermion acting as dark matter. Unlike a conventional singlet scalar setup, we assume that the singlet scalar does not acquire a vacuum expectation value at zero temperature. This decouples the scalar mixing angle from the Higgs-portal quartic coupling responsible for the strong first-order electroweak phase transition, allowing it to coexist with current collider and direct-detection constraints. The Higgs-singlet mixing is generated independently through a trilinear portal interaction. We check theoretical consistency conditions, various LHC limits on heavy scalar resonances, dark matter relic abundance, and direct detection bounds to delineate the viable parameter space. We perform a detailed analysis of the electroweak phase transition and show that a strong first-order transition is realized for a selected set of benchmark points. We further compute the resulting stochastic gravitational wave spectra and find that several scenarios yield signals potentially observable at future space-based interferometers. Our results establish a unified and testable framework that connects collider phenomenology, first-order electroweak phase transition, and the resulting production of gravitational waves, along with the dark matter phenomenology, all within a simple renormalizable extension of the Standard Model.

Recent evidence from the ATLAS Collaboration for the rare decay $H \rightarrow μ^{+}μ^{-}$ provides a unique window into the Higgs boson's coupling to second-generation fermions. In this work, we investigate how this signal can probe physics beyond the Standard Model by computing the branching ratios $B(H_{i} \rightarrow μ^{+}μ^{-})$ for the three CP-even Higgs bosons $H_{1,2,3}$ in the broken-phase Next-to-Two-Higgs-Doublet Model (N2HDM). We incorporate one-loop radiative corrections and analyze deviations from the Standard Model prediction due to modified Yukawa couplings, scalar mixing, and singlet--doublet interactions. By confronting our results with the ATLAS signal strength $μ= 1.4 \pm 0.4$, we identify viable regions of the N2HDM parameter space, characterized by tan$β$, the singlet vacuum expectation value, and scalar masses, and assess the model's capacity to explain potential enhancements in the dimuon channel. Our study demonstrates that precision measurements of $H \rightarrow μ^{+}μ^{-}$ serve as a powerful tool to test extended Higgs sectors and uncover new physics at current and future colliders.

Analyses of type Ia supernovae have helped us shed light on the existence and nature of dark energy. Most of these analyses have relied on Bayesian techniques. In this work, we rely on resampling techniques to analyse supernova data. In particular, we use the generalised least squares method together with Jackknife and Bootstrap techniques to estimate parameters of $Λ$CDM, flat $Λ$CDM, $w$CDM, flat $w$CDM, and flat $w_0\,w_a$CDM models from the recent PantheonPlus and SH0ES data. For completeness, we also perform Bayesian analysis using Markov chain Monte Carlo (MCMC) and nested sampling algorithms, and compare the results. We note that resampling techniques can help highlight the limitations of the data. For instance, we see that the Jackknife method estimates a strong positive correlation between $h$ and $M$ and higher standard deviations for both. This may have significant implications for the Hubble tension. We conclude with a discussion of our results.

There is a long-acknowledged deficiency of bright red giants relative to fainter old stars within a few arc seconds of Sgr A*. We explore whether this could be due to tidal stripping by the central black hole. This requires putting the stars onto highly eccentric orbits, for which we evaluate diffusion by both scalar resonant and non-resonant relaxation of the orbital angular momentum. We conclude that tidal stripping does not discriminate sufficiently between main-sequence and red giant stars. While the tidal loss cone increases with stellar radius, the rate of diffusion into the loss cone increases only logarithmically, whereas the lifetime on the red giant branch decreases more rapidly than $R_*^{-1}$. In agreement with previous studies, we find that stellar collisions are a more likely explanation for the deficiency of bright red giants relative to fainter ones.

As the recurrent nova T Coronae Borealis (T CrB) approaches its next predicted thermonuclear eruption, it is currently exhibiting a "super-active state" (SAS) characterized by enhanced multiwavelength emission similar to the behavior recorded prior to the 1946 outburst. We present a multiwavelength analysis of the SAS and the subsequent "faint state" using observations from HST, Swift, NuSTAR, and XMM-Newton. Our results indicate that the SAS was driven by an increase in the mass accretion rate, which caused the accretion disk's boundary layer to become optically thick. A weighted least squares regression analysis quantifies the evolution of the accretion components, displaying a highly significant (4.5$σ$) increase in the luminosity of the optically thin cooling flow (L$_{cf}$) and a marginal (2.58$σ$) decrease in the optically thick boundary layer luminosity (L$_{bb}$) as the system transitioned into the faint state. We find that this dimming is consistent with an intrinsic change in the accretion flow rather than dust obscuration, supported by the lack of infrared excess and the stability of the 2175 Å feature. Additionally, a time-series analysis using autoregressive modeling to account for correlated red noise revealed no significant periodicities, thereby disputing the previously reported $\sim$6000 s signal. These findings suggest that the pre-outburst evolution of T CrB is characterized by significant changes in the accretion disk structure and boundary layer, providing a self-consistent physical framework for the system's behavior as it approaches eruption.

Dynamical modelling is a fundamental tool for measuring galaxy masses and density profiles in the era of large integral-field spectroscopic surveys and Bayesian inference. Solutions based on the Jeans equations are popular due to their robustness and computational efficiency. However, traditional semi-analytic Jeans solvers often require restrictive assumptions about the velocity anisotropy to remain computationally tractable. This paper presents a new spectral solver for the axisymmetric Jeans equations designed to overcome these limitations. I first illustrate, using orbit integrations in realistic potentials, that spherical alignment of the velocity ellipsoid is a physically well-motivated approximation for galaxy modelling. The new method employs a spectral technique to solve the Jeans partial differential equations directly. Two design choices are critical for accuracy and speed: (i) solving for the slowly-varying velocity dispersion rather than the rapidly varying pressure, and (ii) imposing a Robin boundary condition to enforce the asymptotic decay on a finite domain. This formulation supports arbitrary anisotropy distributions beta(r, theta) while simultaneously increasing computational speed by orders of magnitude compared to standard high-accuracy quadratures. Validated against exact analytic benchmarks, the solver recovers intrinsic moments with sub-percent accuracy. The implementation will be included in the public JamPy package and is structured to be optimally suited for massive parallelization on specialized hardware such as GPUs, enabling the rigorous exploration of complex parameter spaces.

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

Magnetars are extreme neutron stars powered by ultra-strong magnetic fields ($\sim10^{14}$ Gauss) and are compelling engines for some of the most powerful extragalactic transients such as Super Luminous Supernovae, Gamma-Ray Bursts, and Fast Radio Bursts. Yet their formation rate relative to ordinary neutron stars remains uncertain, often precluding direct comparisons with the rates of these extragalactic transients. Furthermore, magnetars have been recently shown to be evolutionarily related to other neutron star classes, complicating the estimate of the exact magnetar fraction within the neutron star population. We study the magnetar birth fraction in core-collapse supernovae using pulsar population synthesis of all isolated neutron star classes in our Galaxy, incorporating self-consistently the Galactic dynamical evolution, spin-down and magneto-thermal evolution. This approach allows us to derive strong constraints from small close-to-complete observational samples. In particular, looking at the age-limited young ($<$2 kyr) neutron star population in the Milky Way we find 24 detected young neutron stars, with only 10 of them (41%) being classical rotational powered pulsars, while the others (59%) are either magnetars or central compact objects, the latter believed to be equally magnetically powered. We further compare the results with the nearby volume-limited class ($<$500 pc) of X-ray Dim Isolated Neutron stars, old nearby magnetars. We conclude that the observed population of isolated neutron stars in the Galaxy can be reproduced only by assuming a core-collapse supernova rate larger than two, and a larger magnetar fraction than previously inferred. By assuming a bimodal initial magnetic field ($B_0$) distribution at birth, we find that the magnetar class peaks between $B_0\sim 1-2.5\times10^{14}$ Gauss and represents on average $\sim50$% of the entire neutron star population.

Tunnelling reactions of molecules embedded on cryogenic noble-gas matrices are being used in fundamental studies of how reactivity varies with the nature of the supposedly inert matrix as well as pointers to the chemistry occurring in the interstellar medium on ice-grains. To these ends we present chemical kinetic rate constants for the \textit{cis} to \textit{trans} isomerisation of seleno-, thio- and monomeric formic acids and that of their three dimeric species, based on multidimensional calculations in the gas-phase, from 10~K to 300~K as a guide to the matrix reactions.

We use radiation-hydrodynamical simulations to investigate the formation and synthetic X-ray emission of hot bubbles within planetary nebulae (PNe) driven by the powerful winds of H-deficient, [Wolf-Rayet]([WR])-type stars. Our models, based on {\sc mesa} stellar evolution tracks for 1--3 M$_{\odot}$ progenitors, adopt a recent mass-loss rate prescription for [WR] stars and incorporate the enhanced radiative cooling of their C-rich material, comparing the results against standard H-rich PN models. The enhanced mass-loss in the [WR] models leads to an accelerated post-AGB evolution and a subsequent delay in hot bubble formation compared to their H-rich counterparts, as suggested by a previous work. By computing synthetic X-ray spectra that account for the mixed H-rich and H-deficient gas phases, we find that models incorporating [WR] winds exhibit significantly higher X-ray luminosities ($L_\mathrm{X}$) than their H-rich counterparts, but the emissivity-weighted plasma temperature of the X-ray-emitting gas converge to values of $T_\mathrm{X} = [1-3] \times 10^{6}$~K, regardless of whether the system follows a [WR]-type or an H-rich post-AGB evolutionary path. Our results reinforce previous suggestions that mixing is a key mechanism in generating the observed soft X-ray emission even for PN hosting [WR] central stars.

