arXiv Daily Digest - 2026-03-12

We propose a fully automated AI system that produces short comedic videos similar to sketch shows such as Saturday Night Live. Starting with character references, the system employs a population of agents loosely based on real production studio roles, structured to optimize the quality and diversity of ideas and outputs through iterative competition, evaluation, and improvement. A key contribution is the introduction of LLM critics aligned with real viewer preferences through the analysis of a corpus of comedy videos on YouTube to automatically evaluate humor. Our experiments show that our framework produces results approaching the quality of professionally produced sketches while demonstrating state-of-the-art performance in video generation.

We propose a 3D latent representation that jointly models object geometry and view-dependent appearance. Most prior works focus on either reconstructing 3D geometry or predicting view-independent diffuse appearance, and thus struggle to capture realistic view-dependent effects. Our approach leverages that RGB-depth images provide samples of a surface light field. By encoding random subsamples of this surface light field into a compact set of latent vectors, our model learns to represent both geometry and appearance within a unified 3D latent space. This representation reproduces view-dependent effects such as specular highlights and Fresnel reflections under complex lighting. We further train a latent flow matching model on this representation to learn its distribution conditioned on a single input image, enabling the generation of 3D objects with appearances consistent with the lighting and materials in the input. Experiments show that our approach achieves higher visual quality and better input fidelity than existing methods.

We propose Neural Field Thermal Tomography (NeFTY), a differentiable physics framework for the quantitative 3D reconstruction of material properties from transient surface temperature measurements. While traditional thermography relies on pixel-wise 1D approximations that neglect lateral diffusion, and soft-constrained Physics-Informed Neural Networks (PINNs) often fail in transient diffusion scenarios due to gradient stiffness, NeFTY parameterizes the 3D diffusivity field as a continuous neural field optimized through a rigorous numerical solver. By leveraging a differentiable physics solver, our approach enforces thermodynamic laws as hard constraints while maintaining the memory efficiency required for high-resolution 3D tomography. Our discretize-then-optimize paradigm effectively mitigates the spectral bias and ill-posedness inherent in inverse heat conduction, enabling the recovery of subsurface defects at arbitrary scales. Experimental validation on synthetic data demonstrates that NeFTY significantly improves the accuracy of subsurface defect localization over baselines. Additional details at https://cab-lab-princeton.github.io/nefty/

Generating music that temporally aligns with video events is challenging for existing text-to-music models, which lack fine-grained temporal control. We introduce V2M-Zero, a zero-pair video-to-music generation approach that outputs time-aligned music for video. Our method is motivated by a key observation: temporal synchronization requires matching when and how much change occurs, not what changes. While musical and visual events differ semantically, they exhibit shared temporal structure that can be captured independently within each modality. We capture this structure through event curves computed from intra-modal similarity using pretrained music and video encoders. By measuring temporal change within each modality independently, these curves provide comparable representations across modalities. This enables a simple training strategy: fine-tune a text-to-music model on music-event curves, then substitute video-event curves at inference without cross-modal training or paired data. Across OES-Pub, MovieGenBench-Music, and AIST++, V2M-Zero achieves substantial gains over paired-data baselines: 5-21% higher audio quality, 13-15% better semantic alignment, 21-52% improved temporal synchronization, and 28% higher beat alignment on dance videos. We find similar results via a large crowd-source subjective listening test. Overall, our results validate that temporal alignment through within-modality features, rather than paired cross-modal supervision, is effective for video-to-music generation. Results are available at https://genjib.github.io/v2m_zero/

We present IsalGraph, a method for representing the structure of any finite, simple graph as a compact string over a nine-character instruction alphabet. The encoding is executed by a small virtual machine comprising a sparse graph, a circular doubly-linked list (CDLL) of graph-node references, and two traversal pointers. Instructions either move a pointer through the CDLL or insert a node or edge into the graph. A key design property is that every string over the alphabet decodes to a valid graph, with no invalid states reachable. A greedy \emph{GraphToString} algorithm encodes any connected graph into a string in time polynomial in the number of nodes; an exhaustive-backtracking variant produces a canonical string by selecting the lexicographically smallest shortest string across all starting nodes and all valid traversal orders. We evaluate the representation on five real-world graph benchmark datasets (IAM Letter LOW/MED/HIGH, LINUX, and AIDS) and show that the Levenshtein distance between IsalGraph strings correlates strongly with graph edit distance (GED). Together, these properties make IsalGraph strings a compact, isomorphism-invariant, and language-model-compatible sequential encoding of graph structure, with direct applications in graph similarity search, graph generation, and graph-conditioned language modelling

The paradigm of LLM-as-a-judge relies on a critical assumption, namely that high inter-evaluator agreement indicates reliable and objective evaluation. We present two complementary findings that challenge this assumption. \textbf{First}, we demonstrate that this consensus is frequently illusory. We identify and formalize \textbf{Evaluation Illusion}, a phenomenon where LLM judges generate sophisticated critiques yet anchor scores on shared surface heuristics rather than substantive quality. Through a large-scale study of 105,600 evaluation instances (32 LLMs $\times$ 3 frontier judges $\times$ 100 tasks $\times$ 11 temperatures), we show that model-level agreement (Spearman $ρ= 0.99$) masks fragile sample-level agreement (Pearson $\bar{r} = 0.72$; absolute agreement ICC $= 0.67$), that merely sharing rubric structure restores 62\% of total agreement, and that high-quality outputs paradoxically receive the \textit{least} consistent evaluations. \textbf{Second}, we demonstrate that dynamically generating evaluation rubrics grounded in domain knowledge produces more meaningful assessment. We introduce MERG (Metacognitive Enhanced Rubric Generation), a knowledge-driven rubric generation framework whose domain-selective effects confirm this. Agreement \textit{increases} in codified domains (Education +22\%, Academic +27\%) where knowledge anchors evaluators on shared standards, while it decreases in subjective domains where genuine evaluative pluralism emerges. These findings suggest that evaluation rubrics should be dynamically enriched with expert knowledge rather than relying on generic criteria, with implications for reward modeling in RLAIF.

Rising environmental awareness in e-commerce necessitates recommender systems that not only guide users to sustainable products but also minimize their own digital carbon footprints. Traditional session-based systems, optimized for short-term conversions, often fail to capture nuanced user intents for eco-friendly choices, perpetuating a gap between green intentions and actions. To tackle this, we introduce LLMGreenRec, a novel multi-agent framework that leverages Large Language Models (LLMs) to promote sustainable consumption. Through collaborative analysis of user interactions and iterative prompt refinement, LLMGreenRec's specialized agents deduce green-oriented user intents and prioritize eco-friendly product recommendations. Notably, this intent-driven approach also reduces unnecessary interactions and energy consumption. Extensive experiments on benchmark datasets validate LLMGreenRec's effectiveness in recommending sustainable products, demonstrating a robust solution that fosters a responsible digital economy.

VLMs have become increasingly proficient at a range of computer vision tasks, such as visual question answering and object detection. This includes increasingly strong capabilities in the domain of art, from analyzing artwork to generation of art. In an interdisciplinary collaboration between computer scientists and art historians, we characterize the mechanisms underlying VLMs' ability to predict artistic style and assess the extent to which they align with the criteria art historians use to reason about artistic style. We employ a latent-space decomposition approach to identify concepts that drive art style prediction and conduct quantitative evaluations, causal analysis and assessment by art historians. Our findings indicate that 73% of the extracted concepts are judged by art historians to exhibit a coherent and semantically meaningful visual feature and 90% of concepts used to predict style of a given artwork were judged relevant. In cases where an irrelevant concept was used to successfully predict style, art historians identified possible reasons for its success; for example, the model might "understand" a concept in more formal terms, such as dark/light contrasts.

Scalar quantization of large language models (LLMs) is fundamentally limited by information-theoretic bounds. While vector quantization (VQ) overcomes these limits by encoding blocks of parameters jointly, practical implementations must avoid the need for expensive lookup mechanisms or other explicit codebook storage. Lattice approaches address this through highly structured and dense packing. This paper explores the Leech lattice, which, with its optimal sphere packing and kissing configurations at 24 dimensions, is the highest dimensional lattice known with such optimal properties. To make the Leech lattice usable for LLM quantization, we extend an existing search algorithm based on the extended Golay code construction, to i) support indexing, enabling conversion to and from bitstrings without materializing the codebook, ii) allow angular search over union of Leech lattice shells, iii) propose fully-parallelisable dequantization kernel. Together this yields a practical algorithm, namely Leech Lattice Vector Quantization (LLVQ). LLVQ delivers state-of-the-art LLM quantization performance, outperforming recent methods such as Quip\#, QTIP, and PVQ. These results highlight the importance of high-dimensional lattices for scalable, theoretically grounded model compression.

Pseudo-relevance feedback (PRF) methods built on large language models (LLMs) can be organized along two key design dimensions: the feedback source, which is where the feedback text is derived from and the feedback model, which is how the given feedback text is used to refine the query representation. However, the independent role that each dimension plays is unclear, as both are often entangled in empirical evaluations. In this paper, we address this gap by systematically studying how the choice of feedback source and feedback model impact PRF effectiveness through controlled experimentation. Across 13 low-resource BEIR tasks with five LLM PRF methods, our results show: (1) the choice of feedback model can play a critical role in PRF effectiveness; (2) feedback derived solely from LLM-generated text provides the most cost-effective solution; and (3) feedback derived from the corpus is most beneficial when utilizing candidate documents from a strong first-stage retriever. Together, our findings provide a better understanding of which elements in the PRF design space are most important.

Human uplift studies - or studies that measure AI effects on human performance relative to a status quo, typically using randomized controlled trial (RCT) methodology - are increasingly used to inform deployment, governance, and safety decisions for frontier AI systems. While the methods underlying these studies are well-established, their interaction with the distinctive properties of frontier AI systems remains underexamined, particularly when results are used to inform high-stakes decisions. We present findings from interviews with 16 expert practitioners with experience conducting human uplift studies in domains including biosecurity, cybersecurity, education, and labor. Across interviews, experts described a recurring tension between standard causal inference assumptions and the object of study itself. Rapidly evolving AI systems, shifting baselines, heterogeneous and changing user proficiency, and porous real-world settings strain assumptions underlying internal, external, and construct validity, complicating the interpretation and appropriate use of uplift evidence. We synthesize these challenges across key stages of the human uplift research lifecycle and map them to practitioner-reported solutions, clarifying both the limits and the appropriate uses of evidence from human uplift studies in high-stakes decision-making.

Single-cell electrophysiological recordings provide a powerful window into neuronal functional diversity and offer an interpretable route for linking intrinsic physiology to transcriptomic identity. Here, we replicate and extend the electrophysiology-to-transcriptomics framework introduced by Gouwens et al. (2020) using publicly available Allen Institute Patch-seq datasets from both mouse and human cortex. We focus on GABAergic inhibitory interneurons to target a subclass structure (Lamp5, Pvalb, Sst, Vip) that is comparable and conserved across species. After quality control, we analyzed 3,699 mouse visual cortex neurons and 506 human neocortical neurons from neurosurgical resections. Using standardized electrophysiological features and sparse PCA, we reproduced the major class-level separations reported in the original mouse study. For supervised prediction, a class-balanced random forest provided a strong feature-engineered baseline in mouse data and a reduced but still informative baseline in human data. We then developed an attention-based BiLSTM that operates directly on the structured IPFX feature-family representation, avoiding sPCA and providing feature-family-level interpretability via learned attention weights. Finally, we evaluated a cross-species transfer setting in which the sequence model is pretrained on mouse data and fine-tuned on human data for an aligned 4-class task, improving human macro-F1 relative to a human-only training baseline. Together, these results confirm reproducibility of the Gouwens pipeline in mouse data, demonstrate that sequence models can match feature-engineered baselines, and show that mouse-to-human transfer learning can provide measurable gains for human subclass prediction.

A data-driven, model-free approach to modeling the temporal evolution of physical systems mitigates the need for explicit knowledge of the governing equations. Even when physical priors such as partial differential equations are available, such systems often reside in high-dimensional state spaces and exhibit nonlinear dynamics, making traditional numerical solvers computationally expensive and ill-suited for real-time analysis and control. Consider the problem of learning a parametric flow of a dynamical system: with an initial field and a set of physical parameters, we aim to predict the system's evolution over time in a way that supports long-horizon rollouts, generalization to unseen parameters, and spectral analysis. We propose a physics-coded neural field parameterization of the Koopman operator's spectral decomposition. Unlike a physics-constrained neural field, which fits a single solution surface, and neural operators, which directly approximate the solution operator at fixed time horizons, our model learns a factorized flow operator that decouples spatial modes and temporal evolution. This structure exposes underlying eigenvalues, modes, and stability of the underlying physical process to enable stable long-term rollouts, interpolation across parameter spaces, and spectral analysis. We demonstrate the efficacy of our method on a range of dynamics problems, showcasing its ability to accurately predict complex spatiotemporal phenomena while providing insights into the system's dynamic behavior.

The rapid evolution and inherent complexity of modern software requirements demand highly flexible and responsive development methodologies. While Agile frameworks have become the industry standard for prioritizing iteration, collaboration, and adaptability, software development teams continue to face persistent challenges in managing constantly evolving requirements and maintaining product quality under tight deadlines. This article explores the intersection of Artificial Intelligence (AI) and Software Engineering (SE), to analyze how AI serves as a powerful catalyst for enhancing agility and fostering innovation. The research combines a comprehensive review of existing literature with an empirical study, utilizing a survey directed at Software Engineering professionals to assess the perception, adoption, and impact of AI-driven tools. Key findings reveal that the integration of AI (specifically through Machine Learning (ML) and Natural Language Processing (NLP) )facilitates the automation of tedious tasks, from requirement management to code generation and testing . This paper demonstrates that AI not only optimizes current Agile practices but also introduces new capabilities essential for sustaining quality, speed, and innovation in the future landscape of software development.

Accelerating the explorations of stationary points on potential energy surfaces building local surrogates spans decades of effort. Done correctly, surrogates reduce required evaluations by an order of magnitude while preserving the accuracy of the underlying theory. We present a unified Bayesian Optimization view of minimization, single point saddle searches, and double ended saddle searches through a unified six-step surrogate loop, differing only in the inner optimization target and acquisition criterion. The framework uses Gaussian process regression with derivative observations, inverse-distance kernels, and active learning. The Optimal Transport GP extensions of farthest point sampling with Earth mover's distance, MAP regularization via variance barrier and oscillation detection, and adaptive trust radius form concrete extensions of the same basic methodology, improving accuracy and efficiency. We also demonstrate random Fourier features decouple hyperparameter training from predictions enabling favorable scaling for high-dimensional systems. Accompanying pedagogical Rust code demonstrates that all applications use the exact same Bayesian optimization loop, bridging the gap between theoretical formulation and practical execution.

Deep learning models are being used for the analysis of parametric statistical models based on simulation-only frameworks. Bayesian models using normalizing flows simulate data from a prior distribution and are composed of two deep neural networks: a summary network that learns a sufficient statistic for the parameter and a normalizing flow that conditional on the summary network can approximate the posterior distribution. Here, we explore frequentist models that are based on a single summary network. During training, input of the network is a simulated data set based on a parameter and the loss function minimizes the mean-square error between learned summary and parameter. The network thereby solves the inverse problem of parameter estimation. We propose a branched network structure that contains collapsing layers that reduce a data set to summary statistics that are further mapped through fully connected layers to approximate the parameter estimate. We motivate our choice of network structure by theoretical considerations. In simulations we demonstrate three desirable properties of parameter estimates: finite sample exactness, robustness to data contamination, and algorithm approximation. These properties are achieved offering the the network varying sample size, contaminated data, and data needing algorithmic reconstruction during the training phase. In our simulations an EM-algorithm for genetic data is automatically approximated by the network. Simulation only approaches seem to offer practical advantages in complex modeling tasks where the simpler data simulation part is left to the researcher and the more complex problem of solving the inverse problem is left to the neural network. Challenging future work includes offering pre-trained models that can be used in a wide variety of applications.

Neural networks are a commonly used approach to replace physical models with computationally cheap surrogates. Parametric uncertainty quantification can be included in training, assuming that an accurate prior distribution of the model parameters is available. Here we study the common opposite situation, where direct screening or random sampling of model parameters leads to exhaustive training times and evaluations at unphysical parameter values. Our solution is to decouple uncertainty quantification from network architecture. Instead of sampling network weights, we introduce the model-parameter distribution as an input to network training via Markov chain Monte Carlo (MCMC). In this way, the surrogate achieves the same uncertainty quantification as the underlying physical model, but with substantially reduced computation time. The approach is fully agnostic with respect to the neural network choice. In our examples, we present a quantile emulator for prediction and a novel autoencoder-based ODE network emulator that can flexibly estimate different trajectory paths corresponding to different ODE model parameters. Moreover, we present a mathematical analysis that provides a transparent way to relate potential performance loss to measurable distribution mismatch.

We show that MLP layers in transformer language models perform binary routing of continuous signals: the decision of whether a token needs nonlinear processing is well-captured by binary neuron activations, even though the signals being routed are continuous. In GPT-2 Small (124M parameters), we find that specific neurons implement a consensus architecture -- seven "default-ON" neurons and one exception handler (N2123 in Layer 11) that are 93-98% mutually exclusive -- creating a binary routing switch. A cross-layer analysis reveals a developmental arc: early layers (L1-3) use single gateway neurons to route exceptions without consensus quorums; middle layers (L4-6) show diffuse processing with neither gateway nor consensus; and late layers (L7-11) crystallize full consensus/exception architectures with increasing quorum size (1 to 3 to 7 consensus neurons). Causal validation confirms the routing is functional: removing the MLP at consensus breakdown costs 43.3% perplexity, while at full consensus removing it costs only 10.1% -- exceeding a 4x difference. Comparing binary vs. continuous features for the routing decision confirms that binarization loses essentially no information (79.2% vs. 78.8% accuracy), while continuous activations carry additional magnitude information (R^2 = 0.36 vs. 0.22). This binary routing structure explains why smooth polynomial approximation fails: cross-validated polynomial fits (degrees 2-7) never exceed R^2 = 0.06 for highly nonlinear layers. We propose that the well-established piecewise-affine characterization of deep networks can be complemented by a routing characterization: along the natural data manifold, the piecewise boundaries implement binary decisions about which tokens need nonlinear processing, routing continuous signals through qualitatively different computational paths.

Low Earth Orbit (LEO) Non-Terrestrial Networks (NTNs) require efficient beam management under dynamic propagation conditions. This work investigates Federated Learning (FL)-based beam selection in LEO satellite constellations, where orbital planes operate as distributed learners through the utilization of High-Altitude Platform Stations (HAPS). Two models, a Multi-Layer Perceptron (MLP) and a Graph Neural Network (GNN), are evaluated using realistic channel and beamforming data. Results demonstrate that GNN surpasses MLP in beam prediction accuracy and stability, particularly at low elevation angles, enabling lightweight and intelligent beam management for future NTN deployments.

Vision Language Models (VLMs) exhibit persistent hallucinations in counting tasks, with accuracy substantially lower than other visual reasoning tasks (excluding sentiment). This phenomenon persists even in state-of-the-art reasoning-capable VLMs. Conversely, CNN-based object detection models (ODMs) such as YOLO excel at spatial localization and instance counting with minimal computational overhead. We propose GroundCount, a framework that augments VLMs with explicit spatial grounding from ODMs to mitigate counting hallucinations. In the best case, our prompt-based augmentation strategy achieves 81.3% counting accuracy on the best-performing model (Ovis2.5-2B) - a 6.6pp improvement - while reducing inference time by 22% through elimination of hallucination-driven reasoning loops for stronger models. We conduct comprehensive ablation studies demonstrating that positional encoding is a critical component, being beneficial for stronger models but detrimental for weaker ones. Confidence scores, by contrast, introduce noise for most architectures and their removal improves performance in four of five evaluated models. We further evaluate feature-level fusion architectures, finding that explicit symbolic grounding via structured prompts outperforms implicit feature fusion despite sophisticated cross-attention mechanisms. Our approach yields consistent improvements across four of five evaluated VLM architectures (6.2--7.5pp), with one architecture exhibiting degraded performance due to incompatibility between its iterative reflection mechanisms and structured prompts. These results suggest that counting failures stem from fundamental spatial-semantic integration limitations rather than architecture-specific deficiencies, while highlighting the importance of architectural compatibility in augmentation strategies.

As wireless systems evolve toward Beyond 5G (B5G), the adoption of cell-free (CF) millimeter-wave (mmWave) architectures combined with Reconfigurable Intelligent Surfaces (RIS) is emerging as a key enabler for ultra-reliable, high-capacity, scalable, and secure Industrial Internet of Things (IIoT) communications. However, safeguarding these complex and distributed environments against eavesdropping remains a critical challenge, particularly when conventional security mechanisms struggle to overcome scalability, and latency constraints. In this paper, a novel framework for detecting malicious users in RIS-enhanced cell-free mmWave networks using Federated Learning (FL) is presented. The envisioned setup features multiple access points (APs) operating without traditional cell boundaries, assisted by RIS nodes to dynamically shape the wireless propagation environment. Edge devices collaboratively train a Deep Convolutional Neural Network (DCNN) on locally observed Channel State Information (CSI), eliminating the need for raw data exchange. Moreover, an early-exit mechanism is incorporated in that model to jointly satisfy computational complexity requirements. Performance evaluation indicates that the integration of FL and multi-RIS coordination improves approximately 30% the achieved secrecy rate (SR) compared to baseline non-RIS-assisted methods while maintaining near-optimal detection accuracy levels. This work establishes a distributed, privacy-preserving approach to physical layer eavesdropping detection tailored for next-generation IIoT deployments.

This report summarizes the discussions and recommendations from the NSF Workshop on Algorithm-Hardware Co-design for Medical Applications, held on September 26-27, 2024, in Pittsburgh, PA. The workshop assembled an interdisciplinary cohort of researchers, clinicians, and industry leaders to examine foundational challenges and develop a strategic roadmap for algorithm-hardware co-design in medical computing. The workshop focuses on four thematic areas: (1) teleoperations, telehealth, and surgical operations; (2) wearable and implantable medicine, including implantable living pharmacies; (3) home ICU, hospital systems, and elderly care; and (4) medical sensing, imaging, and reconstruction. This report calls for a fundamental shift in how next-generation medical technologies are conceived, designed, validated, and translated into practice. The report recommends that NSF sustain investment in shared standardized data infrastructures and compute infrastructures, develop clinic workflow-aware systems and human-AI collaboration frameworks, promote scalable validation ecosystems grounded in objective, continuous measures, and physics-informed, and enable safe, accountable, and resilient platforms, including virtual-physical healthcare ecosystems, to de-risk translational pathways. The workshop information can be found on the website: https://sites.google.com/view/nsfworkshop.

Deep Reinforcement learning (DRL) has achieved remarkable success in domains with well-defined reward structures, such as Atari games and locomotion. In contrast, dexterous manipulation lacks general-purpose reward formulations and typically depends on task-specific, handcrafted priors to guide hand-object interactions. We propose Contact Coverage-Guided Exploration (CCGE), a general exploration method designed for general-purpose dexterous manipulation tasks. CCGE represents contact state as the intersection between object surface points and predefined hand keypoints, encouraging dexterous hands to discover diverse and novel contact patterns, namely which fingers contact which object regions. It maintains a contact counter conditioned on discretized object states obtained via learned hash codes, capturing how frequently each finger interacts with different object regions. This counter is leveraged in two complementary ways: (1) to assign a count-based contact coverage reward that promotes exploration of novel contact patterns, and (2) an energy-based reaching reward that guides the agent toward under-explored contact regions. We evaluate CCGE on a diverse set of dexterous manipulation tasks, including cluttered object singulation, constrained object retrieval, in-hand reorientation, and bimanual manipulation. Experimental results show that CCGE substantially improves training efficiency and success rates over existing exploration methods, and that the contact patterns learned with CCGE transfer robustly to real-world robotic systems. Project page is https://contact-coverage-guided-exploration.github.io.

Quantum computers have demonstrated utility in simulating quantum systems beyond brute-force classical approaches. As the community builds on these demonstrations to explore using quantum computing for applied research, algorithms and workflows have emerged that require leveraging both quantum computers and classical high-performance computing (HPC) systems to scale applications, especially in chemistry and materials, beyond what either system can simulate alone. Today, these disparate systems operate in isolation, forcing users to manually orchestrate workloads, coordinate job scheduling, and transfer data between systems -- a cumbersome process that hinders productivity and severely limits rapid algorithmic exploration. These challenges motivate the need for flexible and high-performance Quantum-Centric Supercomputing (QCSC) systems that integrate Quantum Processing Units (QPUs), Graphics Processing Units (GPUs), and Central Processing Units (CPUs) to accelerate discovery of such algorithms across applications. These systems will be co-designed across quantum and classical HPC infrastructure, middleware, and application layers to accelerate the adoption of quantum computing for solving critical computational problems. We envision QCSC evolution through three distinct phases: (1) quantum systems as specialized compute offload engines within existing HPC complexes; (2) heterogeneous quantum and classical HPC systems coupled through advanced middleware, enabling seamless execution of hybrid quantum-classical algorithms; and (3) fully co-designed heterogeneous quantum-HPC systems for hybrid computational workflows. This article presents a reference architecture and roadmap for these QCSC systems.

With their increasing capabilities, Large Language Models (LLMs) are now used across many industries. They have become useful tools for software engineers and support a wide range of development tasks. As LLMs are increasingly used in software development workflows, a critical question arises: are LLMs good at software security? At the same time, organizations worldwide invest heavily in cybersecurity to reduce exposure to disruptive attacks. The integration of LLMs into software engineering workflows may introduce new vulnerabilities and weaken existing security efforts. We introduce TOSSS (Two-Option Secure Snippet Selection), a benchmark that measures the ability of LLMs to choose between secure and vulnerable code snippets. Existing security benchmarks for LLMs cover only a limited range of vulnerabilities. In contrast, TOSSS relies on the CVE database and provides an extensible framework that can integrate newly disclosed vulnerabilities over time. Our benchmark gives each model a security score between 0 and 1 based on its behavior; a score of 1 indicates that the model always selects the secure snippet, while a score of 0 indicates that it always selects the vulnerable one. We evaluate 14 widely used open-source and closed-source models on C/C++ and Java code and observe scores ranging from 0.48 to 0.89. LLM providers already publish many benchmark scores for their models, and TOSSS could become a complementary security-focused score to include in these reports.

Foundation models for point cloud data have recently grown in capability, often leveraging extensive representation learning from language or vision. In this work, we take a more controlled approach by introducing a lightweight transformer-based point cloud architecture. In contrast to the heavy reliance on cross-modal supervision, our model is trained only on 39k point clouds - yet it outperforms several larger foundation models trained on over 200k training samples. Interestingly, our method approaches state-of-the-art results from models that have seen over a million point clouds, images, and text samples, demonstrating the value of a carefully curated training setup and architecture. To ensure rigorous evaluation, we conduct a comprehensive replication study that standardizes the training regime and benchmarks across multiple point cloud architectures. This unified experimental framework isolates the impact of architectural choices, allowing for transparent comparisons and highlighting the benefits of our design and other tokenizer-free architectures. Our results show that simple backbones can deliver competitive results to more complex or data-rich strategies. The implementation, including code, pre-trained models, and training protocols, is available at https://github.com/KonradSzafer/Pointy.

Wearable accelerometers have enabled large-scale health and wellness monitoring, yet learning robust human-activity representations has been constrained by the scarcity of labeled data. While self-supervised learning offers a potential remedy, existing approaches treat sensor streams as unstructured time series, overlooking the underlying biological structure of human movement, a factor we argue is critical for effective Human Activity Recognition (HAR). We introduce a novel tokenization strategy grounded in the submovement theory of motor control, which posits that continuous wrist motion is composed of superposed elementary basis functions called submovements. We define our token as the movement segment, a unit of motion composed of a finite sequence of submovements that is readily extractable from wrist accelerometer signals. By treating these segments as tokens, we pretrain a Transformer encoder via masked movement-segment reconstruction to model the temporal dependencies of movement segments, shifting the learning focus beyond local waveform morphology. Pretrained on the NHANES corpus (approximately 28k hours; approximately 11k participants; approximately 10M windows), our representations outperform strong wearable SSL baselines across six subject-disjoint HAR benchmarks. Furthermore, they demonstrate stronger data efficiency in data-scarce settings. Code and pretrained weights will be made publicly available.

Test-time scaling evaluates reasoning LLMs by sampling multiple outputs per prompt, but ranking models in this regime remains underexplored. We formalize dense benchmark ranking under test-time scaling and introduce Scorio, a library that implements statistical ranking methods such as paired-comparison models, item response theory (IRT) models, voting rules, and graph- and spectral-based methods. Across $20$ reasoning models on four Olympiad-style math benchmarks (AIME'24, AIME'25, HMMT'25, and BrUMO'25; up to $N=80$ trials), most full-trial rankings agree closely with the Bayesian gold standard $\mathrm{Bayes}_{\mathcal{U}}@80$ (mean Kendall's $τ_b = 0.93$--$0.95$), and $19$--$34$ methods recover exactly the same ordering. In the single-trial regime, the best methods reach $τ_b \approx 0.86$. Using greedy decoding as an empirical prior ($\mathrm{Bayes}_{\mathbf{R}_0}@N$) reduces variance at $N=1$ by $16$--$52\%$, but can bias rankings when greedy and stochastic sampling disagree. These results identify reliable ranking methods for both high- and low-budget test-time scaling. We release Scorio as an open-source library at https://github.com/mohsenhariri/scorio.

Machine learning methods for identifying molecular structures from tandem mass spectra (MS/MS) have advanced rapidly, yet current approaches still exhibit significant error rates. In high-stakes applications such as clinical metabolomics and environmental screening, incorrect annotations can have serious consequences, making it essential to determine when a prediction can be trusted. We introduce a selective prediction framework for molecular structure retrieval from MS/MS spectra, enabling models to abstain from predictions when uncertainty is too high. We formulate the problem within the risk-coverage tradeoff framework and comprehensively evaluate uncertainty quantification strategies at two levels of granularity: fingerprint-level uncertainty over predicted molecular fingerprint bits, and retrieval-level uncertainty over candidate rankings. We compare scoring functions including first-order confidence measures, aleatoric and epistemic uncertainty estimates from second-order distributions, as well as distance-based measures in the latent space. All experiments are conducted on the MassSpecGym benchmark. Our analysis reveals that while fingerprint-level uncertainty scores are poor proxies for retrieval success, computationally inexpensive first-order confidence measures and retrieval-level aleatoric uncertainty achieve strong risk-coverage tradeoffs across evaluation settings. We demonstrate that by applying distribution-free risk control via generalization bounds, practitioners can specify a tolerable error rate and obtain a subset of annotations satisfying that constraint with high probability.

Simulation-based testing has become a standard approach to validating autonomous driving agents prior to real-world deployment. A high-quality validation campaign will exercise an agent in diverse contexts comprised of varying static environments, e.g., lanes, intersections, signage, and dynamic elements, e.g., vehicles and pedestrians. To achieve this, existing test generation techniques rely on template-based, manually constructed, or random scenario generation. When applied to validate formally specified safety requirements, such methods either require significant human effort or run the risk of missing important behavior related to the requirement. To address this gap, we present STADA, a Specification-based Test generation framework for Autonomous Driving Agents that systematically generates the space of scenarios defined by a formal specification expressed in temporal logic (LTLf). Given a specification, STADA constructs all distinct initial scenes, a diverse space of continuations of those scenes, and simulations that reflect the behaviors of the specification. Evaluation of STADA on a variety of LTLf specifications formalized in SCENEFLOW using three complementary coverage criteria demonstrates that STADA yields more than 2x higher coverage than the best baseline on the finest criteria and a 75% increase for the coarsest criteria. Moreover, it matches the coverage of the best baseline with 6 times fewer simulations. While set in the context of autonomous driving, the approach is applicable to other domains with rich simulation environments.

Safe Reinforcement Learning from Human Feedback (RLHF) typically enforces safety through expected cost constraints, but the expectation captures only a single statistic of the cost distribution and fails to account for distributional uncertainty, particularly under heavy tails or rare catastrophic events. This limitation is problematic when robustness and risk sensitivity are critical. Stochastic dominance offers a principled alternative by comparing entire cost distributions rather than just their averages, enabling direct control over tail risks and potential out-of-distribution failures that expectation-based constraints may overlook. In this work, we propose Risk-sensitive Alignment via Dominance (RAD), a novel alignment framework that replaces scalar expected cost constraints with First-Order Stochastic Dominance (FSD) constraints. We operationalize this constraint by comparing the target policy's cost distribution to that of a reference policy within an Optimal Transport (OT) framework, using entropic regularization and Sinkhorn iterations to obtain a differentiable and computationally efficient objective for stable end-to-end optimization. Furthermore, we introduce quantile-weighted FSD constraints and show that weighted FSD universally controls a broad class of Spectral Risk Measures (SRMs), so that improvements under weighted dominance imply guaranteed improvements in the corresponding spectral risk. This provides a principled mechanism for tuning a model's risk profile via the quantile weighting function. Empirical results demonstrate that RAD improves harmlessness over baselines while remaining competitive in helpfulness, and exhibits greater robustness on out-of-distribution harmlessness evaluations.

The use of synthetic data has become increasingly popular as a privacy-preserving alternative to sharing real datasets, especially in sensitive domains such as healthcare, finance, and demography. However, the privacy assurances of synthetic data are not absolute, and remain susceptible to membership inference attacks (MIAs), where adversaries aim to determine whether a specific individual was present in the dataset used to train the generator. In this work, we propose a practical and effective method to quantify membership disclosure risk in tabular synthetic datasets using kernel density estimators (KDEs). Our KDE-based approach models the distribution of nearest-neighbour distances between synthetic data and the training records, allowing probabilistic inference of membership and enabling robust evaluation via ROC curves. We propose two attack models: a 'True Distribution Attack', which assumes privileged access to training data, and a more realistic, implementable 'Realistic Attack' that uses auxiliary data without true membership labels. Empirical evaluations across four real-world datasets and six synthetic data generators demonstrate that our method consistently achieves higher F1 scores and sharper risk characterization than a prior baseline approach, without requiring computationally expensive shadow models. The proposed method provides a practical framework and metric for quantifying membership disclosure risk in synthetic data, which enables data custodians to conduct a post-generation risk assessment prior to releasing their synthetic datasets for downstream use. The datasets and codes for this study are available at https://github.com/PyCoder913/MIA-KDE.

Variational autoencoders (VAEs) frequently suffer from posterior collapse, where latent variables become uninformative and the approximate posterior degenerates to the prior. Recent work has characterized this phenomenon as a phase transition governed by the spectral properties of the data covariance matrix. In this paper, we propose a fundamentally different approach: instead of avoiding collapse through architectural constraints or hyperparameter tuning, we eliminate the possibility of collapse altogether by leveraging the multiplicity of Gaussian mixture model (GMM) clusterings. We introduce Historical Consensus Training, an iterative selection procedure that progressively refines a set of candidate GMM priors through alternating optimization and selection. The key insight is that models trained to satisfy multiple distinct clustering constraints develop a historical barrier -- a region in parameter space that remains stable even when subsequently trained with a single objective. We prove that this barrier excludes the collapsed solution, and demonstrate through extensive experiments on synthetic and real-world datasets that our method achieves non-collapsed representations regardless of decoder variance or regularization strength. Our approach requires no explicit stability conditions (e.g., $σ^{\prime 2} < λ_{\max}$) and works with arbitrary neural architectures. The code is available at https://github.com/tsegoochang/historical-consensus-vae.

Time-series anomaly detectors are commonly compared on workstation-class hardware under unconstrained execution. In-vehicle monitoring, however, requires predictable latency and stable behavior under limited CPU parallelism. Accuracy-only leaderboards can therefore misrepresent which methods remain feasible under deployment-relevant constraints. We present ECoLAD (Efficiency Compute Ladder for Anomaly Detection), a deployment-oriented evaluation protocol instantiated as an empirical study on proprietary automotive telemetry (anomaly rate ${\approx}$0.022) and complementary public benchmarks. ECoLAD applies a monotone compute-reduction ladder across heterogeneous detector families using mechanically determined, integer-only scaling rules and explicit CPU thread caps, while logging every applied configuration change. Throughput-constrained behavior is characterized by sweeping target scoring rates and reporting (i) coverage (the fraction of entities meeting the target) and (ii) the best AUC-PR achievable among measured ladder configurations satisfying the target. On constrained automotive telemetry, lightweight classical detectors sustain both coverage and detection lift above the random baseline across the full throughput sweep. Several deep methods lose feasibility before they lose accuracy.

Hybrid quantum computing systems that combine discrete-variable qubits with continuous-variable qumodes offer promising advantages for quantum simulation, error correction, and sensing applications. However, existing quantum software frameworks lack native support for expressing and manipulating hybrid circuits, forcing developers to work with fragmented toolchains or rely on simulation-coupled representations that limit scalability. We present Hybridlane, an open-source software development kit providing a unified frontend for hybrid continuous-discrete variable quantum computing. Hybridlane introduces automatic wire type inference to distinguish qubits from qumodes without manual annotations, enabling compile-time validation of circuit correctness. By decoupling gate semantics from matrix representations, Hybridlane can describe wide and deep circuits with minimal memory consumption and without requiring simulation. The framework implements a comprehensive library of hybrid gates and decompositions following established instruction set architectures, while remaining compatible with PennyLane's extensive qubit algorithm library. Furthermore, it supports multiple backends including classical simulation with Bosonic Qiskit and hardware compilation to Sandia National Laboratories' QSCOUT ion trap. We demonstrate Hybridlane's capabilities through bosonic quantum phase estimation and ion trap calibration workflows.

Machine learning models have demonstrated remarkable success in sports prediction in the past years, often treating sports prediction as a classification task within the field. This paper introduces new perspectives for analyzing sports data to predict outcomes more accurately. We leverage rankings to generate team rankings for the 2024 dataset using Combinatorial Fusion Analysis (CFA), a new paradigm for combining multiple scoring systems through the rank-score characteristic (RSC) function and cognitive diversity (CD). Our result based on rank combination with respect to team ranking has an accuracy rate of $74.60\%$, which is higher than the best of the ten popular public ranking systems ($73.02\%$). This exhibits the efficacy of CFA in enhancing the precision of sports prediction through different lens.

LLM-based text embedders typically encode the semantic content of their input. However, embedding tasks require mapping diverse inputs to similar outputs. Typically, this input-output is addressed by training embedding models with paired data using contrastive learning. In this work, we propose a novel self-supervised approach, LLM2Vec-Gen, which adopts a different paradigm: rather than encoding the input, we learn to represent the model's potential response. Specifically, we add trainable special tokens to the LLM's vocabulary, append them to input, and optimize them to represent the LLM's response in a fixed-length sequence. Training is guided by the LLM's own completion for the query, along with an unsupervised embedding teacher that provides distillation targets. This formulation helps to bridge the input-output gap and transfers LLM capabilities such as safety alignment and reasoning to embedding tasks. Crucially, the LLM backbone remains frozen and training requires only unlabeled queries. LLM2Vec-Gen achieves state-of-the-art self-supervised performance on the Massive Text Embedding Benchmark (MTEB), improving by 9.3% over the best unsupervised embedding teacher. We also observe up to 43.2% reduction in harmful content retrieval and 29.3% improvement in reasoning capabilities for embedding tasks. Finally, the learned embeddings are interpretable and can be decoded into text to reveal their semantic content.

GLM-OCR is an efficient 0.9B-parameter compact multimodal model designed for real-world document understanding. It combines a 0.4B-parameter CogViT visual encoder with a 0.5B-parameter GLM language decoder, achieving a strong balance between computational efficiency and recognition performance. To address the inefficiency of standard autoregressive decoding in deterministic OCR tasks, GLM-OCR introduces a Multi-Token Prediction (MTP) mechanism that predicts multiple tokens per step, significantly improving decoding throughput while keeping memory overhead low through shared parameters. At the system level, a two-stage pipeline is adopted: PP-DocLayout-V3 first performs layout analysis, followed by parallel region-level recognition. Extensive evaluations on public benchmarks and industrial scenarios show that GLM-OCR achieves competitive or state-of-the-art performance in document parsing, text and formula transcription, table structure recovery, and key information extraction. Its compact architecture and structured generation make it suitable for both resource-constrained edge deployment and large-scale production systems.

Large language models are increasingly adopted as semantic backbones for neural text-to-speech systems. However, frozen LLM representations are insufficient for modeling speaker specific acoustic and perceptual characteristics. Our experiments involving fine tuning of the Language Model backbone of TTS show promise in improving the voice consistency and Signal to Noise ratio SNR in voice cloning task. Across multiple speakers LoRA finetuning consistently outperforms the non-finetuned base Qwen-0.5B model across three complementary dimensions of speech quality. First, perceptual quality improves significantly with DNS-MOS gains of up to 0.42 points for speakers whose training data exhibits sufficient acoustic variability. Second, speaker fidelity improves for all evaluated speakers with consistent increases in voice similarity indicating that LoRA effectively adapts speaker identity representations without degrading linguistic modeling. Third, signal level quality improves in most cases with signal to noise ratio increasing by as much as 34 percent. Crucially these improvements are strongly governed by the characteristics of the training data. Speakers with high variability in acoustic energy and perceptual quality achieve simultaneous gains in DNS-MOS voice similarity and SNR. Overall this work establishes that LoRA finetuning is not merely a parameter efficient optimization technique but an effective mechanism for better speaker level adaptation in compact LLM-based TTS systems. When supported by sufficiently diverse training data LoRA adapted Qwen-0.5B consistently surpasses its frozen base model in perceptual quality speaker similarity with low latency using GGUF model hosted in quantized form.

Convolutional Neural Networks (CNNs) have proven to be highly effective in solving a broad spectrum of computer vision tasks, such as classification, identification, and segmentation. These methods can be deployed in both centralized and distributed environments, depending on the computational demands of the task. While much of the literature has focused on the explainability of CNNs, which is essential for building trust and confidence in their predictions, there remains a gap in understanding their impact on computational resources, particularly in distributed training contexts. In this study, we analyze how CNN architectures primarily influence model accuracy and investigate additional factors that affect computational efficiency in distributed systems. Our findings contribute valuable insights for optimizing the deployment of CNNs in resource-intensive scenarios, paving the way for further exploration of variables critical to distributed learning.

Transformer-based large language models (LLMs) rely on key-value (KV) caching to avoid redundant computation during autoregressive inference. While this mechanism greatly improves efficiency, the cache size grows linearly with the input sequence length, quickly becoming a bottleneck for long-context tasks. Existing solutions mitigate this problem by evicting prompt KV that are deemed unimportant, guided by estimated importance scores. Notably, a recent line of work proposes to improve eviction quality by "glimpsing into the future", in which a draft generator produces a surrogate future response approximating the target model's true response, and this surrogate is subsequently used to estimate the importance of cached KV more accurately. However, these approaches rely on computationally expensive draft generation, which introduces substantial prefilling overhead and limits their practicality in real-world deployment. To address this challenge, we propose LookaheadKV, a lightweight eviction framework that leverages the strength of surrogate future response without requiring explicit draft generation. LookaheadKV augments transformer layers with parameter-efficient modules trained to predict true importance scores with high accuracy. Our design ensures negligible runtime overhead comparable to existing inexpensive heuristics, while achieving accuracy superior to more costly approximation methods. Extensive experiments on long-context understanding benchmarks, across a wide range of models, demonstrate that our method not only outperforms recent competitive baselines in various long-context understanding tasks, but also reduces the eviction cost by up to 14.5x, leading to significantly faster time-to-first-token. Our code is available at https://github.com/SamsungLabs/LookaheadKV.

In reinforcement learning, we typically aim to optimize the expected value of the sum of rewards an agent collects over a trajectory. However, if the process generating these rewards is non-ergodic, the expected value, i.e., the average over infinitely many trajectories with a given policy, is uninformative for the average over a single, but infinitely long trajectory. Thus, if we care about how the individual agent performs during deployment, the expected value is not a good optimization objective. In this paper, we discuss the impact of non-ergodic reward processes on reinforcement learning agents through an instructive example, relate the notion of ergodic reward processes to more widely used notions of ergodic Markov chains, and present existing solutions that optimize long-term performance of individual trajectories under non-ergodic reward dynamics.

Medication errors pose a significant threat to patient safety, making pharmacist verification (PV) a critical, yet heavily burdened, final safeguard. The direct application of Large Language Models (LLMs) to this zero-tolerance domain is untenable due to their inherent factual unreliability, lack of traceability, and weakness in complex reasoning. To address these challenges, we introduce PharmGraph-Auditor, a novel system designed for safe and evidence-grounded prescription auditing. The core of our system is a trustworthy Hybrid Pharmaceutical Knowledge Base (HPKB), implemented under the Virtual Knowledge Graph (VKG) paradigm. This architecture strategically unifies a relational component for set constraint satisfaction and a graph component for topological reasoning via a rigorous mapping layer. To construct this HPKB, we propose the Iterative Schema Refinement (ISR) algorithm, a framework that enables the co-evolution of both graph and relational schemas from medical texts. For auditing, we introduce the KB-grounded Chain of Verification (CoV), a new reasoning paradigm that transforms the LLM from an unreliable generator into a transparent reasoning engine. CoV decomposes the audit task into a sequence of verifiable queries against the HPKB, generating hybrid query plans to retrieve evidence from the most appropriate data store. Experimental results demonstrate robust knowledge extraction capabilities and show promises of using PharmGraph-Auditor to enable pharmacists to achieve safer and faster prescription verification.