We present measurements of the cool molecular gas mass around the nuclei of two gas-rich mergers, III Zw 035 and IRAS F01364-1042, whose enclosed masses (M$_\mathrm{enc}$) within the central 40-80 pc would be overmassive if attributed entirely to the supermassive black hole mass (SMBH) and compared to SMBH-galaxy scaling relations. Our gas mass measurements are derived from Atacama Large Millimeter/submillimeter Array (ALMA) Band 6 long-baseline observations of CO(J=2-1) and 230 GHz continuum emission at 14-20 pc resolution, which probes below the resolving limit of the previous black hole mass measurements. Subtracting molecular gas mass from these enclosed masses is not enough to reconcile with BH-galaxy relationships, but independently measuring M$_\mathrm{enc}$ using the cold CO(2-1) gas does shift the black holes down to their expected values. Still, these ALMA data reveal respective molecular gas masses of $\sim$3$\times$10$^7$ to $\sim$6$\times$10$^8$ M$_\odot$ within 70 pc of these black holes, which could challenge some black hole accretion models that assume nuclear gas like this has no angular momentum.

Although galaxy evolution is governed by the interplay between baryonic physics and dark matter halo assembly, how halo properties shape observed galaxies remains unclear. With current challenges in measuring halo properties, the stellar-to-total dynamical mass relation is introduced as an alternative metric sensitive to the dark matter content within galaxies. We explore how spatially resolved stellar population properties vary across this relation using optical IFS data and photometry from 265 CALIFA galaxies. Spatially resolved ages and metallicities, [M/H], are derived using a Bayesian framework fed with a library of model spectra based on stochastic star formation and metallicity histories and dust attenuation. We study these properties in terms of both stellar and total dynamical mass, with the latter being enclosed mass within three effective radii from Jeans dynamical modeling. We find that ages and [M/H] measured at different annuli depend on both stellar and total mass, yet showing distinct radial trends. While the dependence of age on total mass is more prominent in the outskirts, that of [M/H] is significant in the inner parts. This behavior is reflected in the stellar population profiles and gradients, more strongly for age and connected to morphology. Intermediate-mass early-types have higher stellar-to-total mass ratios and flatter age profiles with older ages, and steep negative [M/H] profiles, whereas later-types have lower stellar-to-total mass ratios, negative age profiles with younger ages and shallower negative [M/H] profiles. Moreover, at fixed stellar mass galaxies have more negative age gradients and shallower [M/H] ones as total mass increases. Our results show that total dynamical mass is linked to systematic variations in stellar populations and radial gradients at fixed stellar mass, suggesting a relevant role of dark matter halos in shaping galaxy properties

On December 6, 2024, 1ES 1959+650, one of the X-ray brightest blazars known, underwent a high-amplitude X-ray outburst detected by SVOM, the first such discovery with this mission. The source was subsequently monitored with SVOM and Swift from December 2024 to March 2025. We report the detection and multi-wavelength follow-up of this event, and describe the temporal and spectral evolution observed during the campaign. Data from SVOM/MXT, SVOM/ECLAIRs, and Swift/XRT were analyzed with log-parabola models to track flux and spectral variability. The source was detected in a bright state over the 0.3-50 keV range. During the three months of monitoring, the X-ray flux varied significantly, showing episodes of spectral hardening at high flux levels. The spectral curvature evolved more irregularly and did not show a clear trend with flux. A shift of the Spectral Energy Distribution (SED) synchrotron peak to higher energies is seen when the flux increases. This constitutes the first blazar outburst discovered in X-rays by SVOM. The coordinated follow-up with Swift provided continuous coverage of the flare and highlights the strong complementarity of the two missions for time-domain studies of blazars. The flare shows no clear signatures of either Fermi I or Fermi II acceleration, suggesting a mixed Fermi I/II scenario.

The presence of multi-phase outflows in low ionisation nuclear emission-line regions (LINERs) has been confirmed to be frequent, but the mechanisms that launch them are still under study. We aim to explore the connections between the ionised gas outflow, radio continuum structures and X-ray emission detected in the LINER NGC4438. We analyse L, C and X-band images (from 1.4 to 12 GHz) of the LINER NGC4438, combining high-resolution data from enhanced Multi Element Radio Linked Interferometer Network (e-MERLIN) and Karl G Jansky Very Large Array (VLA). We produce radio flux, spectral index maps, and an energetic model that allows us to characterise the source. We incorporate optical integral field spectroscopy (IFS) data (GTC/MEGARA) and Chandra X-ray data, with comparable resolution, to better trace the outflow, the AGN and their potential connection. We present new L, C, and X-band high-resolution, high-sensitivity radio images and spectral-index maps that probe $\sim$ 25 pc scales in NGC 4438. These data reveal a close morphological correspondence between the radio structures and the ionised gas bubble. Using a spatially resolved energetic model based on radio flux and spectral index, we disentangle the compact AGN emission from the extended bubble for the first time, establishing their distinct physical origins. We measure a kinetic power of $\sim 5\times 10^{44}$ erg s$^{-1}$ for the radio bubble, exceeding the power of the ionised outflow by more than three orders of magnitude. Our multi-wavelength analysis indicates that NGC 4438 is undergoing jet-mode feedback, where a low-luminosity, weakly collimated jet impacts the dense northern interstellar medium. This interaction drives shock-ionised gas, produces a moderate velocity outflow that removes material from the region, and generates thermal X-ray emission coincident with the radio and H$α$ cavity.

IRAS23226-3843 has been identified as a highly variable Seyfert galaxy and even as a changing-look active galactic nucleus based on optical spectra. Here we present follow-up observations - taken over the past five years - for examining the ongoing photometric and spectral variations in this remarkable galaxy. We carried out SWIFT observations of IRAS23226-3843 together with new optical spectra taken in 2023 and 2024. In parallel we investigate ASAS-SN photometric data from 2014 till 2025. IRAS23226-3843 stayed on a high continuum flux level in the X-ray as well as in the optical since a historic outburst in 2019. However, it shows strong short-term variations on timescales of a few months. Densely sampled ASAS-SN V-band continuum data from 2014 till 2025 confirm that behavior. IRAS23226-3843 switched from a clear Seyfert 1 type in December 2019 to a Seyfert 1.9/2 type in July 2020 based on its optical spectra. Afterward, it again became a Seyfert 1 type with symmetric broad single-peaked Balmer line profiles in January 2023. These spectra prove the repeating changing-look character of the galaxy.IRAS23226-3843 exhibits extreme high Balmer decrements Ha\Hb based on their broad line components. The Balmer decrement values are on the order of 10. IRAS23226-3843 successively showed all types of broad line Balmer profiles during the past 25 years over periods of many years: asymmetric single-peaked, double-peaked, as well as single-peaked and symmetric profiles in addition to its Seyfert 1.9/2 transition. These variations are not clearly correlated with continuum and line intensity variations.

We present the discovery of HE0144-4657, an ultra metal-poor, CNO-enhanced star dynamically associated with the Helmi Stream disrupted dwarf-galaxy remnant. This star was first identified as a carbon-enhanced, metal-poor star candidate from the Hamburg/ESO objective-prism survey, then followed up with medium- and high-resolution spectroscopy. At [Fe/H]=-4.11, HE0144-4657 is the lowest metallicity star found in a stellar stream to date. Its chemistry is consistent with field halo stars in the same metallicity regime, and the light-element (atomic number Z<=30) chemical abundance pattern suggests that HE0144-4657 is a bona-fide second-generation star with a possible Population III progenitor in the 50Msun mass range with low explosion energy. One possible scenario for the origin of HE0144-4657 is that it was formed in an ultra-faint dwarf galaxy accreted by the Helmi Stream progenitor system before merging with the Milky Way. This discovery provides further evidence for the extragalactic origin of carbon-enhanced ultra metal-poor stars in the Milky Way and for the specific environments conducive to their formation.

Massive star-forming clumps are a prominent feature of high-redshift galaxies and are thought to trace gravitational fragmentation, feedback, and bulge growth in gas-rich disks. We present a statistical analysis of clump-like structures in $\sim$3600 galaxies spanning $2 \lesssim z \lesssim 8$ from deep JWST/NIRCam imaging in the JADES GOODS--South field. Clumps are identified as residual features after subtracting smooth Sérsic profiles, enabling a uniform, rest-frame optical census of sub-galactic structure. We characterize their physical properties, size--mass relations, and spatial distributions to constrain models of sub-galactic structure formation and evolution. We find that clumps in our sample are typically low-mass ($10^{\sim7-8}M_\odot$), actively star-forming, and show diverse gas-phase metallicity, dust attenuation, and stellar population properties. Their sizes and average pairwise separations increase with cosmic time (toward lower redshift), consistent with inside-out disk growth. The clump mass function follows a power law with slope $α= -1.50_{-0.17}^{+0.19}$, consistent with fragmentation in turbulent disks. We find a deficit of relatively young clumps near galaxy centers and a radial transition in the size--mass relation: outer clumps exhibit steeper, near-virial slopes ($R_{\rm e}\propto M_*^{\sim 0.3}$), while inner clumps follow flatter trends ($R_{\rm e}\propto M_*^{\sim 0.2}$), consistent with structural evolution via migration or disruption. These results provide new constraints on the formation, survival, and dynamical evolution of clumps, highlighting their role in shaping galaxy morphology during the peak of cosmic star formation.

We present nitrogen abundance measurements based on the low-ionisation [NII]6583 emission line for 588 galaxies between 1.5<z<7.0 from the JWST Advanced Deep Extragalactic Survey (JADES). We detect the temperature-sensitive [OIII]4363 auroral line in 40 galaxies in our sample, affording $T_e$-based abundances for this subset. We find that the average N/O abundance ratio in our low-metallicity sample is at least 0.1 dex higher than z ~ 0 samples. In particular, we find significant scatter toward high N/O, with five galaxies being identified with enhanced nitrogen abundances (log(N/O)>-1.1) at low-metallicity (12+log(O/H)<8.0) from $T_e$-based measurements. Meanwhile, applying strong-line abundance measurements to the remainder of our sample reveals a further 14 candidate galaxies passing these abundance cuts, implying that around 13 % of 12+log(O/H)<8.0 galaxies at these redshifts are nitrogen-enhanced at this level. We find that N/O abundance in low-metallicity systems correlates with SFR, surface density of SFR, and surface density of stellar mass at high redshift, while only in high-metallicity systems does a correlation with stellar mass emerge. Despite healthy representation of these `moderately nitrogen-enhanced' galaxies (-1.1<log(N/O)<-0.6), no galaxies in our low-metallicity sample are identified as having log(N/O)>-0.6, abundances that are typical of high-redshift NIII]- and NIV]-emitters. This demonstrates that the extreme nitrogen enhancements seen in some NIII]- and NIV]-emitters are only attained during the most extreme starbursts. This suggests that these elevated abundances are caused by enrichment from young massive stars in extreme environments and that the impact of this enrichment pathway is milder, though still important, for high-redshift systems on the star-forming main sequence.