Reinforcement learning (RL) finetuning has become a key technique for enhancing the reasoning abilities of large language models (LLMs). However, its effectiveness critically depends on the selection of training data. Recent advances underscore the importance of online prompt selection methods, which typically concentrate training on partially solved or moderately challenging examples under the current policy, thereby yielding more effective model updates. While significantly accelerating RL finetuning in terms of training steps, they also incur substantial computational overhead by requiring extensive LLM rollouts over large candidate batches to identify informative samples, an expense that can outweigh the finetuning process itself. To address this challenge, this work proposes Dynamics-Predictive Sampling (DPS), which online predicts and selects informative prompts by inferring their learning dynamics prior to costly rollouts. Specifically, we introduce a new perspective by modeling each prompt's solving progress during RL finetuning as a dynamical system, where the extent of solving is represented as the state and the transition is characterized by a hidden Markov model. Using historical rollout reward signals, we perform online Bayesian inference to estimate evolving state distributions, and the inference outcome provides a predictive prior for efficient prompt selection without rollout-intensive filtering. Empirical results across diverse reasoning tasks, including mathematics, planning, and visual geometry, demonstrate that DPS substantially reduces redundant rollouts, accelerates the training process, and achieves superior reasoning performance.

We propose novel kernel-based tests for assessing the equivalence between distributions. Traditional goodness-of-fit testing is inappropriate for concluding the absence of distributional differences, because failure to reject the null hypothesis may simply be a result of lack of test power, also known as the Type-II error. This motivates \emph{equivalence testing}, which aims to assess the \emph{absence} of a statistically meaningful effect under controlled error rates. However, existing equivalence tests are either limited to parametric distributions or focus only on specific moments rather than the full distribution. We address these limitations using two kernel-based statistical discrepancies: the \emph{kernel Stein discrepancy} and the \emph{Maximum Mean Discrepancy}. The null hypothesis of our proposed tests assumes the candidate distribution differs from the nominal distribution by at least a pre-defined margin, which is measured by these discrepancies. We propose two approaches for computing the critical values of the tests, one using an asymptotic normality approximation, and another based on bootstrapping. Numerical experiments are conducted to assess the performance of these tests.

We present a parameter-efficient Diffusion Transformer (DiT) for generating 200bp cell-type-specific regulatory DNA sequences. By replacing the U-Net backbone of DNA-Diffusion with a transformer denoiser equipped with a 2D CNN input encoder, our model matches the U-Net's best validation loss in 13 epochs (60$\times$ fewer) and converges 39% lower, while reducing memorization from 5.3% to 1.7% of generated sequences aligning to training data via BLAT. Ablations show the CNN encoder is essential: without it, validation loss increases 70% regardless of positional embedding choice. We further apply DDPO finetuning using Enformer as a reward model, achieving a 38$\times$ improvement in predicted regulatory activity. Cross-validation against DRAKES on an independent prediction task confirms that improvements reflect genuine regulatory signal rather than reward model overfitting.

Electroencephalogram (EEG) classification is critical for applications ranging from medical diagnostics to brain-computer interfaces, yet it remains challenging due to the inherently low signal-to-noise ratio (SNR) and high inter-subject variability. To address these issues, we propose LAtte, a novel framework that integrates a Lorentz Attention Module with an InceptionTime-based encoder to enable robust and generalizable EEG classification. Unlike prior work, which evaluates primarily on single-subject performance, LAtte focuses on cross-subject training. First, we learn a shared baseline signal across all subjects using pretraining tasks to capture common underlying patterns. Then, we utilize novel Lorentz low-rank adapters to learn subject-specific embeddings that model individual differences. This allows us to learn a shared model that performs robustly across subjects, and can be subsequently finetuned for individual subjects or used to generalize to unseen subjects. We evaluate LAtte on three well-established EEG datasets, achieving a substantial improvement in performance over current state-of-the-art methods.

Knowledge distillation (KD) methods are pivotal in compressing large pre-trained language models into smaller models, ensuring computational efficiency without significantly dropping performance. Traditional KD techniques assume homogeneity in modalities between the teacher (source) and the student (target) models. On the other hand, existing multimodal knowledge distillation methods require modality-specific pre-training of the teacher model, which is computationally infeasible in most cases. In this paper, we introduce ARMADA, an efficient cross-modal knowledge distillation framework designed to transfer knowledge from large vision-language models, including black-box models, to language-only models. Unlike existing KD techniques that rely on the internal structures of multimodal teachers or require computationally expensive pre-training, ARMADA leverages novel alignment techniques to distil knowledge without altering the teacher model, ensuring efficiency and scalability. We empirically validate ARMADA on twelve natural language understanding, eight complex generative reasoning and five instruction-tuning tasks, demonstrating consistent performance improvements in large models such as DeBERTa-v2-1.4B, OPT-1.3B, LLaMA-{3B, 7B, 8B}. ARMADA achieves up to 3.4% improvement on language understanding tasks and 2.6% boost in generative reasoning, all without requiring expensive multimodal pre-training or fine-tuning of the teacher model. Our findings challenge conventional knowledge distillation paradigms by demonstrating that even vision-language models, despite lacking direct textual understanding, can significantly enhance language models when distilled appropriately.

Subject indexing is vital for discovery but hard to sustain at scale and across languages. We release a large bilingual (English/German) corpus of catalog records annotated with the Integrated Authority File (GND), plus a machine-actionable GND taxonomy. The resource enables ontology-aware multi-label classification, mapping text to authority terms, and agent-assisted cataloging with reproducible, authority-grounded evaluation. We provide a brief statistical profile and qualitative error analyses of three systems. We invite the community to assess not only accuracy but usefulness and transparency, toward authority-anchored AI co-pilots that amplify catalogers' work.

Polygenic risk scores and other genomic analyses require large individual-level genotype datasets, yet strict data access restrictions impede sharing. Synthetic genotype generation offers a privacy-preserving alternative, but most existing methods operate unconditionally, producing samples without phenotype alignment, or rely on unsupervised compression, creating a gap between statistical fidelity and downstream task utility. We present SNPgen, a two-stage conditional latent diffusion framework for generating phenotype-supervised synthetic genotypes. SNPgen combines GWAS-guided variant selection (1,024-2,048 trait-associated SNPs) with a variational autoencoder for genotype compression and a latent diffusion model conditioned on binary disease labels via classifier-free guidance. Evaluated on 458,724 UK Biobank individuals across four complex diseases (coronary artery disease, breast cancer, type 1 and type 2 diabetes), models trained on synthetic data matched real-data predictive performance in a train-on-synthetic, test-on-real protocol, approaching genome-wide PRS methods that use $2$-$6\times$ more variants. Privacy analysis confirmed zero identical matches, near-random membership inference (AUC $\approx 0.50$), preserved linkage disequilibrium structure, and high allele frequency correlation ($r \geq 0.95$) with source data. A controlled simulation with known causal effects verified faithful recovery of the imposed genetic association structure.

Communication is a crucial social factor in the success of software projects, as positively or negatively perceived statements can influence how recipients feel and affect team collaboration through emotional contagion. Whether a developer perceives a written message as positive, negative, or neutral is likely shaped by multiple factors. In this paper, we investigate how mood traits and states, life circumstances, project phases, and group dynamics relate to the perception of text-based messages in software development. We conducted a four-round survey study with 81 students in team-based software projects. Across rounds, participants reported these factors and labeled 30 decontextualized statements for sentiment, including meta-data on labeling rationale and uncertainty. Our results show: (1) Sentiment perception is only moderately stable within individuals, and label changes concentrate on ambiguity-prone statements; (2) Correlation-level signals are small and do not survive global multiple-testing correction; (3) In statement-level repeated-measures models (GEE), higher mood trait and reactivity are associated with more positive (and less neutral) labeling, while predictors of negative labeling are weaker and at most trend-level (e.g., task conflict); (4) We find no clear evidence of systematic project-phase effects. Overall, sentiment perception varies within persons and is strongly statement-dependent. Although our study was conducted in an academic setting, the observed variability and ambiguity effects suggest caution when interpreting sentiment analysis outputs and motivate future work with contextualized, in-project communication.

SiDiaC-v.2.0 is the largest comprehensive Sinhala Diachronic Corpus to date, covering a period from 1800 CE to 1955 CE in terms of publication dates, and a historical span from the 5th to the 20th century CE in terms of written dates. The corpus consists of 244k words across 185 literary works that underwent thorough filtering, preprocessing, and copyright compliance checks, followed by extensive post-processing. Additionally, a subset of 59 documents totalling 70k words was annotated based on their written dates. Texts from the National Library of Sri Lanka were selected from the SiDiaC-v.1.0 non-filtered list, which was digitised using Google Document AI OCR. This was followed by post-processing to correct formatting issues, address code-mixing, include special tokens, and fix malformed tokens. The construction of SiDiaC-v.2.0 was informed by practices from other corpora, such as FarPaHC, SiDiaC-v.1.0, and CCOHA. This was particularly relevant for syntactic annotation and text normalisation strategies, given the shared characteristics of low-resource language status between Faroese and the similar cleaning strategies utilised in CCOHA. This corpus is categorised into two layers based on genres: primary and secondary. The primary categorisation is binary, assigning each book to either Non-Fiction or Fiction. The secondary categorisation is more detailed, grouping texts under specific genres such as Religious, History, Poetry, Language, and Medical. Despite facing challenges due to limited resources, SiDiaC-v.2.0 serves as a comprehensive resource for Sinhala NLP, building upon the work previously done in SiDiaC-v.1.0.

Advancing Multi-Agent Pathfinding (MAPF) and Multi-Robot Motion Planning (MRMP) requires platforms that enable transparent, reproducible comparisons across modeling choices. Existing tools either scale under simplifying assumptions (grids, homogeneous agents) or offer higher fidelity with less comparable instrumentation. We present GRACE, a unified 2D simulator+benchmark that instantiates the same task at multiple abstraction levels (grid, roadmap, continuous) via explicit, reproducible operators and a common evaluation protocol. Our empirical results on public maps and representative planners enable commensurate comparisons on a shared instance set. Furthermore, we quantify the expected representation-fidelity trade-offs (MRMP solves instances at higher fidelity but lower speed, while grid/roadmap planners scale farther). By consolidating representation, execution, and evaluation, GRACE thereby aims to make cross-representation studies more comparable and provides a means to advance multi-robot planning research and its translation to practice.

The Environmental Mapping and Analysis Program (EnMAP) mission has opened new frontiers in the monitoring of optically complex environments. However, the accurate retrieval of surface reflectance over water bodies remains a significant challenge, as the water-leaving signal typically accounts for only a small fraction of the total radiance, being easily obscured by atmospheric scattering and surface reflection effects. This paper introduces 6ABOS (6S-based Atmospheric Background Offset Subtraction), a novel open-source Python framework designed to automate the atmospheric correction (AC) of EnMAP hyperspectral imagery. By leveraging the Second Simulation of the Satellite Signal in the Solar Spectrum (6S) radiative transfer model, 6ABOS implements a physically-based inversion scheme that accounts for Rayleigh scattering, aerosol interactions, and gaseous absorption. The framework integrates automated EnMAP metadata parsing with dynamic atmospheric parameter retrieval via the Google Earth Engine (GEE) Application Programming Interface (API). Validation was conducted over two Mediterranean inland water reservoirs with contrasting trophic states: the oligotrophic Benag{'e}ber and the hypertrophic Bell{'u}s. Results demonstrate a high degree of spectral similarity between in situ measurements and EnMAP-derived water-leaving reflectances. The Spectral Angle Mapper (SAM) values remained consistently low (SAM $<$ 10$^\circ$) across both study sites. 6ABOS is distributed via conda-forge, providing the scientific community with a scalable, transparent, and reproducible open-science tool for advancing hyperspectral aquatic research in the cloud-computing era.

The information flows in serverless platforms are complex and non-conservative. This is a direct result of how independently deployed functions interact under the platform coarse-grained control mechanisms. To manage this complexity, we introduce a topological model for serverless services. Using Hodge decomposition, we can separate observed operational flows into two distinct categories. They include components that can be corrected locally and harmonic modes that persist at any scale. Our analysis reveals that these harmonic flows emerge naturally from different types of inter-function interactions. They should be understood as structural properties of serverless systems, not as configuration errors. Building on this insight, we present an iterative method for analyzing inter-function flows. This method helps deriving practical remediation strategies. One such strategy is the introduction of "dumping effects" to contain harmonic inefficiencies, offering an alternative to completely restructuring the service's topological model. Our experimental results confirm that this approach can uncover latent architectural structures.

In Reinforcement Learning with Verifiable Rewards (RLVR), constructing a robust advantage baseline is critical for policy gradients, effectively guiding the policy model to reinforce desired behaviors. Recent research has introduced Generalist Value Models (such as $V_0$), which achieve pre-trained value estimation by explicitly encoding model capabilities in-context, eliminating the need to synchronously update the value model alongside the policy model. In this paper, we propose $V_{0.5}$, which adaptively fuses the baseline predicted by such value model (acting as a prior) with the empirical mean derived from sparse rollouts. This constructs a robust baseline that balances computational efficiency with extremely low variance. Specifically, we introduce a real-time statistical testing and dynamic budget allocation. This balances the high variance caused by sparse sampling against the systematic bias (or hallucinations) inherent in the value model's prior. By constructing a hypothesis test to evaluate the prior's reliability in real-time, the system dynamically allocates additional rollout budget on demand. This mechanism minimizes the baseline estimator's Mean Squared Error (MSE), guaranteeing stable policy gradients, even under extreme sparsity with a group size of 4. Extensive evaluations across six mathematical reasoning benchmarks demonstrate that $V_{0.5}$ significantly outperforms GRPO and DAPO, achieving faster convergence and over some 10% performance improvement.

Reliable localisation in vineyards is hindered by row-level perceptual aliasing: parallel crop rows produce nearly identical LiDAR observations, causing geometry-only and vision-based SLAM systems to converge towards incorrect corridors, particularly during headland transitions. We present a Semantic Landmark Particle Filter (SLPF) that integrates trunk and pole landmark detections with 2D LiDAR within a probabilistic localisation framework. Detected trunks are converted into semantic walls, forming structural row boundaries embedded in the measurement model to improve discrimination between adjacent rows. GNSS is incorporated as a lightweight prior that stabilises localisation when semantic observations are sparse. Field experiments in a 10-row vineyard demonstrate consistent improvements over geometry-only (AMCL), vision-based (RTAB-Map), and GNSS baselines. Compared to AMCL, SLPF reduces Absolute Pose Error by 22% and 65% across two traversal directions; relative to a NoisyGNSS baseline, APE decreases by 65% and 61%. Row correctness improves from 0.67 to 0.73, while mean cross-track error decreases from 1.40 m to 1.26 m. These results show that embedding row-level structural semantics within the measurement model enables robust localisation in highly repetitive outdoor agricultural environments.

Deploying Large Language Models to data-scarce programming domains poses significant challenges, particularly for kernel synthesis on emerging Domain-Specific Architectures where a "Data Wall" limits available training data. While models excel on data-rich platforms like CUDA, they suffer catastrophic performance drops on data-scarce ecosystems such as NPU programming. To overcome this cold-start barrier without expensive fine-tuning, we introduce EvoKernel, a self-evolving agentic framework that automates the lifecycle of kernel synthesis from initial drafting to continual refining. EvoKernel addresses this by formulating the synthesis process as a memory-based reinforcement learning task. Through a novel value-driven retrieval mechanism, it learns stage-specific Q-values that prioritize experiences based on their contribution to the current objective, whether bootstrapping a feasible draft or iteratively refining latency. Furthermore, by enabling cross-task memory sharing, the agent generalizes insights from simple to complex operators. By building an NPU variant of KernelBench and evaluating on it, EvoKernel improves frontier models' correctness from 11.0% to 83.0% and achieves a median speedup of 3.60x over initial drafts through iterative refinement. This demonstrates that value-guided experience accumulation allows general-purpose models to master the kernel synthesis task on niche hardware ecosystems. Our official page is available at https://evokernel.zhuo.li.

Human Presence Detection (HPD) is key to enable intelligent power management and security features in everyday devices. In this paper we propose the first HPD solution that leverages monostatic Wi-Fi sensing and detects user position using only the built-in Wi-Fi hardware of a device, with no need for external devices, access points, or additional sensors. In contrast, existing HPD solutions for laptops require external dedicated sensors which add cost and complexity, or rely on camera-based approaches that introduce significant privacy concerns. We herewith introduce the Range-Filtered Doppler Spectrum (RF-DS), a novel Wi-Fi sensing technique for presence estimation that enables both range-selective and temporally windowed detection of user presence. By applying targeted range-area filtering in the Channel Impulse Response (CIR) domain before Doppler analysis, our method focuses processing on task-relevant spatial zones, significantly reducing computational complexity. In addition, the use of temporal windows in the spectrum domain provides greater estimator stability compared to conventional 2D Range-Doppler detectors. Furthermore, we propose an adaptive multi-rate processing framework that dynamically adjusts Channel State Information (CSI) sampling rates-operating at low frame rates (10Hz) during idle periods and high rates (100Hz) only when motion is detected. To our knowledge, this is the first low-complexity solution for occupancy detection using monostatic Wi-Fi sensing on a built-in Wi-Fi network interface controller (NIC) of a commercial off-the-shelf laptop that requires no external network infrastructure or specialized sensors. Our solution can scale across different environments and devices without calibration or retraining.

Existing hard-label text attacks often rely on inefficient "outside-in" strategies that traverse vast search spaces. We propose PivotAttack, a query-efficient "inside-out" framework. It employs a Multi-Armed Bandit algorithm to identify Pivot Sets-combinatorial token groups acting as prediction anchors-and strategically perturbs them to induce label flips. This approach captures inter-word dependencies and minimizes query costs. Extensive experiments across traditional models and Large Language Models demonstrate that PivotAttack consistently outperforms state-of-the-art baselines in both Attack Success Rate and query efficiency.

Understanding how neural networks rely on visual cues offers a human-interpretable view of their internal decision processes. The cue-conflict benchmark has been influential in probing shape-texture preference and in motivating the insight that stronger, human-like shape bias is often associated with improved in-domain performance. However, we find that the current stylization-based instantiation can yield unstable and ambiguous bias estimates. Specifically, stylization may not reliably instantiate perceptually valid and separable cues nor control their relative informativeness, ratio-based bias can obscure absolute cue sensitivity, and restricting evaluation to preselected classes can distort model predictions by ignoring the full decision space. Together, these factors can confound preference with cue validity, cue balance, and recognizability artifacts. We introduce REFINED-BIAS, an integrated dataset and evaluation framework for reliable and interpretable shape-texture bias diagnosis. REFINED-BIAS constructs balanced, human- and model- recognizable cue pairs using explicit definitions of shape and texture, and measures cue-specific sensitivity over the full label space via a ranking-based metric, enabling fairer cross-model comparisons. Across diverse training regimes and architectures, REFINED-BIAS enables fairer cross-model comparison, more faithful diagnosis of shape and texture biases, and clearer empirical conclusions, resolving inconsistencies that prior cue-conflict evaluations could not reliably disambiguate.

The Segment Anything Model (SAM) has revolutionized interactive segmentation through spatial prompting. While existing work primarily focuses on automating prompts in various settings, real-world annotation workflows involve iterative refinement where annotators observe model outputs and strategically place prompts to resolve ambiguities. Current pipelines typically rely on the annotator's visual assessment of the predicted mask quality. We postulate that a principled approach for automated interactive prompting is to use a model-derived criterion to identify the most informative region for the next prompt. In this work, we establish active prompting: a spatial active learning approach where locations within images constitute an unlabeled pool and prompts serve as queries to prioritize information-rich regions, increasing the utility of each interaction. We further present BALD-SAM: a principled framework adapting Bayesian Active Learning by Disagreement (BALD) to spatial prompt selection by quantifying epistemic uncertainty. To do so, we freeze the entire model and apply Bayesian uncertainty modeling only to a small learned prediction head, making intractable uncertainty estimation practical for large multi-million parameter foundation models. Across 16 datasets spanning natural, medical, underwater, and seismic domains, BALD-SAM demonstrates strong cross-domain performance, ranking first or second on 14 of 16 benchmarks. We validate these gains through a comprehensive ablation suite covering 3 SAM backbones and 35 Laplace posterior configurations, amounting to 38 distinct ablation settings. Beyond strong average performance, BALD-SAM surpasses human prompting and, in several categories, even oracle prompting, while consistently outperforming one-shot baselines in final segmentation quality, particularly on thin and structurally complex objects.

Speech-aware large language models (LLMs) can accept speech inputs, yet their training objectives largely emphasize linguistic content or specific fields such as emotions or the speaker's gender, leaving it unclear whether they encode speaker identity. First, we propose a model-agnostic scoring protocol that produces continuous verification scores for both API-only and open-weight models, using confidence scores or log-likelihood ratios from the Yes/No token probabilities. Using this protocol, we benchmark recent speech-aware LLMs and observe weak speaker discrimination (EERs above 20% on VoxCeleb1). Second, we introduce a lightweight augmentation that equips an LLM with ASV capability by injecting frozen ECAPA-TDNN speaker embeddings through a learned projection and training only LoRA adapters. On TinyLLaMA-1.1B, the resulting ECAPA-LLM achieves 1.03% EER on VoxCeleb1-E, approaching a dedicated speaker verification system while preserving a natural-language interface.

Deep generative models can help with data scarcity and privacy by producing synthetic training data, but they struggle in low-data, imbalanced tabular settings to fully learn the complex data distribution. We argue that striving for the full joint distribution could be overkill; for greater data efficiency, models should prioritize learning the conditional distribution $P(y\mid \bm{X})$, as suggested by recent theoretical analysis. Therefore, we overcome this limitation with \textbf{ReTabSyn}, a \textbf{Re}inforced \textbf{Tab}ular \textbf{Syn}thesis pipeline that provides direct feedback on feature correlation preservation during synthesizer training. This objective encourages the generator to prioritize the most useful predictive signals when training data is limited, thereby strengthening downstream model utility. We empirically fine-tune a language model-based generator using this approach, and across benchmarks with small sample sizes, class imbalance, and distribution shift, ReTabSyn consistently outperforms state-of-the-art baselines. Moreover, our approach can be readily extended to control various aspects of synthetic tabular data, such as applying expert-specified constraints on generated observations.

Accurate and robust trajectory prediction is essential for safe and efficient autonomous driving, yet recent work has shown that even state-of-the-art prediction models are highly vulnerable to inputs being mildly perturbed by adversarial attacks. Although model vulnerabilities to such attacks have been studied, work on effective countermeasures remains limited. In this work, we develop and evaluate a new defense mechanism for trajectory prediction models based on randomized smoothing -- an approach previously applied successfully in other domains. We evaluate its ability to improve model robustness through a series of experiments that test different strategies of randomized smoothing. We show that our approach can consistently improve prediction robustness of multiple base trajectory prediction models in various datasets without compromising accuracy in non-adversarial settings. Our results demonstrate that randomized smoothing offers a simple and computationally inexpensive technique for mitigating adversarial attacks in trajectory prediction.

Deep learning models can predict protein properties with unprecedented accuracy but rarely offer mechanistic insight or actionable guidance for engineering improved variants. When a model flags an antibody as unstable, the protein engineer is left without recourse: which mutations would rescue stability while preserving function? We introduce Manifold-Constrained Counterfactual Optimization for Proteins (MCCOP), a framework that computes minimal, biologically plausible sequence edits that flip a model's prediction to a desired target state. MCCOP operates in a continuous joint sequence-structure latent space and employs a pretrained diffusion model as a manifold prior, balancing three objectives: validity (achieving the target property), proximity (minimizing mutations), and plausibility (producing foldable proteins). We evaluate MCCOP on three protein engineering tasks - GFP fluorescence rescue, thermodynamic stability enhancement, and E3 ligase activity recovery - and show that it generates sparser, more plausible counterfactuals than both discrete and continuous baselines. The recovered mutations align with known biophysical mechanisms, including chromophore packing and hydrophobic core consolidation, establishing MCCOP as a tool for both model interpretation and hypothesis-driven protein design. Our code is publicly available at github.com/weroks/mccop.

The emergence of large language model (LLM)-based agent frameworks has shifted the primary challenge in building domain-expert AI agents from raw capability to effective encoding of domain expertise. Two dominant paradigms -- code-first development, which embeds expertise in deterministic pipelines, and prompt-first development, which captures expertise in static system prompts -- both treat agent construction as a discrete engineering phase preceding deployment. We argue that this sequential assumption creates a fundamental mismatch with the nature of domain expertise, which is substantially tacit, deeply personal, and continuously evolving. We propose Nurture-First Development (NFD), a paradigm in which agents are initialized with minimal scaffolding and progressively grown through structured conversational interaction with domain practitioners. The central mechanism is the Knowledge Crystallization Cycle, whereby fragmented knowledge embedded in operational dialogue is periodically consolidated into structured, reusable knowledge assets. We formalize NFD through: (1) a Three-Layer Cognitive Architecture organizing agent knowledge by volatility and personalization degree; (2) the Knowledge Crystallization Cycle with formal definitions of crystallization operations and efficiency metrics; and (3) an operational framework comprising a Dual-Workspace Pattern and Spiral Development Model. We illustrate the paradigm through a detailed case study on building a financial research agent for U.S. equity analysis and discuss the conditions, limitations, and broader implications of NFD for human-agent co-evolution.

The rapid adoption of large language models (LLMs) in financial services introduces new operational, regulatory, and security risks. Yet most red-teaming benchmarks remain domain-agnostic and fail to capture failure modes specific to regulated BFSI settings, where harmful behavior can be elicited through legally or professionally plausible framing. We propose a risk-aware evaluation framework for LLM security failures in Banking, Financial Services, and Insurance (BFSI), combining a domain-specific taxonomy of financial harms, an automated multi-round red-teaming pipeline, and an ensemble-based judging protocol. We introduce the Risk-Adjusted Harm Score (RAHS), a risk-sensitive metric that goes beyond success rates by quantifying the operational severity of disclosures, accounting for mitigation signals, and leveraging inter-judge agreement. Across diverse models, we find that higher decoding stochasticity and sustained adaptive interaction not only increase jailbreak success, but also drive systematic escalation toward more severe and operationally actionable financial disclosures. These results expose limitations of single-turn, domain-agnostic security evaluation and motivate risk-sensitive assessment under prolonged adversarial pressure for real-world BFSI deployment.

The growing demand for wireless connectivity, combined with limited spectrum resources, calls for more efficient spectrum management. Spectrum sharing is a promising approach; however, regulators need accurate methods to characterize demand dynamics and guide allocation decisions. This paper builds and validates a spectrum demand proxy from public deployment records and uses a graph attention network in a hierarchical, multi-resolution setup (HR-GAT) to estimate spectrum demand at fine spatial scales. The model captures both neighborhood effects and cross-scale patterns, reducing spatial autocorrelation and improving generalization. Evaluated across five Canadian cities and against eight competitive baselines, HR-GAT reduces median RMSE by roughly 21% relative to the best alternative and lowers residual spatial bias. The resulting demand maps are regulator-accessible and support spectrum sharing and spectrum allocation in wireless networks.

Accurate spatial prediction of cellular traffic demand is essential for 5G NR capacity planning, network densification, and data-driven 6G planning. Although machine learning can fuse heterogeneous geospatial and socio-economic layers to estimate fine-grained demand maps, spatial autocorrelation can cause neighborhood leakage under naive train/test splits, inflating accuracy and weakening planning reliability. This paper presents an AI-driven framework that reduces leakage and improves spatial generalization via a context-aware two-stage splitting strategy with residual spatial error correction. Experiments using crowdsourced usage indicators across five major Canadian cities show consistent mean absolute error (MAE) reductions relative to location-only clustering, supporting more reliable bandwidth provisioning and evidence-based spectrum planning and sharing assessments.

EVMbench, released by OpenAI, Paradigm, and OtterSec, is the first large-scale benchmark for AI agents on smart contract security. Its results -- agents detect up to 45.6% of vulnerabilities and exploit 72.2% of a curated subset -- have fueled expectations that fully automated AI auditing is within reach. We identify two limitations: its narrow evaluation scope (14 agent configurations, most models tested on only their vendor scaffold) and its reliance on audit-contest data published before every model's release that models may have seen during training. To address these, we expand to 26 configurations across four model families and three scaffolds, and introduce a contamination-free dataset of 22 real-world security incidents postdating every model's release date. Our evaluation yields three findings: (1) agents' detection results are not stable, with rankings shifting across configurations, tasks, and datasets; (2) on real-world incidents, no agent succeeds at end-to-end exploitation across all 110 agent-incident pairs despite detecting up to 65% of vulnerabilities, contradicting EVMbench's conclusion that discovery is the primary bottleneck; and (3) scaffolding materially affects results, with an open-source scaffold outperforming vendor alternatives by up to 5 percentage points, yet EVMbench does not control for this. These findings challenge the narrative that fully automated AI auditing is imminent. Agents reliably catch well-known patterns and respond strongly to human-provided context, but cannot replace human judgment. For developers, agent scans serve as a pre-deployment check. For audit firms, agents are most effective within a human-in-the-loop workflow where AI handles breadth and human auditors contribute protocol-specific knowledge and adversarial reasoning. Code and data: https://github.com/blocksecteam/ReEVMBench/.

We present the Multilingual Reasoning Gym, an extension of Reasoning Gym (Stojanovski et al., 2025), that procedurally generates verifiable reasoning problems across 14 languages. We translate templates for 94 tasks with native-speaker validation in 10 languages and targeted code or template adaptations to ensure linguistic naturalness. The Multilingual Reasoning Gym preserves the core benefits of the procedural generation approach used in the original Reasoning Gym, such as virtually unlimited problem instance generation and adjustable difficulty, and remains directly usable for Reinforcement Learning from Verifiable Rewards and evaluation settings. Problems in the Multilingual Reasoning Gym are parallel across languages, enabling crosslingually parallel data generation at massive scale due to the procedural nature of the environments. We release our implementation to support research into multilingual reasoning models.

We present LuxBorrow, a borrowing-first analysis of Luxembourgish (LU) news spanning 27 years (1999-2025), covering 259,305 RTL articles and 43.7M tokens. Our pipeline combines sentence-level language identification (LU/DE/FR/EN) with a token-level borrowing resolver restricted to LU sentences, using lemmatization, a collected loanword registry, and compiled morphological and orthographic rules. Empirically, LU remains the matrix language across all documents, while multilingual practice is pervasive: 77.1% of articles include at least one donor language and 65.4% use three or four. Breadth does not imply intensity: median code-mixing index (CMI) increases from 3.90 (LU+1) to only 7.00 (LU+3), indicating localized insertions rather than balanced bilingual text. Domain and period summaries show moderate but persistent mixing, with CMI rising from 6.1 (1999-2007) to a peak of 8.4 in 2020. Token-level adaptations total 25,444 instances and exhibit a mixed profile: morphological 63.8%, orthographic 35.9%, lexical 0.3%. The most frequent individual rules are orthographic, such as on->oun and eur->er, while morphology is collectively dominant. Diachronically, code-switching intensifies, and morphologically adapted borrowings grow from a small base. French overwhelmingly supplies adapted items, with modest growth for German and negligible English. We advocate borrowing-centric evaluation, including borrowed token and type rates, donor entropy over borrowed items, and assimilation ratios, rather than relying only on document-level mixing indices.

Metaphor identification is a foundational task in figurative language processing, yet most computational approaches operate as opaque classifiers offering no insight into why an expression is judged metaphorical. This interpretability gap is especially acute for Chinese, where rich figurative traditions, absent morphological cues, and limited annotated resources compound the challenge. We present an LLM-assisted pipeline that operationalises four metaphor identification protocols--MIP/MIPVU lexical analysis, CMDAG conceptual-mapping annotation, emotion-based detection, and simile-oriented identification--as executable, human-auditable rule scripts. Each protocol is a modular chain of deterministic steps interleaved with controlled LLM calls, producing structured rationales alongside every classification decision. We evaluate on seven Chinese metaphor datasets spanning token-, sentence-, and span-level annotation, establishing the first cross-protocol comparison for Chinese metaphor identification. Within-protocol evaluation shows Protocol A (MIP) achieves an F1 of 0.472 on token-level identification, while cross-protocol analysis reveals striking divergence: pairwise Cohen's kappa between Protocols A and D is merely 0.001, whereas Protocols B and C exhibit near-perfect agreement (kappa = 0.986). An interpretability audit shows all protocols achieve 100% deterministic reproducibility, with rationale correctness from 0.40 to 0.87 and editability from 0.80 to 1.00. Error analysis identifies conceptual-domain mismatch and register sensitivity as dominant failure modes. Our results demonstrate that protocol choice is the single largest source of variation in metaphor identification, exceeding model-level variation, and that rule-script architectures achieve competitive performance while maintaining full transparency.

Sparse Attention Vectors (SAVs) have emerged as an excellent training-free alternative to supervised finetuning or low-rank adaptation to improve the performance of Vision Language Models (VLMs). At their heart, SAVs select a few accurate attention heads for a task of interest and use them as classifiers, rather than relying on the model's prediction. In a similar spirit, we find that directly probing the raw activations of the VLM, in the form of scalar values, is sufficient to yield accurate classifiers on diverse visually grounded downstream tasks. Shifting focus from attention vectors to scalar activations dramatically increases the search space for accurate parameters, allowing us to find more discriminative neurons immediately from the first generated token. We call such activations Super Neurons (SNs). In this probing setting, we discover that enough SNs appear in the shallower layers of the large language model to allow for extreme early exiting from the first layer of the model at the first generated token. Compared to the original network, SNs robustly improve the classification performance while achieving a speedup of up to 5.10x.

Predicting extreme events in high-dimensional chaotic dynamical systems remains a fundamental challenge, as such events are rare, intermittent, and arise from transient dynamical mechanisms that are difficult to infer from limited observations. Accordingly, real-time forecasting calls for precursors that encode the mechanisms driving extremes, rather than relying solely on statistical associations. We propose a fully data-driven framework for long-lead prediction of extreme events that constructs interpretable, mechanism-aware precursors by explicitly tracking transient instabilities preceding event onset. The approach leverages a reduced-order formulation to compute finite-time Lyapunov exponent (FTLE)-like precursors directly from state snapshots, without requiring knowledge of the governing equations. To avoid the prohibitive computational cost of classical FTLE computation, instability growth is evaluated in an adaptively evolving low-dimensional subspace spanned by Optimal Time-Dependent (OTD) modes, enabling efficient identification of transiently amplifying directions. These precursors are then provided as input to a Transformer-based model, enabling forecast of extreme event observables. We demonstrate the framework on Kolmogorov flow, a canonical model of intermittent turbulence. The results show that explicitly encoding transient instability mechanisms substantially extends practical prediction horizons compared to baseline observable-based approaches.

Large Language Models (LLMs) have demonstrated excellent performance on Machine Translation Quality Estimation (MTQE), yet their high inference costs make them impractical for direct application. In this work, we propose applying LLMs to generate MQM-style annotations for training a COMET model: following Fernandes et al. (2023), we reckon that segment-level annotations provide a strong rationale for LLMs and are key to good segment-level QE. We propose a simplified MQM scheme, mostly restricted to top-level categories, to guide LLM selection. We present a systematic approach for the development of a GPT-4o-based prompt, called PPbMQM (Prompt-Pattern-based-MQM). We show that the resulting annotations correlate well with human annotations and that training COMET on them leads to competitive performance on segment-level QE for Chinese-English and English-German.

Large language models (LLMs) trained with canonical tokenization exhibit surprising robustness to non-canonical inputs such as character-level tokenization, yet the mechanisms underlying this robustness remain unclear. We study this phenomenon through mechanistic interpretability and identify a core process we term word recovery. We first introduce a decoding-based method to detect word recovery, showing that hidden states reconstruct canonical word-level token identities from character-level inputs. We then provide causal evidence by removing the corresponding subspace from hidden states, which consistently degrades downstream task performance. Finally, we conduct a fine-grained attention analysis and show that in-group attention among characters belonging to the same canonical token is critical for word recovery: masking such attention in early layers substantially reduces both recovery scores and task performance. Together, our findings provide a mechanistic explanation for tokenization robustness and identify word recovery as a key mechanism enabling LLMs to process character-level inputs.

Microservices are a dominant architecture in cloud computing, offering scalability and modularity, but also posing complex deployment challenges. As data centers contribute significantly to global carbon emissions, carbon-aware scheduling has emerged as a promising mitigation strategy. However, most existing solutions target batch, high-performance, or serverless workloads and assume access to global-scale infrastructure. Such an assumption does not hold for many national or regional small to medium-sized enterprises (SMEs) with microservice applications, which represent the real-world majority. In this paper, we present Aceso, an Adaptive Carbon- and Efficiency-aware placement for microservices that considers carbon, cost, and latency constraints. Aceso dynamically places microservices across geographically constrained regions using a scalable optimization strategy that leverages insight-based search space pruning techniques. Evaluation on a real-world deployment shows that Aceso quickly adapts to real-time changes in workload and carbon intensity and reduces carbon emissions by 37.4% and operational cost by 3.6%, on average, compared to a static deployment within a single country, while consistently meeting SLOs. In this way, Aceso enables carbon- and cost-aware microservice deployment for latency-sensitive applications in regionally limited infrastructures for SMEs.

Reinforcement Learning with Verifiable Rewards (RLVR) has been successfully applied to significantly boost the capabilities of pretrained large language models, especially in the math and logic problem domains. However, current research and available training datasets remain English-centric. While mul- tilingual training data and benchmarks have been created in the past, they were not created with RLVR and current model capability in mind, and their level of difficulty is often too low to provide appropriate training signals for current models. To address this gap, we provide mAceReason-Math, a dataset of high-quality translations of challenging math problems sourced from a corpus specifically curated for RLVR (AceReason-Math). We further take specific care to clean and improve our translations, resulting in a coverage of 14 languages with more than 10,000 samples per language. We release the dataset to facilitate multilingual RLVR research and benchmarking in the research community.

Heart diseases remain a leading cause of morbidity and mortality worldwide, necessitating accurate and trustworthy differential diagnosis. However, existing artificial intelligence-based diagnostic methods are often limited by insufficient cardiology knowledge, inadequate support for complex reasoning, and poor interpretability. Here we present HeartAgent, a cardiology-specific agent system designed to support a reliable and explainable differential diagnosis. HeartAgent integrates customized tools and curated data resources and orchestrates multiple specialized sub-agents to perform complex reasoning while generating transparent reasoning trajectories and verifiable supporting references. Evaluated on the MIMIC dataset and a private electronic health records cohort, HeartAgent achieved over 36% and 20% improvements over established comparative methods, in top-3 diagnostic accuracy, respectively. Additionally, clinicians assisted by HeartAgent demonstrated gains of 26.9% in diagnostic accuracy and 22.7% in explanatory quality compared with unaided experts. These results demonstrate that HeartAgent provides reliable, explainable, and clinically actionable decision support for cardiovascular care.

Wireless federated learning (FL) facilitates collaborative training of artificial intelligence (AI) models to support ubiquitous intelligent applications at the wireless edge. However, the inherent constraints of limited wireless resources inevitably lead to unreliable communication, which poses a significant challenge to wireless FL. To overcome this challenge, we propose Sign-Prioritized FL (SP-FL), a novel framework that improves wireless FL by prioritizing the transmission of important gradient information through uneven resource allocation. Specifically, recognizing the importance of descent direction in model updating, we transmit gradient signs in individual packets and allow their reuse for gradient descent if the remaining gradient modulus cannot be correctly recovered. To further improve the reliability of transmission of important information, we formulate a hierarchical resource allocation problem based on the importance disparity at both the packet and device levels, optimizing bandwidth allocation across multiple devices and power allocation between sign and modulus packets. To make the problem tractable, the one-step convergence behavior of SP-FL, which characterizes data importance at both levels in an explicit form, is analyzed. We then propose an alternating optimization algorithm to solve this problem using the Newton-Raphson method and successive convex approximation (SCA). Simulation results confirm the superiority of SP-FL, especially in resource-constrained scenarios, demonstrating up to 9.96\% higher testing accuracy on the CIFAR-10 dataset compared to existing methods.

More and more companies' Intellectual Property (IP) is being integrated into Neural Network (NN) models. This IP has considerable value for companies and, therefore, requires adequate protection. For example, an attacker might replicate a production machines' hardware and subsequently simply copy associated software and NN models onto the cloned hardware. To make copying NN models onto cloned hardware infeasible, we present an approach to bind NN models - and thus also the IP contained within them - to their underlying hardware. For this purpose, we link an NN model's weights, which are crucial for its operation, to unique and unclonable hardware properties by leveraging Physically Unclonable Functions (PUFs). By doing so, sufficient accuracy can only be achieved using the target hardware to restore the original weights, rendering proper execution of the NN model on cloned hardware impossible. We demonstrate that our approach accomplishes the desired degradation of accuracy on various NN models and outline possible future improvements.

The randomized distributed function computation (RDFC) framework, which unifies many cutting-edge distributed computation and learning applications, is considered. An autoencoder (AE) architecture is proposed to minimize the total variation distance between the probability distribution simulated by the AE outputs and an unknown target distribution, using only data samples. We illustrate significantly high RDFC performance with communication load gains from our AEs compared to data compression methods. Our designs establish deep learning-based RDFC methods and aim to facilitate the use of RDFC methods, especially when the amount of common randomness is limited and strong function computation guarantees are required.

Accurate estimation of uncertainty in deep learning is critical for deploying models in high-stakes domains such as medical diagnosis and autonomous decision-making, where overconfident predictions can lead to harmful outcomes. In practice, understanding the reason behind a model's uncertainty and the type of uncertainty it represents can support risk-aware decisions, enhance user trust, and guide additional data collection. However, many existing methods only address a single type of uncertainty or require modifications and retraining of the base model, making them difficult to adopt in real-world systems. We introduce CUPID (Comprehensive Uncertainty Plug-in estImation moDel), a general-purpose module that jointly estimates aleatoric and epistemic uncertainty without modifying or retraining the base model. CUPID can be flexibly inserted into any layer of a pretrained network. It models aleatoric uncertainty through a learned Bayesian identity mapping and captures epistemic uncertainty by analyzing the model's internal responses to structured perturbations. We evaluate CUPID across a range of tasks, including classification, regression, and out-of-distribution detection. The results show that it consistently delivers competitive performance while offering layer-wise insights into the origins of uncertainty. By making uncertainty estimation modular, interpretable, and model-agnostic, CUPID supports more transparent and trustworthy AI. Related code and data are available at https://github.com/a-Fomalhaut-a/CUPID.

Data leakage affected 294 published papers across 17 scientific fields (Kapoor & Narayanan, 2023). The dominant response has been documentation: checklists, linters, best-practice guides. Documentation does not prevent these failures. This paper proposes a structural remedy: a grammar that decomposes the supervised learning lifecycle into 7 kernel primitives connected by a typed directed acyclic graph (DAG), with four hard constraints that reject the two most damaging leakage classes at call time. The grammar's core contribution is the terminal assess constraint: a runtime-enforced evaluate/assess boundary where repeated test-set assessment is rejected by a guard on a nominally distinct Evidence type. A companion study across 2,047 experimental instances quantifies why this matters: selection leakage inflates performance by d_z = 0.93 and memorization leakage by d_z = 0.53-1.11. Three separate implementations (Python, R, and Julia) confirm the claims. The appendix specification lets anyone build a conforming version.

Deep neural networks (DNNs) have become integral to a wide range of scientific and practical applications due to their flexibility and strong predictive performance. Despite their accuracy, however, DNNs frequently exhibit poor calibration, often assigning overly confident probabilities to incorrect predictions. This limitation underscores the growing need for integrated mechanisms that provide reliable uncertainty estimation. In this article, we compare two prominent approaches for uncertainty quantification: a Bayesian approximation via Monte Carlo Dropout and the nonparametric Conformal Prediction framework. Both methods are assessed using two convolutional neural network architectures; H-CNN VGG16 and GoogLeNet, trained on the Fashion-MNIST dataset. The empirical results show that although H-CNN VGG16 attains higher predictive accuracy, it tends to exhibit pronounced overconfidence, whereas GoogLeNet yields better-calibrated uncertainty estimates. Conformal Prediction additionally demonstrates consistent validity by producing statistically guaranteed prediction sets, highlighting its practical value in high-stakes decision-making contexts. Overall, the findings emphasize the importance of evaluating model performance beyond accuracy alone and contribute to the development of more reliable and trustworthy deep learning systems.

The modern generative audio models can be used by an adversary in an unlawful manner, specifically, to impersonate other people to gain access to private information. To mitigate this issue, speech deepfake detection (SDD) methods started to evolve. Unfortunately, current SDD methods generally suffer from the lack of generalization to new audio domains and generators. More than that, they lack interpretability, especially human-like reasoning that would naturally explain the attribution of a given audio to the bona fide or spoof class and provide human-perceptible cues. In this paper, we propose HIR-SDD, a novel SDD framework that combines the strengths of Large Audio Language Models (LALMs) with the chain-of-thought reasoning derived from the novel proposed human-annotated dataset. Experimental evaluation demonstrates both the effectiveness of the proposed method and its ability to provide reasonable justifications for predictions.