Dust attenuation modifies the observed spectral energy distribution (SED), leading to biases in the properties inferred from integrated SED fitting. As spatially resolved SED modeling becomes feasible for large high-redshift samples, it is increasingly important to assess how dust attenuation affects resolved mass estimates. We evaluate the impact of dust attenuation on stellar mass estimates derived from integrating spatially resolved SED fitting results. We perform spatially resolved and integrated SED fitting on a sample of 3408 galaxies at $3 \leq z < 9$ from the GOODS South field, combining deep NIRCam from the JWST Advanced Deep Extragalactic Survey (JADES) and HST/ACS imaging from GOODS and CANDELS. We compare galaxy-integrated properties derived from fitting the summed SED with those obtained from spatially resolved SED modeling. Using a two-component dust attenuation model with a variable slope, we investigate how the dust attenuation slope, A(V), and stellar population properties contribute to discrepancies in the resulting stellar mass estimates. Resolved stellar masses are systematically higher than integrated estimates, with a median offset of +0.24 dex. Resolved analyses recover higher dust attenuations ($ΔA(V)\approx +0.08$ mag), lower birth cloud fractions ($Δμ\approx -0.28$), and grayer attenuation curves ($Δδ_{\mathrm{ISM}} = +0.08$), arising from preferential sampling of compact star-forming regions. Integrated fits underestimate stellar ages by $\sim23\%$ at $z < 5$ and 31$\%$ at $z \gtrsim 5$. The stellar mass offset correlates strongly with the age difference and the attenuation slope difference, indicating that age-dependent outshining and spatially varying dust geometry are primary drivers of the discrepancy between resolved and integrated stellar masses.

In the local Universe, supermassive black hole (SMBH) masses strongly correlate with their host-galaxies' stellar masses ($M_{*}$), but galaxies hosting faint AGN recently found by JWST may deviate from this relation. To constrain the M$_{\text{BH}}$-M$_{*}$ relation at high redshift, we performed AGN-host image decomposition for 17 low-luminosity AGN galaxies at $z$\,$\sim$\,4--6 using NIRCam images in the JADES GOODS-N field. These sources are identified as AGNs from broad H$α$ emission lines detected by the CONGRESS and FRESCO surveys. We used \textsc{galfit+MCMC} to fit spatial profiles in 7 wide-band images and detected extended emission in 9 sources out of 17. The close spatial alignment between the extended components and the AGN centers indicates that this emission likely originates from the host galaxies. These sources are extended at 0.9--2.0~$μ$m, suggesting significant host-galaxy light in the rest-frame UV. For the sources with the host detection, the stellar mass inferred based on image decomposition result can be 1-2 dex lower than the results without image decomposition. The BH-to-stellar mass ratio spans $M_{\text{BH}}/M_\ast$\,$\sim$\,0.01--1.48, placing them well above the local $M_{\text{BH}}$--$M_\ast$ relation. In contrast, the host-galaxy size--mass relation broadly agrees with previous measurements. Our results suggest that the host galaxies of these faint AGN are either genuinely under-massive compared to their black hole masses, or too compact to be spatially resolved.

Using JWST NIRCam imaging and grism spectroscopy from the JWST Advanced Deep Extragalactic Survey (JADES) Origins Fields, we report spectroscopic redshift measurements of 1,445 emission-line galaxies at $z=0-9$. Within this sample, we identify two prominent galaxy protoclusters at $z = 3.47$ and 3.69, each anchored by massive dusty star-forming galaxies (DSFGs). In the vicinity of these systems, we discover seven background galaxies at $z=3.6 - 6$ that simultaneously exhibit strong rest-frame optical emission lines (e.g., [O III] and H$α$) and unusually reddened UV-to-optical continua. We attribute this reddening to dust extinction arising from the circumgalactic medium (CGM) of the foreground DSFGs at projected separations of 7-30 kpc. We infer a high dust column density ($\gtrsim 10^{-1}$ Msun/kpc^2), substantially exceeding those measured in low-redshift halos and those predicted by hydrodynamical simulations like IllustrisTNG and FIRE-2. The steep extinction curves, comparable to or steeper than that of the SMC, indicate a dominant population of small dust grains in the high-redshift CGM. We conclude that DSFGs at this epoch host large reservoirs of dusty CGM enriched to solar metallicity. These extended dust components are largely invisible to (sub-)millimeter interferometers such as ALMA because of their low surface brightness. We discuss the physical processes in dust transport that might be key to reproducing our observations, including galaxy mergers, cool-phase gas outflows, dust shattering, sputtering and radiation pressure. Finally, we caution that foreground CGM dust extinction may redden background galaxies at intermediate redshifts to mimic Lyman-break galaxies at $z\gtrsim10$.

We report a galaxy overdensity candidate at $z\approx 10.5$ in the JWST Advanced Deep Extragalactic Survey (JADES). This overdensity contains 18 galaxies with consistent photometric redshifts and robust F115W dropouts within 8 comoving Mpc in projection. The galaxy number density is four times higher than the field expectation, accounting for one-third of comparably bright galaxies and nearly 50% of the total star formation rate at $10<z_\mathrm{phot}<12$ in the GOODS-S field. Two compact members of the overdensity show potential Balmer breaks suggestive of evolved stellar populations or little red dots (LRDs). One-third of galaxies have close companions or substructures within 1 kpc at consistent photometric redshifts, implying more frequent interactions in an overdense environment. Most galaxies have stellar masses of 0.6-3$\times10^8$ $M_\odot$, half-light radii of $\sim$200 pc, and star formation rates of $\sim$5 $M_\odot \mathrm{yr^{-1}}$, with no significant deviation from typical high-redshift scaling relations. We find tentative evidence for a spatially varying Ly$α$ transmission inferred photometrically, consistent with an emerging ionized bubble. This overdensity provides a rare opportunity for probing the environmental impact on galaxy evolution and the onset of cosmic reionization within the first 500 Myr.

We present a sample of 2081 sources selected at photometric redshift $z_{\mathrm{phot}} > 8$ across the JADES DR5 data release in GOODS-S and GOODS-N over a total area of 469 square arcmin. These sources range from $M_{\mathrm{UV}} = -22$ to $M_{\mathrm{UV}} = -16$, with 19 objects at $z_{\mathrm{phot}} > 14$. We estimate the UV slopes for the full sample from fits to the photometry and find evidence for a steepening of the relationship between the UV continuum slope and $M_{\mathrm{UV}}$ to higher redshifts, a result that differs from prior analyses of brighter samples in the literature. We provide evidence that over one quarter of our sources have evidence for being morphologically extended, with many galaxies showing multiple bright knots or clumps even out to $z \sim 13 - 14$, an indication of how galaxies at Cosmic Dawn are growing and evolving. We discuss JADES-GN+189.15982+62.28899, a GOODS-N F200W dropout galaxy at $z_{\mathrm{phot}} \sim 15 - 18$ which has been observed spectroscopically with JWST/NIRSpec in prism mode, resulting in a very low signal-to-noise spectrum that is consistent with the photometry and rules out a number of low-redshift solutions for the source. Finally, we use a subsample of 123 objects in our sample with spectroscopic redshifts to explore the usage of alternate fitting templates and a prescription for Ly-$α$ damping wing absorption, finding that both produce significant improvements to the estimated photometric redshifts.

Wisps are among the most prominent scattered light artifacts in JWST/NIRCam imaging. They often appear in certain regions of the detectors and contaminate observations at surface-brightness levels relevant for faint-source photometry. We introduce a new subtraction method that uses the non-negative matrix factorization (NMF) algorithm to model and remove wisps. Using deep NIRCam observations from the JWST Advanced Deep Extragalactic Survey (JADES) and other programs, we construct multi-component, filter- and detector-specific wisp templates that capture the wisp structures and their exposure-to-exposure morphological variations. Wisps in individual exposures are represented as non-negative linear combinations of these templates, consistent with their additive nature and reducing degeneracies relative to single-template scaling. Compared to existing approaches, our method delivers lower residual root mean square in wisp-affected regions and reduces photometric bias and scatter to levels consistent with clean detector areas. The NMF wisp templates are readily applicable to other datasets and are publicly released to support future NIRCam extragalactic surveys.

We present morphological parameters and their uncertainties for all sources detected in JWST/NIRCam imaging in GOODS-N and GOODS-S from the JWST Advanced Deep Extragalactic Survey (JADES) catalogs. We model the surface brightness profiles of these sources with single-component Sérsic profiles, performing Bayesian inference of galaxy structural parameters. We fit each of the $>10^5$ sources with every available JWST/NIRCam wide-band filter individually, amounting to over 3 million Sérsic profiles computed. We provide catalogs of this morphological information, building one of the largest extragalactic morphological datasets to date, which we share alongside imaging and photometry from the JADES Data Release 5. With this information, we analyze the rest-frame optical redshift evolution of the effective radius and the surface luminosity density within a radius of 1 kiloparsec, $Σ_{\text{1 kpc}}$, for 24,692 galaxies at $z>1$. We find $r_{\text{eff}} \propto (1+z)^{-0.635 \pm 0.013}$ kpc, while $Σ_{\text{1 kpc}}$ is relatively constant across time. Additionally, we explore bulge-disk decomposition on a subset of 8,390 galaxies in the JADES deep imaging covering the Hubble Ultra Deep Field, finding the effective radius of the bulge-components to increase marginally with time, whereas the disk-component sizes evolve as $r_{\text{eff,disk}} \propto (1+z)^{-1.091 \pm 0.043}$. Future work modeling multi-component surface brightness profiles will enable further analysis of the morphological evolution of galaxies across cosmic time.