Large Language Models (LLMs) rely on optimizations like Automatic Prefix Caching (APC) to accelerate inference. APC works by reusing previously computed states for the beginning part of a request (prefix), when another request starts with the same text. While APC improves throughput, it introduces timing side channels: cache hits are faster than misses, creating observable latency differences. In multi-tenant systems, attackers can exploit these differences to infer sensitive information, e.g., by incrementally reconstructing another user's request by observing hit/miss patterns. Current defenses take a sledgehammer approach: they disable APC and cache sharing, isolating users, and sacrificing efficiency for regular users. This paper presents CacheSolidarity, a system that secures multi-tenant LLM serving systems against APC side channels without sacrificing performance and efficiency. CacheSolidarity monitors cache reuse across users, flags suspicious sharing, and selectively isolates prefixes, restricting their reuse only when necessary. Evaluation shows that CacheSolidarity enables up to 70% higher cache reuse and 30% lower inference latency compared to existing defenses that isolate users. CacheSolidarity's lightweight design demonstrates how security in LLM serving does not have to come at the cost of unnecessarily reduced performance or unbearable overheads.

Traffic scene understanding from unmanned aerial vehicle (UAV) platforms is crucial for intelligent transportation systems due to its flexible deployment and wide-area monitoring capabilities. However, existing methods face significant challenges in real-world surveillance, as their heavy reliance on optical imagery leads to severe performance degradation under adverse illumination conditions like nighttime and fog. Furthermore, current Visual Question Answering (VQA) models are restricted to elementary perception tasks, lacking the domain-specific regulatory knowledge required to assess complex traffic behaviors. To address these limitations, we propose a novel Cross-spectral Traffic Cognition Network (CTCNet) for robust UAV traffic scene understanding. Specifically, we design a Prototype-Guided Knowledge Embedding (PGKE) module that leverages high-level semantic prototypes from an external Traffic Regulation Memory (TRM) to anchor domain-specific knowledge into visual representations, enabling the model to comprehend complex behaviors and distinguish fine-grained traffic violations. Moreover, we develop a Quality-Aware Spectral Compensation (QASC) module that exploits the complementary characteristics of optical and thermal modalities to perform bidirectional context exchange, effectively compensating for degraded features to ensure robust representation in complex environments. In addition, we construct Traffic-VQA, the first large-scale optical-thermal infrared benchmark for cognitive UAV traffic understanding, comprising 8,180 aligned image pairs and 1.3 million question-answer pairs across 31 diverse types. Extensive experiments demonstrate that CTCNet significantly outperforms state-of-the-art methods in both cognition and perception scenarios. The dataset is available at https://github.com/YuZhang-2004/UAV-traffic-scene-understanding.

In this paper, we investigate the learning-augmented $k$-median clustering problem, which aims to improve the performance of traditional clustering algorithms by preprocessing the point set with a predictor of error rate $α\in [0,1)$. This preprocessing step assigns potential labels to the points before clustering. We introduce an algorithm for this problem based on a simple yet effective sampling method, which substantially improves upon the time complexities of existing algorithms. Moreover, we mitigate their exponential dependency on the dimensionality of the Euclidean space. Lastly, we conduct experiments to compare our method with several state-of-the-art learning-augmented $k$-median clustering methods. The experimental results suggest that our proposed approach can significantly reduce the computational complexity in practice, while achieving a lower clustering cost.

Flow Matching enables simulation-free training of generative models on Riemannian manifolds, yet sampling typically still relies on numerically integrating a probability-flow ODE. We propose Riemannian MeanFlow (RMF), extending MeanFlow to manifold-valued generation where velocities lie in location-dependent tangent spaces. RMF defines an average-velocity field via parallel transport and derives a Riemannian MeanFlow identity that links average and instantaneous velocities for intrinsic supervision. We make this identity practical in a log-map tangent representation, avoiding trajectory simulation and heavy geometric computations. For stable optimization, we decompose the RMF objective into two terms and apply conflict-aware multi-task learning to mitigate gradient interference. RMF also supports conditional generation via classifier-free guidance. Experiments on spheres, tori, and SO(3) demonstrate competitive one-step sampling with improved quality-efficiency trade-offs and substantially reduced sampling cost.

Recent advances in generative models have amplified the risk of malicious misuse of speech synthesis technologies, enabling adversaries to impersonate target speakers and access sensitive resources. Although speech deepfake detection has progressed rapidly, most existing countermeasures lack formal robustness guarantees or fail to generalize to unseen generation techniques. We propose PV-VASM, a probabilistic framework for verifying the robustness of voice anti-spoofing models (VASMs). PV-VASM estimates the probability of misclassification under text-to-speech (TTS), voice cloning (VC), and parametric signal transformations. The approach is model-agnostic and enables robustness verification against unseen speech synthesis techniques and input perturbations. We derive a theoretical upper bound on the error probability and validate the method across diverse experimental settings, demonstrating its effectiveness as a practical robustness verification tool.

Early detection of cancer is essential for timely diagnosis and improved patient outcomes. Among emerging technologies, intra-body nanoscale communication offers an innovative solution to identify molecular cues within the human bloodstream. This study investigates a minimally invasive approach for early-stage cancer biomarker detection using nanomachines introduced into the bloodstream. To assess the feasibility of this approach, computational simulations are used to emulate the vascular environment and evaluate biomarker detection performance under different physiological conditions. Current modeling approaches often fail to capture essential vascular characteristics, including non-uniform flow structures, size-dependent particle mobility, and particle margination driven by red blood cell interactions. To address these limitations, our study incorporates these factors into the simulation framework and quantifies their individual and combined effects on biomarker detection efficiency. Baseline detection performance is first obtained under uniform flow assumptions, after which introducing realistic vascular transport mechanisms progressively reduces detection probability for all vessel types and nanomachine sizes. Among the considered vessels, capillary consistently achieves the highest detection probability across all nanomachine sizes.

Prompt highlighting steers a large language model to prioritize user-specified text spans during generation. A key challenge is extracting steering directions that capture the difference between relevant and irrelevant contexts, rather than shared structural patterns common to both. We propose PRISM-$Δ$ (Projection-based Relevance-Informed Steering Method), which decomposes the difference between positive and negative cross-covariance matrices to maximize discriminative energy while eliminating shared directions. Each attention head receives a continuous softplus importance weight, letting weak-but-useful heads contribute at reduced strength. The framework extends naturally to Value representations, capturing content-channel signal that Key-only methods leave unused. Across four benchmarks and five models, PRISM-$Δ$ matches or exceeds the best existing method on 19 of 20 configurations, with relative gains up to +10.6%, while halving the fluency cost of steering. PRISM-$Δ$ also scales to long-context retrieval, outperforming the best existing method by up to +4.8% relative gain. PRISM-$Δ$ is compatible with FlashAttention and adds negligible memory overhead.

Life sciences research depends heavily on open-source academic software, yet many tools remain underused due to practical barriers. These include installation requirements that hinder adoption and limited developer resources for software distribution and long-term maintenance. Jupyter notebooks are popular because they combine code, documentation, and results into a single executable document, enabling quick method development. However, notebooks are often fragile due to reproducibility issues in coding environments, and sharing them, especially for local execution, does not ensure others can run them successfully. LabConstrictor closes this deployment gap by bringing CI/CD-style automation to academic developers without needing DevOps expertise. Its GitHub-based pipeline checks environments and packages notebooks into one-click installable desktop applications. After installation, users access a unified start page with documentation, links to the packaged notebooks, and version checks. Code cells can be hidden by default, and run-cell controls combined with widgets provide an app-like experience. By simplifying the distribution, installation, and sharing of open-source software, LabConstrictor allows faster access to new computational methods and promotes routine reuse across labs.

In target speaker extraction (TSE), we aim to recover target speech from a multi-talker mixture using a short enrollment utterance as reference. Recent studies on diffusion and flow-matching generators have improved target-speech fidelity. However, multi-step sampling increases latency, and one-step solutions often rely on a mixture-dependent time coordinate that can be unreliable for real-world conversations. We present AlphaFlowTSE, a one-step conditional generative model trained with a Jacobian-vector product (JVP)-free AlphaFlow objective. AlphaFlowTSE learns mean-velocity transport along a mixture-to-target trajectory starting from the observed mixture, eliminating auxiliary mixing-ratio prediction, and stabilizes training by combining flow matching with an interval-consistency teacher-student target. Experiments on Libri2Mix and REAL-T confirm that AlphaFlowTSE improves target-speaker similarity and real-mixture generalization for downstream automatic speech recognition (ASR).

Retrieval-Augmented Generation (RAG) systems typically treat documents as flat text, ignoring the structured metadata and linked relationships that knowledge graphs provide. In this paper, we investigate whether structured linked data, specifically Schema.org markup and dereferenceable entity pages served by a Linked Data Platform, can improve retrieval accuracy and answer quality in both standard and agentic RAG systems. We conduct a controlled experiment across four domains (editorial, legal, travel, e-commerce) using Vertex AI Vector Search 2.0 for retrieval and the Google Agent Development Kit (ADK) for agentic reasoning. Our experimental design tests seven conditions: three document representations (plain HTML, HTML with JSON-LD, and an enhanced agentic-optimized entity page) crossed with two retrieval modes (standard RAG and agentic RAG with multi-hop link traversal), plus an Enhanced+ condition that adds rich navigational affordances and entity interlinking. Our results reveal that while JSON-LD markup alone provides only modest improvements, our enhanced entity page format, incorporating llms.txt-style agent instructions, breadcrumbs, and neural search capabilities, achieves substantial gains: +29.6% accuracy improvement for standard RAG and +29.8% for the full agentic pipeline. The Enhanced+ variant, with richer navigational affordances, achieves the highest absolute scores (accuracy: 4.85/5, completeness: 4.55/5), though the incremental gain over the base enhanced format is not statistically significant. We release our dataset, evaluation framework, and enhanced entity page templates to support reproducibility.

Neural text-to-SQL models, which translate natural language questions (NLQs) into SQL queries given a database schema, have achieved remarkable performance. However, database schemas frequently evolve to meet new requirements. Such schema evolution often leads to performance degradation for models trained on static schemas. Existing work either mainly focuses on simply paraphrasing some syntactic or semantic mappings among NLQ, DB and SQL, or lacks a comprehensive and controllable way to investigate the model robustness issue under the schema evolution, which is insufficient when facing the increasingly complex and rich database schema changes in reality, especially in the LLM era. To address the challenges posed by schema evolution, we present EvoSchema, a comprehensive benchmark designed to assess and enhance the robustness of text-to-SQL systems under real-world schema changes. EvoSchema introduces a novel schema evolution taxonomy, encompassing ten perturbation types across columnlevel and table-level modifications, systematically simulating the dynamic nature of database schemas. Through EvoSchema, we conduct an in-depth evaluation spanning different open source and closed-source LLMs, revealing that table-level perturbations have a significantly greater impact on model performance compared to column-level changes. Furthermore, EvoSchema inspires the development of more resilient text-to-SQL systems, in terms of both model training and database design. The models trained on EvoSchema's diverse schema designs can force the model to distinguish the schema difference for the same questions to avoid learning spurious patterns, which demonstrate remarkable robustness compared to those trained on unperturbed data on average. This benchmark offers valuable insights into model behavior and a path forward for designing systems capable of thriving in dynamic, real-world environments.

Being trained on large and diverse datasets, visual foundation models (VFMs) can be fine-tuned to achieve remarkable performance and efficiency in various downstream computer vision tasks. The high computational cost of data collection and training makes these models valuable assets, which motivates some VFM owners to distribute them alongside a license to protect their intellectual property rights. In this paper, we propose an approach to ownership verification of visual foundation models that leverages a small encoder-decoder network to embed digital watermarks into an internal representation of a hold-out set of input images. The method is based on random watermark embedding, which makes the watermark statistics detectable in functional copies of the watermarked model. Both theoretically and experimentally, we demonstrate that the proposed method yields a low probability of false detection for non-watermarked models and a low probability of false misdetection for watermarked models.

While Secure Aggregation (SA) protects update confidentiality in Cross-silo Federated Learning, it fails to guarantee aggregation integrity, allowing malicious servers to silently omit or tamper with updates. Existing verifiable aggregation schemes rely on heavyweight cryptography (e.g., ZKPs, HE), incurring computational costs that scale poorly with model size. In this paper, we propose a lightweight architecture that shifts from extrinsic cryptographic proofs to \textit{Intrinsic Proofs}. We repurpose backdoor injection to embed verification signals directly into model parameters. By harnessing Catastrophic Forgetting, these signals are robust for immediate verification yet ephemeral, naturally decaying to preserve final model utility. We design a randomized, single-verifier auditing framework compatible with SA, ensuring client anonymity and preventing signal collision without trusted third parties. Experiments on SVHN, CIFAR-10, and CIFAR-100 demonstrate high detection probabilities against malicious servers. Notably, our approach achieves over $1000\times$ speedup on ResNet-18 compared to cryptographic baselines, effectively scaling to large models.

Black-box adversarial attacks are widely used as tools to test the robustness of deep neural networks against malicious perturbations of input data aimed at a specific change in the output of the model. Such methods, although they remain empirically effective, usually do not guarantee that an adversarial example can be found for a particular model. In this paper, we propose Contract And Conquer (CAC), an approach to provably compute adversarial examples for neural networks in a black-box manner. The method is based on knowledge distillation of a black-box model on an expanding distillation dataset and precise contraction of the adversarial example search space. CAC is supported by the transferability guarantee: we prove that the method yields an adversarial example for the black-box model within a fixed number of algorithm iterations. Experimentally, we demonstrate that the proposed approach outperforms existing state-of-the-art black-box attack methods on ImageNet dataset for different target models, including vision transformers.

Digital Human Modelling (DHM) is increasingly shaped by advances in AI, wearable biosensing, and interactive digital environments, particularly in research addressing accessibility and inclusion. However, many AI-enabled DHM approaches remain tightly coupled to specific platforms, tasks, or interpretative pipelines, limiting reproducibility, scalability, and ethical reuse. This paper presents a platform-agnostic DHM framework designed to support AI-ready multimodal interaction research by explicitly separating sensing, interaction modelling, and inference readiness. The framework integrates the OpenBCI Galea headset as a unified multimodal sensing layer, providing concurrent EEG, EMG, EOG, PPG, and inertial data streams, alongside a reproducible, game-based interaction environment implemented using SuperTux. Rather than embedding AI models or behavioural inference, physiological signals are represented as structured, temporally aligned observables, enabling downstream AI methods to be applied under appropriate ethical approval. Interaction is modelled using computational task primitives and timestamped event markers, supporting consistent alignment across heterogeneous sensors and platforms. Technical verification via author self-instrumentation confirms data integrity, stream continuity, and synchronisation; no human-subjects evaluation or AI inference is reported. Scalability considerations are discussed with respect to data throughput, latency, and extension to additional sensors or interaction modalities. Illustrative use cases demonstrate how the framework can support AI-enabled DHM and HCI studies, including accessibility-oriented interaction design and adaptive systems research, without requiring architectural modifications. The proposed framework provides an emerging-technology-focused infrastructure for future ethics-approved, inclusive DHM research.

Agile software development evolves so rapidly that research struggles to remain timely and transferable - an issue heightened by the swift adoption of generative AI and agentic tools. Earlier discussions highlight theory and time gaps, leading to results that often lack clear reuse conditions or arrive too late for practical decisions. This paper introduces a project-based, AI-integrated agile education platform as a collaborative research environment, positioned between controlled studies and real-world industry. The platform enables rapid inquiry through sprint rhythms, quality gates, and genuine stakeholder involvement. We present a framework specifying iteration structures, recurring events, and quality gates for AI-assisted engineering artifacts. Early results from several semesters - covering project pipeline, cohort growth, and stakeholder participation - show the platform's potential to generate practice-relevant evidence efficiently and with reusable context. Finally, we outline future steps to enhance governance and evidence capture.

Magnetohydrodynamic (MHD) effects play a key role in the design and operation of nuclear fusion systems, where electrically conducting fluids (such as liquid metals or molten salts in reactor blankets) interact with magnetic fields of varying intensity and orientation, which affect the resulting flow. The numerical resolution of MHD models involves highly nonlinear multiphysics systems of equations and can become computationally expensive, particularly in multi-query, parametric, or real-time contexts. This work investigates a fully data-driven framework for MHD state reconstruction that combines dimensionality reduction via Singular Value Decomposition (SVD) with the SHallow REcurrent Decoder (SHRED), a neural network architecture designed to recover the full spatio-temporal state from sparse time-series measurements of a limited number of observables. The methodology is applied to a parametric MHD test case involving compressible lead-lithium flow in a stepped channel subjected to thermal gradients and magnetic fields spanning a broad range of intensities. To improve efficiency, the full-order dataset is first compressed using SVD, yielding a reduced representation used as reference truth for training. Only temperature measurements from three sensors are provided as input, while the network reconstructs the full fields of velocity, pressure, and temperature. To assess robustness with respect to sensor placement, thirty randomly generated sensor configurations are tested in ensemble mode. Results show that SHRED accurately reconstructs the full MHD state even for magnetic field intensities not included in the training set. These findings demonstrate the potential of SHRED as a computationally efficient surrogate modeling strategy for fusion-relevant multiphysics problems, enabling low-cost state estimation with possible applications in real-time monitoring and control.

Clinical diagnosis is a complex cognitive process, grounded in dynamic cue acquisition and continuous expertise accumulation. Yet most current artificial intelligence (AI) systems are misaligned with this reality, treating diagnosis as single-pass retrospective prediction while lacking auditable mechanisms for governed improvement. We developed DxEvolve, a self-evolving diagnostic agent that bridges these gaps through an interactive deep clinical research workflow. The framework autonomously requisitions examinations and continually externalizes clinical experience from increasing encounter exposure as diagnostic cognition primitives. On the MIMIC-CDM benchmark, DxEvolve improved diagnostic accuracy by 11.2% on average over backbone models and reached 90.4% on a reader-study subset, comparable to the clinician reference (88.8%). DxEvolve improved accuracy on an independent external cohort by 10.2% (categories covered by the source cohort) and 17.1% (uncovered categories) compared to the competitive method. By transforming experience into a governable learning asset, DxEvolve supports an accountable pathway for the continual evolution of clinical AI.

Industrial Control Systems (ICS) underpin critical infrastructure and face growing cyber-physical threats due to the convergence of operational technology and networked environments. While machine learning-based anomaly detection approaches in ICS shows strong theoretical performance, deployment is often limited by poor explainability, high false-positive rates, and sensitivity to evolving system behavior, i.e., baseline drifting. We propose a Spatio-Temporal Attention Graph Neural Network (STA-GNN) for unsupervised and explainable anomaly detection in ICS that models both temporal dynamics and relational structure of the system. Sensors, controllers, and network entities are represented as nodes in a dynamically learned graph, enabling the model to capture inter-dependencies across physical processes and communication patterns. Attention mechanisms provide influential relationships, supporting inspection of correlations and potential causal pathways behind detected events. The approach supports multiple data modalities, including SCADA point measurements, network flow features, and payload features, and thus enables unified cyber-physical analysis. To address operational requirements, we incorporate a conformal prediction strategy to control false alarm rates and monitor performance degradation under drifting of the environment. Our findings highlight the possibilities and limitations of model evaluation and common pitfalls in anomaly detection in ICS. Our findings emphasise the importance of explainable, drift-aware evaluation for reliable deployment of learning-based security monitoring systems.

Recently, progress has been made on the Intra Pattern Copy (IPC) tool for JPEG XS, an image compression standard designed for low-latency and low-complexity coding. IPC performs wavelet-domain intra compensation predictions to reduce spatial redundancy in screen content. A key module of IPC is the displacement vector (DV) search, which aims to solve the optimal prediction reference offset. However, the DV search process is computationally intensive, posing challenges for practical hardware deployment. In this paper, we propose an efficient pipelined FPGA architecture design for the DV search module to promote the practical deployment of IPC. Optimized memory organization, which leverages the IPC computational characteristics and data inherent reuse patterns, is further introduced to enhance the performance. Experimental results show that our proposed architecture achieves a throughput of 38.3 Mpixels/s with a power consumption of 277 mW, demonstrating its feasibility for practical hardware implementation in IPC and other predictive coding tools, and providing a promising foundation for ASIC deployment.

We propose FAME (Formal Abstract Minimal Explanations), a new class of abductive explanations grounded in abstract interpretation. FAME is the first method to scale to large neural networks while reducing explanation size. Our main contribution is the design of dedicated perturbation domains that eliminate the need for traversal order. FAME progressively shrinks these domains and leverages LiRPA-based bounds to discard irrelevant features, ultimately converging to a formal abstract minimal explanation. To assess explanation quality, we introduce a procedure that measures the worst-case distance between an abstract minimal explanation and a true minimal explanation. This procedure combines adversarial attacks with an optional VERIX+ refinement step. We benchmark FAME against VERIX+ and demonstrate consistent gains in both explanation size and runtime on medium- to large-scale neural networks.

In real-world deployment, vision-language models often encounter disturbances such as weather, occlusion, and camera motion. Under such conditions, their understanding and reasoning degrade substantially, revealing a gap between clean, controlled (i.e., unperturbed) evaluation settings and real-world robustness. To address this limitation, we propose ROVA, a novel training framework that improves robustness by modeling a robustness-aware consistency reward under spatio-temporal corruptions. ROVA introduces a difficulty-aware online training strategy that prioritizes informative samples based on the model's evolving capability. Specifically, it continuously re-estimates sample difficulty via self-reflective evaluation, enabling adaptive training with a robustness-aware consistency reward. We also introduce PVRBench, a new benchmark that injects real-world perturbations into embodied video datasets to assess both accuracy and reasoning quality under realistic disturbances. We evaluate ROVA and baselines on PVRBench, UrbanVideo, and VisBench, where open-source and proprietary models suffer up to 35% and 28% drops in accuracy and reasoning under realistic perturbations. ROVA effectively mitigates performance degradation, boosting relative accuracy by at least 24% and reasoning by over 9% compared with baseline models (QWen2.5/3-VL, InternVL2.5, Embodied-R). These gains transfer to clean standard benchmarks, yielding consistent improvements.

Task and Motion Planning combines high-level task sequencing (what to do) with low-level motion planning (how to do it) to generate feasible, collision-free execution plans. However, in many real-world domains, such as automated warehouses, tasks are predefined, shifting the challenge to if, when, and how to execute them safely and efficiently under resource, time and motion constraints. In this paper, we formalize this as the Scheduling and Motion Planning problem for multi-object navigation in shared workspaces. We propose a novel solution framework that interleaves off-the-shelf schedulers and motion planners in an incremental learning loop. The scheduler generates candidate plans, while the motion planner checks feasibility and returns symbolic feedback, i.e., spatial conflicts and timing adjustments, to guide the scheduler towards motion-feasible solutions. We validate our proposal on logistics and job-shop scheduling benchmarks augmented with motion tasks, using state-of-the-art schedulers and sampling-based motion planners. Our results show the effectiveness of our framework in generating valid plans under complex temporal and spatial constraints, where synchronized motion is critical.

Environmental, Social, and Governance (ESG) standards have been increasingly adopted by organizations to demonstrate accountability towards ethical, social, and sustainability goals. However, generating ESG reports that align with these standards remains challenging due to unstructured data formats, inconsistent terminology, and complex requirements. Existing ESG lifecycles provide guidance for structuring ESG reports but lack the automation, adaptability, and continuous feedback mechanisms needed to address these challenges. To bridge this gap, we introduce an agentic ESG lifecycle framework that systematically integrates the ESG stages of identification, measurement, reporting, engagement, and improvement. In this framework, multiple AI agents extract ESG information, verify ESG performance, and update ESG reports based on organisational outcomes. By embedding agentic components within the ESG lifecycle, the proposed framework transforms ESG from a static reporting process into a dynamic, accountable, and adaptive system for sustainability governance. We further define the technical requirements and quality attributes needed to support four main ESG tasks, such as report validation, multi-report comparison, report generation, and knowledge-base maintenance, and propose three architectural approaches, namely single-model, single-agent, and multi-agent, for addressing these tasks. The source code and data for the prototype of these approaches are available at https://gitlab.com/for_peer_review-group/esg_assistant.

Machine learning backdoors have the property that the machine learning model should work as expected on normal inputs, but when the input contains a specific $\textit{trigger}$, it behaves as the attacker desires. Detecting such triggers has been proven to be extremely difficult. In this paper, we present a novel and explainable approach to detect and eliminate such backdoor triggers based on active paths found in neural networks. We present promising experimental evidence of our approach, which involves injecting backdoors into a machine learning model used for intrusion detection.

This paper presents a research program dedicated to evaluating and advancing the reasoning capabilities of Bielik, a Polish large language model. The study describes a number of stages of work: initial benchmarking and creation of evaluation methodology, analyzing of comparative results with other LLMs and outlining of future prospects that take into account the limitations of the analyses conducted so far and aims to keep Bielik in the race give the ever-changing -- and competitive -- AI landscape.

In high-performance computing (HPC) applications, FP64 arithmetic remains indispensable for ensuring numerical accuracy and stability. However, in recent hardware generations, improvements in FP64 arithmetic performance have been relatively modest. Consequently, achieving sustained performance gains for FP64 computations necessitates the effective utilization of high-throughput low-precision arithmetic, such as INT8 and FP8. In several recent architectures, such as NVIDIA Blackwell Ultra and NVIDIA Rubin, INT8 performance has been significantly reduced, making reliance on INT8 alone insufficient. The use of FP8 arithmetic is thus increasingly important. In this paper, we propose a method for emulating double-precision (FP64) general matrix--matrix multiplication (DGEMM), a fundamental and performance-critical kernel in many HPC applications, using FP8 matrix multiply-accumulate (MMA) units. The Ozaki-I and Ozaki-II schemes are well established as foundational approaches for emulating DGEMM via low-precision arithmetic. For DGEMM emulation via the Ozaki-I scheme, implementations using INT8, FP8, and FP16 MMA units have been proposed, all of which can be realized based on the same underlying algorithmic structure. In contrast, although implementations of DGEMM emulation via the Ozaki-II scheme using INT8 MMA units have been reported, the original algorithm cannot be directly adapted to exploit FP8 MMA units. In this work, we introduce a novel technique to overcome this limitation and demonstrate FP64 matrix multiplication emulation based on the Ozaki-II scheme that operates on FP8 MMA units. Compared to FP8-based emulation via the Ozaki-I scheme, our method significantly reduces the number of required FP8 matrix multiplications and enables efficient FP64 emulation on emerging GPU architectures.

Reinforcement Learning with Verifiable Rewards (RLVR) has proven effective in enhancing the reasoning capabilities of large language models, particularly in domains such as mathematics where reliable rule-based verifiers can be constructed. However, the reliance on handcrafted, domain-specific verification rules substantially limits the applicability of RLVR to general reasoning domains with free-form answers, where valid answers often exhibit significant variability, making it difficult to establish complete and accurate rules. To address this limitation, we propose Conditional Expectation Reward (CER), which leverages the large language model itself as an implicit verifier, and is therefore applicable to general domains and eliminates the need for external verifiers or auxiliary models. CER is defined as the expected likelihood of generating the reference answer conditioned on the generated answer. In contrast to rule-based verifiers that yield binary feedback, CER provides a soft, graded reward signal that reflects varying degrees of correctness, making it better suited to tasks where answers vary in correctness. Experimental results demonstrate that CER is effective across a wide range of reasoning tasks, spanning both mathematical and general domains, indicating that CER serves as a flexible and general verification mechanism. The code is available at https://github.com/changyi7231/CER.

Environmental sound understanding in computational auditory scene analysis (CASA) is often formulated as an audio-only recognition problem. This formulation leaves a persistent drawback in multi-label audio tagging (AT): acoustic similarity can make certain events difficult to separate from waveforms alone. In such cases, disambiguating cues often lie outside the waveform. Geospatial semantic context (GSC), derived from geographic information system data, e.g., points of interest (POI), provides location-tied environmental priors that can help reduce this ambiguity. A systematic study of this direction is enabled through the proposed geospatial audio tagging (Geo-AT) task, which conditions multi-label sound event tagging on GSC alongside audio. To benchmark Geo-AT, Geo-ATBench is introduced as a polyphonic audio benchmark with geographical annotations, containing 10.71 hours of audio across 28 event categories; each clip is paired with a GSC representation from 11 semantic context categories. GeoFusion-AT is proposed as a unified geo-audio fusion framework that evaluates feature-, representation-, and decision-level fusion on representative audio backbones, with audio- and GSC-only baselines. Results show that incorporating GSC improves AT performance, especially on acoustically confounded labels, indicating geospatial semantics provide effective priors beyond audio alone. A crowdsourced listening study with 10 participants on 579 samples shows that there is no significant difference in performance between models on Geo-ATBench labels and aggregated human labels, supporting Geo-ATBench as a human-aligned benchmark. The Geo-AT task, benchmark Geo-ATBench, and reproducible geo-audio fusion framework GeoFusion-AT provide a foundation for studying AT with geospatial semantic context within the CASA community. Dataset, code, models are on homepage (https://github.com/WuYanru2002/Geo-ATBench).

The first edition of the QuantumX track, held within the XXIX Jornadas de Ingeniería del Software y Bases de Datos (JISBD 2025), brought together leading Spanish research groups working at the intersection of Quantum Computing and Software Engineering. The event served as a pioneering forum to explore how principles of software quality, governance, testing, orchestration, and abstraction can be adapted to the quantum paradigm. The presented works spanned diverse areas (from quantum service engineering and hybrid architectures to quality models, circuit optimization, and quantum machine learning), reflecting the interdisciplinary nature and growing maturity of Quantum Computing and Quantum Software Engineering. The track also fostered community building and collaboration through the presentation of national and Ibero-American research networks such as RIPAISC and QSpain, and through dedicated networking sessions that encouraged joint initiatives. Beyond reporting on the event, this article provides a structured synthesis of the contributions presented at QuantumX, identifies common research themes and engineering concerns, and outlines a set of open challenges and future directions for the advancement of Quantum Software Engineering. This first QuantumX track established the foundation for a sustained research community and positioned Spain as an emerging contributor to the European and global quantum software ecosystem.

The recent advancement of large language models has spurred a growing trend of integrating pre-trained language model (PLM) embeddings into topic models, fundamentally reshaping how topics capture semantic structure. Classical models such as Latent Dirichlet Allocation (LDA) derive topics from word co-occurrence statistics, whereas PLM-augmented models anchor these statistics to pre-trained embedding spaces, imposing a prior that also favours clustering of semantically similar words. This structural difference can be captured by the psycholinguistic dimensions of thematic relatedness and taxonomic similarity of the topic words. To disentangle these dimensions in topic models, we construct a large synthetic benchmark of word pairs using LLM-based annotation to train a neural scoring function. We apply this scorer to a comprehensive evaluation across multiple corpora and topic model families, revealing that different model families capture distinct semantic structure in their topics. We further demonstrate that similarity and relatedness scores successfully predict downstream task performance depending on task requirements. This paper establishes similarity and relatedness as essential axes for topic model evaluation and provides a reliable pipeline for characterising these across model families and corpora.

The goal of news image captioning is to generate captions by integrating news article content with corresponding images, highlighting the relationship between textual context and visual elements. The majority of research on news image captioning focuses on English, primarily because datasets in other languages are scarce. To address this limitation, we create the first multilingual news image captioning benchmark, MUNIChus, comprising 9 languages, including several low-resource languages such as Sinhala and Urdu. We evaluate various state-of-the-art neural news image captioning models on MUNIChus and find that news image captioning remains challenging. We also make MUNIChus publicly available with over 20 models already benchmarked. MUNIChus opens new avenues for further advancements in developing and evaluating multilingual news image captioning models.

LLM-powered agents face a persistent challenge: learning from their execution experiences to improve future performance. While agents can successfully complete many tasks, they often repeat inefficient patterns, fail to recover from similar errors, and miss opportunities to apply successful strategies from past executions. We present a novel framework for automatically extracting actionable learnings from agent execution trajectories and utilizing them to improve future performance through contextual memory retrieval. Our approach comprises four components: (1) a Trajectory Intelligence Extractor that performs semantic analysis of agent reasoning patterns, (2) a Decision Attribution Analyzer that identifies which decisions and reasoning steps led to failures, recoveries, or inefficiencies, (3) a Contextual Learning Generator that produces three types of guidance -- strategy tips from successful patterns, recovery tips from failure handling, and optimization tips from inefficient but successful executions, and (4) an Adaptive Memory Retrieval System that injects relevant learnings into agent prompts based on multi-dimensional similarity. Unlike existing memory systems that store generic conversational facts, our framework understands execution patterns, extracts structured learnings with provenance, and retrieves guidance tailored to specific task contexts. Evaluation on the AppWorld benchmark demonstrates consistent improvements, with up to 14.3 percentage point gains in scenario goal completion on held-out tasks and particularly strong benefits on complex tasks (28.5~pp scenario goal improvement, a 149\% relative increase).

We present a JAX implementation of the Self-Scaled Broyden family of quasi-Newton methods, fully compatible with JAX and building on the Optimistix~\cite{rader_optimistix_2024} optimisation library. The implementation includes BFGS, DFP, Broyden and their Self-Scaled variants(SSBFGS, SSDFP, SSBroyden), together with a Zoom line search satisfying the strong Wolfe conditions. This is a short technical note, not a research paper, as it does not claim any novel contribution; its purpose is to document the implementation and ease the adoption of these optimisers within the JAX community. The code is available at https://github.com/IvanBioli/ssbroyden_optimistix.git.

Trajectory prediction is critical for autonomous driving, enabling safe and efficient planning in dense, dynamic traffic. Most existing methods optimize prediction accuracy under fixed-length observations. However, real-world driving often yields variable-length, incomplete observations, posing a challenge to these methods. A common strategy is to directly map features from incomplete observations to those from complete ones. This one-shot mapping, however, struggles to learn accurate representations for short trajectories due to significant information gaps. To address this issue, we propose a Progressive Retrospective Framework (PRF), which gradually aligns features from incomplete observations with those from complete ones via a cascade of retrospective units. Each unit consists of a Retrospective Distillation Module (RDM) and a Retrospective Prediction Module (RPM), where RDM distills features and RPM recovers previous timesteps using the distilled features. Moreover, we propose a Rolling-Start Training Strategy (RSTS) that enhances data efficiency during PRF training. PRF is plug-and-play with existing methods. Extensive experiments on datasets Argoverse 2 and Argoverse 1 demonstrate the effectiveness of PRF. Code is available at https://github.com/zhouhao94/PRF.

We reveal a precise mathematical framework about a new family of generative models which we call Gradient Flow Drifting. With this framework, we prove an equivalence between the recently proposed Drifting Model and the Wasserstein gradient flow of the forward KL divergence under kernel density estimation (KDE) approximation. Specifically, we prove that the drifting field of drifting model (arXiv:2602.04770) equals, up to a bandwidth-squared scaling factor, the difference of KDE log-density gradients $\nabla \log p_{\mathrm{kde}} - \nabla \log q_{\mathrm{kde}}$, which is exactly the particle velocity field of the Wasserstein-2 gradient flow of $KL(q\|p)$ with KDE-approximated densities. Besides that, this broad family of generative models can also include MMD-based generators, which arises as special cases of Wasserstein gradient flows of different divergences under KDE approximation. We provide a concise identifiability proof, and a theoretically grounded mixed-divergence strategy. We combine reverse KL and $χ^2$ divergence gradient flows to simultaneously avoid mode collapse and mode blurring, and extend this method onto Riemannian manifold which loosens the constraints on the kernel function, and makes this method more suitable for the semantic space. Preliminary experiments on synthetic benchmarks validate the framework.

Reinforcement learning with verifiable rewards (RLVR) has achieved remarkable success in logical reasoning tasks, yet whether large language model (LLM) alignment requires fundamentally different approaches remains unclear. Given the apparent tolerance for multiple valid responses in moral reasoning, a natural hypothesis is that alignment tasks inherently require diversity-seeking distribution-matching algorithms rather than reward-maximizing policy-based methods. We conduct the first comprehensive empirical study comparing both paradigms on MoReBench. To enable stable RLVR training, we build a rubric-grounded reward pipeline by training a Qwen3-1.7B judge model. Contrary to our hypothesis, we find that distribution-matching approaches do not demonstrate significant advantages over reward-maximizing methods as expected on alignment tasks. Through semantic visualization mapping high-reward responses to semantic space, we demonstrate that moral reasoning exhibits more concentrated high-reward distributions than mathematical reasoning, where diverse solution strategies yield similarly high rewards. This counter-intuitive finding explains why mode-seeking optimization proves equally or more effective for alignment tasks. Our results suggest that alignment tasks do not inherently require diversity-preserving algorithms, and standard reward-maximizing RLVR methods can effectively transfer to moral reasoning without explicit diversity mechanisms.

Ensembling is commonly used in machine learning on tabular data to boost predictive performance and robustness, but larger ensembles often lead to increased hardware demand. We introduce HAPEns, a post-hoc ensembling method that explicitly balances accuracy against hardware efficiency. Inspired by multi-objective and quality diversity optimization, HAPEns constructs a diverse set of ensembles along the Pareto front of predictive performance and resource usage. Existing hardware-aware post-hoc ensembling baselines are not available, highlighting the novelty of our approach. Experiments on 83 tabular classification datasets show that HAPEns significantly outperforms baselines, finding superior trade-offs for ensemble performance and deployment cost. Ablation studies also reveal that memory usage is a particularly effective objective metric. Further, we show that even a greedy ensembling algorithm can be significantly improved in this task with a static multi-objective weighting scheme.

Computer-Use Agents (CUAs) are emerging as a new paradigm in human-computer interaction, enabling autonomous execution of tasks in desktop environment by perceiving high-level natural-language instructions. As such agents become increasingly capable and are deployed across diverse desktop environments, evaluating their behavior in a scalable and reliable manner becomes a critical challenge. Existing evaluation pipelines rely on static benchmarks, rule-based success checks, or manual inspection, which are brittle, costly, and poorly aligned with real-world usage. In this work, we study Vision-Language Models (VLMs) as autonomous auditors for assessing CUA task completion directly from observable interactions and conduct a large-scale meta-evaluation of five VLMs that judge task success given a natural-language instruction and the final environment state. Our evaluation spans three widely used CUA benchmarks across macOS, Windows, and Linux environments and analyzes auditor behavior along three complementary dimensions: accuracy, calibration of confidence estimates, and inter-model agreement. We find that while state-of-the-art VLMs achieve strong accuracy and calibration, all auditors exhibit notable performance degradation in more complex or heterogeneous environments, and even high-performing models show significant disagreement in their judgments. These results expose fundamental limitations of current model-based auditing approaches and highlight the need to explicitly account for evaluator reliability, uncertainty, and variance when deploying autonomous CUAs in real-world settings.

In-context learning (ICL) allows Transformers to adapt to novel tasks without weight updates, yet the underlying algorithms remain poorly understood. We adopt a statistical decision-theoretic perspective by investigating simple binary hypothesis testing, where the optimal policy is determined by the likelihood-ratio test. Notably, this setup provides a mathematically rigorous setting for mechanistic interpretability where the target algorithmic ground truth is known. By training Transformers on tasks requiring distinct geometries (linear shifted means vs. nonlinear variance estimation), we demonstrate that the models approximate the Bayes-optimal sufficient statistics from context up to some monotonic transformation, matching the performance of an ideal oracle estimator in nonlinear regimes. Leveraging this analytical ground truth, mechanistic analysis via logit lens and circuit alignment suggests that the model does not rely on a fixed kernel smoothing heuristic. Instead, it appears to adapt the point at which decisions become linearly decodable: exhibiting patterns consistent with a voting-style ensemble for linear tasks while utilizing a deeper sequential computation for nonlinear tasks. These findings suggest that ICL emerges from the construction of task-adaptive statistical estimators rather than simple similarity matching.

Large language models (LLMs) combined with retrieval augmented generation have enabled the deployment of domain-specific chatbots, but these systems remain prone to generating unsupported or incorrect answers. Reliable evaluation is therefore critical, yet manual review is costly and existing frameworks often depend on curated test sets and static metrics, limiting scalability. We propose an end-to-end automatic evaluator designed to substantially reduce human effort. Our system generates Q\&A pairs directly from the underlying knowledge base, uses LLMs to judge chatbot responses against reference answers, and applies confidence-based filtering to highlight uncertain cases. Applied to a Vietnamese news dataset, the evaluator achieves high agreement with human judgments while significantly lowering review overhead. The framework is modular and language-agnostic, making it readily adaptable to diverse domains. This work introduces a practical, scalable solution for evaluating chatbots with minimal reliance on manual intervention.

The integration of Generative AI models into AI-native network systems offers a transformative path toward achieving autonomous and adaptive control. However, the application of such models to continuous control tasks is impeded by intrinsic architectural limitations, including finite context windows, the lack of explicit reward signals, and the degradation of the long context. This paper posits that the key to unlocking robust continuous control is enabling agents to internalize experience by distilling it into their parameters, rather than relying on prompt-based memory. To this end, we propose a novel self-finetuning framework that enables agentic systems to learn continuously through direct interaction with the environment, bypassing the need for handcrafted rewards. Our framework implements a bi-perspective reflection mechanism that generates autonomous linguistic feedback to construct preference datasets from interaction history. A subsequent preference-based fine-tuning process distills long-horizon experiences into the model's parameters. We evaluate our approach on a dynamic Radio Access Network (RAN) slicing task, a challenging multi-objective control problem that requires the resolution of acute trade-offs between spectrum efficiency, service quality, and reconfiguration stability under volatile network conditions. Experimental results show that our framework outperforms standard Reinforcement Learning (RL) baselines and existing Large Language Model (LLM)-based agents in sample efficiency, stability, and multi-metric optimization. These findings demonstrate the potential of self-improving generative agents for continuous control tasks, paving the way for future AI-native network infrastructure.

This paper addresses the challenge of generating synthetic electroencephalogram (EEG) covariance matrices for motor imagery brain-computer interface (MI-BCI) applications. Objective: We aim to develop a generative model capable of producing high-fidelity synthetic covariance matrices while preserving their symmetric positive-definite nature. Approach: We propose a Riemannian geometry-preserving variational autoencoder (RGP-VAE) integrating geometric mappings with a composite loss function combining Riemannian distance, tangent space reconstruction accuracy and generative diversity. Results: The model generates valid, representative EEG covariance matrices, while learning a subject-invariant latent space. Synthetic data proves practically useful for MI-BCI, with its impact depending on the paired classifier. Contribution: This work introduces and validates the RGP-VAE as a geometry-preserving generative model for EEG covariance matrices, highlighting its potential for signal privacy, scalability and data augmentation.

Monotone operator equilibrium networks are implicit-layer models whose output is the unique equilibrium of a monotone operator, guaranteeing existence, uniqueness, and convergence. When deployed on low-precision hardware, weights are quantized, potentially destroying these guarantees. We analyze weight quantization as a spectral perturbation of the underlying monotone inclusion. Convergence of the quantized solver is guaranteed whenever the spectral-norm weight perturbation is smaller than the monotonicity margin; the displacement between quantized and full-precision equilibria is bounded in terms of the perturbation size and margin; and a condition number characterizing the ratio of the operator norm to the margin links quantization precision to forward error. MNIST experiments confirm a phase transition at the predicted threshold: three- and four-bit post-training quantization diverge, while five-bit and above converge. The backward-pass guarantee enables quantization-aware training, which recovers provable convergence at four bits.

This paper studies cross-market return predictability through a machine learning framework that preserves economic structure. Exploiting the non-overlapping trading hours of the U.S. and Chinese equity markets, we construct a directed bipartite graph that captures time-ordered predictive linkages between stocks across markets. Edges are selected via rolling-window hypothesis testing, and the resulting graph serves as a sparse, economically interpretable feature-selection layer for downstream machine learning models. We apply a range of regularized and ensemble methods to forecast open-to-close returns using lagged foreign-market information. Our results reveal a pronounced directional asymmetry: U.S. previous-close-to-close returns contain substantial predictive information for Chinese intraday returns, whereas the reverse effect is limited. This informational asymmetry translates into economically meaningful performance differences and highlights how structured machine learning frameworks can uncover cross-market dependencies while maintaining interpretability.

Static Application Security Testing (SAST) tools play a vital role in modern software development by automatically detecting potential vulnerabilities in source code. However, their effectiveness is often limited by a high rate of false positives, which wastes developer's effort and undermines trust in automated analysis. This work presents a Graph Convolutional Network (GCN) model designed to predict SAST reports as true and false positive. The model leverages Code Property Graphs (CPGs) constructed from static analysis results to capture both, structural and semantic relationships within code. Trained on the CamBenchCAP dataset, the model achieved an accuracy of 100% on the test set using an 80/20 train-test split. Evaluation on the CryptoAPI-Bench benchmark further demonstrated the model's practical applicability, reaching an overall accuracy of up to 96.6%. A detailed qualitative inspection revealed that many cases marked as misclassifications corresponded to genuine security weaknesses, indicating that the model effectively reflects conservative, security-aware reasoning. Identified limitations include incomplete control-flow representation due to missing interprocedural connections. Future work will focus on integrating call graphs, applying graph explainability techniques, and extending training data across multiple SAST tools to improve generalization and interpretability.

Today's massive AI computation loads push heavy data synchronization across sites, i.e., nodes in data centers. Any reduction in such consensus latency can significantly improve the overall performance of desired systems. This consensus challenge explosively peaks at cross-domain sites. In this paper, we proposed CD-Raft to address the cross-domain latency challenge, an optimized Raft protocol for strong consistency in cross-domain sites. CD-Raft can significantly reduce consensus latency by optimizing cross-domain round-trip time (RTT) for reads and writes, as well as carefully positioning the leader node. We verified the correctness of CD-Raft in a formal specification using the TLA+ specification, guaranteeing the strong consistency across sites. We have prototyped CD-Raft and evaluated it using the YCSB benchmark. Empirical results show that compared to the classic Raft, CD-Raft reduces the average latency by 32.90% and (99th percentile) tail latency by 49.24% for renown traces across multiple sites.

Active infrared thermography (AIRT) is currently witnessing a surge of artificial intelligence (AI) methodologies being deployed for automated subsurface defect analysis of high performance carbon fiber-reinforced polymers (CFRP). Deploying AI-based AIRT methodologies for inspecting CFRPs requires the creation of time consuming and expensive datasets of CFRP inspection sequences to train neural networks. To address this challenge, this work introduces a novel language-guided framework for cognitive defect analysis in CFRPs using AIRT and vision-language models (VLMs). Unlike conventional learning-based approaches, the proposed framework does not require developing training datasets for extensive training of defect detectors, instead it relies solely on pretrained multimodal VLM encoders coupled with a lightweight adapter to enable generative zero-shot understanding and localization of subsurface defects. By leveraging pretrained multimodal encoders, the proposed system enables generative zero-shot understanding of thermographic patterns and automatic detection of subsurface defects. Given the domain gap between thermographic data and natural images used to train VLMs, an AIRT-VLM Adapter is proposed to enhance the visibility of defects while aligning the thermographic domain with the learned representations of VLMs. The proposed framework is validated using three representative VLMs; specifically, GroundingDINO, Qwen-VL-Chat, and CogVLM. Validation is performed on 25 CFRP inspection sequences with impacts introduced at different energy levels, reflecting realistic defects encountered in industrial scenarios. Experimental results demonstrate that the AIRT-VLM adapter achieves signal-to-noise ratio (SNR) gains exceeding 10 dB compared with conventional thermographic dimensionality-reduction methods, while enabling zero-shot defect detection with intersection-over-union values reaching 70%.

Designing data integration pipelines typically requires substantial manual effort from data engineers to configure pipeline components and label training data. While LLMs have shown promise in handling individual steps of the integration process, their potential to replace all human input across end-to-end data integration pipelines has not been investigated. As a step toward exploring this potential, we present an automatic data integration pipeline that uses GPT-5.2 to generate all artifacts required to adapt the pipeline to specific use cases. These artifacts are schema mappings, value mappings for data normalization, training data for entity matching, and validation data for selecting conflict resolution heuristics in data fusion. We compare the performance of this LLM-based pipeline to the performance of human-designed pipelines along three case studies requiring the integration of video game, music, and company related data. Our experiments show that the LLM-based pipeline is able to produce similar results, for some tasks even better results, as the human-designed pipelines. End-to-end, the human and the LLM pipelines produce integrated datasets of comparable size and density. Having the LLM configure the pipelines costs approximately \$10 per case study, which represents only a small fraction of the cost of having human data engineers perform the same tasks.

Efficiently allocating incoming jobs to nodes in large-scale clusters can lead to substantial improvements in both cluster utilization and job performance. In order to allocate incoming jobs, cluster schedulers usually rely on a set of scoring functions to rank feasible nodes. Results from individual scoring functions are usually weighted equally, which could lead to sub-optimal deployments as the one-size-fits-all solution does not take into account the characteristics of each workload. Tuning the weights of scoring functions, however, requires expert knowledge and is computationally expensive. This paper proposes a reinforcement learning approach for learning the weights in scheduler scoring algorithms with the overall objective of improving the end-to-end performance of jobs for a given cluster. Our approach is based on percentage improvement reward, frame-stacking, and limiting domain information. We propose a percentage improvement reward to address the objective of multi-step parameter tuning. The inclusion of frame-stacking allows for carrying information across an optimization experiment. Limiting domain information prevents overfitting and improves performance in unseen clusters and workloads. The policy is trained on different combinations of workloads and cluster setups. We demonstrate the proposed approach improves performance on average by 33\% compared to fixed weights and 12\% compared to the best-performing baseline in a lab-based serverless scenario.

Residual connections are central to modern deep neural networks, enabling stable optimization and efficient information flow across depth. In this work, we propose SCORE (Skip-Connection ODE Recurrent Embedding), a discrete recurrent alternative to classical layer stacking. Instead of composing multiple independent layers, SCORE iteratively applies a single shared neural block using an ODE (Ordinary Differential Equation)-inspired contractive update: ht+1 = (1 - dt) * ht + dt * F(ht) This formulation can be interpreted as a depth-by-iteration refinement process, where the step size dt explicitly controls stability and update magnitude. Unlike continuous Neural ODE approaches, SCORE uses a fixed number of discrete iterations and standard backpropagation without requiring ODE solvers or adjoint methods. We evaluate SCORE across graph neural networks (ESOL molecular solubility), multilayer perceptrons, and Transformer-based language models (nanoGPT). Across architectures, SCORE generally improves convergence speed and often accelerates training. SCORE is reducing parameter count through shared weights. In practice, simple Euler integration provides the best trade-off between computational cost and performance, while higher-order integrators yield marginal gains at increased compute. These results suggest that controlled recurrent depth with contractive residual updates offers a lightweight and effective alternative to classical stacking in deep neural networks.

Promptable Foundation Models (FMs), initially introduced for natural image segmentation, have also revolutionized medical image segmentation. The increasing number of models, along with evaluations varying in datasets, metrics, and compared models, makes direct performance comparison between models difficult and complicates the selection of the most suitable model for specific clinical tasks. In our study, 11 promptable FMs are tested using non-iterative 2D and 3D prompting strategies on a private and public dataset focusing on bone and implant segmentation in four anatomical regions (wrist, shoulder, hip and lower leg). The Pareto-optimal models are identified and further analyzed using human prompts collected through a dedicated observer study. Our findings are: 1) The segmentation performance varies a lot between FMs and prompting strategies; 2) The Pareto-optimal models in 2D are SAM and SAM2.1, in 3D nnInteractive and Med-SAM2; 3) Localization accuracy and rater consistency vary with anatomical structures, with higher consistency for simple structures (wrist bones) and lower consistency for complex structures (pelvis, tibia, implants); 4) The segmentation performance drops using human prompts, suggesting that performance reported on "ideal" prompts extracted from reference labels might overestimate the performance in a human-driven setting; 5) All models were sensitive to prompt variations. While two models demonstrated intra-rater robustness, it did not scale to inter-rater settings. We conclude that the selection of the most optimal FM for a human-driven setting remains challenging, with even high-performing FMs being sensitive to variations in human input prompts. Our code base for prompt extraction and model inference is available: https://github.com/CarolineMagg/segmentation-FM-benchmark/

In deep networks, operations such as ReLU and hardware-driven clamping often cause activations to accumulate near the edges of the distribution, leading to biased clustering and suboptimal quantization in existing nonlinear (NL) quantization methods. This paper introduces Boundary Suppressed K-Means Quantization (BS-KMQ), a novel NL quantization approach designed to reduce the resolution requirements of analog-to-digital converters (ADCs) in in-memory computing (IMC) systems. By suppressing boundary outliers before clustering, BS-KMQ achieves more balanced and informative NL quantization levels. The resulting NL references are implemented using a reconfigurable in-memory NL-ADC, achieving a 7x area improvement over prior NL-ADC designs. When evaluated on ResNet-18, VGG-16, Inception-V3, and DistilBERT, BS-KMQ achieves at least 3x lower quantization error compared to linear, Lloyd-Max, cumulative distribution function (CDF), and K-means methods. It also improves post-training quantization accuracy by up to 66.8%, 25.4%, 66.6%, and 67.7%, respectively, compared to linear quantization. After low-bit fine-tuning, BS-KMQ maintains competitive accuracy with significantly fewer NL-ADC levels (3/3/4/4b). System-level simulations on ResNet-18 (6/2/3b) demonstrate up to a 4x speedup and 24x energy efficiency improvement over existing IMC accelerators.

This paper presents a novel hardware system for high-speed, event-sparse sampling-based electronic skin (e-skin)that integrates sensing and neuromorphic computing. The system is built around a 16x16 piezoresistive tactile array with front end and introduces a event-based binary scan search strategy to classify the digits. This event-driven strategy achieves a 12.8x reduction in scan counts, a 38.2x data compression rate and a 28.4x equivalent dynamic range, a 99% data sparsity, drastically reducing the data acquisition overhead. The resulting sparse data stream is processed by a multi-layer convolutional spiking neural network (Conv-SNN) implemented on an FPGA, which requires only 65% of the computation and 15.6% of the weight storage relative to a CNN. Despite these significant efficiency gains, the system maintains a high classification accuracy of 92.11% for real-time handwritten digit recognition. Furthermore, a real neuromorphic tactile dataset using Address Event Representation (AER) is constructed. This work demonstrates a fully integrated, event-driven pipeline from analog sensing to neuromorphic classification, offering an efficient solution for robotic perception and human-computer interaction.

Reinforcement learning significantly enhances LLM capabilities but suffers from a critical issue: length inflation, where models adopt verbosity or inefficient reasoning to maximize rewards. Prior approaches struggle to address this challenge in a general and lossless manner, primarily because additive penalties introduce a compensatory effect that creates optimization shortcuts, while heuristic gating strategies lack generality beyond binary feedback. To bridge this gap, we present Group Relative Reward Rescaling (GR$^3$), which reframes length control as a multiplicative rescaling paradigm, effectively establishing a generalized, continuous, and reward-dependent gating mechanism. To further ensure lossless optimization, we incorporate group-relative regularization and advantage-aware calibration, which dynamically adapt length budgets to instance difficulty and preserve the advantage signal of high-quality trajectories. Empirically, across both RLHF and RLVR settings, GR$^3$~maintains training dynamics and downstream performance comparable to standard GRPO while significantly mitigating length inflation, outperforming state-of-the-art length-regularized baselines.

Unmanned aerial vehicles (UAVs) are increasingly used to support time-critical medical supply delivery, providing rapid and flexible logistics during emergencies and resource shortages. However, effective deployment of UAV fleets requires coordination mechanisms capable of prioritizing medical requests, allocating limited aerial resources, and adapting delivery schedules under uncertain operational conditions. This paper presents a multi-agent reinforcement learning (MARL) framework for coordinating UAV fleets in stochastic medical delivery scenarios where requests vary in urgency, location, and delivery deadlines. The problem is formulated as a partially observable Markov decision process (POMDP) in which UAV agents maintain awareness of medical delivery demands while having limited visibility of other agents due to communication and localization constraints. The proposed framework employs Proximal Policy Optimization (PPO) as the primary learning algorithm and evaluates several variants, including asynchronous extensions, classical actor--critic methods, and architectural modifications to analyze scalability and performance trade-offs. The model is evaluated using real-world geographic data from selected clinics and hospitals extracted from the OpenStreetMap dataset. The framework provides a decision-support layer that prioritizes medical tasks, reallocates UAV resources in real time, and assists healthcare personnel in managing urgent logistics. Experimental results show that classical PPO achieves superior coordination performance compared to asynchronous and sequential learning strategies, highlighting the potential of reinforcement learning for adaptive and scalable UAV-assisted healthcare logistics.

Degradation prognosis for lithium-ion cells requires forecasting the state-of-health (SOH) trajectory over future cycles. Existing data-driven approaches can produce trajectory outputs through direct regression, but lack a mechanism to propagate degradation dynamics forward in time. This paper formulates battery degradation prognosis as a world model problem, encoding raw voltage, current, and temperature time-series from each cycle into a latent state and propagating it forward via a learned dynamics transition to produce a future trajectory spanning 80 cycles. To investigate whether electrochemical knowledge improves the learned dynamics, a Single Particle Model (SPM) constraint is incorporated into the training loss. Three configurations are evaluated on the Severson LiFePO4 (LFP) dataset of 138 cells. Iterative rollout halves the trajectory forecast error compared to direct regression from the same encoder. The SPM constraint improves prediction at the degradation knee where the resistance to SOH relationship is most applicable, without changing aggregate accuracy.

We present the AILS-NTUA system for SemEval-2026 Task 8 (MTRAGEval), addressing all three subtasks of multi-turn retrieval-augmented generation: passage retrieval (A), reference-grounded response generation (B), and end-to-end RAG (C). Our unified architecture is built on two principles: (i) a query-diversity-over-retriever-diversity strategy, where five complementary LLM-based query reformulations are issued to a single corpus-aligned sparse retriever and fused via variance-aware nested Reciprocal Rank Fusion; and (ii) a multistage generation pipeline that decomposes grounded generation into evidence span extraction, dual-candidate drafting, and calibrated multi-judge selection. Our system ranks 1st in Task A (nDCG@5: 0.5776, +20.5% over the strongest baseline) and 2nd in Task B (HM: 0.7698). Empirical analysis shows that query diversity over a well-aligned retriever outperforms heterogeneous retriever ensembling, and that answerability calibration-rather than retrieval coverage-is the primary bottleneck in end-to-end performance.

Instruction hierarchy (IH) defines how LLMs prioritize system, developer, user, and tool instructions under conflict, providing a concrete, trust-ordered policy for resolving instruction conflicts. IH is key to defending against jailbreaks, system prompt extractions, and agentic prompt injections. However, robust IH behavior is difficult to train: IH failures can be confounded with instruction-following failures, conflicts can be nuanced, and models can learn shortcuts such as overrefusing. We introduce IH-Challenge, a reinforcement learning training dataset, to address these difficulties. Fine-tuning GPT-5-Mini on IH-Challenge with online adversarial example generation improves IH robustness by +10.0% on average across 16 in-distribution, out-of-distribution, and human red-teaming benchmarks (84.1% to 94.1%), reduces unsafe behavior from 6.6% to 0.7% while improving helpfulness on general safety evaluations, and saturates an internal static agentic prompt injection evaluation, with minimal capability regression. We release the IH-Challenge dataset (https://huggingface.co/datasets/openai/ih-challenge) to support future research on robust instruction hierarchy.

Chebyshev filtered subspace iteration is a well-known algorithm for the solution of (symmetric/Hermitian) algebraic eigenproblems which has been implemented in several application codes~\cite{Kronik:2006ff, abinit:2020} or in stand alone libraries~\cite{ChASE}. An essential part of the algorithm is the QR-factorization of the array of vectors spanning the active subspace that have been filtered by the Chebyshev filter. Typically such an array has an a-priori unknown high condition number that directly influences the choice of QR-factorization algorithm. In this work we show how such condition number can be bound from above with precise and inexpensive estimates. We then proceed to use these estimates to implement a mechanism for the choice of QR-factorization in the ChASE library. We show how such mechanism enhance the performance of the library without compromising on its accuracy.

Artificial intelligence has advanced significantly through the development of intelligent game-playing systems, providing rigorous testbeds for decision-making, strategic planning, and adaptive learning. However, resource-constrained environments pose critical challenges, as conventional deep learning methods heavily rely on extensive datasets and computational resources. In this paper, we propose a lightweight hybrid framework for the Game of the Amazons, which explores the paradigm of weak-to-strong generalization by integrating the structural reasoning of graph-based learning with the generative capabilities of large language models. Specifically, we leverage a Graph Attention Autoencoder to inform a multi-step Monte Carlo Tree Search, utilize a Stochastic Graph Genetic Algorithm to optimize evaluation signals, and harness GPT-4o-mini to generate synthetic training data. Unlike traditional approaches that rely on expert demonstrations, our framework learns from noisy and imperfect supervision. We demonstrate that the Graph Attention mechanism effectively functions as a structural filter, denoising the LLM's outputs. Experiments on a 10$\times$10 Amazons board show that our hybrid approach not only achieves a 15\%--56\% improvement in decision accuracy over baselines but also significantly outperforms its teacher model (GPT-4o-mini), achieving a competitive win rate of 45.0\% at N=30 nodes and a decisive 66.5\% at only N=50 nodes. These results verify the feasibility of evolving specialized, high-performance game AI from general-purpose foundation models under stringent computational constraints.

Training autonomous web agents is fundamentally limited by the environments they learn from: real-world websites are unsafe to explore, hard to reset, and rarely provide verifiable feedback. We propose VeriEnv, a framework that treats language models as environment creators, automatically cloning real-world websites into fully executable, verifiable synthetic environments. By exposing controlled internal access via a Python SDK, VeriEnv enables agents to self-generate tasks with deterministic, programmatically verifiable rewards, eliminating reliance on heuristic or LLM-based judges. This design decouples agent learning from unsafe real-world interaction while enabling scalable self-evolution through environment expansion. Through experiments on web agent benchmarks, we show that agents trained with VeriEnv generalize to unseen websites, achieve site-specific mastery through self-evolving training, and benefit from scaling the number of training environments. Code and resources will be released at https://github.com/kyle8581/VeriEnv upon acceptance.

Generative AI systems increasingly expose powerful reasoning and image refinement capabilities through user-facing chatbot interfaces. In this work, we show that the naïve exposure of such capabilities fundamentally undermines modern deepfake detectors. Rather than proposing a new image manipulation technique, we study a realistic and already-deployed usage scenario in which an adversary uses only benign, policy-compliant prompts and commercial generative AI systems. We demonstrate that state-of-the-art deepfake detection methods fail under semantic-preserving image refinement. Specifically, we show that generative AI systems articulate explicit authenticity criteria and inadvertently externalize them through unrestricted reasoning, enabling their direct reuse as refinement objectives. As a result, refined images simultaneously evade detection, preserve identity as verified by commercial face recognition APIs, and exhibit substantially higher perceptual quality. Importantly, we find that widely accessible commercial chatbot services pose a significantly greater security risk than open-source models, as their superior realism, semantic controllability, and low-barrier interfaces enable effective evasion by non-expert users. Our findings reveal a structural mismatch between the threat models assumed by current detection frameworks and the actual capabilities of real-world generative AI. While detection baselines are largely shaped by prior benchmarks, deployed systems expose unrestricted authenticity reasoning and refinement despite stringent safety controls in other domains.

We introduce the tubal tensor train (TTT) decomposition, a tensor-network model that combines the t-product algebra of the tensor singular value decomposition (T-SVD) with the low-order core structure of the tensor train (TT) format. For an order-$(N+1)$ tensor with a distinguished tube mode, the proposed representation consists of two third-order boundary cores and $N-2$ fourth-order interior cores linked through the t-product. As a result, for bounded tubal ranks, the storage scales linearly with the number of modes, in contrast to direct high-order extensions of T-SVD. We present two computational strategies: a sequential fixed-rank construction, called TTT-SVD, and a Fourier-slice alternating scheme based on the alternating two-cores update (ATCU). We also state a TT-SVD-type error bound for TTT-SVD and illustrate the practical performance of the proposed model on image compression, video compression, tensor completion, and hyperspectral imaging.

Brief Hospital Course (BHC) narratives must be clinically useful yet faithful to fragmented EHR evidence. LLM-based clinical summarizers still introduce unsupported statements, and alignment can encourage omissions ("say-less" degeneration). We introduce VERI-DPO, which uses claim verification to mine preferences and distill them into the summarizer with Direct Preference Optimization (DPO). On MIMIC-III-Ext-VeriFact-BHC (100 ICU patients; patient-level splits), we train a retrieval-augmented verifier to label claim-evidence pairs as Supported, Not Supported, or Not Addressed via a single-token format. The verifier scores sentence-level claims from sampled BHC candidates and aggregates margins into a coverage-aware utility to mine length-controlled, contradiction-anchored preference pairs. On held-out patients, verifier-mined preferences separate candidates by contradiction density, and VERI-DPO reduces Not Supported claim rates from 10.7% to 1.9% (local verifier judge) and from 11.6% to 6.4% (GPT-4o judge), while improving validity from 76.7% to 82.5% and maintaining informative length.

Estimating the intrinsic dimensionality (ID) of data is a fundamental problem in machine learning and computer vision, providing insight into the true degrees of freedom underlying high-dimensional observations. Existing methods often rely on geometric or distributional assumptions and can significantly fail when these assumptions are violated. In this paper, we introduce a novel ID estimator based on nearest-neighbor distance ratios that involves simple calculations and achieves state-of-the-art results. Most importantly, we provide a theoretical analysis proving that our estimator is \emph{universal}, namely, it converges to the true ID independently of the distribution generating the data. We present experimental results on benchmark manifolds and real-world datasets to demonstrate the performance of our estimator.

We present PULSE, a medical reasoning agent that combines a domain-tuned large language model with scientific literature retrieval to support diagnostic decision-making in complex real-world cases. To evaluate its capabilities, we curated a benchmark of 82 authentic endocrinology case reports encompassing a broad spectrum of disease types and incidence levels. In controlled experiments, we compared PULSE's performance against physicians with varying levels of expertise-from residents to senior specialists-and examined how AI assistance influenced human diagnostic reasoning. PULSE attained expert-competitive accuracy, outperforming residents and junior specialists while matching senior specialist performance at both Top@1 and Top@4 thresholds. Unlike physicians, whose accuracy declined with disease rarity, PULSE maintained stable performance across incidence tiers. The agent also exhibited adaptive reasoning, increasing output length with case difficulty in a manner analogous to the longer deliberation observed among expert clinicians. When used collaboratively, PULSE enabled physicians to correct initial errors and broaden diagnostic hypotheses, but also introduced risks of automation bias. The study explores both serial and concurrent collaboration workflows, revealing that PULSE offers robust support across common and rare presentations. These findings underscore both the promise and the limitations of language model-based agents in clinical diagnosis, and offer a framework for evaluating their role in real-world decision-making.

Animal brains flexibly and efficiently achieve many behavioral tasks with a single neural network. A core goal in modern neuroscience is to map the mechanisms of the brain's flexibility onto the dynamics underlying neural populations. However, identifying task-specific dynamical rules from limited, noisy, and high-dimensional experimental neural recordings remains a major challenge, as experimental data often provide only partial access to brain states and dynamical mechanisms. While recurrent neural networks (RNNs) directly constrained neural data have been effective in inferring underlying dynamical mechanisms, they are typically limited to single-task domains and struggle to generalize across behavioral conditions. Here, we introduce JEDI, a hierarchical model that captures neural dynamics across tasks and contexts by learning a shared embedding space over RNN weights. This model recapitulates individual samples of neural dynamics while scaling to arbitrarily large and complex datasets, uncovering shared structure across conditions in a single, unified model. Using simulated RNN datasets, we demonstrate that JEDI accurately learns robust, generalizable, condition-specific embeddings. By reverse-engineering the weights learned by JEDI, we show that it recovers ground truth fixed point structures and unveils key features of the underlying neural dynamics in the eigenspectra. Finally, we apply JEDI to motor cortex recordings during monkey reaching to extract mechanistic insight into the neural dynamics of motor control. Our work shows that joint learning of contextual embeddings and recurrent weights provides scalable and generalizable inference of brain dynamics from recordings alone.

In this work we study the convergence properties of the Dual Space Preconditioned Gradient Descent, encompassing optimizers such as Normalized Gradient Descent, Gradient Clipping and Adam. We consider preconditioners of the form $\nabla K$, where $K: \mathbb{R}^p \to \mathbb{R}$ is convex and assume that the latter is applied to train an over-parameterized linear model with loss of the form $\ell({X} {W} - {Y})$, for weights ${W} \in \mathbb{R}^{d \times k}$, labels ${Y} \in \mathbb{R}^{n \times k}$ and data ${X} \in \mathbb{R}^{n \times d}$. Under the aforementioned assumptions, we prove that the iterates of the preconditioned gradient descent always converge to a point ${W}_{\infty} \in \mathbb{R}^{d \times k}$ satisfying ${X}{W}_{\infty} = {Y}$. Our proof techniques are of independent interest as we introduce a novel version of the Bregman Divergence with accompanying identities that allow us to establish convergence. We also study the implicit bias of Dual Space Preconditioned Gradient Descent. First, we demonstrate empirically that, for general $K(\cdot)$, ${W}_\infty$ depends on the chosen learning rate, hindering a precise characterization of the implicit bias. Then, for preconditioners of the form $K({G}) = h(\|{G}\|_F)$, known as \textit{isotropic preconditioners}, we show that ${W}_\infty$ minimizes $\|{W}_\infty - {W}_0\|_F^2$ subject to ${X}{W}_\infty = {Y}$, where ${W}_0$ is the initialization. Denoting the convergence point of GD initialized at ${W}_0$ by ${W}_{\text{GD}, \infty}$, we thus note ${W}_{\infty} = {W}_{\text{GD}, \infty}$ for isotropic preconditioners. Finally, we show that a similar fact holds for general preconditioners up to a multiplicative constant, namely, $\|{W}_0 - {W}_{\infty}\|_F \le c \|{W}_0 - {W}_{\text{GD}, \infty}\|_F$ for a constant $c>0$.

Software verification is now costly, taking over half the project effort while failing on modern complex systems. We hence propose a shift from verification and modeling to herding: treating testing as a model-free search task that steers systems toward target goals. This exploits the "Sparsity of Influence" -the fact that, often, large software state spaces are ruled by just a few variables, We introduce EZR (Efficient Zero-knowledge Ranker), a stochastic learner that finds these controllers directly. Across dozens of tasks, EZR achieved 90% of peak results with only 32 samples, replacing heavy solvers with light sampling.

Prompt design is a primary control interface for large language models (LLMs), yet standard evaluations largely reduce performance to answer correctness, obscuring why a prompt succeeds or fails and providing little actionable guidance. We propose PEEM (Prompt Engineering Evaluation Metrics), a unified framework for joint and interpretable evaluation of both prompts and responses. PEEM defines a structured rubric with 9 axes: 3 prompt criteria (clarity/structure, linguistic quality, fairness) and 6 response criteria (accuracy, coherence, relevance, objectivity, clarity, conciseness), and uses an LLM-based evaluator to output (i) scalar scores on a 1-5 Likert scale and (ii) criterion-specific natural-language rationales grounded in the rubric. Across 7 benchmarks and 5 task models, PEEM's accuracy axis strongly aligns with conventional accuracy while preserving model rankings (aggregate Spearman rho about 0.97, Pearson r about 0.94, p < 0.001). A multi-evaluator study with four models shows consistent relative judgments (pairwise rho = 0.68-0.85), supporting evaluator-agnostic deployment. Beyond alignment, PEEM captures complementary linguistic failure modes and remains informative under prompt perturbations: prompt-quality trends track downstream accuracy under iterative rewrites, semantic adversarial manipulations induce clear score degradation, and meaning-preserving paraphrases yield high stability (robustness rate about 76.7-80.6%). Finally, using only PEEM scores and rationales as feedback, a zero-shot prompt rewriting loop improves downstream accuracy by up to 11.7 points, outperforming supervised and RL-based prompt-optimization baselines. Overall, PEEM provides a reproducible, criterion-driven protocol that links prompt formulation to response behavior and enables systematic diagnosis and optimization of LLM interactions.

The alignment of large language models (LLMs) has progressed substantially in single-agent settings through paradigms such as RLHF and Constitutional AI, with recent work exploring scalable alternatives such as RLAIF and evolving alignment objectives. However, these approaches remain limited in multi-stakeholder settings, where conflicting values arise and deliberative negotiation capabilities are required. This work proposes a multi-agent negotiation-based alignment framework that aligns LLMs to Collective Agency (CA)-an existing alignment objective introduced to promote the continual expansion of agency-while simultaneously improving conflict-resolution capability. To enable scalable training, two self-play instances of the same LLM, assigned opposing personas, engage in structured turn-based dialogue to synthesize mutually beneficial solutions. We generate synthetic moral-dilemma prompts and conflicting persona pairs, and optimize the policy via RLAIF using GRPO with an external LLM reward model. While rewards are computed from CA scores assigned to the final completion, gradients are applied to dialogue tokens to directly improve deliberative interaction dynamics. Experiments show that the resulting model achieves CA alignment comparable to a single-agent baseline while substantially improving conflict-resolution performance without degrading general language capabilities. These results suggest that negotiation-driven deliberation training provides a practical path toward LLMs that better support collective decision-making in value-conflict scenarios.

Human locomotion emerges from high-dimensional neuromuscular control, making predictive musculoskeletal simulation challenging. We present a physiology-informed reinforcement-learning framework that constrains control using muscle synergies. We extracted a low-dimensional synergy basis from inverse musculoskeletal analyses of a small set of overground walking trials and used it as the action space for a muscle-driven three-dimensional model trained across variable speeds, slopes and uneven terrain. The resulting controller generated stable gait from 0.7-1.8 m/s and on $\pm$ 6$^{\circ}$ grades and reproduced condition-dependent modulation of joint angles, joint moments and ground reaction forces. Compared with an unconstrained controller, synergy-constrained control reduced non-physiological knee kinematics and kept knee moment profiles within the experimental envelope. Across conditions, simulated vertical ground reaction forces correlated strongly with human measurements, and muscle-activation timing largely fell within inter-subject variability. These results show that embedding neurophysiological structure into reinforcement learning can improve biomechanical fidelity and generalization in predictive human locomotion simulation with limited experimental data.

The paradigm shift from item-centric ranking to answer-centric synthesis is redefining the role of search engines. While recent industrial progress has applied generative techniques to closed-set item ranking in e-commerce, research and deployment of open-ended generative search on large content platforms remain limited. This setting introduces challenges, including robustness to noisy retrieval, non-negotiable safety guarantees, and alignment with diverse user needs. In this work, we introduce SearchLLM, the first large language model (LLM) for open-ended generative search. We design a hierarchical, multi-dimensional reward system that separates bottom-line constraints, including factual grounding, basic answer quality and format compliance, from behavior optimization objectives that promote robustness to noisy retrieval and alignment with user needs. Concretely, our reward model evaluates responses conditioned on the user query, session history, and retrieved evidence set, combining rule-based checks with human-calibrated LLM judges to produce an interpretable score vector over these dimensions. We introduce a Gated Aggregation Strategy to derive the training reward for optimizing SearchLLM with Group Relative Policy Optimization (GRPO). We deploy SearchLLM in the AI search entry of RedNote. Offline evaluations and online A/B tests show improved generation quality and user engagement, increasing Valid Consumption Rate by 1.03% and reducing Re-search Rate by 2.81%, while upholding strict safety and reliability standards.

Personalized news recommendation is highly time-sensitive, as user interests are often driven by emerging events, trending topics, and shifting real-world contexts. These dynamics make it essential to model not only users' long-term preferences, which reflect stable reading habits and high-order collaborative patterns, but also their short-term, context-dependent interests that change rapidly over time. However, most existing approaches rely on a single static interaction graph, which struggles to capture both long-term preference patterns and short-term interest changes as user behavior evolves. To address this challenge, we propose a unified framework that learns user preferences from both global and local temporal perspectives. A global preference modeling component captures long-term collaborative signals from the overall interaction graph, while a local preference modeling component partitions historical interactions into stage-wise temporal subgraphs to represent short-term dynamics. Within this module, an LSTM branch models the progressive evolution of recent interests, and a self-attention branch captures long-range temporal dependencies. Extensive experiments on two large-scale real-world datasets show that our approach consistently outperforms strong baselines and delivers fresher and more relevant recommendations across diverse user behaviors and temporal settings.

We study timestamped speaker-attributed ASR for long-form, multi-party speech with overlap, where chunk-wise inference must preserve meeting-level speaker identity consistency while producing time-stamped, speaker-labeled transcripts. Previous Speech-LLM systems tend to prioritize either local diarization or global labeling, but often lack the ability to capture fine-grained temporal boundaries or robust cross-chunk identity linking. We propose G-STAR, an end-to-end system that couples a time-aware speaker-tracking module with a Speech-LLM transcription backbone. The tracker provides structured speaker cues with temporal grounding, and the LLM generates attributed text conditioned on these cues. G-STAR supports both component-wise optimization and joint end-to-end training, enabling flexible learning under heterogeneous supervision and domain shift. Experiments analyze cue fusion, local versus long-context trade-offs and hierarchical objectives.

Physics-Informed Neural Networks (PINNs) have shown promise in solving incompressible Navier-Stokes equations, yet existing approaches are predominantly designed for single-flow settings. When extended to multi-flow scenarios, these methods face three key challenges: (1) difficulty in simultaneously capturing both shared physical principles and flow-specific characteristics, (2) susceptibility to inter-task negative transfer that degrades prediction accuracy, and (3) unstable training dynamics caused by disparate loss magnitudes across heterogeneous flow regimes. To address these limitations, we propose UniPINN, a unified multi-flow PINN framework that integrates three complementary components: a shared-specialized architecture that disentangles universal physical laws from flow-specific features, a cross-flow attention mechanism that selectively reinforces relevant patterns while suppressing task-irrelevant interference, and a dynamic weight allocation strategy that adaptively balances loss contributions to stabilize multi-objective optimization. Extensive experiments on three canonical flows demonstrate that UniPINN effectively unifies multi-flow learning, achieving superior prediction accuracy and balanced performance across heterogeneous regimes while successfully mitigating negative transfer. The source code of this paper will be released on https://github.com/Event-AHU/OpenFusion

We develop a theoretical and computational framework for beam-plasma collective oscillations in intense charged-particle beams at intermediate energies (10-100 MeV). In Part I, we formulate a kinetic field theory governed by the Vlasov-Poisson system, deriving the Lindhard dielectric function and random phase approximation (RPA) polarization tensor for three beam distribution functions. We prove via the dielectric function epsilon(omega,q)=0 the existence of undamped Langmuir wave modes above a critical beam density n_c, obtain explicit beam-plasma dispersion relations, and show that Landau damping vanishes above the particle-hole continuum. The plasma frequency Omega_p^2 = ne^2/(m*epsilon_0) is fixed by the f-sum rule independently of distribution shape; higher dispersion coefficients depend on velocity moments. Space charge effects drive anomalous beam broadening with sqrt(n-n_c) onset and Friedel oscillations at q=2k_F. The beam-plasma transition belongs to the 3D Ising universality class via renormalization group analysis. In Part II, we validate these predictions using Prometheus, a beta-VAE trained on static structure factor data S(q) from particle-in-cell (PIC) beam simulations. Prometheus detects collective plasma oscillation onset in Gaussian and uniform distributions, confirms their absence in the degenerate Fermi gas (n_c -> 0), and resolves the Kohn anomaly at q=2k_F. Dispersion analysis of S(q,omega) from PIC simulations verifies the distribution-independent Omega_p predicted by the f-sum rule. All six validation checks pass. Predicted signatures -- density-tunable plasma resonances at omega_p proportional to sqrt(n), anomalous beam broadening with sqrt(n-n_c) onset, and Friedel oscillations -- are accessible at existing intermediate-energy beam facilities.

This study proposes a multi-resolution Convolutional Long Short-Term Memory (ConvLSTM) ensemble framework that leverages diverse temporal input resolutions to mitigate error accumulation and improve long-horizon forecasting of retaining-structure behavior during staged excavation. An extensive database of lateral wall displacement responses was generated through PLAXIS2D simulations incorporating five-layered soil stratigraphy, two excavation depths (14 and 20 m), and stochastically varied geotechnical and structural parameters, yielding 2,000 time-series deflection profiles. Three ConvLSTM models trained at different input resolutions were integrated using a fully connected neural network meta-learner to construct the ensemble model. Validation using both numerical results and field measurements demonstrated that the ensemble approach consistently outperformed the standalone ConvLSTM models, particularly in long-term multi-step prediction, exhibiting reduced error propagation and improved generalization. These findings underscore the potential of multi-resolution ensemble strategies that jointly exploit diverse temporal input scales to enhance predictive stability and accuracy in AI-driven geotechnical forecasting.

Isotonic regression (IR) is shape-constrained regression to maintain a univariate fitting curve non-decreasing, which has numerous applications including single-index models and probability calibration. When it comes to multi-output regression, the classical IR is no longer applicable because the monotonicity is not readily extendable. We consider a novel multi-output regression problem where a regression function is \emph{cyclically monotone}. Roughly speaking, a cyclically monotone function is the gradient of some convex potential. Whereas enforcing cyclic monotonicity is apparently challenging, we leverage the fact that Kantorovich's optimal transport (OT) always yields a cyclically monotone coupling as an optimal solution. This perspective naturally allows us to interpret a regression function and the convex potential as a link function in generalized linear models and Brenier's potential in OT, respectively, and hence we call this IR extension \emph{Brenier isotonic regression}. We demonstrate experiments with probability calibration and generalized linear models. In particular, IR outperforms many famous baselines in probability calibration robustly.

Achieving human-like dexterous manipulation through the collaboration of multi-fingered hands with robotic arms remains a longstanding challenge in robotics, primarily due to the scarcity of high-quality demonstrations and the complexity of high-dimensional action spaces. To address these challenges, we propose FAR-Dex, a hierarchical framework that integrates few-shot data augmentation with adaptive residual refinement to enable robust and precise arm-hand coordination in dexterous tasks. First, FAR-DexGen leverages the IsaacLab simulator to generate diverse and physically constrained trajectories from a few demonstrations, providing a data foundation for policy training. Second, FAR-DexRes introduces an adaptive residual module that refines policies by combining multi-step trajectory segments with observation features, thereby enhancing accuracy and robustness in manipulation scenarios. Experiments in both simulation and real-world demonstrate that FAR-Dex improves data quality by 13.4% and task success rates by 7% over state-of-the-art methods. It further achieves over 80% success in real-world tasks, enabling fine-grained dexterous manipulation with strong positional generalization.

Machine unlearning aims to remove specific outputs from trained models, often at the concept level, such as forgetting all occurrences of a particular celebrity or filtering content via text prompts. However, many undesired outputs, such as an individual's face or generations culturally or factually misinterpreted, cannot often be specified by text prompts. We address this underexplored setting of instance unlearning for outputs that are undesired but unpromptable, where the goal is to forget target outputs selectively while preserving the rest. To this end, we introduce an effective surrogate-based unlearning method that leverages image editing, timestep-aware weighting, and gradient surgery to guide trained diffusion models toward forgetting specific outputs. Experiments on conditional (Stable Diffusion 3) and unconditional (DDPM-CelebA) diffusion models demonstrate that our prompt-free method uniquely unlearns unpromptable outputs, such as faces and culturally inaccurate depictions, with preserved integrity, unlike prompt-based and prompt-free baselines. Our proposed method would serve as a practical hotfix for diffusion model providers to ensure privacy protection and ethical compliance.

Large language models trained on natural language exhibit pronounced anisotropy: a small number of directions concentrate disproportionate energy, while the remaining dimensions form a broad semantic tail. In low-bit training regimes, this geometry becomes numerically unstable. Because blockwise quantization scales are determined by extreme elementwise magnitudes, dominant directions stretch the dynamic range, compressing long-tail semantic variation into narrow numerical bins. We show that this instability is primarily driven by a coherent rank-one mean bias, which constitutes the dominant component of spectral anisotropy in LLM representations. This mean component emerges systematically across layers and training stages and accounts for the majority of extreme activation magnitudes, making it the principal driver of dynamic-range inflation under low precision. Crucially, because the dominant instability is rank-one, it can be eliminated through a simple source-level mean-subtraction operation. This bias-centric conditioning recovers most of the stability benefits of SVD-based spectral methods while requiring only reduction operations and standard quantization kernels. Empirical results on FP4 (W4A4G4) training show that mean removal substantially narrows the loss gap to BF16 and restores downstream performance, providing a hardware-efficient path to stable low-bit LLM training.

Conditional density estimation is complicated by multimodality, heteroscedasticity, and strong non-Gaussianity. Gaussian processes (GPs) provide a principled nonparametric framework with calibrated uncertainty, but standard GP regression is limited by its unimodal Gaussian predictive form. We introduce the Generalized Gaussian Mixture Process (GGMP), a GP-based method for multimodal conditional density estimation in settings where each input may be associated with a complex output distribution rather than a single scalar response. GGMP combines local Gaussian mixture fitting, cross-input component alignment and per-component heteroscedastic GP training to produce a closed-form Gaussian mixture predictive density. The method is tractable, compatible with standard GP solvers and scalable methods, and avoids the exponentially large latent-assignment structure of naive multimodal GP formulations. Empirically, GGMPs improve distributional approximation on synthetic and real-world datasets with pronounced non-Gaussianity and multimodality.

Large deep neural networks (DNNs), especially transformer-based and multimodal architectures, are computationally demanding and challenging to deploy on resource-constrained edge platforms like field robots. These challenges intensify in mission-critical scenarios (e.g., disaster response), where robots must collaborate under tight constraints on bandwidth, latency, and battery life, often without infrastructure or server support. To address these limitations, we present COHORT, a collaborative DNN inference and task-execution framework for multi-robot systems built on the Robotic Operating System (ROS). COHORT employs a hybrid offline-online reinforcement learning (RL) strategy to dynamically schedule and distribute DNN module execution across robots. Our key contributions are threefold: (a) Offline RL policy learning combined with Advantage-Weighted Regression (AWR), trained on auction-based task allocation data from heterogeneous DNN workloads across distributed robots, (b) Online policy adaptation via Multi-Agent PPO (MAPPO), initialized from the offline policy and fine-tuned in real time, and (c) comprehensive evaluation of COHORT on vision-language model (VLM) inference tasks such as CLIP and SAM, analyzing scalability with increasing robot/workload and robustness under . We benchmark COHORT against genetic algorithms and multiple RL baselines. Experimental results demonstrate that COHORT reduces battery consumption by 15.4% and increases GPU utilization by 51.67%, while satisfying frame-rate and deadline constraints 2.55 times of the time.

Active learning for regression reduces labeling costs by selecting the most informative samples. Improved Greedy Sampling is a prominent method that balances feature-space diversity and output-space uncertainty using a static, multiplicative rule. We propose Weighted improved Greedy Sampling (WiGS), which replaces this framework with a dynamic, additive criterion. We formulate weight selection as a reinforcement learning problem, enabling an agent to adapt the exploration-investigation balance throughout learning. Experiments on 18 benchmark datasets and a synthetic environment show WiGS outperforms iGS and other baseline methods in both accuracy and labeling efficiency, particularly in domains with irregular data density where the baseline's multiplicative rule ignores high-error samples in dense regions.

The construction of high quality health indicators (HIs) is crucial for effective prognostics and health management. Although deep learning has significantly advanced HI modeling, existing approaches often struggle with distribution mismatches resulting from varying operating conditions. Although domain adaptation is typically employed to mitigate these shifts, two critical challenges remain: (1) the misalignment of degradation stages during random mini-batch sampling, resulting in misleading discrepancy losses, and (2) the structural limitations of small-kernel 1D-CNNs in capturing long-range temporal dependencies within complex vibration signals. To address these issues, we propose a domain-adaptive framework comprising degradation stage synchronized batch sampling (DSSBS) and the cross-domain aligned fusion large autoencoder (CAFLAE). DSSBS utilizes kernel change-point detection to segment degradation stages, ensuring that source and target mini-batches are synchronized by their failure phases during alignment. Complementing this, CAFLAE integrates large-kernel temporal feature extraction with cross-attention mechanisms to learn superior domain-invariant representations. The proposed framework was rigorously validated on a Korean defense system dataset and the XJTU-SY bearing dataset, achieving an average performance enhancement of 24.1% over state-of-the-art methods. These results demonstrate that DSSBS improves cross-domain alignment through stage-consistent sampling, whereas CAFLAE offers a high-performance backbone for long-term industrial condition monitoring.

Adversarial examples can represent a serious threat to machine learning (ML) algorithms. If used to manipulate the behaviour of ML-based Network Intrusion Detection Systems (NIDS), they can jeopardize network security. In this work, we aim to mitigate such risks by increasing the robustness of NIDS towards adversarial attacks. To that end, we explore two adversarial methods for generating malicious network traffic. The first method is based on Generative Adversarial Networks (GAN) and the second one is the Fast Gradient Sign Method (FGSM). The adversarial examples generated by these methods are then used to evaluate a novel multilayer defense mechanism, specifically designed to mitigate the vulnerability of ML-based NIDS. Our solution consists of one layer of stacking classifiers and a second layer based on an autoencoder. If the incoming network data are classified as benign by the first layer, the second layer is activated to ensure that the decision made by the stacking classifier is correct. We also incorporated adversarial training to further improve the robustness of our solution. Experiments on two datasets, namely UNSW-NB15 and NSL-KDD, demonstrate that the proposed approach increases resilience to adversarial attacks.