JADES Data Release 5 (DR5) photometric catalogs and describes the methodologies used for source detection, deblending, photometry, uncertainty estimation, and catalog curation. The catalogs are constructed from 35 space-based imaging mosaics obtained with JWST/NIRCam, JWST/MIRI, HST/ACS, and HST/WFC3, combining approximately 1250 hours of JADES imaging with extensive additional public JWST and HST observations in the GOODS fields. Sources are identified using custom signal-to-noise-based detection and deblending algorithms optimized for the depth, resolution, and complex point-spread-function structure of JWST imaging. Source centroids, shapes, and photometric apertures are determined using a new fast two-dimensional Gaussian regression method applied to detection-image profiles. We provide forced circular-aperture photometry, ellipsoidal Kron photometry, and curve-of-growth measurements for every source in every band. We introduce a new pixel-level regression framework to model photometric uncertainties as a function of aperture size and local mosaic properties, accounting for correlated noise in heterogeneous JWST mosaics. Photometric redshifts are computed using template-based fitting applied to both small-aperture photometry on unconvolved images and Kron photometry on common-PSF mosaics. The JADES DR5 catalogs supersede previous JADES photometric releases, and are publicly released through the Mikulski Archive for Space Telescopes and an interactive web interface.

Medium to ultra-deep mid-infrared imaging surveys with the James Webb Space Telescope (JWST)'s Mid-Infrared Instrument (MIRI) are reframing our view of the early Universe, from the emergence of ultra-red dusty and quiescent galaxies to the epoch of reionization to the first galaxies. Here we present the MIRI coordinated parallels component of the JADES program, which obtained ultra-deep (155 ks) imaging at $7.7 μ$m over $\sim10$ arcmin$^2$ as well as medium depth ($\sim5-15$ ks) imaging at $7.7, 12.8$, and $15 μ$m over $\sim36$, 25, and 22 arcmin$^2$, respectively, in the GOODS-S and GOODS-N fields. This paper describes the data reduction, which combines the official JWST Calibration Pipeline with custom steps to optimize flagging of warm/hot pixels and optimize background subtraction. We further introduce a new step to address artifacts caused by persistence from saturating sources. The final, fully reduced JADES/MIRI mosaics are being released as part of JADES Data Release 5, along with prior-based forced photometry using NIRCam detection images, providing critical rest-frame near-infrared and optical constraints on early galaxy populations.

We present the Near Infrared Camera (NIRCam) imaging products of the fifth data release (DR5) of the James Webb Space Telescope (JWST) Advanced Deep Extragalactic Survey (JADES). The JADES survey is one of the most ambitious programs yet conducted on JWST, producing deep infrared imaging and multiobject spectroscopy on the GOODS-S and GOODS-N extragalactic deep fields in order to explore galaxies to the earliest epoch. Here we describe the NIRCam data reduction procedures that result in deep and well-characterized mosaics in up to 18 filters covering 469 arcmin$^2$, with 250 arcmin$^2$ having at least 8 filters of coverage. This release contains the full NIRCam imaging of JADES, over 800 JWST mission hours, as well as co-reductions of 19 other programs in these two premier deep fields. We perform detailed tests on the final data products, thereby characterizing the photometric properties, point-spread function, and astrometric alignment. We release mosaics for individual programs (or epochs, depending on scheduling) and the mosaics combining data from all programs in order to facilitate photometric variability studies and the deepest possible photometry.

The thermal Sunyaev-Zel'dovich (tSZ) effect provides a powerful probe of the thermal pressure of ionised gas in galaxy clusters and the cosmic web; constrained simulations reconstruct the mass and velocity fields of the local Universe. We explore how these two may be mutually informative: the tSZ signal provides a benchmark for assessing the fidelity of constrained simulations, and constrained simulations contribute information on the positions, total masses and density profiles of cosmic web structures for use in tSZ studies. We focus on cluster predictions in the Bayesian Origin Reconstruction from Galaxies (BORG) paradigm, introducing CSiBORG-Manticore, a new state-of-the-art suite of digital twins -- data-constrained posterior simulations whose initial conditions are inferred via Bayesian forward modelling. We develop a framework for scoring constrained simulations on their ability to match measured Compton-$y$ maps from Planck for cluster cutouts, and use it to demonstrate improvement from previous BORG reconstructions. We further validate halo masses against weak-lensing-calibrated X-ray masses from eROSITA. We also show how high-fidelity digital twins offer a practical route to extracting additional information from tSZ data through a novel calibration of the mass-observable relation, and provide a complementary framework to purely statistical analyses of Compton-$y$ maps. This paves the way for integrating the large-scale structure information inherent in constrained simulations into the study of CMB secondary anisotropies.

Intrinsic alignments are a major astrophysical contaminant for next generation large-sky surveys like Euclid and LSST. Large hydrodynamic simulations are crucial for informing the alignment modelling for these surveys. We measure position-position and position-shape correlations of a Luminous Red Galaxy sample from the FLAMINGO suite of hydrodynamical simulations, measuring the alignment signal for more than 4.9 million galaxies at redshift 0. We jointly model the clustering and alignment correlations to provide the tightest constraints on the alignment amplitude to date from a hydrodynamic simulation. We find that both the Non-Linear Alignment (NLA) and the more complex Tidal Alignment Tidal Torquing (TATT) models provide good fits to the data. We compare the measured $A_1$ amplitude to observational data and find good agreement. We measure the dependence of the NLA and TATT free parameters on halo mass. We also introduce a mass-dependent TATT model, TATT-M, by finding empirical relations between the halo mass and the TATT parameters. This allows us to fit TATT with only one parameter, $A_1$, with $A_2/A_1$ being a constant and $A_{1δ}/A_1$ being a function of halo mass. Using a Bayesian approach, we find that TATT-M is very strongly preferred by the data over NLA. Using the baryonic feedback variations of the FLAMINGO simulation suite, we test whether the TATT parameters are sensitive to feedback. Variations in AGN and supernova feedback do not significantly change the alignment amplitude beyond the change associated with the dependence of galaxy stellar mass on the strength of feedback. Our results inform the IA modelling for upcoming surveys by providing guidance on model choices, priors and sensitivities to feedback.

Reconstructing images of the radio sky from incomplete Fourier information is a key challenge in radio astronomy. In this work, we present a method for radio interferometeic image reconstruction using a data-driven prior for the radio sky based on denoising diffusion probabilistic models (DDPMs). We first train a DDPM on radio galaxy observations from the VLA FIRST survey. We create simulated VLA, EHT, and ALMA observations of radio galaxies, then use an unsupervised posterior sampling method called Denoising Diffusion Restoration Models (DDRM) to reconstruct the corresponding images, using our DDPM as a prior. Our approach is agnostic to the measured radio interferometric data and naturally incorporates the physics of the measurement process. We are able to reconstruct images with very high fidelity PSNR>60, a marked improvement over CLEAN and similar image reconstruction methods using conditional DDPMs

Deuterated molecules are a useful diagnostic tool to probe the evolution and the kinematics in the earliest stages of star formation. Due to the low temperatures and high densities in the centre of pre-stellar cores, the deuterium fraction is enhanced by several orders of magnitude. We study the distribution of the emission and the deuteration of the two carbon chains HC3N and CH3CCH throughout the pre-stellar core L1544. We analyse emission maps of CH3CCH, CH2DCCH, CH3CCD, HC3N, HCC13CN, and DC3N, observed with the IRAM 30m single-dish radio telescope. We use non-LTE radiative transfer calculations, combined with chemical modelling of the molecular abundances, to constrain physical parameters of the observed species. Following this, we derive the column density and deuteration maps. We find D-fractions of N(DC3N)/N(HC3N)=0.04-0.07, N(CH2DCCH)/N(CH3CCH)=0.09-0.15, and N(CH3CCD)/N(CH3CCH)=0.07-0.09. The deuteration of HC3N appears homogeneous across the core, with widespread D-fraction values above 0.06, tracing intermediate-density gas in the outer layers of the core. CH3CCD is most efficiently formed in the higher-density regions towards the core centre, while the D-fraction of CH2DCCH traces a local density enhancement in the north-east of the core, coinciding with the CH3OH emission peak. The results suggest that gas-phase reactions dominate the formation and deuteration of both HC3N and CH3CCH in L1544, with spatial variations driven by physical structure, density and external radiation. The significantly higher D-fraction of CH2DCCH compared to CH3CCD and a tentative gradient with higher values in the north suggest different deuteration mechanisms for the two functional groups. Similarities between the CH2DCCH emission and CH2DOH might indicate an additional deuteration pathway of CH3CCH on the surfaces of dust grains, as observed for H2CO.

Motivated by recent claims of a possible redshift dependence in late-Universe determinations of the Hubble constant ($H_0$), we test the robustness of this behaviour using multiple cosmological probes. We perform a joint redshift-binned analysis of $H_0$ across eight bins using late-Universe probes -- Pantheon+ SNe~Ia, DESI BAO, cosmic chronometers, and water megamasers -- under three cosmological frameworks: flat $Λ$CDM, CPL, and Padé cosmography. Under a common baseline scheme, all three models show a qualitatively similar, low-amplitude variation in the per-bin $H_0$ estimates. A simple Fourier-like parametrization captures this behaviour, but the amplitude differs from zero only at a marginal significance of about $1.71$--$1.94\,σ$, with similar behaviour observed across all three cosmological frameworks. We then investigate the robustness and possible origin of this feature. Alternative binning schemes preserve its qualitative form, whereas single-probe per-bin fits (SNe-only, CC-only, BAO-only) yield ratios $H_{0,i}/H_{0,\mathrm{global}}$ mostly consistent with unity and do not reproduce the pronounced drift seen in the joint baseline constraints. Finally, by comparing different global versus piecewise-constant configurations for $\{H_0,Ω_m,M,r_d\}$, we find that a baseline-like oscillatory pattern re-emerges only when multiple degenerate parameter combinations are allowed to vary across bins, while it is strongly suppressed when only $H_0$ is bin-dependent. Taken together, these results indicate that the apparent oscillatory behaviour of $H_0(z)$ in late-time arises from known parameter degeneracies and does not constitute robust evidence for a genuine redshift evolution.