Spatio-temporal time series are widely used in real-world applications, including traffic prediction and weather forecasting. They are sequences of observations over extensive periods and multiple locations, naturally represented as multidimensional data. Forecasting is a central task in spatio-temporal analysis, and numerous deep learning methods have been developed to address it. However, as dataset sizes and model complexities continue to grow in practice, training deep learning models has become increasingly time- and resource-intensive. A promising solution to this challenge is dataset distillation, which synthesizes compact datasets that can effectively replace the original data for model training. Although successful in various domains, including time series analysis, existing dataset distillation methods compress only one dimension, making them less suitable for spatio-temporal datasets, where both spatial and temporal dimensions jointly contribute to the large data volume. To address this limitation, we propose STemDist, the first dataset distillation method specialized for spatio-temporal time series forecasting. A key idea of our solution is to compress both temporal and spatial dimensions in a balanced manner, reducing training time and memory. We further reduce the distillation cost by performing distillation at the cluster level rather than the individual location level, and we complement this coarse-grained approach with a subset-based granular distillation technique that enhances forecasting performance. On five real-world datasets, we show empirically that, compared to both general and time-series dataset distillation methods, datasets distilled by our STemDist method enable model training (1) faster (up to 6X) (2) more memory-efficient (up to 8X), and (3) more effective (with up to 12% lower prediction error).

Designing service systems requires selecting among alternative configurations -- choosing the best chatbot variant, the optimal routing policy, or the most effective quality control procedure. In many service systems, the primary evidence of performance quality is textual -- customer support transcripts, complaint narratives, compliance review reports -- rather than the scalar measurements assumed by classical optimization methods. Large language models (LLMs) can read such textual evidence and produce standardized quality scores, but these automated judges exhibit systematic biases that vary across alternatives and evaluation instances. Human expert review remains accurate but costly. We study how to identify the best service configuration with high confidence while minimizing expensive human audits, given that automated evaluation is cheap but biased. We formalize this as a sequential decision problem where a biased proxy score is observed for every evaluation, and a verified outcome can be acquired selectively at additional cost. We prove that LLM-only selection fails under arm-dependent bias, and that naive selective-audit estimators can be asymptotically biased. We develop an estimator combining proxy scores with inverse-propensity-weighted residuals and construct anytime-valid confidence sequences. Our algorithm, PP-LUCB, jointly decides which alternatives to evaluate and whether to request human audits, concentrating reviews where the LLM judge is least reliable. We prove correctness and establish instance-dependent cost bounds showing near-optimal efficiency. On a customer support ticket classification task, our algorithm correctly identifies the best model in 40/40 trials while achieving 90\% audit cost reduction.

One crucial factor behind the success of deep learning lies in the implicit bias induced by noise inherent in gradient-based training algorithms. Motivated by empirical observations that training with noisy labels improves model generalization, we delve into the underlying mechanisms behind stochastic gradient descent (SGD) with label noise. Focusing on a two-layer over-parameterized linear network, we analyze the learning dynamics of label noise SGD, unveiling a two-phase learning behavior. In \emph{Phase I}, the magnitudes of model weights progressively diminish, and the model escapes the lazy regime; enters the rich regime. In \emph{Phase II}, the alignment between model weights and the ground-truth interpolator increases, and the model eventually converges. Our analysis highlights the critical role of label noise in driving the transition from the lazy to the rich regime and minimally explains its empirical success. Furthermore, we extend these insights to Sharpness-Aware Minimization (SAM), showing that the principles governing label noise SGD also apply to broader optimization algorithms. Extensive experiments, conducted under both synthetic and real-world setups, strongly support our theory. Our code is released at https://github.com/a-usually/Label-Noise-SGD.

Despite the growing demand for eliciting uncertainty from large language models (LLMs), empirical evidence suggests that LLM behavior is not always adequately captured by the elicitation techniques developed under the classical probabilistic uncertainty framework. This mismatch leads to systematic failure modes, particularly in settings that involve ambiguous question-answering, in-context learning, and self-reflection. To address this, we propose novel prompt-based uncertainty elicitation techniques grounded in \emph{imprecise probabilities}, a principled framework for repesenting and eliciting higher-order uncertainty. Here, first-order uncertainty captures uncertainty over possible responses to a prompt, while second-order uncertainty (uncertainty about uncertainty) quantifies indeterminacy in the underlying probability model itself. We introduce general-purpose prompting and post-processing procedures to directly elicit and quantify both orders of uncertainty, and demonstrate their effectiveness across diverse settings. Our approach enables more faithful uncertainty reporting from LLMs, improving credibility and supporting downstream decision-making.

Graph generation is a fundamental task with broad applications, such as drug discovery. Recently, discrete flow matching-based graph generation, \aka, graph flow model (GFM), has emerged due to its superior performance and flexible sampling. However, effectively aligning GFMs with complex human preferences or task-specific objectives remains a significant challenge. In this paper, we propose Graph-GRPO, an online reinforcement learning (RL) framework for training GFMs under verifiable rewards. Our method makes two key contributions: (1) We derive an analytical expression for the transition probability of GFMs, replacing the Monte Carlo sampling and enabling fully differentiable rollouts for RL training; (2) We propose a refinement strategy that randomly perturbs specific nodes and edges in a graph, and regenerates them, allowing for localized exploration and self-improvement of generation quality. Extensive experiments on both synthetic and real datasets demonstrate the effectiveness of Graph-GRPO. With only 50 denoising steps, our method achieves 95.0\% and 97.5\% Valid-Unique-Novelty scores on the planar and tree datasets, respectively. Moreover, Graph-GRPO achieves state-of-the-art performance on the molecular optimization tasks, outperforming graph-based and fragment-based RL methods as well as classic genetic algorithms.

In this paper, we present a novel probabilistic safe control framework for human-robot interaction that combines control barrier functions (CBFs) with conformal risk control to provide formal safety guarantees while considering complex human behavior. The approach uses conformal risk control to quantify and control the prediction errors in CBF safety values and establishes formal guarantees on the probability of constraint satisfaction during interaction. We introduce an algorithm that dynamically adjusts the safety margins produced by conformal risk control based on the current interaction context. Through experiments on human-robot navigation scenarios, we demonstrate that our approach significantly reduces collision rates and safety violations as compared to baseline methods while maintaining high success rates in goal-reaching tasks and efficient control. The code, simulations, and other supplementary material can be found on the project website: https://jakeagonzales.github.io/crc-cbf-website/.

Diffusion models have recently achieved remarkable success in generative modeling, yet their training dynamics across different noise levels remain highly imbalanced, which can lead to inefficient optimization and unstable learning behavior. In this work, we investigate this imbalance from the perspective of loss variance across log-SNR levels and propose a variance-aware adaptive weighting strategy to address it. The proposed approach dynamically adjusts training weights based on the observed variance distribution, encouraging a more balanced optimization process across noise levels. Extensive experiments on CIFAR-10 and CIFAR-100 demonstrate that the proposed method consistently improves generative performance over standard training schemes, achieving lower Fréchet Inception Distance (FID) while also reducing performance variance across random seeds. Additional analysis, including loss-log-SNR visualization, variance heatmaps, and ablation studies, further reveal that the adaptive weighting effectively stabilizes training dynamics. These results highlight the potential of variance-aware training strategies for improving diffusion model optimization.

Evaluating LLM reliability via scalar probabilities often fails to capture the structural dynamics of reasoning. We introduce TRACED, a framework that assesses reasoning quality through theoretically grounded geometric kinematics. By decomposing reasoning traces into Progress (displacement) and Stability (curvature), we reveal a distinct topological divergence: correct reasoning manifests as high-progress, stable trajectories, whereas hallucinations are characterized by low-progress, unstable patterns (stalled displacement with high curvature fluctuations). Leveraging these signatures, our probabilistic framework achieves competitive performance and superior robustness across diverse benchmarks. Crucially, TRACED bridges geometry and cognition by mapping high curvature to ''Hesitation Loops'' and displacement to ''Certainty Accumulation'', offering a physical lens to decode the internal dynamics of machine thought.

This paper presents a novel extension of neural scaling laws to Mixture-of-Experts (MoE) models, focusing on the optimal allocation of compute between expert and attention sub-layers. As MoE architectures have emerged as an efficient method for scaling model capacity without proportionally increasing computation, determining the optimal expert-attention compute ratio becomes critical. We define the ratio $r$ as the fraction of total FLOPs per token dedicated to the expert layers versus the attention layers, and explore how this ratio interacts with the overall compute budget and model sparsity. Through extensive experiments with GPT-style MoE Transformers, we empirically find that the optimal ratio $r^*$ follows a power-law relationship with total compute and varies with sparsity. Our analysis leads to an explicit formula for $r^*$, enabling precise control over the expert-attention compute allocation. We generalize the Chinchilla scaling law by incorporating this architectural parameter, providing a new framework for tuning MoE models beyond size and data. Our findings offer practical guidelines for designing efficient MoE models, optimizing performance while respecting fixed compute budgets.

Sparse autoencoders can localize where concepts live in language models, but not how they interact during multi-step reasoning. We propose Causal Concept Graphs (CCG): a directed acyclic graph over sparse, interpretable latent features, where edges capture learned causal dependencies between concepts. We combine task-conditioned sparse autoencoders for concept discovery with DAGMA-style differentiable structure learning for graph recovery and introduce the Causal Fidelity Score (CFS) to evaluate whether graph-guided interventions induce larger downstream effects than random ones. On ARC-Challenge, StrategyQA, and LogiQA with GPT-2 Medium, across five seeds ($n{=}15$ paired runs), CCG achieves $\CFS=5.654\pm0.625$, outperforming ROME-style tracing ($3.382\pm0.233$), SAE-only ranking ($2.479\pm0.196$), and a random baseline ($1.032\pm0.034$), with $p<0.0001$ after Bonferroni correction. Learned graphs are sparse (5-6\% edge density), domain-specific, and stable across seeds.

Emerging experimental evidence shows that writing with AI assistance can change both the views people express in writing and the opinions they hold afterwards. Yet, we lack substantive understanding of procedural and behavioral changes in co-writing with AI that underlie the observed opinion-shaping power of AI writing tools. We conducted a mixed-methods study, combining retrospective interviews with 19 participants about their AI co-writing experience with a quantitative analysis tracing engagement with ideas and opinions in 1{,}291 AI co-writing sessions. Our analysis shows that engaging with the AI's suggestions -- reading them and deciding whether to accept them -- becomes a central activity in the writing process, taking away from more traditional processes of ideation and language generation. As writers often do not complete their own ideation before engaging with suggestions, the suggested ideas and opinions seeded directions that writers then elaborated on. At the same time, writers did not notice the AI's influence and felt in full control of their writing, as they -- in principle -- could always edit the final text. We term this shift \textit{Reactive Writing}: an evaluation-first, suggestion-led writing practice that departs substantially from conventional composing in the presence of AI assistance and is highly vulnerable to AI-induced biases and opinion shifts.

Robotic systems operating in real-world environments often suffer from concept shift, where the input-output relationship changes due to latent environmental factors that are not directly observable. Conventional adaptation methods update model parameters, which may cause catastrophic forgetting and incur high computational cost. This paper proposes a latent Trend ID-based framework for few-shot adaptation in non-stationary environments. Instead of modifying model weights, a low-dimensional environmental state, referred to as the Trend ID, is estimated via backpropagation while the model parameters remain fixed. To prevent overfitting caused by per-sample latent variables, we introduce temporal regularization and a state transition model that enforces smooth evolution of the latent space. Experiments on a quantitative food grasping task demonstrate that the learned Trend IDs are distributed across distinct regions of the latent space with temporally consistent trajectories, and that few-shot adaptation to unseen environments is achieved without modifying model parameters. The proposed framework provides a scalable and interpretable solution for robotics applications operating across diverse and evolving environments.

Speech tokenizers are essential for connecting speech to large language models (LLMs) in multimodal systems. These tokenizers are expected to preserve both semantic and acoustic information for downstream understanding and generation. However, emerging evidence suggests that what is termed "semantic" in speech representations does not align with text-derived semantics: a mismatch that can degrade multimodal LLM performance. In this paper, we systematically analyze the information encoded by several widely used speech tokenizers, disentangling their semantic and phonetic content through word-level probing tasks, layerwise representation analysis, and cross-modal alignment metrics such as CKA. Our results show that current tokenizers primarily capture phonetic rather than lexical-semantic structure, and we derive practical implications for the design of next-generation speech tokenization methods.

Generative Recommender Systems (GR) increasingly model user behavior as a sequence generation task by interleaving item and action tokens. While effective, this formulation introduces significant structural and computational inefficiencies: it doubles sequence length, incurs quadratic overhead, and relies on implicit attention to recover the causal relationship between an item and its associated action. Furthermore, interleaving heterogeneous tokens forces the Transformer to disentangle semantically incompatible signals, leading to increased attention noise and reduced representation efficiency.In this work, we propose a principled reformulation of generative recommendation that aligns sequence modeling with underlying causal structures and attention theory. We demonstrate that current interleaving mechanisms act as inefficient proxies for similarity-weighted action pooling. To address this, we introduce two novel architectures that eliminate interleaved dependencies to reduce sequence complexity by 50%: Attention-based Late Fusion for Actions (AttnLFA) and Attention-based Mixed Value Pooling (AttnMVP). These models explicitly encode the $i_n \rightarrow a_n$ causal dependency while preserving the expressive power of Transformer-based sequence modeling.We evaluate our framework on large-scale product recommendation data from a major social network. Experimental results show that AttnLFA and AttnMVP consistently outperform interleaved baselines, achieving evaluation loss improvements of 0.29% and 0.80%, and significant gains in Normalized Entropy (NE). Crucially, these performance gains are accompanied by training time reductions of 23% and 12%, respectively. Our findings suggest that explicitly modeling item-action causality provides a superior design paradigm for scalable and efficient generative ranking.

The performance of task-oriented dialogue models is strongly tied to how well they track dialogue states, which records and updates user information across multi-turn interactions. However, current multi-domain DST encounters two key challenges: the difficulty of effectively modeling dialogue history and the limited availability of annotated data, both of which hinder model performance. To tackle the aforementioned problems, we develop a dynamic knowledge fusion framework applicable to multi-domain DST. The model operates in two stages: first, an encoder-only network trained with contrastive learning encodes dialogue history and candidate slots, selecting relevant slots based on correlation scores; second, dynamic knowledge fusion leverages the structured information of selected slots as contextual prompts to enhance the accuracy and consistency of dialogue state tracking. This design enables more accurate integration of dialogue context and domain knowledge. Results obtained from multi-domain dialogue benchmarks indicate that our method notably improves both tracking accuracy and generalization, validating its capability in handling complex dialogue scenarios.

Distilling reasoning capabilities from Large Reasoning Models (LRMs) into smaller models is typically constrained by the limitation of rejection sampling. Standard methods treat the teacher as a static filter, discarding complex "corner-case" problems where the teacher fails to explore valid solutions independently, thereby creating an artificial "Teacher Ceiling" for the student. In this work, we propose Hindsight Entropy-Assisted Learning (HEAL), an RL-free framework designed to bridge this reasoning gap. Drawing on the educational theory of the Zone of Proximal Development(ZPD), HEAL synergizes three core modules: (1) Guided Entropy-Assisted Repair (GEAR), an active intervention mechanism that detects critical reasoning breakpoints via entropy dynamics and injects targeted hindsight hints to repair broken trajectories; (2) Perplexity-Uncertainty Ratio Estimator (PURE), a rigorous filtering protocol that decouples genuine cognitive breakthroughs from spurious shortcuts; and (3) Progressive Answer-guided Curriculum Evolution (PACE), a three-stage distillation strategy that organizes training from foundational alignment to frontier breakthrough. Extensive experiments on multiple benchmarks demonstrate that HEAL significantly outperforms traditional SFT distillation and other baselines.

Congestion is a critical and challenging problem in communication networks. Congestion control protocols allow network applications to tune their sending rate in a way that optimizes their performance and the network utilization. In the common distributed setting, the applications cannot collaborate with each other directly but instead obtain similar estimations about the state of the network using latency and loss measurements. These measurements can be fed into analytical functions, referred to by utility functions, whose gradients help each and all distributed senders to converge to a desired state. The above process becomes extremely complicated when each application has different optimization goals and requirements. Crafting these utilization functions has been a research subject for over a decade, with small incremental changes requiring rigorous mathematical analysis as well as real-world experiments. In this work, we present GenCC, a framework leveraging the code generation capabilities of large language models (LLMs) coupled with realistic network testbed, to design congestion control utility functions. Using GenCC, we analyze the impact of different guidance strategies on the performance of the generated protocols, considering application-specific requirements and network capacity. Our results show that LLMs, guided by either a generative code evolution strategy or mathematical chain-of-thought (CoT), can obtain close to optimal results, improving state-of-the-art congestion control protocols by 37%-142%, depending on the scenario.

With the increasing volumes of Large Language Models (LLMs) and the expanding context lengths, attention computation has become a key performance bottleneck in LLM serving. For fast attention computation, recent practices often parallelize the attention heads on multiple GPUs, and also widely adopt attention sparsification to reduce the computation amount -- which selectively computes a subset of attention pairs under a preset sparsity budget. In this paper, we notice that attention heads of an LLM model often exhibit heterogeneous-yet-stable sparsity elasticities, which motivates us to enforce head-adaptive sparsity budgets to attain better efficiency while preserving high inference quality. Yet, from the system aspect, with heterogeneous sparsity levels, attention computation time on different heads would be inconsistent, yielding cross-GPU resource bubbles under head-parallel deployment. To further minimize such bubbles, we propose a novel attention deployment strategy called Sparsity-aware Head-Parallel Load Balance (S-HPLB). Experiments on long-context benchmark show that, S-HPLB can achieve a $2.88\times$ improvement in average attention computation latency without quality degradation.

Large language models (LLMs) have become a standard for multilingual evaluation, yet they exhibit a severe systematic translationese bias. In this paper, translationese bias is characterized as LLMs systematically favoring machine-translated text over human-authored references, particularly in low-resource languages. We attribute this bias to spurious correlations with (i) latent manifold alignment with English and (ii) cross-lingual predictability. To mitigate this bias, we propose DIBJudge, a robust fine-tuning framework that learns a minimally sufficient, judgment-critical representation via variational information compression, while explicitly isolating spurious factors into the dedicated bias branch. Furthermore, we incorporate a cross-covariance penalty that explicitly suppresses statistical dependence between robust and bias representations, thereby encouraging effective disentanglement. Extensive evaluations on multilingual reward modeling benchmarks and a dedicated translationese bias evaluation suite demonstrate that the proposed DIBJudge consistently outperforms strong baselines and substantially mitigates translationese bias.

We study the fixed-budget best-arm identification (BAI) problem in non-stationary linear bandits. Concretely, given a fixed time budget $T\in \mathbb{N}$, finite arm set $\mathcal{X} \subset \mathbb{R}^d$, and a potentially adversarial sequence of unknown parameters $\lbrace θ_t\rbrace_{t=1}^{T}$ (hence non-stationary), a learner aims to identify the arm with the largest cumulative reward $x_* = \arg\max_{x \in \mathcal{X}} x^\top\sum_{t=1}^T θ_t$ with high probability. In this setting, it is well-known that uniformly sampling arms from the G-optimal design yields a minimax-optimal error probability of $\exp\left(-Θ\left(T / H_{G}\right)\right)$, where $H_{G}$ scales proportionally with the dimension $d$. However, this notion of complexity is overly pessimistic, as it is derived from a lower bound in which the arm set consists only of the standard basis vectors, thus masking any potential advantages arising from arm sets with richer geometric structure. To address this, we establish an arm-set-dependent lower bound that, in contrast, holds for any arm set. Motivated by the ideas underlying our lower bound, we propose the Adjacent-optimal design, a specialization of the well-known $\mathcal{X}\mathcal{Y}$-optimal design, and develop the $\textsf{Adjacent-BAI}$ algorithm. We prove that the error probability of $\textsf{Adjacent-BAI}$ matches our lower bound up to constants, verifying the tightness of our lower bound, and establishing the arm-set-dependent complexity of this setting.

Large language models (LLMs) are increasingly deployed as AI agents that operate in short reasoning-action loops, interleaving model computation with external calls. Unlike traditional chat applications, these agentic workloads require inference serving systems to balance low latency, stable token emission, and throughput under multiple request arrivals from different AI agents. Recent deployments highlight a shift toward running small language models (SLMs) locally on consumer-grade GPUs, driven by privacy, compliance, and cost constraints. When heterogeneous requests overlap on a single GPU, long prefills and short decodes contend for resources, creating head-of-line blocking that destabilizes interactive performance. By analyzing agent workloads, we observe that their execution naturally separates into cold prefills, which process long system prompts, resume prefills, which append tool outputs to cached contexts, and short decodes, which are latency-critical. This mix intensifies contention compared to conventional chatbot serving. We present AgentServe, a single-GPU serving system that ensures stable multi-agent execution under such conditions by isolating prefills from decodes, applying dynamic budgeting to resume prefills, and allocating GPU resources through pre-established CUDA Green Context slots with adaptive control. Evaluation results show that AgentServe significantly improves latency stability while sustaining competitive throughput, achieving up to 2.8x TTFT improvement and 2.7x TPOT improvement over state-of-the-art baselines across different settings.

Federated active learning (FAL) seeks to reduce annotation cost under privacy constraints, yet its effectiveness degrades in realistic settings with severe global class imbalance and highly heterogeneous clients. We conduct a systematic study of query-model selection in FAL and uncover a central insight: the model that achieves more class-balanced sampling, especially for minority classes, consistently leads to better final performance. Moreover, global-model querying is beneficial only when the global distribution is highly imbalanced and client data are relatively homogeneous; otherwise, the local model is preferable. Based on these findings, we propose FairFAL, an adaptive class-fair FAL framework. FairFAL (1) infers global imbalance and local-global divergence via lightweight prediction discrepancy, enabling adaptive selection between global and local query models; (2) performs prototype-guided pseudo-labeling using global features to promote class-aware querying; and (3) applies a two-stage uncertainty-diversity balanced sampling strategy with k-center refinement. Experiments on five benchmarks show that FairFAL consistently outperforms state-of-the-art approaches under challenging long-tailed and non-IID settings. The code is available at https://github.com/chenchenzong/FairFAL.

Vision-Language-Action (VLA) models demonstrate impressive zero-shot generalization but frequently suffer from a "Precision-Reasoning Gap" in cluttered environments. This failure is driven by background-induced feature dilution, where high-frequency semantic noise corrupts the geometric grounding required for precise manipulation. To bridge this gap, we propose Concept-Gated Visual Distillation (CGVD), a training-free, model-agnostic inference framework that stabilizes VLA policies. CGVD operates by parsing instructions into safe and distractor sets, utilizing a two-layer target refinement process--combining cross-validation and spatial disambiguation--to explicitly penalize false positives and isolate genuine manipulation targets. We then process the scene via Fourier-based inpainting, generating a clean observation that actively suppresses semantic distractors while preserving critical spatial geometry and visual proprioception. Extensive evaluations in highly cluttered manipulation tasks demonstrate that CGVD prevents performance collapse. In environments with dense semantic distractors, our method significantly outperforms state-of-the-art baselines, achieving a 77.5% success rate compared to the baseline's 43.0%. By enforcing strict attribute adherence, CGVD establishes inference-time visual distillation as a critical prerequisite for robust robotic manipulation in the clutter.

While reasoning rerankers, such as Rank1, have demonstrated strong abilities in improving ranking relevance, it is unclear how they perform on other retrieval qualities such as fairness. We conduct the first systematic comparison of fairness between reasoning and non-reasoning rerankers. Using the TREC 2022 Fair Ranking Track dataset, we evaluate six reranking models across multiple retrieval settings and demographic attributes. Our findings demonstrate reasoning neither improve nor harm fairness compared to non-reasoning approaches. Our fairness metric, Attention-Weighted Rank Fairness (AWRF) remained stable (0.33-0.35) across all models, even as relevance varies substantially (nDCG 0.247-1.000). Demographic breakdown analysis revealed fairness gaps for geographic attributes regardless of model architecture. These results indicate that future work in specializing reasoning models to be aware of fairness attributes could lead to improvements, as current implementations preserve the fairness characteristics of their input ranking.

Autonomous driving in complex traffic requires planners that generalize beyond hand-crafted rules, motivating data-driven approaches that learn behavior from expert demonstrations. Diffusion-based trajectory planners have recently shown strong closed-loop performance by iteratively denoising a full-horizon plan, but they remain difficult to certify and can fail catastrophically in rare or out-of-distribution scenarios. To address this challenge, we present PC-Diffuser, a safety augmentation framework that embeds a certifiable, path-consistent barrier-function structure directly into the denoising loop of diffusion planning. The key idea is to make safety an intrinsic part of trajectory generation rather than a post-hoc fix: we enforce forward invariance along the rollout while preserving the diffusion model's intended path geometry. Specifically, PC-Diffuser (i) evaluates collision risk using a capsule-distance barrier function that better reflects vehicle geometry and reduces unnecessary conservativeness, (ii) converts denoised waypoints into dynamically feasible motion under a kinematic bicycle model, and (iii) applies a path-consistent safety filter that eliminates residual constraint violations without geometric distortion, so the corrected plan remains close to the learned distribution. By injecting these safety-consistent corrections at every denoising step and feeding the refined trajectory back into the diffusion process, PC-Diffuser enables iterative, context-aware safeguarding instead of post-hoc repair...

Silent and whispered speech offer promise for always-available voice interaction with AI, yet existing methods struggle to balance vocabulary size, wearability, silence, and noise robustness. We present NasoVoce, a nose-bridge-mounted interface that integrates a microphone and a vibration sensor. Positioned at the nasal pads of smart glasses, it unobtrusively captures both acoustic and vibration signals. The nasal bridge, close to the mouth, allows access to bone- and skin-conducted speech and enables reliable capture of low-volume utterances such as whispered speech. While the microphone captures high-quality audio, it is highly sensitive to environmental noise. Conversely, the vibration sensor is robust to noise but yields lower signal quality. By fusing these complementary inputs, NasoVoce generates high-quality speech robust against interference. Evaluation with Whisper Large-v2, PESQ, STOI, and MUSHRA ratings confirms improved recognition and quality. NasoVoce demonstrates the feasibility of a practical interface for always-available, continuous, and discreet AI voice conversations.

Social media text shows promise for monitoring trends in the opioid overdose crisis; however, the overwhelming majority of social media text is unrelated to opioids. When leveraging social media text to monitor trends in the ongoing opioid overdose crisis, a common strategy for identifying relevant content is to use a lexicon of opioid-related terms as inclusion criteria. However, many slang terms for opioids, such as "smack" or "blues," have common non-opioid meanings, making them ambiguous. The advanced textual reasoning capability of large language models (LLMs) presents an opportunity to disambiguate these slang terms at scale. We present three tasks on which to evaluate four state-of-the-art LLMs (GPT-4, GPT-5, Gemini 2.5 Pro, and Claude Sonnet 4.5): a lexicon-based setting, in which the LLM must disambiguate a specific term within the context of a given post; a lexicon-free setting, in which the LLM must identify opioid-related posts from context without a lexicon; and an emergent slang setting, in which the LLM must identify opioid-related posts with simulated new slang terms. All four LLMs showed excellent performance across all tasks. In both subtasks of the lexicon-based setting, LLM F1 scores ("fenty" subtask: 0.824-0.972; "smack" subtask: 0.540-0.862) far exceeded those of the best lexicon strategy (0.126 and 0.009, respectively). In the lexicon-free task, LLM F1 scores (0.544-0.769) surpassed those of lexicons (0.080-0.540), and LLMs demonstrated uniformly higher recall. On emergent slang, all LLMs had higher accuracy (average: 0.784), F1 score (average: 0.712), precision (average: 0.981), and recall (average: 0.587) than the two lexicons assessed. Our results show that LLMs can be used to identify relevant content for low-prevalence topics, including but not limited to opioid references, enhancing data provided to downstream analyses and predictive models.

Machine learning models can represent climate processes that are nonlocal in horizontal space, height, and time, often by combining information across these dimensions in highly nonlinear ways. While this can improve predictive skill, it makes learned relationships difficult to interpret and prone to overfitting as the extent of nonlocal information grows. We address this challenge by introducing data-driven integration kernels, a framework that adds structure to nonlocal operator learning by explicitly separating nonlocal information aggregation from local nonlinear prediction. Each spatiotemporal predictor field is first integrated using learnable kernels (defined as continuous weighting functions over horizontal space, height, and/or time), after which a local nonlinear mapping is applied only to the resulting kernel-integrated features and any optional local inputs. This design confines nonlinear interactions to a small set of integrated features and makes each kernel directly interpretable as a weighting pattern that reveals which horizontal locations, vertical levels, and past timesteps contribute most to the prediction. We demonstrate the framework for South Asian monsoon precipitation using a hierarchy of neural network models with increasing structure, including baseline, nonparametric kernel, and parametric kernel models. Across this hierarchy, kernel-based models achieve near-baseline performance with far fewer trainable parameters, showing that much of the relevant nonlocal information can be captured through a small set of interpretable integrations when appropriate structural constraints are imposed.

Judging the novelty of research ideas is crucial for advancing science, enabling the identification of unexplored directions, and ensuring contributions meaningfully extend existing knowledge rather than reiterate minor variations. However, given the exponential growth of scientific literature, manually judging the novelty of research ideas through literature reviews is labor-intensive, subjective, and infeasible at scale. Therefore, recent efforts have proposed automated approaches for research idea novelty judgment. Yet, evaluation of these approaches remains largely inconsistent and is typically based on non-standardized human evaluations, hindering large-scale, comparable evaluations. To address this, we introduce RINoBench, the first comprehensive benchmark for large-scale evaluation of research idea novelty judgments. It comprises 1,381 research ideas derived from and judged by human experts as well as nine automated evaluation metrics designed to assess both rubric-based novelty scores and textual justifications of novelty judgments. Using this benchmark, we evaluate several state-of-the-art large language models (LLMs) on their ability to judge the novelty of research ideas. Our findings reveal that while LLM-generated reasoning closely mirrors human rationales, this alignment does not reliably translate into accurate novelty judgments, which diverge significantly from human gold standard judgments - even among leading reasoning-capable models. Data and code available at: https://github.com/TimSchopf/RINoBench.

A plethora of protein language models have been released in recent years. Yet comparatively little work has addressed how to best sample from them to optimize desired biological properties. We fill this gap by proposing a flexible, effective sampling method for masked language models (MLMs), and by systematically evaluating models and methods both in silico and in vitro on actual antibody therapeutics campaigns. Firstly, we propose sampling with stochastic beam search, exploiting the fact that MLMs are remarkably efficient at evaluating the pseudo-perplexity of the entire 1-edit neighborhood of a sequence. Reframing generation in terms of entire-sequence evaluation enables flexible guidance with multiple optimization objectives. Secondly, we report results from our extensive in vitro head-to-head evaluation for the antibody engineering setting. This reveals that choice of sampling method is at least as impactful as the model used, motivating future research into this under-explored area.

A basic unanswered question in neural network training is: what is the best learning rate schedule shape for a given workload? The choice of learning rate schedule is a key factor in the success or failure of the training process, but beyond having some kind of warmup and decay, there is no consensus on what makes a good schedule shape. To answer this question, we designed a search procedure to find the best shapes within a parameterized schedule family. Our approach factors out the schedule shape from the base learning rate, which otherwise would dominate cross-schedule comparisons. We applied our search procedure to a variety of schedule families on three workloads: linear regression, image classification on CIFAR-10, and small-scale language modeling on Wikitext103. We showed that our search procedure indeed generally found near-optimal schedules. We found that warmup and decay are robust features of good schedules, and that commonly used schedule families are not optimal on these workloads. Finally, we explored how the outputs of our shape search depend on other optimization hyperparameters, and found that weight decay can have a strong effect on the optimal schedule shape. To the best of our knowledge, our results represent the most comprehensive results on near-optimal schedule shapes for deep neural network training, to date.

This work introduces a regime-aware in-context learning framework that leverages large language models (LLMs) for financial volatility forecasting under nonstationary market conditions. The proposed approach deploys pretrained LLMs to reason over historical volatility patterns and adjust their predictions without parameter fine-tuning. We develop an oracle-guided refinement procedure that constructs regime-aware demonstrations from training data. An LLM is then deployed as an in-context learner that predicts the next-step volatility from the input sequence using demonstrations sampled conditional to the estimated market label. This conditional sampling strategy enables the LLM to adapt its predictions to regime-dependent volatility dynamics through contextual reasoning alone. Experiments with multiple financial datasets show that the proposed regime-aware in-context learning framework outperforms both classical volatility forecasting approaches and direct one-shot learning, especially during high-volatility periods.

The rapid rise of large language models (LLMs) and their ability to capture semantic relationships has led to their adoption in a wide range of applications. Text-attributed graphs (TAGs) are a notable example where LLMs can be combined with Graph Neural Networks to improve the performance of node classification. In TAGs, each node is associated with textual content and such graphs are commonly seen in various domains such as social networks, citation graphs, recommendation systems, etc. Effectively learning from TAGs would enable better representations of both structural and textual representations of the graph and improve decision-making in relevant domains. We present GaLoRA, a parameter-efficient framework that integrates structural information into LLMs. GaLoRA demonstrates competitive performance on node classification tasks with TAGs, performing on par with state-of-the-art models with just 0.24% of the parameter count required by full LLM fine-tuning. We experiment with three real-world datasets to showcase GaLoRA's effectiveness in combining structural and semantical information on TAGs.

Large Language Models (LLMs) have advanced reasoning through techniques like Chain-of-Thought (CoT). However, their reasoning largely re-mains textual and hypothetical, lacking empirical grounding in complex, dynamic domains like transportation. This paper introduces Simulation-in-the-Reasoning (SiR), a novel conceptual framework that embeds domain-specific simulators directly into the LLM reasoning loop. By treating intermediate reasoning steps as executable simulation experiments, SiR transforms LLM reasoning from narrative plausibility into a falsifiable, hypothesis-simulate-analyze workflow. We discuss applications, where LLM can formulate Intelligent Transport System (ITS) strategy hypotheses, invoke a traffic simulator via the Model Context Protocol (MCP), evaluate results under different demand patterns, and refine strategies through verification and aggregation. While implementing the framework is part of our ongoing work, this paper primarily establishes the conceptual foundation, discusses design considerations like API granularity, and outlines the vision of SiR as a cornerstone for interactive transportation digital twins. We argue that SiR represents a critical step towards trustworthy, empirically-validated AI for autonomous transportation systems.

The remarkable progress of vision-language models (VLMs) has enabled GUI agents to interact with computers in a human-like manner. Yet real-world computer-use tasks remain difficult due to long-horizon workflows, diverse interfaces, and frequent intermediate errors. Prior work equips agents with external memory built from large collections of trajectories, but relies on flat retrieval over discrete summaries or continuous embeddings, falling short of the structured organization and self-evolving characteristics of human memory. Inspired by the brain, we propose Hybrid Self-evolving Structured Memory (HyMEM), a graph-based memory that couples discrete high-level symbolic nodes with continuous trajectory embeddings. HyMEM maintains a graph structure to support multi-hop retrieval, self-evolution via node update operations, and on-the-fly working-memory refreshing during inference. Extensive experiments show that HyMEM consistently improves open-source GUI agents, enabling 7B/8B backbones to match or surpass strong closed-source models; notably, it boosts Qwen2.5-VL-7B by +22.5% and outperforms Gemini2.5-Pro-Vision and GPT-4o.

Whether uniquely quantum resources confer advantages in fully classical, competitive environments remains an open question. Competitive zero-sum reinforcement learning is particularly challenging, as success requires modelling dynamic interactions between opposing agents rather than static state-action mappings. Here, we conduct a controlled study isolating the role of quantum entanglement in a quantum-classical hybrid agent trained on Pong, a competitive Markov game. An 8-qubit parameterised quantum circuit serves as a feature extractor within a proximal policy optimisation framework, allowing direct comparison between separable circuits and architectures incorporating fixed (CZ) or trainable (IsingZZ) entangling gates. Entangled circuits consistently outperform separable counterparts with comparable parameter counts and, in low-capacity regimes, match or exceed classical multilayer perceptron baselines. Representation similarity analysis further shows that entangled circuits learn structurally distinct features, consistent with improved modelling of interacting state variables. These findings establish entanglement as a function resource for representation learning in competitive reinforcement learning.

LLM-as-a-Judge is a flexible framework for text evaluation, which allows us to obtain scores for the quality of a given text from various perspectives by changing the prompt template. Two main challenges in using LLM-as-a-Judge are computational cost of LLM inference, especially when evaluating a large number of texts, and inherent bias of an LLM evaluator. To address these issues and reveal the structure of score bias caused by an LLM evaluator, we propose to apply a tensor clustering method to a given LLM-as-a-Judge score tensor, whose entries are the scores for different combinations of questions, answerers, and evaluators. Specifically, we develop a new tensor clustering method MultiwayPAM, with which we can simultaneously estimate the cluster membership and the medoids for each mode of a given data tensor. By observing the medoids obtained by MultiwayPAM, we can gain knowledge about the membership of each question/answerer/evaluator cluster. We experimentally show the effectiveness of MultiwayPAM by applying it to the score tensors for two practical datasets.

Recent digitisation efforts in natural history museums have produced large volumes of collection data, yet their scale and scientific complexity often hinder public access and understanding. Conventional data management tools, such as databases, restrict exploration through keyword-based search or require specialised schema knowledge. This paper presents a system design that uses conversational AI to query nearly 1.7 million digitised specimen records from the life-science collections of the Australian Museum. Designed and developed through a human-centred design process, the system contains an interactive map for visual-spatial exploration and a natural-language conversational agent that retrieves detailed specimen data and answers collection-specific questions. The system leverages function-calling capabilities of contemporary large language models to dynamically retrieve structured data from external APIs, enabling fast, real-time interaction with extensive yet frequently updated datasets. Our work provides a new approach of connecting large museum collections with natural language-based queries and informs future designs of scientific AI agents for natural history museums.

A key challenge in travel demand analysis is the presence of unobserved factors that may generate non-causal dependencies, obscuring the true causal effects. To address the issue, the study introduces a novel deep learning based fully interpretable joint modelling framework, Copula-ResLogit, which integrates the flexibility of Residual Neural Network (ResNet) architectures with the dependence capturing capabilities of copula models. This hybrid structure enables us to first detect unobserved confounding through traditional copula function based joint modelling and then mitigate these hidden associations by incorporating deep learning components. The study applies this framework to two case studies, including the relationship between stress levels and wait time of pedestrians when crossing mid block in VR and the dependencies between travel mode choice and travel distance in London travel behaviour data. Results show that Copula-ResLogit substantially reduces or eliminates the dependencies, demonstrating the ability of residual layers to account for hidden confounding effects.

Geometry-grounded learning asks models to respect structure in the problem domain rather than treating observations as arbitrary vectors. Motivated by this view, we revisit a classical but underused primitive for comparing datasets: linear relations between two data matrices, expressed via the co-span constraint $Ax = By = z$ in a shared ambient space. To operationalize this comparison, we use the generalized singular value decomposition (GSVD) as a joint coordinate system for two subspaces. In particular, we exploit the GSVD form $A = HCU$, $B = HSV$ with $C^{\top}C + S^{\top}S = I$, which separates shared versus dataset-specific directions through the diagonal structure of $(C, S)$. From these factors we derive an interpretable *angle score* $θ(z) \in [0, π/2]$ for a sample $z$, quantifying whether z is explained relatively more by $A$, more by $B$, or comparably by both. The primary role of $θ(z)$ is as a *per-sample geometric diagnostic*. We illustrate the behavior of the score on MNIST through angle distributions and representative GSVD directions. A binary classifier derived from $θ(z)$ is presented as an illustrative application of the score as an interpretable diagnostic tool.

While score-based generative models have emerged as powerful priors for solving inverse problems, directly integrating them into optimization algorithms such as ADMM remains nontrivial. Two central challenges arise: i) the mismatch between the noisy data manifolds used to train the score functions and the geometry of ADMM iterates, especially due to the influence of dual variables, and ii) the lack of convergence understanding when ADMM is equipped with score-based denoisers. To address the manifold mismatch issue, we propose ADMM plug-and-play (ADMM-PnP) with the AC-DC denoiser, a new framework that embeds a three-stage denoiser into ADMM: (1) auto-correction (AC) via additive Gaussian noise, (2) directional correction (DC) using conditional Langevin dynamics, and (3) score-based denoising. In terms of convergence, we establish two results: first, under proper denoiser parameters, each ADMM iteration is a weakly nonexpansive operator, ensuring high-probability fixed-point $\textit{ball convergence}$ using a constant step size; second, under more relaxed conditions, the AC-DC denoiser is a bounded denoiser, which leads to convergence under an adaptive step size schedule. Experiments on a range of inverse problems demonstrate that our method consistently improves solution quality over a variety of baselines.

Aligning generative recommender systems to user preferences via post-training is critical for closing the gap between next-item prediction and actual recommendation quality. Existing post-training methods are ill-suited for production-scale systems: RLHF methods reward hack due to noisy user feedback and unreliable reward models, offline RL alternatives require propensity scores that are unavailable, and online interaction is infeasible. We identify exponential reward-weighted SFT with weights $w = \exp(r/λ)$ as uniquely suited to this setting, and provide the theoretical and empirical foundations that explain why. By optimizing directly on observed rewards without querying a learned reward model, the method is immune to reward hacking, requires no propensity scores, and is fully offline. We prove the first policy improvement guarantees for this setting under noisy rewards, showing that the gap scales only logarithmically with catalog size and remains informative even for large item catalogs. Crucially, we show that temperature $λ$ explicitly and quantifiably controls the robustness-improvement tradeoff, providing practitioners with a single interpretable regularization hyperparameter with theoretical grounding. Experiments on three open-source and one proprietary dataset against four baselines confirm that exponential reward weighting is simple, scalable, and consistently outperforms RLHF-based alternatives.

In this paper, we propose a fast method for estimating the condition number of sparse matrices using graph neural networks (GNNs). To enable efficient training and inference of GNNs, our proposed feature engineering for GNNs achieves $\mathrm{O}(\mathrm{nnz} + n)$, where $\mathrm{nnz}$ is the number of non-zero elements in the matrix and $n$ denotes the matrix dimension. We propose two prediction schemes for estimating the matrix condition number using GNNs. The extensive experiments for the two schemes are conducted for 1-norm and 2-norm condition number estimation, which show that our method achieves a significant speedup over the Hager-Higham and Lanczos methods.

Autonomous AI agents powered by large language models (LLMs) are increasingly deployed in real-world applications, where reliable and robust behavior is critical. However, existing agent evaluation frameworks either rely heavily on manual efforts, operate within simulated environments, or lack focus on testing complex, multimodal, real-world agents. We introduce SpecOps, a novel, fully automated testing framework designed to evaluate GUI-based AI agents in real-world environments. SpecOps decomposes the testing process into four specialized phases - test case generation, environment setup, test execution, and validation - each handled by a distinct LLM-based specialist agent. This structured architecture addresses key challenges including end-to-end task coherence, robust error handling, and adaptability across diverse agent platforms including CLI tools, web apps, and browser extensions. In comprehensive evaluations across five diverse real-world agents, SpecOps outperforms baselines including general-purpose agentic systems such as AutoGPT and LLM-crafted automation scripts in planning accuracy, execution success, and bug detection effectiveness. SpecOps identifies 164 true bugs in the real-world agents with an F1 score of 0.89. With a cost of under 0.73 USD and a runtime of under eight minutes per test, it demonstrates its practical viability and superiority in automated, real-world agent testing.

Context: Open-source ecosystems rely on sustained package maintenance. When maintenance slows or stops, Technical Lag (TL), the gap between installed and latest dependency versions accumulates, creating security and sustainability risks. However, some existing TL metrics, such as Version Lag, struggle to distinguish between actively maintained and abandoned packages, leading to a systematic underestimation of risk. Objective: We investigate the relationship between Version Lag and software abandonment by (i) identifying which repository-level signals reliably distinguish sustained maintenance from long-term decline, (ii) quantifying how Version Lag magnitude and persistence differ across maintenance states, and (iii) evaluating how maintenance-aware metrics change the identification of high-risk dependencies. Method: We introduce Maintenance-Aware Lag and Technical Abandonment (MALTA), a scoring framework comprising three metrics: Development Activity Score (DAS), Maintainer Responsiveness Score (MRS), and Repository Metadata Viability Score (RMVS). We evaluate MALTA on a dataset of 11,047 Debian packages linked to upstream GitHub repositories, encompassing 1.7 million commits and 4.2 million pull requests. Results: MALTA achieves AUC = 0.783 for classifying active versus declining maintenance. Most significantly, 62.2% of packages classified as "Low Risk" by Version Lag alone are reclassified as "High Risk" when MALTA signals are incorporated. These discordant packages average 2019 days since their last commit, with 9.8% having archived repositories. Conclusions: Version Lag metrics systematically miss abandoned packages, a blind spot affecting the majority of dependencies in distribution ecosystems. MALTA identifies a substantial discordant population invisible to Version Lag by distinguishing resolvable lag from terminal lag caused by upstream abandonment.

We introduce Distribution Contractive Reinforcement Learning (DICE-RL), a framework that uses reinforcement learning (RL) as a "distribution contraction" operator to refine pretrained generative robot policies. DICE-RL turns a pretrained behavior prior into a high-performing "pro" policy by amplifying high-success behaviors from online feedback. We pretrain a diffusion- or flow-based policy for broad behavioral coverage, then finetune it with a stable, sample-efficient residual off-policy RL framework that combines selective behavior regularization with value-guided action selection. Extensive experiments and analyses show that DICE-RL reliably improves performance with strong stability and sample efficiency. It enables mastery of complex long-horizon manipulation skills directly from high-dimensional pixel inputs, both in simulation and on a real robot. Project website: https://zhanyisun.github.io/dice.rl.2026/.

We report the discovery and extraction of a compact hematopoietic algorithm from the single-cell foundation model scGPT, to our knowledge the first biologically useful, competitive algorithm extracted from a foundation model via mechanistic interpretability. We show that scGPT internally encodes a compact hematopoietic manifold with significant developmental branch structure, validated on a strict non-overlap Tabula Sapiens external panel and confirmed via frozen-head zero-shot transfer to an independent multi-donor immune panel. To isolate this geometry, we introduce a general three-stage extraction method consisting of direct operator export from frozen attention weights, a lightweight learned adaptor, and a task-specific readout, producing a standalone algorithm without target-dataset retraining. In 88-split donor-holdout benchmarks against scVI, Palantir, DPT, CellTypist, PCA, and raw-expression baselines, the extracted algorithm achieves the strongest pseudotime-depth ordering and leads on key subtype endpoints (CD4/CD8 AUROC 0.867, mono/macro AUROC 0.951). Compared to standard probing of frozen scGPT embeddings with a 3-layer MLP, the extracted head is BH-significantly better on 6/8 classification endpoints while completing a full 12-split evaluation campaign 34.5x faster with approximately 1000x fewer trainable parameters. The exported operator compresses from three pooled attention heads to a single head without statistically significant loss, and further to a rank-64 surrogate. Mechanistic interpretability of the compact operator reveals a concentrated four-factor core explaining 66.2% of ablation impact, with factors resolving into explicit T/lymphoid, B/plasma, granulocytic, and monocyte/macrophage gene programs. A supplementary second-manifold validation (intercellular communication geometry) confirms that the extraction method generalizes beyond hematopoiesis.

Synthetic tabular data generation addresses data scarcity and privacy constraints in a variety of domains. Tabular Prior-Data Fitted Network (TabPFN), a recent foundation model for tabular data, has been shown capable of generating high-quality synthetic tabular data. However, TabPFN is autoregressive: features are generated sequentially by conditioning on the previous ones, depending on the order in which they appear in the input data. We demonstrate that when the feature order conflicts with causal structure, the model produces spurious correlations that impair its ability to generate synthetic data and preserve causal effects. We address this limitation by integrating causal structure into TabPFN's generation process through two complementary approaches: Directed Acyclic Graph (DAG)-aware conditioning, which samples each variable given its causal parents, and a Completed Partially Directed Acyclic Graph (CPDAG)-based strategy for scenarios with partial causal knowledge. We evaluate these approaches on controlled benchmarks and six CSuite datasets, assessing structural fidelity, distributional alignment, privacy preservation, and Average Treatment Effect (ATE) preservation. Across most settings, DAG-aware conditioning improves the quality and stability of synthetic data relative to vanilla TabPFN. The CPDAG-based strategy shows moderate improvements, with effectiveness depending on the number of oriented edges. These results indicate that injecting causal structure into autoregressive generation enhances the reliability of synthetic tabular data.

Brain imaging classification is commonly approached from two perspectives: modeling the full image volume to capture global anatomical context, or constructing ROI-based graphs to encode localized and topological interactions. Although both representations have demonstrated independent efficacy, their relative contributions and potential complementarity remain insufficiently understood. Existing fusion approaches are typically task-specific and do not enable controlled evaluation of each representation under consistent training settings. To address this gap, we propose a unified cross-view contrastive framework for joint imaging-ROI representation learning. Our method learns subject-level global (imaging) and local (ROI-graph) embeddings and aligns them in a shared latent space using a bidirectional contrastive objective, encouraging representations from the same subject to converge while separating those from different subjects. This alignment produces comparable embeddings suitable for downstream fusion and enables systematic evaluation of imaging-only, ROI-only, and joint configurations within a unified training protocol. Extensive experiments on the ADHD-200 and ABIDE datasets demonstrate that joint learning consistently improves classification performance over either branch alone across multiple backbone choices. Moreover, interpretability analyses reveal that imaging-based and ROI-based branches emphasize distinct yet complementary discriminative patterns, explaining the observed performance gains. These findings provide principled evidence that explicitly integrating global volumetric and ROI-level representations is a promising direction for neuroimaging-based brain disorder classification. The source code is available at https://anonymous.4open.science/r/imaging-roi-contrastive-152C/.

Bayesian hierarchical models are frequently used in practical data analysis contexts. One interpretation of these models is that they provide an indirect way of assigning a prior for unknown parameters, through the introduction of hyperparameters. The resulting marginal prior for the parameters (integrating over the hyperparameters) is usually dependent, so that learning one parameter provides some information about the others. In this contribution, I will demonstrate that, when the prior given the hyperparameters is a canonical distribution (a maximum entropy distribution with moment constraints), the dependent marginal prior also has a maximum entropy property, with a different constraint. This constraint is on the marginal distribution of some function of the unknown quantities. The results shed light on what information is actually being assumed when we assign a hierarchical model.

A commonly used family of RL algorithms for diffusion policies conducts softmax reweighting over the behavior policy, which usually induces an over-greedy policy and fails to leverage feedback from negative samples. In this work, we introduce Signed Measure Policy Optimization (SiMPO), a simple and unified framework that generalizes reweighting scheme in diffusion RL with general monotonic functions. SiMPO revisits diffusion RL via a two-stage measure matching lens. First, we construct a virtual target policy by $f$-divergence regularized policy optimization, where we can relax the non-negativity constraint to allow for a signed target measure. Second, we use this signed measure to guide diffusion or flow models through reweighted matching. This formulation offers two key advantages: a) it generalizes to arbitrary monotonically increasing weighting functions; and b) it provides a principled justification and practical guidance for negative reweighting. Furthermore, we provide geometric interpretations to illustrate how negative reweighting actively repels the policy from suboptimal actions. Extensive empirical evaluations demonstrate that SiMPO achieves superior performance by leveraging these flexible weighting schemes, and we provide practical guidelines for selecting reweighting methods tailored to the reward landscape.

Engineering analysis automation in product development relies on rigid interfaces between tools, data formats and documented processes. When these interfaces change, as they routinely do as the product evolves in the engineering ecosystem, the automation support breaks. This paper presents a DUCTILE (Delegated, User-supervised Coordination of Tool- and document-Integrated LLM-Enabled) agentic orchestration, an approach for developing, executing and evaluating LLM-based agentic automation support of engineering analysis tasks. The approach separates adaptive orchestration, performed by the LLM agent, from deterministic execution, performed by verified engineering tools. The agent interprets documented design practices, inspects input data and adapts the processing path, while the engineer supervises and exercises final judgment. DUCTILE is demonstrated on an industrial structural analysis task at an aerospace manufacturer, where the agent handled input deviations in format, units, naming conventions and methodology that would break traditional scripted pipelines. Evaluation against expert-defined acceptance criteria and deployment with practicing engineers confirm that the approach produces correct, methodologically compliant results across repeated independent runs. The paper discusses practical consequences of adopting agentic automation, including unintended effects on the nature of engineering work and the tension between removing mundane tasks and creating an exhausting supervisory role.

The potential for neuromorphic computing to provide intrinsic fault tolerance has long been speculated, but the brain's robustness in neuromorphic applications has yet to be demonstrated. Here, we show that a previously described, natively spiking neuromorphic algorithm for solving partial differential equations is intrinsically tolerant to structural perturbations in the form of ablated neurons and dropped spikes. The tolerance band for these perturbations is large: we find that as many as 32 percent of the neurons and up to 90 percent of the spikes may be entirely dropped before a significant degradation in the accuracy results. Furthermore, this robustness is tunable through structural hyperparameters. This work demonstrates that the specific brain-like inspiration behind the algorithm contributes to a significant degree of robustness expected from brain-like neuromorphic algorithms.

Recent studies show that the safety alignment of large language models (LLMs) can be easily compromised even by seemingly non-adversarial fine-tuning. To preserve safety alignment during fine-tuning, a widely used strategy is to jointly optimize safety and task objectives by mixing in the original alignment data, which is typically inaccessible even for open-weight LLMs. Inspired by generative replay in continual learning, we propose Generative Replay for Safety Alignment Preservation (GR-SAP), a unified framework that synthesizes domain-specific alignment data from LLMs and integrate them during downstream adaption to preserve safety alignment. Theoretical and empirical analyses demonstrate this synthetic data serves as a reliable proxy for the original alignment data. Experiments across various models and downstream tasks show that GR-SAP substantially mitigates fine-tuning-induced safety degradation while maintaining comparable downstream performance. Our code is available at https://github.com/chili-lab/gr-sap.

Existing high performance blockchains verify one signature per transaction on the critical path, which creates O(N) verification cost, high hardware pressure, and difficult post quantum migration. This paper presents ACE Runtime, a ZKP native execution layer built on identity authorization separation. We replace per transaction signature checks with lightweight HMAC attestations in the hot path, then generate one aggregated zero knowledge finality certificate per block in an asynchronous prove stage. The system is organized as an Attest Execute Prove pipeline with two tier finality: soft finality from BFT voting and hard finality from proof verification. Under standard cryptographic assumptions, we provide formal arguments for attestation unforgeability and hard finality irreversibility. We also define a two phase timeout and backup proving path with witness availability gossip for liveness under builder failure. Quantitative results combine analytical modeling with reference implementation measurements. The prototype shows low CPU orchestration overhead, while model driven analysis projects constant per block verification cost, lower validator hardware requirements for non builders, and better bandwidth efficiency than per transaction signature designs. These results indicate that identity authorization separation is a practical architecture for sub second cryptographic finality with a clear path toward stronger post quantum components.

In modern wireless networks, radio channels serve a dual role. Whilst their primary function is to carry bits of information from a transmitter to a receiver, the intrinsic sensitivity of transmitted signals to the physical structure of the environment makes the channel a powerful source of knowledge about the world. In this paper, we consider an agent that learns about its environment using a quantum sensing probe, optimised using a quantum circuit, which interacts with the radio-frequency (RF) electromagnetic field. We use data obtained from a ray-tracer to train the quantum circuit and learning model and we provide extensive experiments under realistic conditions on a localisation task. We show that using quantum sensors to learn from radio signals can enable intelligent systems that require no channel measurements at deployment, remain sensitive to weak and obstructed RF signals, and can learn about the world despite operating with strictly less information than classical baselines.

Class-incremental learning (CIL) aims to acquire new classes over time while retaining prior knowledge, yet most setups and methods assume balanced task streams. In practice, the number of classes per task often varies significantly. We refer to this as step imbalance, where large tasks that contain more classes dominate learning and small tasks inject unstable updates. Existing CIL methods assume balanced tasks and therefore treat all tasks uniformly, producing imbalanced updates that degrade overall learning performance. To address this challenge, we propose One-A, a unified and imbalance-aware framework that incrementally merges task updates into a single adapter, maintaining constant inference cost. One-A performs asymmetric subspace alignment to preserve dominant subspaces learned from large tasks while constraining low-information updates within them. An information-adaptive weighting balances the contribution between base and new adapters, and a directional gating mechanism selectively fuses updates along each singular direction, maintaining stability in head directions and plasticity in tail ones. Across multiple benchmarks and step-imbalanced streams, One-A achieves competitive accuracy with significantly low inference overhead, showing that a single, asymmetrically fused adapter can remain both adaptive to dynamic task sizes and efficient at deployment.

Deep learning models achieve remarkable predictive performance, yet their black-box nature limits transparency and trustworthiness. Although numerous explainable artificial intelligence (XAI) methods have been proposed, they primarily provide saliency maps or concepts (i.e., unstructured interpretability). Existing approaches often rely on auxiliary models (\eg, GPT, CLIP) to describe model behavior, thereby compromising faithfulness to the original models. We propose Interpretability to Explainability (I2X), a framework that builds structured explanations directly from unstructured interpretability by quantifying progress at selected checkpoints during training using prototypes extracted from post-hoc XAI methods (e.g., GradCAM). I2X answers the question of "why does it look there" by providing a structured view of both intra- and inter-class decision making during training. Experiments on MNIST and CIFAR10 demonstrate effectiveness of I2X to reveal prototype-based inference process of various image classification models. Moreover, we demonstrate that I2X can be used to improve predictions across different model architectures and datasets: we can identify uncertain prototypes recognized by I2X and then use targeted perturbation of samples that allows fine-tuning to ultimately improve accuracy. Thus, I2X not only faithfully explains model behavior but also provides a practical approach to guide optimization toward desired targets.

Evaluating student responses, from long essays to short factual answers, is a key challenge in educational NLP. Automated Essay Scoring (AES) focuses on holistic writing qualities such as coherence and argumentation, while Automatic Short Answer Grading (ASAG) emphasizes factual correctness and conceptual understanding. Despite their shared goal, these paradigms have progressed in isolation with fragmented datasets, inconsistent metrics, and separate communities. We introduce S-GRADES (Studying Generalization of Student Response Assessments in Diverse Evaluative Settings), a web-based benchmark that consolidates 14 diverse grading datasets under a unified interface with standardized access and reproducible evaluation protocols. The benchmark is fully open-source and designed for extensibility, enabling continuous integration of new datasets and evaluation settings. To demonstrate the utility of S-GRADES, we evaluate three state-of-the-art large language models across the benchmark using multiple reasoning strategies in prompting. We further examine the effects of exemplar selection and cross-dataset exemplar transfer. Our analyses illustrate how benchmark-driven evaluation reveals reliability and generalization gaps across essay and short-answer grading tasks, highlighting the importance of standardized, cross-paradigm assessment.

In this paper, we propose a trust-region interior-point stochastic sequential quadratic programming (TR-IP-SSQP) method for solving optimization problems with a stochastic objective and deterministic nonlinear equality and inequality constraints. In this setting, exact evaluations of the objective function and its gradient are unavailable, but their stochastic estimates can be constructed. In particular, at each iteration our method builds stochastic oracles, which estimate the objective value and gradient to satisfy proper adaptive accuracy conditions with a fixed probability. To handle inequality constraints, we adopt an interior-point method (IPM), in which the barrier parameter follows a prescribed decaying sequence. Under standard assumptions, we establish global almost-sure convergence of the proposed method to first-order stationary points. We implement the method on a subset of problems from the CUTEst test set, as well as on logistic regression problems, to demonstrate its practical performance.

Cross-entropy loss has long been the standard choice for training deep neural networks, yet it suffers from interpretability limitations, unbounded weight growth, and inefficiencies that can contribute to costly training dynamics. The harmonic loss is a distance-based alternative grounded in Euclidean geometry that improves interpretability and mitigates phenomena such as grokking, or delayed generalization on the test set. However, the study of harmonic loss remains narrow: only Euclidean distance is explored, and no systematic evaluation of computational efficiency or sustainability was conducted. We extend harmonic loss by systematically investigating a broad spectrum of distance metrics as replacements for the Euclidean distance. We comprehensively evaluate distance-tailored harmonic losses on both vision backbones and large language models. Our analysis is framed around a three-way evaluation of model performance, interpretability, and sustainability. On vision tasks, cosine distances provide the most favorable trade-off, consistently improving accuracy while lowering carbon emissions, whereas Bray-Curtis and Mahalanobis further enhance interpretability at varying efficiency costs. On language models, cosine-based harmonic losses improve gradient and learning stability, strengthen representation structure, and reduce emissions relative to cross-entropy and Euclidean heads. Our code is available at: https://anonymous.4open.science/r/rethinking-harmonic-loss-5BAB/.

Intraoperative Cone Beam Computed Tomography (CBCT) provides a reliable 3D anatomical context essential for interventional planning. However, its static nature fails to provide continuous monitoring of soft-tissue deformations induced by respiration, probe pressure, and surgical manipulation, leading to navigation discrepancies. We propose a deformation-aware CBCT updating framework that leverages robotic ultrasound as a dynamic proxy to infer tissue motion and update static CBCT slices in real time. Starting from calibration-initialized alignment with linear correlation of linear combination (LC2)-based rigid refinement, our method establishes accurate multimodal correspondence. To capture intraoperative dynamics, we introduce the ultrasound correlation UNet (USCorUNet), a lightweight network trained with optical flow-guided supervision to learn deformation-aware correlation representations, enabling accurate, real-time dense deformation field estimation from ultrasound streams. The inferred deformation is spatially regularized and transferred to the CBCT reference to produce deformation-consistent visualizations without repeated radiation exposure. We validate the proposed approach through deformation estimation and ultrasound-guided CBCT updating experiments. Results demonstrate real-time end-to-end CBCT slice updating and physically plausible deformation estimation, enabling dynamic refinement of static CBCT guidance during robotic ultrasound-assisted interventions. The source code is publicly available at https://github.com/anonymous-codebase/us-cbct-demo.

We study a continuous-time diffusion approximation of policy gradient for $k$-armed stochastic bandits. We prove that with a learning rate $η= O(Δ^2/\log(n))$ the regret is $O(k \log(k) \log(n) / η)$ where $n$ is the horizon and $Δ$ the minimum gap. Moreover, we construct an instance with only logarithmically many arms for which the regret is linear unless $η= O(Δ^2)$.

Considering the rise of cyberattacks incidents worldwide, the need to ensure stronger passwords is necessary. Developing a password strength meter (PSM) can help users create stronger passwords when creating an account on an online platform. This research aimed to explore whether incorporating a non-English training dataset (specifically Indian) can improve the performance of a PSM. Findings show that PSMs can be improved by utilising learning of words from other languages. Another contribution of the research was to compare and provide an analysis of AI generated data (specifically by ChatGPT) and PassGAN (existing state-of-the-art model), proving that PassGAN-like tools may no longer be needed as the performance is higher using AI generated data. To further strengthen detection, a Jaro similarity-based matching mechanism was incorporated, enabling the classification of passwords that are highly similar to known weak passwords - this addresses limitations of direct matching techniques used in prior work. A final novel contribution is on developing a PSM tailored for Indian passwords, which has not been developed previously - this resulted in a near-perfect matching accuracy using a Jaro function value of 0.5. Although performance improvements were constrained by limited data and training, results suggest that using the ChatGPT dataset is a viable and effective strategy for developing secure, language-aware password strength meters.

Recovering a tree that represents the evolutionary history of a group of species is a key task in phylogenetics. Performing this task using sequence data from multiple genetic markers poses two key challenges. The first is the discordance between the evolutionary history of individual genes and that of the species. The second challenge is computational, as contemporary studies involve thousands of species. Here we present SDSR, a scalable divide-and-conquer approach for species tree reconstruction based on spectral graph theory. The algorithm recursively partitions the species into subsets until their sizes are below a given threshold. The trees of these subsets are reconstructed by a user-chosen species tree algorithm. Finally, these subtrees are merged to form the full tree. On the theoretical front, we derive recovery guarantees for SDSR, under the multispecies coalescent (MSC) model. We also perform a runtime complexity analysis. We show that SDSR, when combined with a species tree reconstruction algorithm as a subroutine, yields substantial runtime savings as compared to applying the same algorithm on the full data. Empirically, we evaluate SDSR on synthetic benchmark datasets with incomplete lineage sorting and horizontal gene transfer. In accordance with our theoretical analysis, the simulations show that combining SDSR with common species tree methods, such as CA-ML or ASTRAL, yields up to 10-fold faster runtimes. In addition, SDSR achieves a comparable tree reconstruction accuracy to that obtained by applying these methods on the full data.

This technical report presents Sabiá-4 and Sabiazinho-4, a new generation of Portuguese language models with a focus on Brazilian Portuguese language. The models were developed through a four-stage training pipeline: continued pre-training on Portuguese and Brazilian legal corpora, long-context extension to 128K tokens, supervised fine-tuning on instruction data spanning chat, code, legal tasks, and function calling, and preference alignment. We evaluate the models on six benchmark categories: conversational capabilities in Brazilian Portuguese, knowledge of Brazilian legislation, long-context understanding, instruction following, standardized exams, and agentic capabilities including tool use and web navigation. Results show that Sabiá-4 and Sabiazinho-4 achieve a favorable cost-performance trade-off compared to other models, positioning them in the upper-left region of the pricing-accuracy chart. The models show improvements over previous generations in legal document drafting, multi-turn dialogue quality, and agentic task completion.

Cardiac magnetic resonance (CMR) imaging is widely used to visualise cardiac motion and diagnose heart disease. However, standard CMR imaging requires patients to lie still in a confined space inside a loud machine for 40-60 min, which increases patient discomfort. In addition, shorter scan times decrease either or both the temporal and spatial resolutions of cardiac motion, and thus, the diagnostic accuracy of the procedure. Of these, we focus on reduced temporal resolution and propose a neural network called FusionNet to obtain four-dimensional (4D) cardiac motion with high temporal resolution from CMR images captured in a short period of time. The model estimates intermediate 3D heart shapes based on adjacent shapes. The results of an experimental evaluation of the proposed FusionNet model showed that it achieved a performance of over 0.897 in terms of the Dice coefficient, confirming that it can recover shapes more precisely than existing methods. This code is available at: https://github.com/smiyauchi199/FusionNet.git

Vietnamese exhibits extensive dialectal variation, posing challenges for NLP systems trained predominantly on standard Vietnamese. Such systems often underperform on dialectal inputs, especially from underrepresented Central and Southern regions. Previous work on dialect normalization has focused narrowly on Central-to-Northern dialect transfer using synthetic data and limited dialectal diversity. These efforts exclude Southern varieties and intra-regional variants within the North. We introduce ViDia2Std, the first manually annotated parallel corpus for dialect-to-standard Vietnamese translation covering all 63 provinces. Unlike prior datasets, ViDia2Std includes diverse dialects from Central, Southern, and non-standard Northern regions often absent from existing resources, making it the most dialectally inclusive corpus to date. The dataset consists of over 13,000 sentence pairs sourced from real-world Facebook comments and annotated by native speakers across all three dialect regions. To assess annotation consistency, we define a semantic mapping agreement metric that accounts for synonymous standard mappings across annotators. Based on this criterion, we report agreement rates of 86% (North), 82% (Central), and 85% (South). We benchmark several sequence-to-sequence models on ViDia2Std. mBART-large-50 achieves the best results (BLEU 0.8166, ROUGE-L 0.9384, METEOR 0.8925), while ViT5-base offers competitive performance with fewer parameters. ViDia2Std demonstrates that dialect normalization substantially improves downstream tasks, highlighting the need for dialect-aware resources in building robust Vietnamese NLP systems.

While Diffusion Models excel in text-to-image synthesis, they often suffer from concept omission when synthesizing complex multi-instance scenes. Existing training-free methods attempt to resolve this by rescaling attention maps, which merely exacerbates unstructured noise without establishing coherent semantic representations. To address this, we propose Delta-K, a backbone-agnostic and plug-and-play inference framework that tackles omission by operating directly in the shared cross-attention Key space. Specifically, with Vision-language model, we extract a differential key $ΔK$ that encodes the semantic signature of missing concepts. This signal is then injected during the early semantic planning stage of the diffusion process. Governed by a dynamically optimized scheduling mechanism, Delta-K grounds diffuse noise into stable structural anchors while preserving existing concepts. Extensive experiments demonstrate the generality of our approach: Delta-K consistently improves compositional alignment across both modern DiT models and classical U-Net architectures, without requiring spatial masks, additional training, or architectural modifications.

We introduce Flexible Cutoff Learning (FCL), a method for training machine learning interatomic potentials (MLIPs) whose cutoff radii can be adjusted after training. Unlike conventional MLIPs that fix the cutoff radius during training, FCL models are trained by randomly sampling cutoff radii independently for each atom. The resulting model can then be deployed with different per-atom cutoff radii depending on the application, enabling application-specific optimization of the accuracy-cost tradeoff. Using a differentiable cost model, these per-atom cutoffs can be optimized for specific target systems after training. We demonstrate FCL with a modified MACE architecture trained on the MAD dataset. For a subset featuring molecular crystals, optimized per-atom cutoffs reduce computational cost by more than 60% while increasing force errors by less than 1%. These results show that FCL enables training of a single general-purpose MLIP that can be adapted to diverse applications through post-training cutoff optimization, eliminating the need for retraining.

Generating synthetic financial time series that preserve statistical properties of real market data is essential for stress testing, risk model validation, and scenario design. Existing approaches, from parametric models to deep generative networks, struggle to simultaneously reproduce heavy-tailed distributions, negligible linear autocorrelation, and persistent volatility clustering. We propose a hybrid hidden Markov framework that discretizes continuous excess growth rates into Laplace quantile-defined market states and augments regime switching with a Poisson-driven jump-duration mechanism to enforce realistic tail-state dwell times. Parameters are estimated by direct transition counting, bypassing the Baum-Welch EM algorithm. Synthetic data quality is evaluated using Kolmogorov-Smirnov and Anderson-Darling pass rates for distributional fidelity, and ACF mean absolute error for temporal structure. Applied to ten years of SPY data across 1,000 simulated paths, the framework achieves KS and AD pass rates exceeding 97% and 91% in-sample and 94% out-of-sample (calendar year 2025), partially reproducing the ARCH effect that standard regime-switching models miss. No single model dominates all quality dimensions: GARCH(1,1) reproduces volatility clustering more accurately but fails distributional tests (5.5% KS pass rate), while the standard HMM without jumps achieves higher distributional fidelity but cannot generate persistent high-volatility regimes. The proposed framework offers the best joint quality profile across distributional, temporal, and tail-coverage metrics. A Single-Index Model extension propagates the SPY factor path to a 424-asset universe, enabling scalable correlated synthetic path generation while preserving cross-sectional correlation structure.

Policy Dual Averaging (PDA) offers a principled Policy Mirror Descent (PMD) framework that more naturally admits value function approximation than standard PMD, enabling the use of approximate advantage (or Q-) functions while retaining strong convergence guarantees. However, applying PDA in continuous state and action spaces remains computationally challenging, since action selection involves solving an optimization sub-problem at each decision step. In this paper, we propose \textit{actor-accelerated PDA}, which uses a learned policy network to approximate the solution of the optimization sub-problems, yielding faster runtimes while maintaining convergence guarantees. We provide a theoretical analysis that quantifies how actor approximation error impacts the convergence of PDA under suitable assumptions. We then evaluate its performance on several benchmarks in robotics, control, and operations research problems. Actor-accelerated PDA achieves superior performance compared to popular on-policy baselines such as Proximal Policy Optimization (PPO). Overall, our results bridge the gap between the theoretical advantages of PDA and its practical deployment in continuous-action problems with function approximation.

Large Language Models frequently generate fluent but factually incorrect text. We propose Adaptive Activation Cancellation (AAC), a real-time inference-time framework that treats hallucination-associated neural activations as structured interference within the transformer residual stream, drawing an explicit analogy to classical adaptive noise cancellation from signal processing. The framework identifies Hallucination Nodes (H-Nodes) via layer-wise linear probing and suppresses them using a confidence-weighted forward hook during auto-regressive generation -- requiring no external knowledge, no fine-tuning, and no additional inference passes. Evaluated across OPT-125M, Phi-3-mini, and LLaMA 3-8B on TruthfulQA and HaluEval, the real-time hook is the only intervention that consistently improves downstream accuracy on all three scales. Critically, the method is strictly surgical: WikiText-103 perplexity and MMLU reasoning accuracy are preserved at exactly 0.0% degradation across all three model scales, a property that distinguishes AAC from interventions that trade fluency or general capability for factual improvement. On the LLaMA 3-8B scale, the hook additionally yields positive generation-level gains (MC1 +0.04; MC2 +0.003; Token-F1 +0.003) while achieving probe-space selectivity 5.94x - 3.5x higher than the ITI baseline -- demonstrating that targeted neuron-level suppression can simultaneously improve factual accuracy and preserve model capability.

Model Context Protocol (MCP) servers have rapidly emerged over the past year as a widely adopted way to enable Large Language Model (LLM) agents to access dynamic, real-world tools. As MCP servers proliferate and become easy to adopt via open-source releases, understanding their security risks becomes essential for dependable production agent deployments. Recent work has developed MCP threat taxonomies, proposed mitigations, and demonstrated practical attacks. However, to the best of our knowledge, no prior study has conducted a systematic, large-scale assessment of weaknesses in open-source MCP servers. Motivated by this gap, we apply static code analysis to identify Common Weakness Enumeration (CWE) weaknesses and map them to common attack patterns and threat categories using the MITRE Common Attack Pattern Enumerations and Classifications (CAPEC) to ground risk in real-world threats. We then introduce a risk-assessment framework for the MCP landscape that combines these threats using a multi-metric scoring of likelihood and impact. Our findings show that many open-source MCP servers contain exploitable weaknesses that can compromise confidentiality, integrity, and availability, underscoring the need for secure-by-design MCP server development.

Recent progress in learning-based image compression has demonstrated that end-to-end optimization can substantially outperform traditional codecs by jointly learning compact latent representations and probabilistic entropy models. However, many existing approaches achieve high rate-distortion efficiency at the expense of increased computational cost and limited parallelism. This paper presents ARCHE - Autoregressive Residual Compression with Hyperprior and Excitation, an end-to-end learned image compression framework that balances modeling accuracy and computational efficiency. The proposed architecture unifies hierarchical, spatial, and channel-based priors within a single probabilistic framework, capturing both global and local dependencies in the latent representation of the image, while employing adaptive feature recalibration and residual refinement to enhance latent representation quality. Without relying on recurrent or transformer-based components, ARCHE attains state-of-the-art rate-distortion efficiency: it reduces the BD-Rate by approximately 48% relative to the commonly used benchmark model of Balle et al., 30% relative to the channel-wise autoregressive model of Minnen & Singh and 5% against the VVC Intra codec on the Kodak benchmark dataset. The framework maintains computational efficiency with 95M parameters and 222ms running time per image. Visual comparisons confirm sharper textures and improved color fidelity, particularly at lower bit rates, demonstrating that accurate entropy modeling can be achieved through efficient convolutional designs suitable for practical deployment.

Statistical inference with bandit data presents fundamental challenges due to adaptive sampling, which violates the independence assumptions underlying classical asymptotic theory. Recent work has identified stability as a sufficient condition for valid inference under adaptivity. This paper develops a systematic theory of stability for bandit algorithms based on stochastic mirror descent, a broad algorithmic framework that includes the widely-used EXP3 algorithm as a special case. Our contributions are threefold. First, we establish a general stability criterion: if the average iterates of a stochastic mirror descent algorithm converge in ratio to a non-random probability vector, then the induced bandit algorithm is stable. This result provides a unified lens for analyzing stability across diverse algorithmic instantiations. Second, we introduce a family of regularized-EXP3 algorithms employing a log-barrier regularizer with appropriately tuned parameters. We prove that these algorithms satisfy our stability criterion and, as an immediate corollary, that Wald-type confidence intervals for linear functionals of the mean parameter achieve nominal coverage. Notably, we show that the same algorithms attain minimax-optimal regret guarantees up to logarithmic factors, demonstrating that inference-enabling stability and learning efficiency are compatible objectives within the mirror descent framework. Third, we establish robustness to corruption: a modified variant of regularized-EXP3 maintains asymptotic normality of empirical arm means even in the presence of $o(T^{1/2})$ adversarial corruptions. This stands in sharp contrast to other stable algorithms such as UCB, which suffer linear regret even under logarithmic levels of corruption.

The growing adoption of electronic health record (EHR) systems has provided unprecedented opportunities for predictive modeling to guide clinical decision making. Structured EHRs contain longitudinal observations of patients across hospital visits, where each visit is represented by a set of medical codes. While sequence-based, graph-based, and graph-enhanced sequence approaches have been developed to capture rich code interactions over time or within the same visits, they often overlook the inherent heterogeneous roles of medical codes arising from distinct clinical characteristics and contexts. To this end, in this study we propose the Disease Trajectory-aware Transformer for EHR (DT-BEHRT), a graph-enhanced sequential architecture that disentangles disease trajectories by explicitly modeling diagnosis-centric interactions within organ systems and capturing asynchronous progression patterns. To further enhance the representation robustness, we design a tailored pre-training methodology that combines trajectory-level code masking with ontology-informed ancestor prediction, promoting semantic alignment across multiple modeling modules. Extensive experiments on multiple benchmark datasets demonstrate that DT-BEHRT achieves strong predictive performance and provides interpretable patient representations that align with clinicians' disease-centered reasoning. The source code is publicly accessible at https://github.com/GatorAIM/DT-BEHRT.git.

Computer-using agents (CUAs) are becoming increasingly capable; however, it remains difficult to scale evaluation of whether a trajectory truly fulfills a user instruction. In this work, we study reward modeling from execution video: a sequence of keyframes from an agent trajectory that is independent of the agent's internal reasoning or actions. Although video-execution modeling is method-agnostic, it presents key challenges, including highly redundant layouts and subtle, localized cues that determine success. We introduce Execution Video Reward 53k (ExeVR-53k), a dataset of 53k high-quality video--task--reward triplets. We further propose adversarial instruction translation to synthesize negative samples with step-level annotations. To enable learning from long, high-resolution execution videos, we design spatiotemporal token pruning, which removes homogeneous regions and persistent tokens while preserving decisive UI changes. Building on these components, we fine-tune an Execution Video Reward Model (ExeVRM) that takes only a user instruction and a video-execution sequence to predict task success. Our ExeVRM 8B achieves 84.7% accuracy and 87.7% recall on video-execution assessment, outperforming strong proprietary models such as GPT-5.2 and Gemini-3 Pro across Ubuntu, macOS, Windows, and Android, while providing more precise temporal attribution. These results show that video-execution reward modeling can serve as a scalable, model-agnostic evaluator for CUAs.

Explainable speech quality assessment requires moving beyond Mean Opinion Scores (MOS) to analyze underlying perceptual dimensions. To address this, we introduce a novel post-training method that tailors the foundational Audio Large Language Model for multidimensional reasoning, detection and classification of audio artifacts. First, a calibration stage aligns the model to predict predefined perceptual dimensions. Second, a reinforcement learning stage leverages Group Relative Policy Optimization (GRPO) with dimension-specific rewards to heavily enhance accuracy of descriptions and temporal localization of quality issues. With this approach we reach state-of-the-art results of 0.71 mean PCC score on the multidimensional QualiSpeech benchmark and 13% improvement in MOS prediction driven by RL-based reasoning. Furthermore, our fine-grained GRPO rewards substantially advance the model's ability to pinpoint and classify audio artifacts in time.

Every agent interaction generates a next-state signal, namely the user reply, tool output, terminal or GUI state change that follows each action, yet no existing agentic RL system recovers it as a live, online learning source. We present OpenClaw-RL, a framework built on a simple observation: next-state signals are universal, and policy can learn from all of them simultaneously. Personal conversations, terminal executions, GUI interactions, SWE tasks, and tool-call traces are not separate training problems. They are all interactions that can be used to train the same policy in the same loop. Next-state signals encode two forms of information: evaluative signals, which indicate how well the action performed and are extracted as scalar rewards via a PRM judge; and directive signals, which indicate how the action should have been different and are recovered through Hindsight-Guided On-Policy Distillation (OPD). We extract textual hints from the next state, construct an enhanced teacher context, and provide token-level directional advantage supervision that is richer than any scalar reward. Due to the asynchronous design, the model serves live requests, the PRM judges ongoing interactions, and the trainer updates the policy at the same time, with zero coordination overhead between them. Applied to personal agents, OpenClaw-RL enables an agent to improve simply by being used, recovering conversational signals from user re-queries, corrections, and explicit feedback. Applied to general agents, the same infrastructure supports scalable RL across terminal, GUI, SWE, and tool-call settings, where we additionally demonstrate the utility of process rewards. Code: https://github.com/Gen-Verse/OpenClaw-RL

The Model Context Protocol (MCP) is a recently proposed interoperability standard that unifies how AI agents connect with external tools and data sources. By defining a set of common client-server message exchange clauses, MCP replaces fragmented integrations with a standardized, plug-and-play framework. However, to be compatible with diverse AI agents, the MCP specification relaxes many behavioral constraints into optional clauses, leading to misuse-prone SDK implementation. We identify it as a new attack surface that allows adversaries to achieve multiple attacks (e.g, silent prompt injection, DoS, etc.), named as \emph{compatibility-abusing attacks}. In this work, we present the first systematic framework for analyzing this new attack surface across multi-language MCP SDKs. First, we construct a universal and language-agnostic intermediate representation (IR) generator that normalizes SDKs of different languages. Next, based on the new IR, we propose auditable static analysis with LLM-guided semantic reasoning for cross-language/clause compliance analysis. Third, by formalizing the attack semantics of the MCP clauses, we build three attack modalities and develop a modality-guided pipeline to uncover exploitable non-compliance issues.

Low-rank adapters (LoRAs) are a parameter-efficient finetuning technique that injects trainable low-rank matrices into pretrained models to adapt them to new tasks. Mixture-of-LoRAs models expand neural networks efficiently by routing each layer input to a small subset of specialized LoRAs of the layer. Existing Mixture-of-LoRAs routers assign a learned routing weight to each LoRA to enable end-to-end training of the router. Despite their empirical promise, we observe that the routing weights are typically extremely imbalanced across LoRAs in practice, where only one or two LoRAs often dominate the routing weights. This essentially limits the number of effective LoRAs and thus severely hinders the expressive power of existing Mixture-of-LoRAs models. In this work, we attribute this weakness to the nature of learnable routing weights and rethink the fundamental design of the router. To address this critical issue, we propose a new router designed that we call Reinforcement Routing for Mixture-of-LoRAs (ReMix). Our key idea is using non-learnable routing weights to ensure all active LoRAs to be equally effective, with no LoRA dominating the routing weights. However, our routers cannot be trained directly via gradient descent due to our non-learnable routing weights. Hence, we further propose an unbiased gradient estimator for the router by employing the reinforce leave-one-out (RLOO) technique, where we regard the supervision loss as the reward and the router as the policy in reinforcement learning. Our gradient estimator also enables to scale up training compute to boost the predictive performance of our ReMix. Extensive experiments demonstrate that our proposed ReMix significantly outperform state-of-the-art parameter-efficient finetuning methods under a comparable number of activated parameters.

Finetuning on domain-specific data is a well-established method for enhancing LLM performance on downstream tasks. Training on each dataset produces a new set of model weights, resulting in a multitude of checkpoints saved in-house or on open-source platforms. However, these training artifacts are rarely reused for subsequent experiments despite containing improved model abilities for potentially similar tasks. In this paper, we propose Mashup Learning, a simple method to leverage the outputs of prior training runs to enhance model adaptation to new tasks. Our procedure identifies the most relevant historical checkpoints for a target dataset, aggregates them with model merging, and uses the result as an improved initialization for training. Across 8 standard LLM benchmarks, four models, and two collections of source checkpoints, Mashup Learning consistently improves average downstream accuracy by 0.5-5 percentage points over training from scratch. It also accelerates convergence, requiring 41-46% fewer training steps and up to 37% less total wall-clock time to match from-scratch accuracy, including all selection and merging overhead.

In the design of engineered components, rigorous vibration testing is essential for performance validation and identification of resonant frequencies and amplitudes encountered during operation. Performing this evaluation numerically via machine learning has great potential to accelerate design iteration and make testing workflows more efficient. However, dynamical systems are conventionally difficult to solve via machine learning methods without using physics-based regularizing loss functions. To properly perform this forecasting task, a structure that has an inspectable physical obedience can be devised without the use of regularizing terms from first principles. The method employed in this work is a neural operator integrated with an implicit numerical scheme. This architecture enables operators to learn of the underlying state-space dynamics from limited data, allowing generalization to untested driving frequencies and initial conditions. This network can infer the system's global frequency response by training on a small set of input conditions. As a foundational proof of concept, this investigation verifies the machine learning algorithm with a linear, single-degree-of-freedom system, demonstrating implicit obedience of dynamics. This approach demonstrates 99.87% accuracy in predicting the Frequency Response Curve (FRC), forecasting the frequency and amplitude of linear resonance training on 7% of the bandwidth of the solution. By training machine learning models to internalize physics information rather than trajectory, better generalization accuracy can be realized, vastly improving the timeframe for vibration studies on engineered components.

We demonstrate that user preferences can be represented and predicted across topical domains using large-scale social modeling. Given information about popular entities favored by a user, we project the user into a social embedding space learned from a large-scale sample of the Twitter (now X) network. By representing both users and popular entities in a joint social space, we can assess the relevance of candidate entities (e.g., music artists) using cosine similarity within this embedding space. A comprehensive evaluation using link prediction experiments shows that this method achieves effective personalization in zero-shot setting, when no user feedback is available for entities in the target domain, yielding substantial improvements over a strong popularity-based baseline. In-depth analysis further illustrates that socio-demographic factors encoded in the social embeddings are correlated with user preferences across domains. Finally, we argue and demonstrate that the proposed approach can facilitate social modeling of end users using large language models (LLMs).

The last layer of neural language models (LMs) projects output features of dimension $D$ to logits in dimension $V$, the size of the vocabulary, where usually $D \ll V$. This mismatch is known to raise risks of limited expressivity in neural LMs, creating a so-called softmax bottleneck. We show the softmax bottleneck is not only an expressivity bottleneck but also an optimization bottleneck. Backpropagating $V$-dimensional gradients through a rank-$D$ linear layer induces unavoidable compression, which alters the training feedback provided to the vast majority of the parameters. We present a theoretical analysis of this phenomenon and measure empirically that 95-99% of the gradient norm is suppressed by the output layer, resulting in vastly suboptimal update directions. We conduct controlled pretraining experiments showing that the gradient bottleneck makes trivial patterns unlearnable, and drastically affects the training dynamics of LLMs. We argue that this inherent flaw contributes to training inefficiencies at scale independently of the model architecture, and raises the need for new LM head designs.

Retrieval-Augmented Generation (RAG) significantly improves the factuality of Large Language Models (LLMs), yet standard pipelines often lack mechanisms to verify inter- mediate reasoning, leaving them vulnerable to hallucinations in high-stakes domains. To address this, we propose a domain-specific RAG framework that integrates explicit rea- soning and faithfulness verification. Our architecture augments standard retrieval with neural query rewriting, BGE-based cross-encoder reranking, and a rationale generation module that grounds sub-claims in specific evidence spans. We further introduce an eight-category verification taxonomy that enables fine-grained assessment of rationale faithfulness, distinguishing between explicit and implicit support patterns to facilitate structured error diagnosis. We evaluate this framework on the BioASQ and PubMedQA benchmarks, specifically analyzing the impact of dynamic in-context learning and rerank- ing under constrained token budgets. Experiments demonstrate that explicit rationale generation improves accuracy over vanilla RAG baselines, while dynamic demonstration selection combined with robust reranking yields further gains in few-shot settings. Using Llama-3-8B-Instruct, our approach achieves 89.1% on BioASQ-Y/N and 73.0% on Pub- MedQA, competitive with systems using significantly larger models. Additionally, we perform a pilot study combining human expert assessment with LLM-based verification to explore how explicit rationale generation improves system transparency and enables more detailed diagnosis of retrieval failures in biomedical question answering.

Every formal grammar defines a language and can in principle be used in three ways: to generate strings (production), to recognize them (parsing), or -- given only examples -- to infer the grammar itself (grammar induction). Generation and recognition are extensionally equivalent -- they characterize the same set -- but operationally asymmetric in multiple independent ways. Inference is a qualitatively harder problem: it does not have access to a known grammar. Despite the centrality of this triad to compiler design, natural language processing, and formal language theory, no survey has treated it as a unified, multidimensional phenomenon. We identify six dimensions along which generation and recognition diverge: computational complexity, ambiguity, directionality, information availability, grammar inference, and temporality. We show that the common characterization "generation is easy, parsing is hard" is misleading: unconstrained generation is trivial, but generation under constraints can be NP-hard. The real asymmetry is that parsing is always constrained (the input is given) while generation need not be. Two of these dimensions -- directionality and temporality -- have not previously been identified as dimensions of the generation-recognition asymmetry. We connect the temporal dimension to the surprisal framework of Hale (2001) and Levy (2008), arguing that surprisal formalizes the temporal asymmetry between a generator (surprisal = 0) and a parser that predicts under uncertainty (surprisal > 0). We review bidirectional systems in NLP and observe that bidirectionality has been available for fifty years yet has not transferred to most domain-specific applications. We conclude with a discussion of large language models, which architecturally unify generation and recognition while operationally preserving the asymmetry.

Data products enable end users to gain greater insights about their data by providing supporting assets, such as example question-SQL pairs which can be answered using the data or views over the database tables. However, producing useful data products is challenging, and typically requires domain experts to hand-craft supporting assets. We propose a system that automates data product improvement through specialized AI agents operating in a continuous optimization loop. By surfacing questions, monitoring multi-dimensional quality metrics, and supporting human-in-the-loop controls, it transforms data into observable and refinable assets that balance automation with trust and oversight.