Nuclear astrophysics aims at unraveling the cosmic origins of chemical elements and the physical processes powering stars. It constitutes a truly multidisciplinary field, that integrates tools, advancements, and accomplishments from theoretical astrophysics, observational astronomy, cosmochemistry, and theoretical and experimental atomic and nuclear physics. For instance, the advent of high-energy astrophysics, facilitated by space-borne observatories, has ushered in a new era, offering a unique, panchromatic view of the universe (i.e., allowing multifrequency observations of stellar events); supercomputers are also playing a pivotal role, furnishing astrophysicists with computational capabilities essential for studying the intricate evolution of stars within a multidimensional framework; cosmochemists, through examination of primitive meteorites, are uncovering tiny fragments of stardust, shedding light on the physical processes operating in stars and on the mechanisms that govern condensation of stellar ejecta into solids; simultaneously, nuclear physicists managed to measure nuclear reactions at (or close to) stellar energies, using both stable and radioactive ion beam facilities. This paper provides a multidisciplinary view on nucleosynthesis accompanying stellar explosions, with a specific focus on thermonuclear supernovae, classical novae, and type I X-ray bursts.

GRB~250704B/EP250704a, identified as a short gamma-ray burst (sGRB), exhibited prolonged X-ray emission following the prompt phase and, in optical and infrared (IR) bands, an unusual one-day plateau succeeded by a rapid decline. This sGRB was observed by multiple satellites and ground-based observatories across the electromagnetic spectrum. This study presents temporal and spectral analyses from radio to gamma-ray frequencies, spanning several observation periods beginning after the trigger and continuing for nearly 2 days. The results of the temporal and spectral analyses of the prompt episode, the extended X-ray component, and the afterglow phase are consistent with a millisecond magnetar undergoing accretion. The long-lasting X-ray emission is attributed to the internal energy dissipation of the magnetar spin-down power, governed by the magnetization parameter; the extended optical/IR plateau to synchrotron afterglow emission with energy injection; and the steep decay to changes in microphysical parameters during the post-jet break phase. The X-ray observations are consistent with the superposition of spin-down luminosity and synchrotron afterglow scenario. These findings suggest that the compact-object remnant is most likely a long-lived magnetar.

Most efforts in the modeling of recurrent novae have centered on the initial phases of the explosion and ejection, overlooking the subsequent interaction of the ejecta, first with the accretion disk orbiting the white dwarf and ultimately with the secondary star. To address this gap, a series of 3D smoothed-particle hydrodynamics simulations was conducted. These simulations explored the dynamic interactions between the nova ejecta, accretion disk, and stellar companion within the framework of the recurrent nova system U Sco. Notably, the simulations incorporate rotation around the system's center of mass. The primary goal of these simulations was to qualitatively examine the impact of various model parameters, including ejecta mass, velocity, and density, as well as the mass and geometry of the accretion disk. Simulations reveal complete disruption and sweeping of the accretion disk orbiting the white dwarf star for models with flared disks and Mejecta/Mdisk larger than 1. In contrast, V-shaped disks with a (constant) high initial density and Mejecta/Mdisk < 1 partially survive the impact with the nova ejecta. A very minor chemical contamination of the secondary star is anticipated in the U Sco case based on the limited impact of nova ejecta particles on the subgiant in all simulations. Minor mass ejection from the subgiant's outer layers is observed during the late-stage collision with ejecta and disk material, with some particles ejected from the binary system and some accreted by the white dwarf.

Recent high-resolution imaging observations of strong lens systems reveal dense low-mass perturbers. We propose a soliton core, whose central density is boosted by a supermassive black hole (SMBH), in the fuzzy dark matter (FDM) model as an efficient perturber in strong gravitational lensing. The higher central density makes it less efficient in the tidal mass loss, and leads to the higher impact in gravitational lensing. We show that the mass profile of a $\sim 10^6M_\odot$ perturber in JVAS B1938+666, which does not resemble any known astronomical object, can be wel explained by a soliton core in the FDM model with the mass of $4\times 10^{-21}$eV hosting an SMBH with the mass of $4\times 10^5M_\odot$. The high mass of the SMBH may be explained by several scenarios that predcit heavy SMBH seeds such as the direct collapse black hole formation and primordial black holes.

Massive stars ($M>8\mathrm{M_\odot}$) emit neutrinos known as pre-supernova (pre-SN) neutrinos through thermal and nuclear interactions for cooling the stellar core during the final stage of stellar evolution. Real-time monitoring of their pre-SN neutrino interaction rate offers a crucial opportunity to issue an early warning to a core-collapse supernova. Some neutrino detectors, including KamLAND and Super-Kamiokande already operate pre-SN alarm systems based on a statistically significant excess of the observed event rate over the expected background. To improve alarm sensitivity, we propose an alarm method which incorporates the time evolution of the observed pre-SN neutrino event rate. The method uses a log likelihood ratio test that references multiple theoretical stellar-evolution models and treats the core collapse time as a nuisance parameter to be profiled over. The performance of the proposed method was evaluated using simulated data for the KamLAND, Super-Kamiokande with dissolved Gadolinium (SK-Gd) and their combined analysis. The results demonstrate a significant improvement in the warning time compared to the conventional rate-only method, while maintaining the same false alarm rate.

In this paper, we review the extended halo material and the circumgalactic medium (CGM), including both dust and gas, and discuss promising science cases that could be realized using the KASI Deep Rolling Imaging Fast Telescope (K-DRIFT). Scattered starlight from cirrus clouds in our Galaxy poses one of the major challenges to studying the low surface brightness features of extragalactic sources. Therefore, it is essential to investigate how to discriminate extragalactic sources from the cirrus cloud features. At the same time, interstellar dust clouds themselves are fundamental to understanding dust properties and the interstellar radiation field, both of which are essential for studies of chemical evolution and star formation in our Galaxy. Measuring the reddening of background sources, such as quasars, with K-DRIFT, which benefits from its broad field of view and accurate background subtraction, allows for effective detection of extended dust in galactic halos, the CGM, and intracluster space. Observations of the H-alpha emission lines can be used to identify signatures of star formation activity within galaxies, as well as the environmental effects acting on them. Galactic winds driven by active galactic nuclei and starbursts can be traced through H-alpha emission. Strong ram pressure stripping effectively removes the interstellar medium (ISM) from galaxies. The stripped ISM becomes ionized or dissociated through mixing with the hot intracluster medium (ICM), forming H-alpha tails. The surface brightness of these H-alpha tails correlates not only with the presence of star formation in the tails but also the mixing stage of the stripped ISM and ICM. The H-alpha survey with K-DRIFT will enable the investigation of the evolutionary stages of ram pressure stripped galaxies in cluster environments, as well as the multiphase gas reservoir around galaxies and in the CGM.

Next-generation surveys such as the Euclid survey, the Legacy Survey of Space and Time (LSST), and the China Space Station Telescope (CSST) survey are expected to discover ~10^5 galaxy-galaxy scale strong gravitational lenses. This motivates the development of scalable and robust lens modeling approaches that can efficiently and reliably learn from wide-field survey datasets before high-resolution follow-up. We design a scalable, Bayesian, Lenstronomy-based pipeline and apply it to a sample of sixteen lens candidates observed with the Parallel Imager for Southern Cosmology Observations (PISCO) on the Magellan telescope. PISCO provides four-band imaging (z, i, r, g) with colours, depth and seeing conditions comparable to LSST. To fully exploit the constraining power of this dataset, our pipeline performs simultaneous multi-band modeling, using a common mass profile across all four bands while allowing independent light profiles in each. This approach leverages color information to provide joint constraints on the lens mass and yields reduced uncertainties compared to single-band analyses. Fifteen out of sixteen PISCO lens candidates are successfully recovered with interpretable lensing configurations, including DESJ0533-2536, the first reported hyperbolic-umbilic galaxy-galaxy scale strong lensing candidate. We further assess how much model complexity can be reliably constrained given the resolution and seeing of PISCO-like data. Overall, our results demonstrate that scalable, multi-band lens modeling of ground-based data can extract meaningful constraints on mass and source morphology, providing a practical pathway to maximize the scientific return from large samples in upcoming surveys.

We discuss a potential survey to map dark matter and the emergence of large-scale structure to redshift z ~ 1.5 (baseline) or z~3.5 (with near-IR extension) using a massively multiplexed spectrograph on a 10m-class telescope, such as the proposed Widefield Spectroscopic Telescope.

We present a comprehensive catalog of 251 potential counterparts for 250 unassociated hard X-ray sources detected in the Swift Burst Alert Telescope (BAT) 150-month hard X-ray survey. Over 150 months of observation, BAT has detected 2339 sources in the 15-150 keV energy range. Among these, 344 do not have a previously identified low-energy counterpart. Our study focuses on the analysis of soft X-ray observations at energies below 10 keV, spatially overlapping with these new Swift-BAT hard X-ray sources. Such observations were taken with Chandra, Swift-XRT, eROSITA, and XMM-Newton. Within the sample of 251 potential counterparts, 94 (37 percent) are identified as active galactic nuclei and 58 (23 percent) as galaxies. The remaining 99 sources (40 percent) include pulsars, cataclysmic variables, and unclassified soft X-ray counterparts in the 0.5-10 keV band. Redshift information is available for 139 out of the 251 sources, and its distribution is in close agreement with the redshift distribution of previous BAT catalogs. We also present the results of a small optical spectroscopy campaign of 9 out of 58 galaxies. The majority of these are classified as Seyfert 2 galaxies at redshifts slightly larger than the median of the BAT AGN sample.