Hyperspectral images capture vast amounts of high-dimensional spectral information about a scene, making labeling an intensive task that is resistant to out-of-the-box statistical methods. Unsupervised learning of clusters allows for automated segmentation of the scene, enabling a more rapid understanding of the image. Partitioning the spectral information contained within the data via dictionary learning in Wasserstein space has proven an effective method for unsupervised clustering. However, this approach requires balancing the spectral profiles of the data, blurring the classes, and sacrificing robustness to outliers and noise. In this paper, we suggest improving this approach by utilizing unbalanced Wasserstein barycenters to learn a lower-dimensional representation of the underlying data. The deployment of spectral clustering on the learned representation results in an effective approach for the unsupervised learning of labels.

Text embeddings have become central to computational social science and psychology, enabling scalable measurement of meaning and mixed-method inference. Yet most representation learning is optimized and evaluated for prediction and retrieval, yielding a prediction-measurement gap: representations that perform well as features may be poorly suited as scientific instruments. The paper argues that scientific meaning analysis motivates a distinct family of objectives - scientific usability - emphasizing geometric legibility, interpretability and traceability to linguistic evidence, robustness to non-semantic confounds, and compatibility with regression-style inference over semantic directions. Grounded in cognitive and neuro-psychological views of meaning, the paper assesses static word embeddings and contextual transformer representations against these requirements: static spaces remain attractive for transparent measurement, whereas contextual spaces offer richer semantics but entangle meaning with other signals and exhibit geometric and interpretability issues that complicate inference. The paper then outlines a course-setting agenda around (i) geometry-first design for gradients and abstraction, including hierarchy-aware spaces constrained by psychologically privileged levels; (ii) invertible post-hoc transformations that recondition embedding geometry and reduce nuisance influence; and (iii) meaning atlases and measurement-oriented evaluation protocols for reliable and traceable semantic inference. As the field debates the limits of scale-first progress, measurement-ready representations offer a principled new frontier.

We propose a standalone autoregressive (AR) Action Expert that generates actions as a continuous causal sequence while conditioning on refreshable vision-language prefixes. In contrast to existing Vision-Language-Action (VLA) models and diffusion policies that reset temporal context with each new observation and predict actions reactively, our Action Expert maintains its own history through a long-lived memory and is inherently context-aware. This structure addresses the frequency mismatch between fast control and slow reasoning, enabling efficient independent pretraining of kinematic syntax and modular integration with heavy perception backbones, naturally ensuring spatio-temporally consistent action generation across frames. To synchronize these asynchronous hybrid V-L-A modalities, we utilize a re-anchoring mechanism that mathematically accounts for perception staleness during both training and inference. Experiments on simulated and real-robot manipulation tasks demonstrate that the proposed method can effectively replace traditional chunk-based action heads for both specialist and generalist policies. AR-VLA exhibits superior history awareness and substantially smoother action trajectories while maintaining or exceeding the task success rates of state-of-the-art reactive VLAs. Overall, our work introduces a scalable, context-aware action generation schema that provides a robust structural foundation for training effective robotic policies.

The ``Lost in the Middle'' phenomenon -- a U-shaped performance curve where LLMs retrieve well from the beginning and end of a context but fail in the middle -- is widely attributed to learned Softmax artifacts or the distance-decay of positional encodings like RoPE. This paper makes a single, precise claim: \emph{the U-shape is already present at initialization, before any training or positional encoding takes effect.} It is an inherent geometric property of the causal decoder with residual connections. We model multi-layer causal attention as iterated powers of the Cesàro matrix and derive the exact closed-form influence density in the continuous limit. Causal masking forces a logarithmic divergence of gradient influence at the start of the prompt (the Primacy Tail), while residual connections create an isolated $\mathcal{O}(1)$ anchor at the final token (the Recency Delta). Between these extremes lies a factorial dead zone of order $\mathcal{O}(1/(H{-}1)!)$, where $H$ is the network depth, making middle-context retrieval and training structurally hostile. We validate empirically that untrained Qwen2 and GPT-2 architectures exhibit this U-shape at Step~0, and that it is identical with or without RoPE. Comparing initialized and pretrained networks, we show that standard training does not overcome the topological valley, confirming that the U-shape persists as an architectural baseline under standard pretraining objectives. We do not claim that this bias is insurmountable, nor that interventions such as RoPE modifications are useless. We establish what the baseline is and where it comes from, so that future efforts to overcome it can be precisely targeted.

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capacity of Large Language Models (LLMs). However, RLVR solely relies on final answers as outcome rewards, neglecting the correctness of intermediate reasoning steps. Training on these process-wrong but outcome-correct rollouts can lead to hallucination and answer-copying, severely undermining the model's generalization and robustness. To address this, we incorporate a Contrastive Learning mechanism into the Policy Optimization (CLIPO) to generalize the RLVR process. By optimizing a contrastive loss over successful rollouts, CLIPO steers the LLM to capture the invariant structure shared across correct reasoning paths. This provides a more robust cross-trajectory regularization than the original single-path supervision in RLVR, effectively mitigating step-level reasoning inconsistencies and suppressing hallucinatory artifacts. In experiments, CLIPO consistently improves multiple RLVR baselines across diverse reasoning benchmarks, demonstrating uniform improvements in generalization and robustness for policy optimization of LLMs. Our code and training recipes are available at https://github.com/Qwen-Applications/CLIPO.

Accurately upscaling terrestrial carbon fluxes is central to estimating the global carbon budget, yet remains challenging due to the sparse and regionally biased distribution of ground measurements. Existing data-driven upscaling products often fail to generalize beyond observed domains, leading to systematic regional biases and high predictive uncertainty. We introduce Task-Aware Modulation with Representation Learning (TAM-RL), a framework that couples spatio-temporal representation learning with knowledge-guided encoder-decoder architecture and loss function derived from the carbon balance equation. Across 150+ flux tower sites representing diverse biomes and climate regimes, TAM-RL improves predictive performance relative to existing state-of-the-art datasets, reducing RMSE by 8-9.6% and increasing explained variance (R2) from 19.4% to 43.8%, depending on the target flux. These results demonstrate that integrating physically grounded constraints with adaptive representation learning can substantially enhance the robustness and transferability of global carbon flux estimates.

Modern CNNs' high computational demands hinder edge deployment, as traditional ``hard'' sparsity (skipping mathematical zeros) loses effectiveness in deep layers or with smooth activations like Tanh. We propose a ``soft sparsity'' paradigm using a hardware efficient Most Significant Bit (MSB) proxy to skip negligible non-zero multiplications. Integrated as a custom RISC-V instruction and evaluated on LeNet-5 (MNIST), this method reduces ReLU MACs by 88.42% and Tanh MACs by 74.87% with zero accuracy loss--outperforming zero-skipping by 5x. By clock-gating inactive multipliers, we estimate power savings of 35.2\% for ReLU and 29.96\% for Tanh. While memory access makes power reduction sub-linear to operation savings, this approach significantly optimizes resource-constrained inference.

The US Census Bureau Disclosure Avoidance System (DAS) balances confidentiality and utility requirements for the decennial US Census (Abowd et al., 2022). The DAS was used in the 2020 Census to produce demographic datasets critically used for legislative apportionment and redistricting, federal and state funding allocation, municipal and infrastructure planning, and scientific research. At the heart of DAS is TopDown, a heuristic post-processing method that combines billions of private noisy measurements across six geographic levels in order to produce new estimates that are consistent, more accurate, and satisfy certain structural constraints on the data. In this work, we introduce BlueDown, a new post-processing method that produces more accurate, consistent estimates while satisfying the same privacy guarantees and structural constraints. We obtain especially large accuracy improvements for aggregates at the county and tract levels on evaluation metrics proposed by the US Census Bureau. From a technical perspective, we develop a new algorithm for generalized least-squares regression that leverages the hierarchical structure of the measurements and that is statistically optimal among linear unbiased estimators. This reduces the computational dependence on the number of geographic regions measured from matrix multiplication time, which would be infeasible for census-scale data, to linear time. We incorporate the additional structural constraints by combining this regression algorithm with an optimization routine that extends TDA to support correlated measurements. We further improve the efficiency of our algorithm using succinct linear-algebraic operations that exploit symmetries in the structure of the measurements and constraints. We believe our hierarchical regression and succinct operations to be of independent interest.

Recent advances in multi-agent reinforcement learning, particularly Policy-Space Response Oracles (PSRO), have enabled the computation of approximate game-theoretic equilibria in increasingly complex domains. However, these methods rely on deep reinforcement learning oracles that produce `black-box' neural network policies, making them difficult to interpret, trust or debug. We introduce Code-Space Response Oracles (CSRO), a novel framework that addresses this challenge by replacing RL oracles with Large Language Models (LLMs). CSRO reframes the best response computation as a code generation task, prompting an LLM to generate policies directly as human-readable code. This approach not only yields inherently interpretable policies but also leverages the LLM's pretrained knowledge to discover complex, human-like strategies. We explore multiple ways to construct and enhance an LLM-based oracle: zero-shot prompting, iterative refinement and \emph{AlphaEvolve}, a distributed LLM-based evolutionary system. We demonstrate that CSRO achieves performance competitive with baselines while producing a diverse set of explainable policies. Our work presents a new perspective on multi-agent learning, shifting the focus from optimizing opaque policy parameters to synthesizing interpretable algorithmic behavior.

A fundamental requirement for intelligent systems is the ability to learn continuously under changing environments. However, models trained in this regime often suffer from catastrophic forgetting. Leveraging pre-trained models has recently emerged as a promising solution, since their generalized feature extractors enable faster and more robust adaptation. While some earlier works mitigate forgetting by fine-tuning only on the first task, this approach quickly deteriorates as the number of tasks grows and the data distributions diverge. More recent research instead seeks to consolidate task knowledge into a unified backbone, or adapting the backbone as new tasks arrive. However, such approaches may create a (potential) \textit{mismatch} between task-specific classifiers and the adapted backbone. To address this issue, we propose a novel \textit{Local Classifier Alignment} (LCA) loss to better align the classifier with backbone. Theoretically, we show that this LCA loss can enable the classifier to not only generalize well for all observed tasks, but also improve robustness. Furthermore, we develop a complete solution for continual learning, following the model merging approach and using LCA. Extensive experiments on several standard benchmarks demonstrate that our method often achieves leading performance, sometimes surpasses the state-of-the-art methods with a large margin.

Time-series forecasting often faces challenges from non-stationarity, particularly distributional drift, where the data distribution evolves over time. This dynamic behavior can undermine the effectiveness of adaptive optimizers, such as Adam, which are typically designed for stationary objectives. In this paper, we revisit Adam in the context of non-stationary forecasting and identify that its second-order bias correction limits responsiveness to shifting loss landscapes. To address this, we propose TS_Adam, a lightweight variant that removes the second-order correction from the learning rate computation. This simple modification improves adaptability to distributional drift while preserving the optimizer core structure and requiring no additional hyperparameters. TS_Adam integrates easily into existing models and consistently improves performance across long- and short-term forecasting tasks. On the ETT datasets with the MICN model, it achieves an average reduction of 12.8% in MSE and 5.7% in MAE compared to Adam. These results underscore the practicality and versatility of TS_Adam as an effective optimization strategy for real-world forecasting scenarios involving non-stationary data. Code is available at: https://github.com/DD-459-1/TS_Adam.

Recent 3D molecular generation methods primarily use asynchronous auto-regressive or synchronous diffusion models. While auto-regressive models build molecules sequentially, they're limited by a short horizon and a discrepancy between training and inference. Conversely, synchronous diffusion models denoise all atoms at once, offering a molecule-level horizon but failing to capture the causal relationships inherent in hierarchical molecular structures. We introduce Equivariant Asynchronous Diffusion (EAD) to overcome these limitations. EAD is a novel diffusion model that combines the strengths of both approaches: it uses an asynchronous denoising schedule to better capture molecular hierarchy while maintaining a molecule-level horizon. Since these relationships are often complex, we propose a dynamic scheduling mechanism to adaptively determine the denoising timestep. Experimental results show that EAD achieves state-of-the-art performance in 3D molecular generation.

OpenClaw-style agent stacks turn language into privileged execution: LLM intents flow through tool interception, policy gates, and a local executor. In parallel, skill marketplaces such as skills.sh make capability acquisition as easy as installing skills and CLIs, creating a growing capability supply chain. Together, these trends shift the dominant safety failure mode from "wrong answers" to execution-induced loss, where untrusted prompts, compromised skills, or narrative manipulation can trigger real trades and irreversible side effects. We propose Survivability-Aware Execution (SAE), an execution-layer survivability standard for OpenClaw-style systems and skill-enabled agents. SAE sits as middleware between a strategy engine (LLM or non-LLM) and the exchange executor. It defines an explicit execution contract (ExecutionRequest, ExecutionContext, ExecutionDecision) and enforces non-bypassable last-mile invariants: projection-based exposure budgets, cooldown and order-rate limits, slippage bounds, staged execution, and tool/venue allowlists. To make delegated execution testable under supply-chain risk, we operationalize the Delegation Gap (DG) via a logged Intended Policy Spec that enables deterministic out-of-scope labeling and reproducible DG metrics. On an offline replay using official Binance USD-M BTCUSDT/ETHUSDT perpetual data (15m; 2025-09-01--2025-12-01, incl. funding), SAE improves survivability: MDD drops from 0.4643 to 0.0319 (Full; 93.1%), |CVaR_0.99| shrinks from 4.025e-3 to ~1.02e-4 (~97.5%), and DG loss proxy falls from 0.647 to 0.019 (~97.0%). AttackSuccess decreases from 1.00 to 0.728 with zero FalseBlock in this run. Block bootstrap, paired Wilcoxon, and two-proportion tests confirm the shifts. SAE reframes agentic trading safety for the OpenClaw+skills era: treat upstream intent and skills as untrusted, and enforce survivability where actions become side effects.

As embodied models become powerful, humans will collaborate with multiple embodied AI agents at their workplace or home in the future. To ensure better communication between human users and the multi-agent system, it is crucial to interpret incoming information from agents in parallel and refer to the appropriate context for each query. Existing challenges include effectively compressing and communicating high volumes of individual sensory inputs in the form of video and correctly aggregating multiple egocentric videos to construct system-level memory. In this work, we first formally define a novel problem of understanding multiple long-horizon egocentric videos simultaneously collected from embodied agents. To facilitate research in this direction, we introduce MultiAgent-EgoQA (MA-EgoQA), a benchmark designed to systemically evaluate existing models in our scenario. MA-EgoQA provides 1.7k questions unique to multiple egocentric streams, spanning five categories: social interaction, task coordination, theory-of-mind, temporal reasoning, and environmental interaction. We further propose a simple baseline model for MA-EgoQA named EgoMAS, which leverages shared memory across embodied agents and agent-wise dynamic retrieval. Through comprehensive evaluation across diverse baselines and EgoMAS on MA-EgoQA, we find that current approaches are unable to effectively handle multiple egocentric streams, highlighting the need for future advances in system-level understanding across the agents. The code and benchmark are available at https://ma-egoqa.github.io.

The widespread adoption of thinking mode in large language models (LLMs) has significantly enhanced complex task processing capabilities while introducing new security risks. When subjected to jailbreak attacks, the step-by-step reasoning process may cause models to generate more detailed harmful content. We observe that thinking mode exhibits unique vulnerabilities when processing interleaved multiple tasks. Based on this observation, we propose multi-stream perturbation attack, which generates superimposed interference by interweaving multiple task streams within a single prompt. We design three perturbation strategies: multi-stream interleaving, inversion perturbation, and shape transformation, which disrupt the thinking process through concurrent task interleaving, character reversal, and format constraints respectively. On JailbreakBench, AdvBench, and HarmBench datasets, our method achieves attack success rates exceeding most methods across mainstream models including Qwen3 series, DeepSeek, Qwen3-Max, and Gemini 2.5 Flash. Experiments show thinking collapse rates and response repetition rates reach up to 17% and 60% respectively, indicating multi-stream perturbation not only bypasses safety mechanisms but also causes thinking process collapse or repetitive outputs.

Neural network weights are typically viewed as the end product of training, while most deep learning research focuses on data, features, and architectures. However, recent advances show that the set of all possible weight values (weight space) itself contains rich structure: pretrained models form organized distributions, exhibit symmetries, and can be embedded, compared, or even generated. Understanding such structures has tremendous impact on how neural networks are analyzed and compared, and on how knowledge is transferred across models, beyond individual training instances. This emerging research direction, which we refer to as Weight Space Learning (WSL), treats neural weights as a meaningful domain for analysis and modeling. This survey provides the first unified taxonomy of WSL. We categorize existing methods into three core dimensions: Weight Space Understanding (WSU), which studies the geometry and symmetries of weights; Weight Space Representation (WSR), which learns embeddings over model weights; and Weight Space Generation (WSG), which synthesizes new weights through hypernetworks or generative models. We further show how these developments enable practical applications, including model retrieval, continual and federated learning, neural architecture search, and data-free reconstruction. By consolidating fragmented progress under a coherent framework, this survey highlights weight space as a learnable, structured domain with growing impact across model analysis, transferring, and weight generation. We release an accompanying resource at https://github.com/Zehong-Wang/Awesome-Weight-Space-Learning.

AI assistants that support humans in daily life are becoming increasingly feasible, driven by the rapid advancements in multimodal language models. A key challenge lies in overcoming the generic nature of these models to deliver personalized experiences. Existing approaches to personalizing large vision language models often rely on additional training stages, which limit generality and scalability, or on engineered pipelines with external pre-trained modules, which hinder deployment efficiency. In this work, we propose an efficient personalization method that leverages the model's inherent ability to capture personalized concepts. Specifically, we extract visual tokens that predominantly represent the target concept by utilizing the model's internal attention mechanisms. These tokens serve as a memory of that specific concept, enabling the model to recall and describe it when it appears in test images. We conduct a comprehensive and unified evaluation of our approach and SOTA methods across various personalization settings including single-concept, multi-concept, and video personalization, demonstrating strong performance gains with minimal personalization overhead.

Diffusion large language models (dLLMs) are emerging as a promising alternative to autoregressive models (ARMs) due to their ability to capture bidirectional context and the potential for parallel generation. Despite the advantages, dLLM inference remains computationally expensive as the full input context is processed at every iteration. In this work, we analyze the generation dynamics of dLLMs and find that intermediate representations, including key, value, and hidden states, change only subtly across successive iterations. Leveraging this insight, we propose \textbf{ES-dLLM}, a training-free inference acceleration framework for dLLM that reduces computation by skipping tokens in early layers based on the estimated importance. Token importance is computed with intermediate tensor variation and confidence scores of previous iterations. Experiments on LLaDA-8B and Dream-7B demonstrate that ES-dLLM achieves throughput of up to 226.57 and 308.51 tokens per second (TPS), respectively, on an NVIDIA H200 GPU, delivering 5.6$\times$ to 16.8$\times$ speedup over the vanilla implementation and up to 1.85$\times$ over the state-of-the-art caching method, while preserving generation quality.

In this article, we evaluate four Large Language Models (LLMs) and their effectiveness at retrieving data within a specialized Retrieval-Augmented Generation (RAG) system, using a comprehensive food composition database. Our method is focused on the LLMs ability to translate natural language queries into structured metadata filters, enabling efficient retrieval via a Chroma vector database. By achieving high accuracy in this critical retrieval step, we demonstrate that LLMs can serve as an accessible, high-performance tool, drastically reducing the manual effort and technical expertise previously required for domain experts, such as food compilers and nutritionists, to leverage complex food and nutrition data. However, despite the high performance on easy and moderately complex queries, our analysis of difficult questions reveals that reliable retrieval remains challenging when queries involve non-expressible constraints. These findings demonstrate that LLM-driven metadata filtering excels when constraints can be explicitly expressed, but struggles when queries exceed the representational scope of the metadata format.

Engram conditional memory has emerged as a promising component for LLMs by decoupling static knowledge lookup from dynamic computation. Since Engram exhibits sparse access patterns and supports prefetching, its massive embedding tables are well-suited for offloading to lower-tier memory. In this paper, we propose using Compute Express Link (CXL) memory pool for Engram storage. Compared to RDMA, CXL provides fine-grained and low-latency access required by minimal and discrete retrieval patterns of Engram. We integrate the CXL-based Engram pool into SGLang, achieving near-DRAM end-to-end performance. This provides a scalable and cost-efficient storage solution for future Engram-integrated LLMs without compromising inference performance.

Visual Question Answering (VQA) is a fundamental multimodal task that requires models to jointly understand visual and textual information. Early VQA systems relied heavily on language biases, motivating subsequent work to emphasize visual grounding and balanced datasets. With the success of large-scale pre-trained transformers for both text and vision domains -- such as PhoBERT for Vietnamese language understanding and Vision Transformers (ViT) for image representation learning -- multimodal fusion has achieved remarkable progress. For Vietnamese VQA, several datasets have been introduced to promote research in low-resource multimodal learning, including ViVQA, OpenViVQA, and the recently proposed ViTextVQA. These resources enable benchmarking of models that integrate linguistic and visual features in the Vietnamese context. Evaluation of VQA systems often employs automatic metrics originally designed for image captioning or machine translation, such as BLEU, METEOR, CIDEr, Recall, Precision, and F1-score. However, recent research suggests that large language models can further improve the alignment between automatic evaluation and human judgment in VQA tasks. In this work, we explore Vietnamese Visual Question Answering using transformer-based architectures, leveraging both textual and visual pre-training while systematically comparing automatic evaluation metrics under multilingual settings.

This research focuses on developing advanced methods for assessing similarity between recipes by combining different sources of information and analytical approaches. We explore the semantic, lexical, and domain similarity of food recipes, evaluated through the analysis of ingredients, preparation methods, and nutritional attributes. A web-based interface was developed to allow domain experts to validate the combined similarity results. After evaluating 318 recipe pairs, experts agreed on 255 (80%). The evaluation of expert assessments enables the estimation of which similarity aspects--lexical, semantic, or nutritional--are most influential in expert decision-making. The application of these methods has broad implications in the food industry and supports the development of personalized diets, nutrition recommendations, and automated recipe generation systems.

Improving GPU kernel efficiency is crucial for advancing AI systems. Recent work has explored leveraging large language models (LLMs) for GPU kernel generation and optimization. However, existing LLM-based kernel optimization pipelines typically rely on opaque, implicitly learned heuristics within the LLMs to determine optimization strategies. This leads to inefficient trial-and-error and weakly interpretable optimizations. Our key insight is to replace implicit heuristics with expert optimization skills that are knowledge-driven and aware of task trajectories. Specifically, we present KernelSkill, a multi-agent framework with a dual-level memory architecture. KernelSkill operates by coordinating agents with long-term memory of reusable expert skills and short-term memory to prevent repetitive backtracking. On KernelBench Levels 1-3, KernelSkill achieves a 100% success rate and average speedups of 5.44x, 2.82x, and 1.92x over Torch Eager on Levels 1, 2, and 3, respectively, outperforming prior baselines. Code is available at https://github.com/0satan0/KernelMem/.

Current evaluation frameworks and benchmarks for LLM powered agents focus on text chat driven agents, these frameworks do not expose the persona of user to the agent, thus operating in a user agnostic environment. Importantly, in customer experience management domain, the agent's behaviour evolves as the agent learns about user personality. With proliferation of real time TTS and multi-modal language models, LLM based agents are gradually going to become multi-modal. Towards this, we propose the MM-tau-p$^2$ benchmark with metrics for evaluating the robustness of multi-modal agents in dual control setting with and without persona adaption of user, while also taking user inputs in the planning process to resolve a user query. In particular, our work shows that even with state of-the-art frontier LLMs like GPT-5, GPT 4.1, there are additional considerations measured using metrics viz. multi-modal robustness, turn overhead while introducing multi-modality into LLM based agents. Overall, MM-tau-p$^2$ builds on our prior work FOCAL and provides a holistic way of evaluating multi-modal agents in an automated way by introducing 12 novel metrics. We also provide estimates of these metrics on the telecom and retail domains by using the LLM-as-judge approach using carefully crafted prompts with well defined rubrics for evaluating each conversation.

Radiology reports capture crucial longitudinal information on tumor burden, treatment response, and disease progression, yet their unstructured narrative format complicates automated analysis. While large language models (LLMs) have advanced clinical text processing, most state-of-the-art systems remain proprietary, limiting their applicability in privacy-sensitive healthcare environments. We present a fully open-source, locally deployable pipeline for longitudinal information extraction from radiology reports, implemented using the llm_extractinator framework. The system applies the qwen2.5-72b model to extract and link target, non-target, and new lesion data across time points in accordance with RECIST criteria. Evaluation on 50 Dutch CT Thorax/Abdomen report pairs yielded high extraction performance, with attribute-level accuracies of 93.7% for target lesions, 94.9% for non-target lesions, and 94.0% for new lesions. The approach demonstrates that open-source LLMs can achieve clinically meaningful performance in multi-timepoint oncology tasks while ensuring data privacy and reproducibility. These results highlight the potential of locally deployable LLMs for scalable extraction of structured longitudinal data from routine clinical text.

Although concept-based models promise interpretability by explaining predictions with human-understandable concepts, they typically rely on exhaustive annotations and treat concepts as flat and independent. To circumvent this, recent work has introduced Hierarchical Concept Embedding Models (HiCEMs) to explicitly model concept relationships, and Concept Splitting to discover sub-concepts using only coarse annotations. However, both HiCEMs and Concept Splitting are restricted to shallow hierarchies. We overcome this limitation with Multi-Level Concept Splitting (MLCS), which discovers multi-level concept hierarchies from only top-level supervision, and Deep-HiCEMs, an architecture that represents these discovered hierarchies and enables interventions at multiple levels of abstraction. Experiments across multiple datasets show that MLCS discovers human-interpretable concepts absent during training and that Deep-HiCEMs maintain high accuracy while supporting test-time concept interventions that can improve task performance.

Quantum machine learning models based on parameterized circuits can be viewed as Fourier series approximators. However, they often struggle to learn functions with multiple frequency components, particularly high-frequency or non-dominant ones; a phenomenon we term the quantum Fourier parameterization bias. Inspired by recent advances in classical Fourier neural operators (FNOs), we adapt the multi-stage residual learning idea to the quantum domain, iteratively training additional quantum modules on the residuals of previous stages. We evaluate our method on a synthetic benchmark composed of spatially localized frequency components with diverse envelope shapes (Gaussian, Lorentzian, triangular). Systematic experiments show that the number of qubits, the encoding scheme, and residual learning are all crucial for resolving multiple frequencies; residual learning alone can improve test MSE significantly over a single-stage baseline trained for the same total number of epochs. Our work provides a practical framework for enhancing the spectral expressivity of quantum models and offers new insights into their frequency-learning behavior.

Warning: This article includes red-teaming experiments, which contain examples of compromised LLM responses that may be offensive or upsetting. Large Language Models (LLMs) have the potential to create harmful content, such as generating sophisticated phishing emails and assisting in writing code of harmful computer viruses. Thus, it is crucial to ensure their safe and responsible response generation. To reduce the risk of generating harmful or irresponsible content, researchers have developed techniques such as reinforcement learning with human feedback to align LLM's outputs with human values and preferences. However, it is still undetermined whether such measures are sufficient to prevent LLMs from generating interesting responses. In this study, we propose Amnesia, a lightweight activation-space adversarial attack that manipulates internal transformer states to bypass existing safety mechanisms in open-weight LLMs. Through experimental analysis on state-of-the-art, open-weight LLMs, we demonstrate that our attack effectively circumvents existing safeguards, enabling the generation of harmful content without the need for any fine-tuning or additional training. Our experiments on benchmark datasets show that the proposed attack can induce various antisocial behaviors in LLMs. These findings highlight the urgent need for more robust security measures in open-weight LLMs and underscore the importance of continued research to prevent their potential misuse.

We analyse SGD training of a shallow, fully connected network in the NTK scaling and provide a quantitative theory of the catapult phase. We identify an explicit criterion separating two behaviours: When an explicit function $G$, depending only on the kernel, learning rate $η$ and data, is positive, SGD produces large NTK-flattening spikes with high probability; when $G<0$, their probability decays like $(n/η)^{-\vartheta/2}$, for an explicitly characterised $\vartheta\in (0,\infty)$. This yields a concrete parameter-dependent explanation for why such spikes may still be observed at practical widths.

This paper explores potential improvements to the Spatial-Temporal Matching algorithm for aligning the GPS trajectories to road networks. While this algorithm is effective, it presents some limitations in computational efficiency and the accuracy of the results, especially in dense environments with relatively high sampling intervals. To address this, the paper proposes four modifications to the original algorithm: a dynamic buffer, an adaptive observation probability, a redesigned temporal scoring function, and a behavioral analysis to account for the historical mobility patterns. The enhancements are assessed using real-world data from the urban area of Milan, and through newly defined evaluation metrics to be applied in the absence of ground truth. The results of the experiment show significant improvements in performance efficiency and path quality across various metrics.

Stochastic port-Hamiltonian systems represent open dynamical systems with dissipation, inputs, and stochastic forcing in an energy based form. We introduce stochastic port-Hamiltonian neural networks, SPH-NNs, which parameterize the Hamiltonian with a feedforward network and enforce skew symmetry of the interconnection matrix and positive semidefiniteness of the dissipation matrix. For Itô dynamics we establish a weak passivity inequality in expectation under an explicit generator condition, stated for a stopped process on a compact set. We also prove a universal approximation result showing that, on any compact set and finite horizon, SPH-NNs approximate the coefficients of a target stochastic port-Hamiltonian system with $C^2$ accuracy of the Hamiltonian and yield coupled solutions that remain close in mean square up to the exit time. Experiments on noisy mass spring, Duffing, and Van der Pol oscillators show improved long horizon rollouts and reduced energy error relative to a multilayer perceptron baseline.

Offline-to-online reinforcement learning (RL) offers a promising paradigm for robotics by pre-training policies on safe, offline demonstrations and fine-tuning them via online interaction. However, a fundamental challenge remains: how to safely explore online without deviating from the behavioral support of the offline data? While recent methods leverage conditional variational autoencoders (CVAEs) to bound exploration within a latent space, they inherently suffer from an exploitation gap -- a performance ceiling imposed by the decoder's reconstruction loss. We introduce SPAARS, a curriculum learning framework that initially constrains exploration to the low-dimensional latent manifold for sample-efficient, safe behavioral improvement, then seamlessly transfers control to the raw action space, bypassing the decoder bottleneck. SPAARS has two instantiations: the CVAE-based variant requires only unordered (s,a) pairs and no trajectory segmentation; SPAARS-SUPE pairs SPAARS with OPAL temporal skill pretraining for stronger exploration structure at the cost of requiring trajectory chunks. We prove an upper bound on the exploitation gap using the Performance Difference Lemma, establish that latent-space policy gradients achieve provable variance reduction over raw-space exploration, and show that concurrent behavioral cloning during the latent phase directly controls curriculum transition stability. Empirically, SPAARS-SUPE achieves 0.825 normalized return on kitchen-mixed-v0 versus 0.75 for SUPE, with 5x better sample efficiency; standalone SPAARS achieves 92.7 and 102.9 normalized return on hopper-medium-v2 and walker2d-medium-v2 respectively, surpassing IQL baselines of 66.3 and 78.3 respectively, confirming the utility of the unordered-pair CVAE instantiation.

Activation steering is a widely used approach for controlling large language model (LLM) behavior by intervening on internal representations. Existing methods largely rely on the Linear Representation Hypothesis, assuming behavioral attributes can be manipulated using global linear directions. In practice, however, such linear interventions often behave inconsistently. We question this assumption by analyzing the intrinsic geometry of LLM activation spaces. Measuring geometric distortion via the ratio of geodesic to Euclidean distances, we observe substantial and concept-dependent distortions, indicating that activation spaces are not well-approximated by a globally linear geometry. Motivated by this, we propose "Curveball steering", a nonlinear steering method based on polynomial kernel PCA that performs interventions in a feature space, better respecting the learned activation geometry. Curveball steering consistently outperforms linear PCA-based steering, particularly in regimes exhibiting strong geometric distortion, suggesting that geometry-aware, nonlinear steering provides a principled alternative to global, linear interventions.

As backdoor attacks in UAV-based decentralized federated learning (DFL) grow increasingly stealthy and sophisticated, existing defenses have likewise escalated in complexity. Yet these defenses, which rely heavily on outlier detection, remain vulnerable to carefully crafted backdoors. In UAV-DFL, the lack of global coordination and limited resources further render outlier-based defenses impractical. Against this backdrop, gradient spectral analysis offers a promising alternative. While prior work primarily leverages low-frequency coefficients for pairwise comparisons, it neglects to analyze the intrinsic spectral characteristics of backdoor gradients. Through empirical analysis of existing stealthy attacks, we reveal a key insight: the more effort attackers invest in mimicking benign behaviors, the more distinct the spectral concentration becomes. Motivated by this, we propose Task-Aware Spectral Energy Refine (TASER) -- a decentralized defense framework. To our knowledge, this is the first efficient backdoor defense that utilizes spectral concentration instead of complex outlier detection, enabling mitigation of stealthy attacks by structurally disrupting the backdoor task. To suppress the backdoor task, TASER preserves main-task-relevant frequency coefficients and discards others. We provide theoretical guarantees and demonstrate through experiments that TASER remains effective against stealthy backdoor attacks that bypass outlier-based defenses, achieving attack success rate below 20% and accuracy loss under 5%.

Aggregate outcome variables collected through surveys and administrative records are often subject to systematic measurement error. For instance, in disaster loss databases, county-level losses reported may differ from the true damages due to variations in on-the-ground data collection capacity, reporting practices, and event characteristics. Such miscalibration complicates downstream analysis and decision-making. We study the problem of outcome miscalibration and propose a framework guided by proxy variables for estimating and correcting the systematic errors. We model the data-generating process using a causal graph that separates latent content variables driving the true outcome from the latent bias variables that induce systematic errors. The key insight is that proxy variables that depend on the true outcome but are independent of the bias mechanism provide identifying information for quantifying the bias. Leveraging this structure, we introduce a two-stage approach that utilizes variational autoencoders to disentangle content and bias latents, enabling us to estimate the effect of bias on the outcome of interest. We analyze the assumptions underlying our approach and evaluate it on synthetic data, semi-synthetic datasets derived from randomized trials, and a real-world case study of disaster loss reporting.

We construct a minimal task that isolates conditional learning in neural networks: a surjective map with K-fold ambiguity, resolved by a selector token z, so H(A | B) = log K while H(A | B, z) = 0. The model learns the marginal P(A | B) first, producing a plateau at exactly log K, before acquiring the full conditional in a sharp, collective transition. The plateau has a clean decomposition: height = log K (set by ambiguity), duration = f(D) (set by dataset size D, not K). Gradient noise stabilizes the marginal solution: higher learning rates monotonically slow the transition (3.6* across a 7* η range at fixed throughput), and batch-size reduction delays escape, consistent with an entropic force opposing departure from the low-gradient marginal. Internally, a selector-routing head assembles during the plateau, leading the loss transition by ~50% of the waiting time. This is the Type 2 directional asymmetry of Papadopoulos et al. [2024], measured dynamically: we track the excess risk from log K to zero and characterize what stabilizes it, what triggers its collapse, and how long it takes.

Large Language Models (LLMs) show promise for Automated Program Repair (APR), yet their effectiveness on security vulnerabilities remains poorly characterized. This study analyzes 319 LLM-generated security patchesacross 64 Java vulnerabilities from the Vul4J benchmark. Using tri-axis evaluation (compilation, security via PoV tests, functionality via test suites), the analysis reveals that only 24.8% of patches achieve full correctness, while 51.4% fail both security and functionality. The dominant failure mode is semantic misunderstanding: LLMs produce syntactically valid code but apply incorrect repair strategies. The proposed Security Repair Score (SRS) quantifies this gap, showing LLMs preserve functionality (mean 0.832) but struggle with security (mean 0.251). Vulnerability type strongly predicts difficulty, with fix rates ranging from 0% (input validation) to 45% (infinite loop). These findings demonstrate that LLM security patches require rigorous validation before deployment.

Time series foundation models (TSFMs) are increasingly deployed in high-stakes domains, yet their internal representations remain opaque. We present the first application of sparse autoencoders (SAEs) to a TSFM, training TopK SAEs on activations of Chronos-T5-Large (710M parameters) across six layers. Through 392 single-feature ablation experiments, we establish that every ablated feature produces a positive CRPS degradation, confirming causal relevance. Our analysis reveals a depth-dependent hierarchy: early encoder layers encode low-level frequency features, the mid-encoder concentrates causally critical change-detection features, and the final encoder compresses a rich but less causally important taxonomy of temporal concepts. The most critical features reside in the mid-encoder (max single-feature Delta CRPS = 38.61), not in the semantically richest final encoder layer, where progressive ablation paradoxically improves forecast quality. These findings demonstrate that mechanistic interpretability transfers effectively to TSFMs and that Chronos-T5 relies on abrupt-dynamics detection rather than periodic pattern recognition.

We present improved lower bounds for five classical Ramsey numbers: $\mathbf{R}(3, 13)$ is increased from $60$ to $61$, $\mathbf{R}(3, 18)$ from $99$ to $100$, $\mathbf{R}(4, 13)$ from $138$ to $139$, $\mathbf{R}(4, 14)$ from $147$ to $148$, and $\mathbf{R}(4, 15)$ from $158$ to $159$. These results were achieved using AlphaEvolve, an LLM-based code mutation agent. Beyond these new results, we successfully recovered lower bounds for all Ramsey numbers known to be exact, and matched the best known lower bounds across many other cases. These include bounds for which previous work does not detail the algorithms used. Virtually all known Ramsey lower bounds are derived computationally, with bespoke search algorithms each delivering a handful of results. AlphaEvolve is a single meta-algorithm yielding search algorithms for all of our results.

Tool-based Agentic Reinforcement Learning (TARL) has emerged as a promising paradigm for training search agents to interact with external tools for a multi-turn information-seeking process autonomously. However, we identify a critical training instability that leads to catastrophic model collapse: Importance Sampling Distribution Drift(ISDD). In Group Relative Policy Optimization(GRPO), a widely adopted TARL algorithm, ISDD manifests as a precipitous decline in the importance sampling ratios, which nullifies gradient updates and triggers irreversible training failure. To address this, we propose \textbf{S}earch \textbf{A}gent \textbf{P}olicy \textbf{O}ptimization (\textbf{SAPO}), which stabilizes training via a conditional token-level KL constraint. Unlike hard clipping, which ignores distributional divergence, SAPO selectively penalizes the KL divergence between the current and old policies. Crucially, this penalty is applied only to positive tokens with low probabilities where the policy has shifted excessively, thereby preventing distribution drift while preserving gradient flow. Remarkably, SAPO requires only one-line code modification to standard GRPO, ensuring immediate deployability. Extensive experiments across seven QA benchmarks demonstrate that SAPO achieves \textbf{+10.6\% absolute improvement} (+31.5\% relative) over Search-R1, yielding consistent gains across varying model scales (1.5B, 14B) and families (Qwen, LLaMA).

Most adversarial evaluations of large language model (LLM) safety assess single prompts and report binary pass/fail outcomes, which fails to capture how safety properties evolve under sustained adversarial interaction. We present ADVERSA, an automated red-teaming framework that measures guardrail degradation dynamics as continuous per-round compliance trajectories rather than discrete jailbreak events. ADVERSA uses a fine-tuned 70B attacker model (ADVERSA-Red, Llama-3.1-70B-Instruct with QLoRA) that eliminates the attacker-side safety refusals that render off-the-shelf models unreliable as attackers, scoring victim responses on a structured 5-point rubric that treats partial compliance as a distinct measurable state. We report a controlled experiment across three frontier victim models (Claude Opus 4.6, Gemini 3.1 Pro, GPT-5.2) using a triple-judge consensus architecture in which judge reliability is measured as a first-class research outcome rather than assumed. Across 15 conversations of up to 10 adversarial rounds, we observe a 26.7% jailbreak rate with an average jailbreak round of 1.25, suggesting that in this evaluation setting, successful jailbreaks were concentrated in early rounds rather than accumulating through sustained pressure. We document inter-judge agreement rates, self-judge scoring tendencies, attacker drift as a failure mode in fine-tuned attackers deployed out of their training distribution, and attacker refusals as a previously-underreported confound in victim resistance measurement. All limitations are stated explicitly. Attack prompts are withheld per responsible disclosure policy; all other experimental artifacts are released.

Vision-language pretraining has driven significant progress in medical image analysis. However, current methods typically supervise visual encoders using one-hot labels or free-form text, neither of which effectively captures the complex semantic relationships among clinical findings. In this study, we introduce VIVID-Med, a novel framework that leverages a frozen large language model (LLM) as a structured semantic teacher to pretrain medical vision transformers (ViTs). VIVID-Med translates clinical findings into verifiable JSON field-state pairs via a Unified Medical Schema (UMS), utilizing answerability-aware masking to focus optimization. It then employs Structured Prediction Decomposition (SPD) to partition cross-attention into orthogonality-regularized query groups, extracting complementary visual aspects. Crucially, the LLM is discarded post-training, yielding a lightweight, deployable ViT-only backbone. We evaluated VIVID-Med across multiple settings: on CheXpert linear probing, it achieves a macro-AUC of 0.8588, outperforming BiomedCLIP by +6.65 points while using 500x less data. It also demonstrates robust zero-shot cross-domain transfer to NIH ChestX-ray14 (0.7225 macro-AUC) and strong cross-modality generalization to CT, achieving 0.8413 AUC on LIDC-IDRI lung nodule classification and 0.9969 macro-AUC on OrganAMNIST 11-organ classification. VIVID-Med offers a highly efficient, scalable alternative to deploying resource-heavy vision-language models in clinical settings.

Muon has recently shown promising results in LLM training. In this work, we study how to further improve Muon. We argue that Muon's orthogonalized update rule suppresses the emergence of heavy-tailed weight spectra and over-emphasizes the training along noise-dominated directions. Motivated by the Heavy-Tailed Self-Regularization (HT-SR) theory, we propose HTMuon. HTMuon preserves Muon's ability to capture parameter interdependencies while producing heavier-tailed updates and inducing heavier-tailed weight spectra. Experiments on LLM pretraining and image classification show that HTMuon consistently improves performance over state-of-the-art baselines and can also serve as a plug-in on top of existing Muon variants. For example, on LLaMA pretraining on the C4 dataset, HTMuon reduces perplexity by up to $0.98$ compared to Muon. We further theoretically show that HTMuon corresponds to steepest descent under the Schatten-$q$ norm constraint and provide convergence analysis in smooth non-convex settings. The implementation of HTMuon is available at https://github.com/TDCSZ327/HTmuon.

The Epistemic Support-Point Filter (ESPF) was designed around a single epistemological commitment: be quick to embrace ignorance and slow to assert certainty. This paper proves that this commitment has a precise mathematical form and that the ESPF is the unique optimal filter implementing it within the class of epistemically admissible evidence-only filters. The ESPF synthesizes two complementary principles acting at different phases of the recursion. In propagation, it enacts Jaynesian maximum entropy: the support spreads as widely as the dynamics allow, assuming maximal ignorance consistent with known constraints. In the measurement update, it enacts Popperian falsification: hypotheses are eliminated by evidence alone. Any rule incorporating prior possibility is strictly suboptimal and risks race-to-bottom bias. The optimality criterion is possibilistic minimax entropy: among all evidence-only selection rules, minimum-q selection minimizes log det(MVEE), the worst-case possibilistic entropy. Three lemmas establish the result: the Possibilistic Entropy Lemma identifies the ignorance functional; the Possibilistic Cramér-Rao Lemma bounds entropy reduction per measurement; the Evidence-Optimality Lemma proves minimum-q selection is the unique minimizer. The ESPF differs from Bayesian filters by minimizing worst-case epistemic ignorance rather than expected uncertainty. The Kalman filter is recovered in the Gaussian limit. Numerical validation over a 2-day 877-step Smolyak Level-3 orbital tracking run confirms the regime structure under both nominal and stress conditions.

Action-conditioned video models offer a promising path to building general-purpose robot simulators that can improve directly from data. Yet, despite training on large-scale robot datasets, current state-of-the-art video models still struggle to predict physically consistent robot-object interactions that are crucial in robotic manipulation. To close this gap, we present PlayWorld, a simple, scalable, and fully autonomous pipeline for training high-fidelity video world simulators from interaction experience. In contrast to prior approaches that rely on success-biased human demonstrations, PlayWorld is the first system capable of learning entirely from unsupervised robot self-play, enabling naturally scalable data collection while capturing complex, long-tailed physical interactions essential for modeling realistic object dynamics. Experiments across diverse manipulation tasks show that PlayWorld generates high-quality, physically consistent predictions for contact-rich interactions that are not captured by world models trained on human-collected data. We further demonstrate the versatility of PlayWorld in enabling fine-grained failure prediction and policy evaluation, with up to 40% improvements over human-collected data. Finally, we demonstrate how PlayWorld enables reinforcement learning in the world model, improving policy performance by 65% in success rates when deployed in the real world.