We present a multi-epoch test of f(T) gravity with nonminimal torsion-matter coupling, combining early- and late-Universe observations. At the MeV scale, Big-Bang Nucleosynthesis constrains the fractional variation of the weak freeze-out temperature, |δτ_f/τ_f|, thereby mapping light-element abundances into limits on deviations from the standard expansion history. At low redshift, we confront the model with type Ia supernovae, baryon acoustic oscillations, and cosmic-chronometer data, which respectively probe distances, the late-time standard ruler, and the Hubble rate. Independent analyses highlight the complementary roles of each dataset, while a joint SNe Ia + BAO + CC fit breaks degeneracies and yields the tightest combined bounds. As an illustration, we examine two representative torsion-modified gravity scenarios: BBN strongly limits large departures from standard cosmology, whereas late-time probes remain compatible with a near-ΛCDM background. This unified approach demonstrates the power of linking early-Universe nuclear physics with precision cosmological observables in assessing torsional extensions of gravity.

Magnetars are now known to be among the most strongly polarized celestial sources in X-rays. Here we report on the $500\,\mathrm{ks}$ observation of the magnetar 1E 1547.0-5408 performed by the Imaging X-ray Polarimetry Explorer (IXPE) in March 2025. The IXPE spectrum is well reproduced by a single thermal component with blackbody temperature $kT_\mathrm{BB}\sim 0.67\,\mathrm{keV}$ and emission radius $R_\mathrm{BB}\sim 1.2\,\mathrm{km}$. The source exhibits a high linear polarization degree in the $2$--$6\,\mathrm{keV}$ band ($\mathrm{PD}=47.7\pm2.9\%$) with polarization angle $\mathrm{PA}=75^\circ.8 \pm 1.^\circ8$, measured West of celestial North. While $\mathrm{PA}$ does not appear to vary with energy, there is some evidence (at the $1σ$ confidence level) of a minimum in $\mathrm{PD}$ between $3$ and $4\,\mathrm{keV}$, compatible with what is expected by partial mode conversion at the vacuum resonance in a magnetized atmosphere. Phase-resolved spectral and polarimetric analyses reveal that X-ray thermal radiation likely originates from a single, fairly small hot spot with a non-uniform temperature distribution. Fitting the phase-dependent $\mathrm{PA}$ measured by IXPE with a rotating vector model (RVM) constrains the source geometry and indicates that both the dipole axis and line-of-sight are misaligned with respect to the spin axis. Under these conditions, the high polarization of the source cannot be regarded as compelling evidence for the presence of vacuum birefringence in the star magnetosphere. Nevertheless, the fact that the RVM successfully reproduces the modulation of the X-ray polarization angle and the behavior of $\mathrm{PD}$ with the energy hint once more to the presence of QED effects in magnetars.

3I/ATLAS is an interstellar object whose activity provides critical insights into its composition and origin. However, due to its orbital geometry, the object is too close to the Sun near perihelion to be observed from the ground, and space-based measurements are therefore required. Here we characterize the water production rate of 3I/ATLAS using SOHO/SWAN Lyman-$α$ observations from 2025 November to December (heliocentric distances 1.4 to 2.2 au) with 3D Monte Carlo modeling. We report a peak post-perihelion water production rate of $Q_{\mathrm{H_2O}} \approx 4 \times 10^{28}$ mol s$^{-1}$, corresponding to a minimum active fraction of $\sim$30\% (assuming a maximum nucleus radius of 2.8 km). Comparison of our post-perihelion measurements with published pre-perihelion results reveals a heliocentric asymmetry, with an $r_h^{-5.9 \pm 0.8}$ scaling for the inbound rise, followed by a shallower $r_h^{-3.3 \pm 0.3}$ scaling during the outbound decline, where $r_h$ is heliocentric distance. The post-perihelion behavior indicates that the water production of 3I/ATLAS was driven primarily by the varying solar insolation acting on a stable active area. Combined with other evidence, including comparison with the hyperactive comet 103P/Hartley 2, our findings suggest that its water production is likely dominated by a distributed source of icy grains. Furthermore, it displayed remarkable stability in the activity with no signs of outbursts or rapid depletion of water production.

Observational studies have made substantial progress in characterizing quenching as a function of stellar mass and environment, but they are often limited in their ability to constrain quenching timescales and to determine the dominant environmental process responsible for the shutting down of star formation. To address this, we combine recent Sloan Digital Sky Survey (SDSS) observations with the SHARK v2.0 semi-analytic model to study the quenching of satellite galaxies in groups and clusters. We generate mock SDSS-like observations to calibrate the hot halo and cold interstellar medium (ISM) gas stripping prescriptions against observed satellite quenched fractions, finding that the previously adopted stripping prescriptions in SHARK v2.0 are too aggressive and overestimate the quenched fraction of satellite galaxies. Reducing the efficiency of both hot and cold gas stripping yields excellent agreement with observations for low- and intermediate-mass satellite galaxies. We use the calibrated model to investigate quenching timescales and find that satellites quench more quickly in clusters compared to groups, with timescales that generally decrease with increasing stellar mass. The long (>2 Gyr) timescales we measure favour hot halo gas removal as the dominant driver of satellite quenching.

We present an upgraded detailed numerical calculations of supernova (SN) interactions with significantly aspherical circumstellar matter (CSM), primarily formed as a disc or bipolar lobes. The circumstellar disc can arise as a result of, for example, mass transfer in a binary, while bipolar lobes can be the result of a violent pre-explosive ejection of matter, similar to the iconic cases of luminous blue variable stars (LBVs). We numerically simulate the radiation-hydrodynamic (RHD) behaviour of interaction processes using a 2D cylindrical version of the RHD code CASTRO. We then calculate light curves, spectral patterns, and polarisation profiles, all up to a relatively long time of two years after an SN shock breakout and from different directions, using the multidimensional Monte Carlo radiation transfer (MC-RT) codes SEDONA and SIROCCO. We calculated a total of five models for the two aforementioned configurations of the surrounding CSM, for stratified density levels, comparing the simulated hydrodynamic behaviour and differences in their observable properties. RHD models exhibit similar behaviour to previous adiabatic models, but with a significantly slower expansion velocity. The calculated light curves show a relatively smooth evolution in SN-disc interaction, and declines and brightening in SN-lobes interaction. Comparing models with real events with a presumed similar physical process provides guidance for selecting a more accurate CSM configuration when simulating real situations.

The dynamics of the solid crust + magnetic field lines of pulsars is a much debated issue, and remains unsettled after 50 years. Some pieces of evidence have emerged to complete and confirm theoretical calculations and expectations. We discuss in the present work an interpretation of the behavior of the ''Crab Twin'' pulsar PSR B0540-69 in terms of the evolution of the magnetic field/quakes, connecting the behavior of the braking index with the underlying platelet drift and sudden discontinuous rearrangement (fast-slip) and long-term ones (slow-slip events), suggested by analogy with existing theoretical picture observed in the Earth's crust. The relationship of this scenario with permanent torque-changing glitches seen in the Crab and other young pulsars, and a set of similar events in the same object and others is addressed. We conclude that this physical approach is in principle consistent with all these sudden events, and point out future work to clarify the whole picture.

Establishing ages for young clusters is key for properly tracking the star formation history of a region. In this paper we investigate a new approach to estimating ages for young populations, based on the well-founded assumption that the initial mass function is the same throughout a star forming cloud. We trial this method for six young clusters in the Perseus star forming region. For all six clusters, we construct new member samples in a homogeneous way using Gaia DR3. We estimate masses by comparing 2MASS photometry to theoretical isochrones, including Monte Carlo simulations to propagate the errors. We compare the mass distributions of the clusters for a range of plausible ages, looking for a combination of ages that results in indistinguishable mass distributions across the region. We find the best fit for ages of 1 Myr for NGC1333+Autochthe, 2 Myr for IC348, 2-3 Myr for Heleus, 3-4 Myr for Mestor, 4-5 Myr for Electryon+Cynurus, and 5-8 Myr for Alcaeus. All other combinations of ages are ruled out by this criterion. The established age sequence is consistent with the relative ages inferred from disc fractions, and broadly aligns with the age sequence determined in previous studies using isochrone fitting. We suggest that this approach can be a useful complement and cross-check to established methods to estimate ages in young populations.

Accreting systems can launch powerful outflows which interact with the surrounding medium. We combine new radio observations of the accreting neutron star X-ray binary (XRB) Circinus X-1 (Cir X-1) with archival radio observations going back 24 years. The $\sim3$ pc scale wide-angle radio and X-ray emitting caps found around Cir X-1 are identified as synchrotron emitting shocks with significant proper motion and morphological evolution on decade timescales. Proper motion measurements of the shocks reveal they are mildly relativistic and decelerating, with apparent velocity of $0.14c\pm0.03c$ at a propagation distance of 2 pc. We demonstrate that these shocks are likely powered by a hidden relativistic ($\gtrsim0.3c$) wide-angle conical outflow launched in $1972\pm3$, in stark contrast to known structures around other XRBs formed by collimated jets over 1000s of years. The minimum time-averaged power of the outflow required to produce the observed synchrotron emission is $\sim0.1L_\text{Edd}$, while the time-averaged power required for the kinetic energy of the shocks is $\sim40 \left(\frac{n}{10^{-2} \text{cm}^{-3}}\right)L_\text{Edd}$, where $n$ is the average ambient medium number density. This reveals the outflow powering the shocks is likely significantly super-Eddington. We measure significant linear polarisation up to $52\pm6\%$ in the shocks demonstrating the presence of an ordered magnetic field of strength $\sim200~μ\text{G}$. We show that the shocks are potential PeVatrons, capable of accelerating electrons to $\sim0.7~\text{PeV}$ and protons to $\sim20~\text{PeV}$, and we estimate the injection and energetic efficiencies of electron acceleration in the shocks. Finally, we predict that next generation gamma-ray facilities may be able to detect hadronic signatures from the shocks.

[Abbreviated] Context. Solar twins, stars whose stellar parameters (Teff, log g, and [M/H]) are very close to the Solar ones, offer a unique opportunity to investigate Galactic archaeology with very high accuracy and precision. However, most previous catalogs of Solar twins contain only a small number of objects (typically a few tens), and their selection functions are poorly characterized. Aims. We aim at building a large catalog of Solar twins from Gaia DR3 GSP-Spec, providing model-driven, rather than data-driven, stellar parameters including ages, together with a well-characterized selection function. Methods. Using stellar parameters from the Gaia DR3 GSP-Spec catalog, we selected Solar-twin candidates whose parameters lie within +- 200 K in Teff, +- 0.2 in log g, and +- 0.1 dex in [M/H] of the Solar values. Candidates unlikely to be genuine Solar twins were removed using Gaia flags and photometric constraints. We determined accurate ages for individual twins with a Bayesian isochrone-projection method, considering three combinations of parameters: Teff, [M/H], and either log g, M_G, or M_Ks. We also constructed a mock catalog to characterize the selection function. Results. Our final GSP-Spec Solar-twin catalog contains 6,594 stars. The mock catalog consisting of 75,588 artificial twins well reproduces the main characteristics of the observed catalog, especially for ages determined with M_G or M_Ks. To demonstrate the usefulness of our catalog, we compared chemical abundances [X/Fe] with age. We statistically confirmed the age--[X/Fe] relations for several species (e.g., Al, Si, Ca, and Y), demonstrating that trends previously identified in small but very high-precision samples persist in a much larger, independent sample. Conclusions. Our study bridges small high-precision Solar-twin samples and large data-driven ones, enabling demographic studies of Solar twins.

Compact galaxy groups are ideal laboratories for studying the effects of interactions between AGN and multiple nearby galaxies. Recent JWST observations of the nearby compact group Stephan's Quintet highlight tidal flows between the interacting galaxies as well as outflows from the active galaxy NGC 7319. To study the kinematics on a large scale throughout the group, we obtained spatially-resolved long-slit spectra of Stephan's Quintet at multiple slit positions with Apache Point Observatory's Kitt Peak Ohio State Multi-Object Spectrograph. We fit multiple Gaussians to the H$α$ $λ$6563 Å and [N II] $λλ$6548, 6583 Å emission lines to isolate the different kinematic components. We used the kinematics to develop the first biconical outflow model of the narrow-line region of NGC 7319. Using a combination of galactic rotation models, biconical outflow models, and kinematic maps of the ionized gas, we disentangled the outflows, rotation, and tidal flows in the group. We found outflow radial velocities up to 550 km s$^{-1}$ peaking at 2.6 kpc from the central supermassive black hole, and a transition from AGN-powered outflows to gravitationally-powered tidal flows at a projected distance between 2.4 -- 6.3 kpc. We performed a line ratio analysis and determined the gas shows Seyfert-like ionization out to 6.3 kpc (projected), which supports our finding that gas outside this radius is predominantly powered by tidal flows. Our separation of kinematic components in Stephan's Quintet will enable future studies of the physical conditions and dynamical forces in the ionized gas to better quantify the feeding and feedback processes of AGN in compact groups.

We present serendipitous simultaneous radio and X-ray light curves of the dwarf Seyfert galaxy UGC 6728 spanning 5 years. The X-ray light curve exhibits a flaring period, followed by a gradual rise and decline. Throughout these events, the X-ray hardness ratio and spectrum do not change significantly. The radio flux is constant, as far as can be determined from its sparse sampling, until the end of the X-ray flare, then decreases by a factor of two by the midpoint of the gradual X-ray rise before returning to baseline at the end of the X-ray decline. We interpret this behavior in light of a similar event recently reported in NGC 2992, in which there is a temporary obscuration of the radio source by a blob of plasma ejected by a magnetic reconnection in the accretion disk. The energetics of the X-ray flare are consistent with those expected from magnetic disk activity. As in NGC 2992, the X-ray spectrum does not evolve during the obscuration event. We also discuss the possibility that the observed phenomena are due to normal AGN coronal flaring and variability, which is plausible but unlikely given the lack of spectral variation.

The nature of radio outflows in radio-quiet AGN remains poorly understood. In this study, we present kpc-scale polarization observations of the Seyfert galaxy NGC\,4151 using the Karl G. Jansky Very Large Array (VLA) in B-array at 3 and 10 GHz. We find that the inferred magnetic (B-) field structures show a stratification: a `spine-sheath'-like structure, with fields perpendicular to the jet direction in the `spine' and parallel in the `sheath', is observed in the higher resolution (0.5 arcsec) image at 10 GHz. In addition, a `wind'-like component with B-fields perpendicular to the radio outflow is observed in the 3 GHz image (resolution 2 arcsec); this feature is prominent along the `receding' (eastern) jet direction. Rotation measure (RM) ranges from $-230$ to 250 rad m$^{-2}$ over the polarized regions, indicating a low-electron-density ($10^{-2}-10^{-3}$ cm$^{-3}$) tenuous medium surrounding the source causing Faraday rotation. A {tentative} RM gradient of $+75$ to $-25$ rad m$^{-2}$ is observed transverse to the northern `wind' component, while a similar gradient with opposite sign is seen across the southern `wind' component, suggestive of a helical magnetic field threading the outflow. Based on an analysis of the available radio and X-ray data, we conclude that the stratified radio outflow in NGC 4151 is magnetically-driven. The bi-conical radio `wind' is found to be massive ($1050-3200 M_\odot$) with a high mass outflow rate ($0.01-0.03$ M$_\odot$ yr$^{-1}$) but low in kinetic power ($<0.01$% of L$_{\rm{bol}}$), making it less impactful for galactic-scale feedback. Our study suggests that radio-quiet AGN may also host magnetically dominant jets and winds, even while their jets are smaller and weaker compared to radio-loud AGN.

Recent work has shown that optical-radio position offsets and radio position variability are inversely correlated with the photometric variability of active galactic nuclei (AGN). A key prediction of these findings is that a reference frame constructed using highly photometrically variable AGN should be more stable than a frame that does not account for variability and that variability can be used to optimally weight all sources in order to maximize frame stability. Using ICRF3 matched to Gaia DR3, we employed a bootstrap method to estimate the multi-epoch stability of frames constructed using AGN selected at varying levels of photometric variability. We fit vector spherical harmonics to the coordinate differences between the three ICRF3 frames (S/X, K, and X/Ka) and Gaia and quantified the statistical dispersion as a function of blazar-like (high variability), quasar-like (low variability), and intermediate-variability class. An S/X reference frame constructed using blazars exceeds the stability of a frame constructed with quasars by a factor of 6 and is twice as stable as the ICRF3 defining sources. At K and X/Ka, a blazar-based frame matches or exceeds the stability of the defining sources by a factor of 1.4 in the case of X/Ka and exceeds the stability of a frame based on quasars by over a factor of 2 in both cases. The smaller improvement at K and X/Ka is likely because sources selected at higher frequency are more likely to be blazars. We derived a variability-based astrometric covariance scaling method that results in factor of 2 reduction in frame distortions and instabilities between S/X and Gaia, with a mild improvement for K but no difference for X/Ka, which is dominated by known distortions. Our results confirm the prediction that an optimal weighting of the link between the optical and radio celestial reference frames is enabled by accounting for photometric variability.

Simulations predict that supermassive black hole binaries (SMBHBs) will exhibit periodic brightness variations that may exceed the stochastic variability intrinsic to active galactic nuclei (AGN). In this paper, we simulate SMBHBs with damped random walk (DRW) AGN variability and an added sinusoidal signal from the orbital motion, and test three methods -- the Generalized Lomb Scargle Periodogram (GLSP), the nested Bayesian sampler (NBS), and the Weighted Wavelet Z-Transform (WWZ) -- to determine which is best at recovering the periodicity. Our simulated light curves follow the properties of the Catalina Real-Time Transient Survey (CRTS), Legacy Survey of Space and Time (LSST), and Zwicky Transient Facility (ZTF) to best inform current and future SMBHB searches. We map a broad range of parameter space and identify which DRW-only light curves best mimic periodicity and pass each method's model selection. The NBS performs best at detecting periodicity and filtering out DRW-only light curves. Combined candidate selection with both the NBS and GLSP significantly reduces false positive rates with marginal impact to true positive rates. With this joint model selection pipeline, we find the lowest false positive rates in ZTF-like simulations and the highest detection rates in LSST-like simulations. Using a modified computation of the False Alarm Probability (FAP) with GLSP, we efficiently triage LSST AGN light curves (~10^7 light curves in ~10-30 hours) and achieve true- and false- positive rates of ~40% and ~0.5%, respectively.

Cosmic voids have increasingly emerged as a powerful cosmological probe. However, their large spatial extent and intrinsically underdense environments make their identification highly sensitive to shot noise, redshift-space distortions (RSD), and observational systematics, particularly for topological and density-based void definitions. We introduce the Back-In-Time Void Finder (BitVF), a novel dynamical and physically motivated algorithm that identifies cosmic voids as regions of negative divergence of the Lagrangian displacement field reconstructed from the present-day tracer distribution. The reconstruction relies on an optimized discrete optimal transport algorithm that recovers the backward-in-time dynamics of tracers, naturally accounting for tracer bias without relying on cosmological assumptions. We validate BitVF against the widely used topological void finder REVOLVER using high-resolution N-body simulations, showing that it produces void catalogs with smoother and more physically motivated density profiles, as well as abundances that are more stable under tracer subsampling and shot noise. We further apply it to realistic DESI-like mock light-cone galaxy catalogs, demonstrating that it intrinsically mitigates redshift-space systematic effects, preserving real-space void size functions more faithfully than topological methods. Modeling RSD, the reconstruction can be combined with a fiducial cosmology and an assumed tracer bias within a bias-corrected Kaiser framework, yielding reconstructed-space void catalogs consistent with real-space statistics across redshift. Its performance is characterized as a function of the main internal parameters, showing an optimal balance between accuracy, computational efficiency, and applicability to stage IV galaxy surveys. BitVF will be publicly released within the CosmoBolognaLib.

Stellar streams, the debris of tidally disrupted satellites, trace their host's gravitational potential and thus probe dark matter halo structure. While six-dimensional phase-space data of Galactic streams enable precise dark matter halo modelling in the Milky Way, streams around external galaxies are typically available only as low surface brightness features without kinematics (i.e. two-dimensional photometric data), providing only weak constraints when considered individually. We present a hierarchical Bayesian framework that infers the population distribution of halo flattening using only projected stream tracks. Streams are forward-modelled in StreaMAX, a new JAX-accelerated particle-spray package that achieves orders of magnitude faster stream generation when compared to traditional methods. For each stream we fit an axisymmetric dark matter halo model and obtain a posterior on the flattening. These posteriors are then combined through hierarchical reweighting to constrain the population distribution. Using mock data, we show that individual fits recover the correct flattening with modest precision and exhibit projection-induced multi-modalities. Nevertheless, aggregating these fits yields accurate and confident constraints on the underlying population distribution of dark matter halo morphologies, clearly distinguishing between oblate, spherical, and prolate populations. The total computational cost scales linearly with sample size. Our results demonstrate that ensembles of purely photometric streams carry sufficient information to constrain dark matter halo shapes in external galaxies at the population level. With the forthcoming samples from Euclid and Rubin/LSST, this approach offers a practical path to population-level inferences of halo morphology without any kinematic measurements.

We study the kinematics of condensations in 30 fields forming high-mass stars with ALMA at a high-resolution of ~0.08'' on average (~230 au). The presence of disks is important for feeding high-mass stars without feedback halting growth as their masses increase. In the search for velocity gradients resembling rotation that can reveal the presence of disks, we analyze the emission of gas tracers in 49 objects using CH$_3$OH, CH$_3$CN, and tentative detections of HNCO and cis-HCOOH. Most of the velocity distributions show velocity gradients indicative of rotation. We reveal a total of 32 disk candidates, the largest sample to date that has been uniformly analyzed at a few hundred au scales in the high-mass regime. Their position-velocity maps are generally asymmetric with one side brighter than the opposite. We successfully fit a power law to the position-velocity maps of the disk candidates and find indices between -0.5 (Keplerian rotation) and -1 (rotation under specific angular momentum conservation) with a median of -0.7. Under Keplerian rotation assumption, we estimate central masses, uncorrected for inclination, ranging between 7 to 45 M$_\odot$. Excluding outliers, the disk candidates are relatively more compact (<200 au) and less massive (<5 M$_\odot$) than previous results at coarser angular resolution. We calculate an average Toomre-$Q$ parameter and find that most are gravitationally unstable (median of 0.5). We conclude that these observations offer the first opportunity to separate the disk and envelope components of hot cores on a statistically significant sample, and confirm that anisotropic collapse plays an role in feeding high-mass (proto)stars.

Radio observations offer a dust-unobscured view of galaxy star formation via the radio continuum-star formation rate (RC--SFR) relation. Emerging evidence of a stellar mass dependence in the RC--SFR relation raises the broader question of how other galaxy properties may influence this relation. In this work, we study the dependence of the global RC--SFR relation on galaxy properties in local ($z\,\leq$\,0.3) star-forming galaxies (SFGs) using the second data release of the LOFAR Two-Metre Sky Survey (LoTSS-DR2). Employing a non-parametric decision-tree regression algorithm, we identify the most important galaxy properties for estimating the radio luminosity using a sample of 18,828 emission-line-classified SFGs based on spectroscopic data from the SDSS-DR8. Along with the spectroscopically obtained SFRs and stellar mass values, we also use SFRs and stellar masses derived using photometric SED-fitting from the \textit{GALEX}--SDSS--\textit{WISE} Legacy Catalogue (GSWLC) for the same sample. We find that a galaxy's SFR is most important for predicting the radio luminosity, followed by the stellar mass, at $>5σ$ significance. Complementing the LoTSS catalogue 150\,MHz flux densities with aperture photometry for the rest of the emission-line classified sample (35,099 galaxies in total), we obtain a new calibration of the RC--SFR relation, which does not change significantly whether we use spectroscopic or photometrically derived SFRs and stellar masses, despite the fact that the methods probe star formation on different characteristic timescales. Our results highlight the utility of decision-tree algorithms for handling censored radio-selected galaxy samples, which will be useful for future spectroscopic surveys of radio sources.

The study of heavily obscured supermassive black hole (SMBH) growth in late-stage galaxy mergers is challenging: column densities $N_{\mathrm{H}}>10^{24},\mathrm{cm}^{-2}$ can block most nuclear emission, leaving significant gaps in the SMBH growth census. Millimeter-wave continuum emission offers a potential window into this obscured phase, as it can trace Active Galactic Nuclei (AGN) activity through mechanisms less affected by dust extinction. In this work, we test whether the observed correlation between millimeter ($\sim200,\mathrm{GHz}$) and hard X-ray (14 - 150,keV) luminosities can be used to plausibly identify hidden AGN in local (Ultra)Luminous Infrared Galaxies (U)LIRGs, including systems hosting confirmed dual AGN. We identify three sources -- one confirmed AGN and two strong candidates -- presenting significant evidence of AGN activity. The confirmed dual AGN lie within $\sim3σ$ of the mm--X-ray correlation, suggesting this relation can be used to identify hidden pairs. By combining the position of each source relative to this correlation with independent star formation rate constraints, we propose a method to disentangle AGN and star formation contributions for sources with measured column densities. While our analysis is based on a small, heterogeneous local sample and relies on empirical scaling relations, these results indicate that millimeter continuum emission may provide a useful complementary diagnostic for obscured SMBH growth. ALMA observations at high angular resolutions are particularly valuable for this approach, while future facilities such as the ngVLA will be essential to test its robustness in larger and more distant samples.

G284.3$-$1.8 is a supernova remnant with a radio shell and thermal X-ray emission. Located near its center is the gamma-ray binary 1FGL J1018.6$-$5856, although the physical association between the two systems is not clear yet. Our X-ray spectroscopy with Suzaku reveals that G284.3$-$1.8 and 1FGL J1018.6$-$5856 have compatible absorption column densities of $N_\mathrm{H} = 6\textrm{--}7 \times 10^{21}~\mathrm{cm}^{-2}$, indicating that the two systems have similar distances. The actual distance is determined as $3~\mathrm{kpc}$ using $\mathrm{^{12}CO}$ ($J=1\textrm{--}0$) data obtained with NANTEN. The X-ray spectrum of G284.3$-$1.8 shows a strong K-shell emission line of Mg, confirming that the earlier claim that the SNR is one of the few Mg-rich SNRs. Comparing recent stellar models taking into account the "shell merger" processes, we find that the obtained Mg-to-Ne mass ratio of $M_\mathrm{Mg}/M_\mathrm{Ne} = 0.73^{+0.07}_{-0.03}$ and Si-to-Mg mass ratio of $M_\mathrm{Si}/M_\mathrm{Mg} = 0.44\pm0.03$ suggest a supernova explosion that would have left a neutron star. The characteristics of 1FGL J1018.6$-$5856, on the other hand, are better explained with a model in which its compact object is neutron star. The present results, therefore, would suggest a possible scenario where G284.3$-$1.8 and 1FGL J1018.6$-$5856 are both remnants of a common supernova explosion although further observational tests are necessary.

Fast Radio Bursts (FRBs) are millisecond-duration radio transients of extragalactic origin. Classifying repeating FRBs is essential for understanding their emission mechanisms, but remains challenging due to their short durations, high variability, and increasing data volume. Traditional methods often rely on subjective criteria and struggle with high-dimensional data. In this study, we apply an unsupervised machine learning framework that combines Uniform Manifold Approximation and Projection (UMAP) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) to eight observed parameters from FRB 20220912A. Our analysis reveals three distinct clusters of bursts with varying spectral and fluence properties. Comparisons with clustering studies on other repeaters show that some of our clusters share similar features with sources such as FRB 20201124A and FRB 121102, suggesting possible common emission mechanisms. We also provide qualitative interpretations for each cluster, highlighting the spectral diversity within a single source. Notably, one cluster shows broadband emission and high fluence, which are typically seen in non-repeating FRBs. This raises the possibility that some non-repeaters may be misclassified repeaters due to observational limitations. Our results demonstrate the utility of machine learning in uncovering intrinsic diversity in FRB emission and provide a foundation for future classification studies.

China's Tianwen-2 mission, launched on 29 May 2025, targets the near-Earth object (469219) Kamo`oalewa, an Earth quasi-satellite trapped in a 1:1 mean-motion resonance with our planet. Determining the origin of Kamo`oalewa is central to understanding the formation pathways and dynamical evolution of Earth's quasi-satellite population. Here we show a strong possibility of main-belt origin for Kamo`oalewa using long-term dynamical simulations. We examine three candidate source regions: the $ν_6$ secular resonance ($ν_6$), the 3:1 mean-motion resonance with Jupiter (3:1J MMR), and the Flora family. A total of 42,825 test particles were integrated over 100 Myr. We find that asteroids from all three regions can be transported onto Kamo`oalewa-like orbits, albeit with markedly different efficiencies. Particles originating near the $ν_6$ show the highest transfer probability (3.31%), followed by the Flora family (2.54%) and the 3:1J MMR (0.39%). We further identify representative dynamical pathways linking these source regions to Earth quasi-satellite orbits. The Tianwen-2 spacecraft is expected to rendezvous with Kamo`oalewa in 2026, performing close-proximity operations and returning samples. The mission will provide decisive observational constraints on the asteroid's composition and physical properties, offering a critical test of its proposed origin.