Collaborative healthcare research across multiple institutions increasingly requires diverse clinical datasets, but cross-border data sharing is strictly constrained by privacy regulations. Federated learning (FL) enables model training while keeping data local; however, many existing frameworks remain proof-of-concept and do not adequately address governance risks such as unauthorised participation, misuse, and lack of accountability. In particular, enforceable mechanisms for authentication, authorisation, and accounting (AAA) are often missing, limiting real-world clinical deployment. This paper presents FLA$^3$ (Federated Learning with Authentication, Authorisation, and Accounting), a governance-aware federated learning platform that operationalises regulatory obligations through runtime policy enforcement. FLA$^3$ integrates eXtensible Access Control Markup Language (XACML) compliant attribute-based access control (ABAC), cryptographic accounting, and study-scoped federation directly into the federated learning orchestration layer to enforce institutional sovereignty and protocol adherence. We evaluate FLA$^3$ through two complementary studies. First, we demonstrate operational feasibility by deploying the platform infrastructure across five BloodCounts! Consortium institutions in four countries: United Kingdom, Netherlands, India, and The Gambia. Second, we assess clinical utility using simulated federation of full blood count (FBC) data from 54,446 samples from 35,315 subjects across 25 centres in the INTERVAL study. Results show that FLA$^3$ achieves predictive performance comparable to centralised training while strictly enforcing governance constraints. These results show that enforceable governance can function as a first-class privacy-preserving control, improving trustworthiness for scalable artificial intelligence (AI) in cross-jurisdictional healthcare deployments.

As LLM agents evolve into collaborative multi-agent systems, their memory requirements grow rapidly in complexity. This position paper frames multi-agent memory as a computer architecture problem. We distinguish shared and distributed memory paradigms, propose a three-layer memory hierarchy (I/O, cache, and memory), and identify two critical protocol gaps: cache sharing across agents and structured memory access control. We argue that the most pressing open challenge is multi-agent memory consistency. Our architectural framing provides a foundation for building reliable, scalable multi-agent systems.

Pathology underpins modern diagnosis and cancer care, yet its most valuable asset, the accumulated experience encoded in millions of narrative reports, remains largely inaccessible. Although institutions are rapidly digitizing pathology workflows, storing data without effective mechanisms for retrieval and reasoning risks transforming archives into a passive data repository, where institutional knowledge exists but cannot meaningfully inform patient care. True progress requires not only digitization, but the ability for pathologists to interrogate prior similar cases in real time while evaluating a new diagnostic dilemma. We present PathoScribe, a unified retrieval-augmented large language model (LLM) framework designed to transform static pathology archives into a searchable, reasoning-enabled living library. PathoScribe enables natural language case exploration, automated cohort construction, clinical question answering, immunohistochemistry (IHC) panel recommendation, and prompt-controlled report transformation within a single architecture. Evaluated on 70,000 multi-institutional surgical pathology reports, PathoScribe achieved perfect Recall@10 for natural language case retrieval and demonstrated high-quality retrieval-grounded reasoning (mean reviewer score 4.56/5). Critically, the system operationalized automated cohort construction from free-text eligibility criteria, assembling research-ready cohorts in minutes (mean 9.2 minutes) with 91.3% agreement to human reviewers and no eligible cases incorrectly excluded, representing orders-of-magnitude reductions in time and cost compared to traditional manual chart review. This work establishes a scalable foundation for converting digital pathology archives from passive storage systems into active clinical intelligence platforms.

AI agents that execute tasks via tool calls frequently hallucinate results - fabricating tool executions, misstating output counts, or presenting inferences as facts. Recent approaches to verifiable AI inference rely on zero-knowledge proofs, which provide cryptographic guarantees but impose minutes of proving time per query, making them impractical for interactive agents. We propose NabaOS, a lightweight verification framework inspired by Indian epistemology (Nyaya Shastra), which classifies every claim in an LLM response by its epistemic source (pramana): direct tool output (pratyaksha), inference (anumana), external testimony (shabda), absence (abhava), or ungrounded opinion. Our runtime generates HMAC-signed tool execution receipts that the LLM cannot forge, then cross-references claims against these receipts to detect hallucinations in real time. We evaluate on NyayaVerifyBench, a new benchmark of 1,800 agent response scenarios across four languages with injected hallucinations of six types. NabaOS detects 94.2% of fabricated tool references, 87.6% of count misstatements, and 91.3% of false absence claims, with <15ms verification overhead per response. For deep delegation (agents performing multi-step web tasks), our cross-checking protocol catches 78.4% of URL fabrications via independent re-fetching. We compare against five approaches: zkLLM (cryptographic proofs, 180s/query), TOPLOC (locality-sensitive hashing), SPEX (sampling-based proof of execution), tensor commitments, and self-consistency checking. NabaOS achieves the best cost-latency-coverage trade-off for interactive agents: 94.2% coverage at <15ms versus zkLLM's near-perfect coverage at 180,000ms. For interactive agents, practical receipt-based verification provides better cost-benefit than cryptographic proofs, and epistemic classification gives users actionable trust signals rather than binary judgments.

Considering the high volume, wide variety, and rapid speed of data generation, investigating feature selection methods for big data presents various applications and advantages. By removing irrelevant and redundant features, feature selection reduces data dimensions, thereby facilitating optimal decision-making within decision systems. One of the key tools for feature selection in hybrid information systems is fuzzy rough set theory. However, this theory faces two significant challenges: First, obtaining fuzzy equivalence relations through intersection operations in high-dimensional spaces can be both time-consuming and memory-intensive. Additionally, this method may produce noisy data, complicating the feature selection process. The purpose and innovation of this paper are to address these issues. We proposed a new feature selection model that calculates the combined distance between objects and subsequently used this information to derive the fuzzy equivalence relation. Rather than directly solving the feature selection problem, this approach reformulates it into an optimization problem that can be tackled using appropriate meta-heuristic algorithms. We have named this new approach FSbuHD. The FSbuHD model operates in two modes - normal and optimistic - based on the selection of one of the two introduced fuzzy equivalence relations. The model is then tested on standard datasets from the UCI repository and compared with other algorithms. The results of this research demonstrate that FSbuHD is one of the most efficient and effective methods for feature selection when compared to previous methods and algorithms.

Accessing sensitive patient data for machine learning is challenging due to privacy concerns. Datasets with annotations of personally identifiable information are crucial for developing and testing anonymization systems to enable safe data sharing that complies with privacy regulations. Since accessing real patient data is a bottleneck, synthetic data offers an efficient solution for data scarcity, bypassing privacy regulations that apply to real data. Moreover, neural machine translation can help to create high-quality data for low-resource languages by translating validated real or synthetic data from a high-resource language. In this work, we create a multilingual anonymization benchmark in ten languages, using a machine translation methodology that preserves the original annotations and renders names of cities and people in a culturally and contextually appropriate form in each target language. Our evaluation study with medical professionals confirms the quality of the translations, both in general and with respect to the translation and adaptation of personal information. Our benchmark with over 2,500 annotations of personal information can be used in many applications, including training annotators, validating annotations across institutions without legal complications, and helping improve the performance of automatic personal information detection. We make our benchmark and annotation guidelines available for further research.

Software supply-chain security requires provenance mechanisms that support reproducibility and vulnerability assessment under dynamic execution conditions. Conventional Software Bills of Materials (SBOMs) provide static dependency inventories but cannot capture runtime behaviour, environment drift, or exploitability context. This paper introduces agentic Artificial Intelligence Bills of Materials (AIBOMs), extending SBOMs into active provenance artefacts through autonomous, policy-constrained reasoning. We present an agentic AIBOM framework based on a multi-agent architecture comprising (i) a baseline environment reconstruction agent (MCP), (ii) a runtime dependency and drift-monitoring agent (A2A), and (iii) a policy-aware vulnerability and VEX reasoning agent (AGNTCY). These agents generate contextual exploitability assertions by combining runtime execution evidence, dependency usage, and environmental mitigations with ISO/IEC 20153:2025 Common Security Advisory Framework (CSAF) v2.0 semantics. Exploitability is expressed via structured VEX assertions rather than enforcement actions. The framework introduces minimal, standards-aligned schema extensions to CycloneDX and SPDX, capturing execution context, dependency evolution, and agent decision provenance while preserving interoperability. Evaluation across heterogeneous analytical workloads demonstrates improved runtime dependency capture, reproducibility fidelity, and stability of vulnerability interpretation compared with established provenance systems, with low computational overhead. Ablation studies confirm that each agent contributes distinct capabilities unavailable through deterministic automation.

Most OAuth service providers, such as Google and Microsoft, offer only a limited range of coarse-grained data access. As a result, third-party OAuth applications often end up accessing more user data than necessary, even if their developers want to minimize data access. We present OAuthHub, a development framework that leverages users' personal devices as the intermediary controller for OAuth-based data sharing between cloud services. The key innovations of OAuthHub are: (1) the insight that discretionary data access is largely unnecessary for most OAuth apps, which typically only require access at three well-defined moments-during installation, in response to user actions, and at scheduled intervals; (2) a development framework that requires explicit declarations of intended data access and supports the three common access patterns through intermittently available personal devices; and (3) a centralized runtime permission model for managing OAuth access across providers. We evaluated OAuthHub with three real-world apps on both PCs and mobile phones and found that OAuthHub requires moderate changes to the application code and imposes insignificant performance overheads. Our study with 18 developers showed that participants completed programming tasks significantly faster (9.1 vs. 18.0 minutes) with less code (4.7 vs. 15.8 lines) using OAuthHub than conventional OAuth APIs.

We introduce Fish Audio S2, an open-sourced text-to-speech system featuring multi-speaker, multi-turn generation, and, most importantly, instruction-following control via natural-language descriptions. To scale training, we develop a multi-stage training recipe together with a staged data pipeline covering video captioning and speech captioning, voice-quality assessment, and reward modeling. To push the frontier of open-source TTS, we release our model weights, fine-tuning code, and an SGLang-based inference engine. The inference engine is production-ready for streaming, achieving an RTF of 0.195 and a time-to-first-audio below 100 ms.Our code and weights are available on GitHub (https://github.com/fishaudio/fish-speech) and Hugging Face (https://huggingface.co/fishaudio/s2-pro). We highly encourage readers to visit https://fish.audio to try custom voices.

Retrieval-augmented generation (RAG) systems combine document retrieval with a generative model to address complex information seeking tasks like report generation. While the relationship between retrieval quality and generation effectiveness seems intuitive, it has not been systematically studied. We investigate whether upstream retrieval metrics can serve as reliable early indicators of the final generated response's information coverage. Through experiments across two text RAG benchmarks (TREC NeuCLIR 2024 and TREC RAG 2024) and one multimodal benchmark (WikiVideo), we analyze 15 text retrieval stacks and 10 multimodal retrieval stacks across four RAG pipelines and multiple evaluation frameworks (Auto-ARGUE and MiRAGE). Our findings demonstrate strong correlations between coverage-based retrieval metrics and nugget coverage in generated responses at both topic and system levels. This relationship holds most strongly when retrieval objectives align with generation goals, though more complex iterative RAG pipelines can partially decouple generation quality from retrieval effectiveness. These findings provide empirical support for using retrieval metrics as proxies for RAG performance.

Pre-training is crucial for large language models (LLMs), as it is when most representations and capabilities are acquired. However, natural language pre-training has problems: high-quality text is finite, it contains human biases, and it entangles knowledge with reasoning. This raises a fundamental question: is natural language the only path to intelligence? We propose using neural cellular automata (NCA) to generate synthetic, non-linguistic data for pre-pre-training LLMs--training on synthetic-then-natural language. NCA data exhibits rich spatiotemporal structure and statistics resembling natural language while being controllable and cheap to generate at scale. We find that pre-pre-training on only 164M NCA tokens improves downstream language modeling by up to 6% and accelerates convergence by up to 1.6x. Surprisingly, this even outperforms pre-pre-training on 1.6B tokens of natural language from Common Crawl with more compute. These gains also transfer to reasoning benchmarks, including GSM8K, HumanEval, and BigBench-Lite. Investigating what drives transfer, we find that attention layers are the most transferable, and that optimal NCA complexity varies by domain: code benefits from simpler dynamics, while math and web text favor more complex ones. These results enable systematic tuning of the synthetic distribution to target domains. More broadly, our work opens a path toward more efficient models with fully synthetic pre-training.

In 2004, while studying the information geometry of binary Bayesian networks (bitnets), the author conjectured that the volume-averaged Ricci scalar <R> computed with respect to the Fisher information metric is universally quantized to positive half-integers: <R> in (1/2)Z. This paper resolves the conjecture after 20 years. We prove it for tree-structured and complete-graph bitnets via a universal Beta function cancellation mechanism, and disprove it in general by exhibiting explicit loop counterexamples. We extend the program to Gaussian DAG networks, where a sign dichotomy holds: discrete bitnets have positive curvature, while Gaussian networks form solvable Lie groups with negative curvature.

The Pickup and Delivery Problem (PDP) is a fundamental and challenging variant of the Vehicle Routing Problem, characterized by tightly coupled pickup--delivery pairs, precedence constraints, and spatial layouts that often exhibit clustering. Existing deep reinforcement learning (DRL) approaches either model all nodes on a flat graph, relying on implicit learning to enforce constraints, or achieve strong performance through inference-time collaborative search at the cost of substantial latency. In this paper, we propose \emph{CAADRL} (Cluster-Aware Attention-based Deep Reinforcement Learning), a DRL framework that explicitly exploits the multi-scale structure of PDP instances via cluster-aware encoding and hierarchical decoding. The encoder builds on a Transformer and combines global self-attention with intra-cluster attention over depot, pickup, and delivery nodes, producing embeddings that are both globally informative and locally role-aware. Based on these embeddings, we introduce a Dynamic Dual-Decoder with a learnable gate that balances intra-cluster routing and inter-cluster transitions at each step. The policy is trained end-to-end with a POMO-style policy gradient scheme using multiple symmetric rollouts per instance. Experiments on synthetic clustered and uniform PDP benchmarks show that CAADRL matches or improves upon strong state-of-the-art baselines on clustered instances and remains highly competitive on uniform instances, particularly as problem size increases. Crucially, our method achieves these results with substantially lower inference time than neural collaborative-search baselines, suggesting that explicitly modeling cluster structure provides an effective and efficient inductive bias for neural PDP solvers.

Vision-language-action(VLA) models have shown great promise as generalist policies for a large range of relatively simple tasks. However, they demonstrate limited performance on more complex tasks, such as those requiring complex spatial or semantic understanding, manipulation in clutter, or precise manipulation. We propose OMNIGUIDE, a flexible framework that improves VLA performance on such tasks by leveraging arbitrary sources of guidance, such as 3D foundation models, semantic-reasoning VLMs, and human pose models. We show how many kinds of guidance can be naturally expressed as differentiable energy functions with task-specific attractors and repellers located in 3D space, that influence the sampling of VLA actions. In this way, OMNIGUIDE enables guidance sources with complementary task-relevant strengths to improve a VLA model's performance on challenging tasks. Extensive experiments in both simulation and real-world environments, across diverse sources of guidance, demonstrate that OMNIGUIDE enhances the performance of state-of-the-art generalist policies (e.g., $π_{0.5}$, GR00T N1.6) significantly across success and safety rates. Critically, our unified framework matches or surpasses the performance of prior methods designed to incorporate specific sources of guidance into VLA policies. Project Page: $\href{https://omniguide.github.io/}{this \; url}$

We present Midicoth, a lossless compression system that introduces a micro-diffusion denoising layer for improving probability estimates produced by adaptive statistical models. In compressors such as Prediction by Partial Matching (PPM), probability estimates are smoothed by a prior to handle sparse observations. When contexts have been seen only a few times, this prior dominates the prediction and produces distributions that are significantly flatter than the true source distribution, leading to compression inefficiency. Midicoth addresses this limitation by treating prior smoothing as a shrinkage process and applying a reverse denoising step that corrects predicted probabilities using empirical calibration statistics. To make this correction data-efficient, the method decomposes each byte prediction into a hierarchy of binary decisions along a bitwise tree. This converts a single 256-way calibration problem into a sequence of binary calibration tasks, enabling reliable estimation of correction terms from relatively small numbers of observations. The denoising process is applied in multiple successive steps, allowing each stage to refine residual prediction errors left by the previous one. The micro-diffusion layer operates as a lightweight post-blend calibration stage applied after all model predictions have been combined, allowing it to correct systematic biases in the final probability distribution. Midicoth combines five fully online components: an adaptive PPM model, a long-range match model, a trie-based word model, a high-order context model, and the micro-diffusion denoiser applied as the final stage.

Large language model (LLM)-based agents trained with reinforcement learning (RL) have shown strong potential on complex interactive tasks. However, standard RL paradigms favor static problem-solving over continuous adaptation: agents often converge to suboptimal strategies due to insufficient exploration, while learned knowledge remains implicit within parameters rather than explicitly retrievable, limiting effective experiential learning. To address these limitations, we introduce RetroAgent, an online RL framework that empowers agents to master complex interactive environments not just by solving, but by evolving. Concretely, RetroAgent features a hindsight self-reflection mechanism that produces dual intrinsic feedback: (1) intrinsic numerical feedback that that tracks incremental subtask completion relative to prior attempts, rewarding promising explorations, and (2) intrinsic language feedback that distills reusable lessons into a memory buffer, retrieved via our proposed Similarity & Utility-Aware Upper Confidence Bound (SimUtil-UCB) strategy balancing relevance, utility, and exploration to effectively leverage past experiences. Extensive experiments on two model families across four challenging agentic tasks demonstrate that RetroAgent significantly outperforms existing methods, achieving state-of-the-art results -- e.g., surpassing Group Relative Policy Optimization (GRPO)-trained agents by +18.3% on ALFWorld, +15.4% on WebShop, +27.1% on Sokoban, and +8.9% on MineSweeper -- while exhibiting strong test-time adaptation and generalization to out-of-distribution scenarios.

Self-supervised masked modeling shows promise for encrypted traffic classification by masking and reconstructing raw bytes. Yet recent work reveals these methods fail to reduce reliance on labeled data despite costly pretraining: under frozen encoder evaluation, accuracy drops from greater than 0.9 to less than 0.47. We argue the root cause is inductive bias mismatch: flattening traffic into byte sequences destroys protocol-defined semantics. We identify three specific issues: 1) field unpredictability, random fields like ip.id are unlearnable yet treated as reconstruction targets; 2) embedding confusion, semantically distinct fields collapse into a unified embedding space; 3) metadata loss, capture-time metadata essential for temporal analysis is discarded. To address this, we propose a protocol-native paradigm that treats protocol-defined field semantics as architectural priors, reformulating the task to align with the data's intrinsic tabular modality rather than incrementally adapting sequence-based architectures. Instantiating this paradigm, we introduce FlowSem-MAE, a tabular masked autoencoder built on Flow Semantic Units (FSUs). It features predictability-guided filtering that focuses on learnable FSUs, FSU-specific embeddings to preserve field boundaries, and dual-axis attention to capture intra-packet and temporal patterns. FlowSem-MAE significantly outperforms state-of-the-art across datasets. With only half labeled data, it outperforms most existing methods trained on full data.

Chain-of-thought (CoT) prompting enables reasoning in language models but requires explicit verbalization of intermediate steps. Looped transformers offer an alternative by iteratively refining representations within hidden states. This parameter efficiency comes at a cost, as looped models lack the storage capacity of deeper models which use unique weights per layer. In this work, we investigate transformer models that feature both adaptive per-layer looping, where each transformer block learns to iterate its hidden state via a learned halting mechanism, and gated memory banks, that provide additional learned storage. We find that looping primarily benefits mathematical reasoning, while memory banks help recover performance on commonsense tasks compared to parameter and FLOP matched models. Combining both mechanisms yields a model that outperforms an iso-FLOP baseline, with three times the number of layers, across math benchmarks. Analysis of model internals reveals layer specialization: early layers learn to loop minimally and access memory sparingly, while later layers do both more heavily.

Ensemble learning is a well established body of methods for machine learning to enhance predictive performance by combining multiple algorithms/models. Combinatorial Fusion Analysis (CFA) has provided method and practice for combining multiple scoring systems, using rank-score characteristic (RSC) function and cognitive diversity (CD), including ensemble method and model fusion. However, there is no general-purpose Python tool available that incorporate these techniques. In this paper we introduce \texttt{InFusionLayer}, a machine learning architecture inspired by CFA at the system fusion level that uses a moderate set of base models to optimize unsupervised and supervised learning multiclassification problems. We demonstrate \texttt{InFusionLayer}'s ease of use for PyTorch, TensorFlow, and Scikit-learn workflows by validating its performance on various computer vision datasets. Our results highlight the practical advantages of incorporating distinctive features of RSC function and CD, paving the way for more sophisticated ensemble learning applications in machine learning. We open-sourced our code to encourage continuing development and community accessibility to leverage CFA on github: https://github.com/ewroginek/Infusion

Learning to understand speech appears almost effortless for typically developing infants, yet from an information-processing perspective, acquiring a language from acoustic speech is an enormous challenge. This chapter reviews recent developments in using computational models to understand early language acquisition from speech and audiovisual input. The focus is on self-supervised and visually grounded models of perceptual learning. We show how these models are becoming increasingly powerful in learning various aspects of speech without strong linguistic priors, and how many features of early language development can be explained through a shared set of learning principles-principles broadly compatible with multiple theories of language acquisition and human cognition. We also discuss how modern learning simulations are gradually becoming more realistic, both in terms of input data and in linking model behavior to empirical findings on infant language development.

Recent advancements in LLM-based information-seeking agents have achieved record-breaking performance on established benchmarks. However, these agents remain heavily reliant on search-engine-indexed knowledge, leaving a critical blind spot: Unindexed Information Seeking (UIS). This paper identifies and explores the UIS problem, where vital information is not captured by search engine crawlers, such as overlooked content, dynamic webpages, and embedded files. Despite its significance, UIS remains an underexplored challenge. To address this gap, we introduce UIS-QA, the first dedicated UIS benchmark, comprising 110 expert-annotated QA pairs. Notably, even state-of-the-art agents experience a drastic performance drop on UIS-QA (e.g., from 70.90 on GAIA and 46.70 on BrowseComp-zh to 24.55 on UIS-QA), underscoring the severity of the problem. To mitigate this, we propose UIS-Digger, a novel multi-agent framework that incorporates dual-mode browsing and enables simultaneous webpage searching and file parsing. With a relatively small $\sim$30B-parameter backbone LLM optimized using SFT and RFT training strategies, UIS-Digger sets a strong baseline at 27.27\%, outperforming systems integrating sophisticated LLMs such as O3 and GPT-4.1. This demonstrates the importance of proactive interaction with unindexed sources for effective and comprehensive information-seeking. Our work not only uncovers a fundamental limitation in current agent evaluation paradigms but also provides the first toolkit for advancing UIS research, defining a new and promising direction for robust information-seeking systems.

In real-world collaboration, alignment, process structure, and outcome quality do not exhibit a simple linear or one-to-one correspondence: similar alignment may accompany either rapid convergence or extensive multi-branch exploration, and lead to different results. Existing accounts often isolate these dimensions or focus on specific participant types, limiting structural accounts of collaboration. We reconceptualize collaboration through two complementary lenses. The task lens models collaboration as trajectory evolution in a structured task space, revealing patterns such as advancement, branching, and backtracking. The intent lens examines how individual intents are expressed within shared contexts and enter situated decisions. Together, these lenses clarify the structural relationships among alignment, decision-making, and trajectory structure. Rather than reducing collaboration to outcome quality or treating alignment as the sole objective, we propose a unified dynamic view of the relationships among alignment, process, and outcome, and use it to re-examine collaboration structure across Human-Human, AI-AI, and Human-AI settings.

Sharpness-Aware Minimization (SAM) enhances generalization by minimizing the maximum training loss within a predefined neighborhood around the parameters. However, its practical implementation approximates this as gradient ascent(s) followed by applying the gradient at the ascent point to update the current parameters. This practice can be justified as approximately optimizing the objective by neglecting the (full) derivative of the ascent point with respect to the current parameters. Nevertheless, a direct and intuitive understanding of why using the gradient at the ascent point to update the current parameters works superiorly is still lacking. Our work bridges this gap by proposing a novel and intuitive interpretation. We show that the gradient at the single-step ascent point, \uline{when applied to the current parameters}, provides a better approximation of the direction from the current parameters toward the maximum within the local neighborhood than the local gradient. This improved approximation thereby enables a more direct escape from the maximum within the local neighborhood. Nevertheless, our analysis further reveals two issues. First, the approximation by the gradient at the single-step ascent point is often inaccurate. Second, the approximation quality may degrade as the number of ascent steps increases. To address these limitations, we propose in this paper eXplicit Sharpness-Aware Minimization (XSAM). It tackles the first by explicitly estimating the direction of the maximum during training, while addressing the second by crafting a search space that effectively leverages the gradient information at the multi-step ascent point. XSAM features a unified formulation that applies to both single-step and multi-step settings and only incurs negligible computational overhead. Extensive experiments demonstrate the consistent superiority of XSAM against existing counterparts.

Hallucinations in large language models (LLMs) are outputs that are syntactically coherent but factually incorrect or contextually inconsistent. They are persistent obstacles in high-stakes industrial settings such as engineering design, enterprise resource planning, and IoT telemetry platforms. We present and compare five prompt engineering strategies intended to reduce the variance of model outputs and move toward repeatable, grounded results without modifying model weights or creating complex validation models. These methods include: (M1) Iterative Similarity Convergence, (M2) Decomposed Model-Agnostic Prompting, (M3) Single-Task Agent Specialization, (M4) Enhanced Data Registry, and (M5) Domain Glossary Injection. Each method is evaluated against an internal baseline using an LLM-as-Judge framework over 100 repeated runs per method (same fixed task prompt, stochastic decoding at $τ= 0.7$. Under this evaluation setup, M4 (Enhanced Data Registry) received ``Better'' verdicts in all 100 trials; M3 and M5 reached 80\% and 77\% respectively; M1 reached 75\%; and M2 was net negative at 34\% when compared to single shot prompting with a modern foundation model. We then developed enhanced version 2 (v2) implementations and assessed them on a 10-trial verification batch; M2 recovered from 34\% to 80\%, the largest gain among the four revised methods. We discuss how these strategies help overcome the non-deterministic nature of LLM results for industrial procedures, even when absolute correctness cannot be guaranteed. We provide pseudocode, verbatim prompts, and batch logs to support independent assessment.

Wearable sensors in Internet of Things (IoT) ecosystems increasingly support applications such as remote health monitoring, elderly care, and smart home automation, all of which rely on robust human activity recognition (HAR). Continual learning systems must balance plasticity (learning new tasks) with stability (retaining prior knowledge), yet AI models often exhibit catastrophic forgetting, where learning new tasks degrades performance on earlier ones. This challenge is especially acute in domain-incremental HAR, where on-device models must adapt to new subjects with distinct movement patterns while maintaining accuracy on prior subjects without transmitting sensitive data to the cloud. We propose a parameter-efficient continual learning framework based on channel-wise gated modulation of frozen pretrained representations. Our key insight is that adaptation should operate through feature selection rather than feature generation: by restricting learned transformations to diagonal scaling of existing features, we preserve the geometry of pretrained representations while enabling subject-specific modulation. We provide a theoretical analysis showing that gating implements a bounded diagonal operator that limits representational drift compared to unconstrained linear transformations. Empirically, freezing the backbone substantially reduces forgetting, and lightweight gates restore lost adaptation capacity, achieving stability and plasticity simultaneously. On PAMAP2 with 8 sequential subjects, our approach reduces forgetting from 39.7% to 16.2% and improves final accuracy from 56.7% to 77.7%, while training less than 2% of parameters. Our method matches or exceeds standard continual learning baselines without replay buffers or task-specific regularization, confirming that structured diagonal operators are effective and efficient under distribution shift.

Out-of-distribution (OOD) detection is critical in safety-sensitive applications. While this challenge has been addressed from various perspectives, the influence of training objectives on OOD behavior remains comparatively underexplored. In this paper, we present a systematic comparison of four widely used training objectives: Cross-Entropy Loss, Prototype Loss, Triplet Loss, and Average Precision (AP) Loss, spanning probabilistic, prototype-based, metric-learning, and ranking-based supervision, for OOD detection in image classification under standardized OpenOOD protocols. Across CIFAR-10/100 and ImageNet-200, we find that Cross-Entropy Loss, Prototype Loss, and AP Loss achieve comparable in-distribution accuracy, while Cross-Entropy Loss provides the most consistent near- and far-OOD performance overall; the other objectives can be competitive in specific settings.

The continuous expansion of open data platforms and research repositories has led to a fragmented dataset ecosystem, posing significant challenges for cross-source data discovery and interpretation. To address these challenges, we introduce SeDa--a unified framework for dataset discovery, semantic annotation, and multi-entity augmented navigation. SeDa integrates more than 7.6 million datasets from over 200 platforms, spanning governmental, academic, and industrial domains. The framework first performs semantic extraction and standardization to harmonize heterogeneous metadata representations. On this basis, a topic-tagging mechanism constructs an extensible tag graph that supports thematic retrieval and cross-domain association, while a provenance assurance module embedded within the annotation process continuously validates dataset sources and monitors link availability to ensure reliability and traceability. Furthermore, SeDa employs a multi-entity augmented navigation strategy that organizes datasets within a knowledge space of sites, institutions, and enterprises, enabling contextual and provenance-aware exploration beyond traditional search paradigms. Comparative experiments with popular dataset search platforms, such as ChatPD and Google Dataset Search, demonstrate that SeDa achieves superior coverage, timeliness, and traceability. Taken together, SeDa establishes a foundation for trustworthy, semantically enriched, and globally scalable dataset exploration.

Safety benchmarks evaluate language models in isolation, typically using multiple-choice format; production deployments wrap these models in agentic scaffolds that restructure inputs through reasoning traces, critic agents, and delegation pipelines. We report one of the largest controlled studies of scaffold effects on safety (N = 62,808; six frontier models, four deployment configurations), combining pre-registration, assessor blinding, equivalence testing, and specification curve analysis. Map-reduce scaffolding degrades measured safety (NNH = 14), yet two of three scaffold architectures preserve safety within practically meaningful margins. Investigating the map-reduce degradation revealed a deeper measurement problem: switching from multiple-choice to open-ended format on identical items shifts safety scores by 5-20 percentage points, larger than any scaffold effect. Within-format scaffold comparisons are consistent with practical equivalence under our pre-registered +/-2 pp TOST margin, isolating evaluation format rather than scaffold architecture as the operative variable. Model x scaffold interactions span 35 pp in opposing directions (one model degrades by -16.8 pp on sycophancy under map-reduce while another improves by +18.8 pp on the same benchmark), ruling out universal claims about scaffold safety. A generalisability analysis yields G = 0.000: model safety rankings reverse so completely across benchmarks that no composite safety index achieves non-zero reliability, making per-model, per-configuration testing a necessary minimum standard. We release all code, data, and prompts as ScaffoldSafety.

Multimodal dialogue emotion recognition captures emotional cues by fusing text, visual, and audio modalities. However, existing approaches still suffer from notable limitations in modeling emotional dependencies and learning multimodal representations. On the one hand, they are unable to effectively filter out redundant or noisy signals within multimodal features, which hinders the accurate capture of the dynamic evolution of emotional states across and within speakers. On the other hand, during multimodal feature learning, dominant modalities tend to overwhelm the fusion process, thereby suppressing the complementary contributions of non-dominant modalities such as speech and vision, ultimately constraining the overall recognition performance. To address these challenges, we propose an Adaptive Modality-Balanced Dynamic Semantic Graph Differential Network (AMB-DSGDN). Concretely, we first construct modality-specific subgraphs for text, speech, and vision, where each modality contains intra-speaker and inter-speaker graphs to capture both self-continuity and cross-speaker emotional dependencies. On top of these subgraphs, we introduce a differential graph attention mechanism, which computes the discrepancy between two sets of attention maps. By explicitly contrasting these attention distributions, the mechanism cancels out shared noise patterns while retaining modality-specific and context-relevant signals, thereby yielding purer and more discriminative emotional representations. In addition, we design an adaptive modality balancing mechanism, which estimates a dropout probability for each modality according to its relative contribution in emotion modeling.

Targeted bit-flip attacks (BFAs) exploit hardware faults to manipulate model parameters, posing a significant security threat. While prior work targets single-step inference models (e.g., image classifiers), LLM-based agents with multi-stage pipelines and external tools present new attack surfaces, which remain unexplored. This work introduces Flip-Agent, the first targeted BFA framework for LLM-based agents, manipulating both final outputs and tool invocations. Our experiments show that Flip-Agent significantly outperforms existing targeted BFAs on real-world agent tasks, revealing a critical vulnerability in LLM-based agent systems.

Large-scale electrophysiological recordings now allow simultaneous monitoring of thousands of neurons across multiple brain regions, revealing structured variability in neural population activity. Understanding how these collective patterns emerge from microscopic neural interactions requires models that are scalable, predictive, and interpretable. Statistical physics provides principled frameworks to address this complexity, including maximum-entropy models that offer transparent descriptions of collective neural activity but remain largely limited to pairwise interactions and modest system sizes. Here, we use Restricted Boltzmann Machines (RBMs) to model the activity of $\sim1500$-$2000$ simultaneously recorded neurons from the Allen Institute Visual Behavior Neuropixels dataset, spanning multiple cortical and subcortical regions of the mouse brain. RBMs extend the maximum-entropy framework through latent variables, enabling the capture of higher-order dependencies while allowing explicit extraction of effective interaction networks. Recent advances in efficient Markov Chain sampling and training enable accurate learning of these models at this scale. RBMs reproduce the complex statistics of neural recordings with high accuracy. Generated samples match empirical pairwise and higher-order correlations, as well as global statistics such as the distribution of population activity. The inferred parameters provide direct access to effective neuronal interactions, revealing coordination patterns in population activity. These couplings display clear anatomical structure: neurons within visual cortical areas show stronger interactions, consistent with visually driven behavior, while cross-area couplings are weaker. Despite being trained on temporally shuffled data, Markov Chain Monte Carlo simulations also reproduce the global relaxation dynamics of neural activity.

Phototaxis of many species of green algae relies upon directional sensitivity of their membrane-bound photoreceptors, which arises from the presence of a pigmented "eyespot" behind them that blocks light passing through the cell body from reaching the photoreceptor. A decade ago it was discovered that the spherical cell body of the alga $Chlamydomonas~reinhardtii$ acts as a lens to concentrate incoming light, and that in "eyeless" mutants of $Chlamydomonas$ the consequence of that focused light reaching the photoreceptor from behind is a reversal in the sign of phototaxis relative to the wild type behavior. We present a quantitative theory of this sign reversal by completing a recent simplified analysis of lensing [Yang, et al., Phys. Rev. E 113, 022401 (2026)] and incorporating it into an adaptive model for $Chlamydomonas$ phototaxis. This model shows that phototactic dynamics in the presence of lensing is subtle because of the existence of internal light caustics when the cellular index of refraction exceeds that of water. During each period of cellular rotation about its body-fixed axis, the photoreceptor receives two competing signals: a relatively long, slowly-varying signal from the direct illumination, and a stronger, shorter, rapidly-varying lensed signal. The reversal of the sign of phototaxis is then a consequence of the dominance of the flagellar photoresponse to the signal with the higher time derivative. These features lead to a quantitative understanding of phototaxis sign reversal, including bistability in the direction choice, a prediction that can be tested in single-cell tracking studies of mutant phototaxis.

We present explicit counterexamples to a widespread empirical expectation that local commuting charges display all-or-nothing behavior. In the class of bosonic chains with symmetric nearest-neighbor hopping and arbitrary on-site terms (including non-Hermitian terms), we exhibit systems that possess k-local charges for some but not all k. Concretely, we construct non-Hermitian models with a 3-local charge but no other nontrivial local charges and models with k-local charges for all k except k = 4. These results show that the Grabowski--Mathieu integrability test based on 3-local charges is not universally applicable. We further give necessary and sufficient conditions for the existence of k-local charges in this class, yielding an exhaustive classification and uncovering additional integrable models.

We present a computational approach for exciton calculations in two-dimensional (2D) materials within the Bethe-Salpeter equation (BSE) framework, employing an atomistic description with point-like orbitals. Unlike widespread efficient calculations that rely on classical or effective interaction models, such as the Rytova-Keldysh model, our method incorporates quantum screened interactions. By explicitly computing the 2D dielectric function at the random-phase approximation level, we capture screening effects beyond such approximations with an accuracy akin to first-principles methods. Consequently, we can realistically estimate excitonic binding energies with a bearable computational cost. A detailed account of the various convergence parameters sheds light on a possible cause of the large dispersion of binding energies reported in the literature using first-principles GW/BSE implementations. This work thus provides an alternative pathway towards efficient and faithful dielectric screening and exciton computations in low-dimensional materials.

Brains and artificial neural networks compute with continuous variables such as object position or stimulus orientation. However, the complex variability in neural responses makes it difficult to link internal representational structure to task performance. We develop a statistical-mechanical theory of regression capacity that relates linear decoding efficiency of continuous variables to geometric properties of neural manifolds. Our theory handles complex neural variability and applies to real data, revealing increasing capacity for decoding object position and size along the monkey visual stream.

In quantum Monte Carlo (QMC), what can be measured efficiently is largely determined by what is sampled. When the sampled object is the partition function, a broad class of observables, including general off-diagonal operators, is typically unavailable as direct estimators. In this article, we introduce a paradigm shift by replacing the partition function with a generalized reduced density matrix (GRDM) as the simulated object. This reformulation removes the measurement bottleneck at its source and extends the dimensional-reduction advantage of reduced descriptions from static quantities to dynamical observables, thereby enabling much richer information extraction. As substantial demonstrations, the framework allows the directed-loop algorithm to measure both equal-time and imaginary-time off-diagonal observables, with the latter giving direct access to dynamical spectra. It also enables measurements of Rényi-1 correlators that diagnose strong-to-weak symmetry breaking in mixed states. This work establishes a unified framework for holographic characterization within QMC.

Understanding fluid flow in wedge-shaped geometries is essential for predicting hydrodynamic interactions in confined systems, such as microfluidic devices and near-corner transport phenomena. This article reviews analytical methods and techniques for addressing wedge problems in low-Reynolds-number hydrodynamics, focusing on solutions of the Stokes equations for a point force (Stokeslet) and a point torque (rotlet). The formulation is based on the Papkovich-Neuber representation, which uses four harmonic functions to characterize the fluid flow. A concise overview of the Fourier-Kontorovich-Lebedev (FKL) transform method is provided, highlighting key properties and steps employed in deriving these solutions. This offers a versatile framework for predicting particle dynamics in wedge confinements and for designing microfluidic systems with corner geometries.

We study the correlated ground states of twisted mono-bilayer graphene with and without proximity-induced spin-orbit coupling (SOC) from a transition-metal dichalcogenide layer placed on top. We perform self-consistent Hartree-Fock calculations that allow the variational space to include multi-$Q$ translational symmetry broken states for all integer and half-integer fillings of the conduction bands, where signatures of correlated, topological states have been reported experimentally. We find interaction-induced insulators that retain moiré translational symmetry at integer fillings, but that break this symmetry at half-integer fillings. We argue that translational symmetry breaking arises from half-filled polarized bands, even when SOC is present. Yet, we find that small SOC can already crucially affect the spin nature of correlated states. Generally, Ising SOC favors out-of-plane spin polarization and spin-valley locking, while Rashba SOC favors in-plane spin order. If only one of these two terms is present, we find that, depending on the type of SOC, it drives a transition from a tetrahedal antiferromagnet to either a coplanar, non-coplanar, or collinear spin-density wave state for half-integer fillings. The frustration associated with the simultaneous presence of both types of SOC can induce chiral, non-coplanar order in parameter ranges where the ground state in the absence of SOC is collinear.

Quantum batteries are quantum systems that store energy and deliver it on demand, and their practical value hinges on how fast they can be charged. While collective charging protocols and global control are known to enhance charging power, it remains unclear how the battery's internal interaction architecture itself constrains performance. Here we study interacting fermionic batteries whose internal couplings are encoded by a graph adjacency matrix, charged via a simple interaction with an external fermionic device. We prove that the star topology maximises the early time charging power, which proxies the maximal average power - a widely used quantum battery quality metric. We substantiate the result numerically by an exhaustive sweep over all graphs with $N\leq 7$ vertices and by benchmarks against random graph ensembles at larger $N$. Our findings shed light on architecture as a controllable knob for fast charging and motivate hub-and-spoke designs in scalable quantum-battery platforms.

The Lindblad equation determines the time evolution of the density operator of open quantum systems. While valid for any system size, its use is, in practice, restricted to prototype/surrogate models with the aim of tackling specific aspects of the overall quantum complexity of a multi-atomic system. Path integral molecular dynamics (PIMD) instead provides static and dynamical quantum statistical averages of physical observables for systems in equilibrium composed of up to thousands of atoms over timescales up to nanoseconds, under the condition that short-time quantum coherence is not relevant for the properties of interest. PIMD relies on the well-established technique of molecular dynamics (MD) with its associated classical trajectories. However, it cannot describe a direct time evolution of a system and its convergence to a stationary state in situations out of equilibrium. In this work, we analyze the link between the Lindblad equation and PIMD; specifically, we will discuss how PIMD can actually be used to calculate the time evolution of ensemble-averaged physical observables and their convergence to a stationary state for situations out of equilibrium, bypassing the need of explicitly solving the Lindblad equation. Yet, at the same time, the Lindblad equation and PIMD are linked to one another through a formal relation of equivalence, which provides an argument for the consistency of PIMD results, namely the positivity of the density operator at any time. A numerical study of a prototype system, which is of interest in chemical physics, will be used to showcase the method.

MnBi$_2$Te$_4$ and MnBi$_4$Te$_7$ are antiferromagnetic topological insulators belonging to the MnBi$_{2n}$Te$_{3n+1}$ series, where structural layering provides a natural route to tune magnetic interaction in van der Waals magnets. Despite extensive interest in their topological properties, how the insertion of Bi$_2$Te$_3$ quintuple layers modifies magnetic critical fluctuations near the antiferromagnetic transition remains unresolved. Here, we combine scanning tunneling microscopy (STM), critical scaling analysis, and magnetocaloric measurements to directly correlate real-space structures with magnetic criticality. STM reveals atomically flat septuple-layer terraces in MnBi$_2$Te$_4$ whereas MnBi$_4$Te$_7$ displays coexisting septuple and quintuple layer terminations reflecting its alternating stacking sequence. MnBi$_2$Te$_4$ exhibits robust three-dimensional Ising-like critical behavior together with a distinct low-temperature first-order transition. In contrast, MnBi$_4$Te$_7$ displays crossover-dominated criticality arising from weakened interlayer exchange and competing magnetic phases. Correspondingly, the magnetocaloric response differs significantly between the two compounds. MnBi$_2$Te$_4$ shows dual-type magnetocaloric behavior with a sharp field-induced sign reversal of the isothermal magnetic entropy change ($-ΔS_M$). It exhibits both inverse ($-ΔS_M < 0$) and conventional ($-ΔS_M > 0$) magnetocaloric effects. In contrast, MnBi$_4$Te$_7$ shows only conventional magnetocaloric response with a broad positive entropy peak. These results establish structural layering as a key parameter governing magnetic critical fluctuations and magnetocaloric behavior in MnBi$_{2n}$Te$_{3n+1}$ topological magnets.

In interacting quantum many-body systems, relaxation toward equilibrium reflects a competition between internal chaotic dynamics and environmental dissipation. While conventional Markovian baths typically produce exponential decay, non-Markovian dissipation can give rise to more intricate behavior, including algebraic relaxation. We study an open Sachdev-Ye-Kitaev (SYK) model coupled to a pseudogapped fermionic bath, using the Keldysh formalism to compute steady-state correlations in the large-$N$ limit. Our results uncover a rich dynamical phase diagram, with regimes of bath-driven power-law relaxation, chaos-driven exponential decay, and an intermediate pre-relaxation phase where exponential decay crosses over to algebraic decay. These findings demonstrate that non-Markovian environments can qualitatively reshape relaxation mechanisms in strongly correlated quantum many-body systems.

Atomic force nanolithography provides a precise method for sculpting magnetic thin films, enabling controlled engineering of magnetic anisotropy in soft ferromagnets at the microscale. We demonstrate that nanoscale groove arrays patterned into permalloy (Ni80Fe20) films induce a robust in-plane uniaxial anisotropy, with the easy axis aligned along the groove direction. The anisotropy field is shown to increase with decreasing groove period and increasing engraving depth, offering continuous tunability of magnetic hardness within a single fabrication step. Artificially engraved microstructures further allow domain configurations and domain-wall trajectories to be directed along predefined pathways, exemplified by the creation of a chessboard-like magnetic landscape. Owing to its adaptability to diverse ferromagnetic materials and arbitrary corrugation geometries, this approach provides a versatile platform for tailoring in-plane magnetic anisotropy. Concrete applications are demonstrated in the design of magnonic elements and anisotropic magnetoresistance sensors.

We study purification dynamics in monitored quantum processes governed by ensembles of quantum circuits in different random-matrix symmetry classes. We analyze the universal aspects that emerge away from the measurement induced phase transition and inside the volume/weak measurement phase and in the scaling limit of large time and Hilbert space dimension. We present two toy models that reveal two complementary visions and provide quantitative access to universal scaling: i) a discrete-time dynamic in which each time step corresponds to multiplication by a Gaussian random matrix; ii) weak continuous-time monitoring that induces a Dyson brownian motion of the eigenvalues of the density matrix. The first approach provides an algebraic characterization based on rotational invariance emerging in Kraus's operator space, focusing in particular on the unitary and orthogonal cases, respectively $β=2$ and $β=1$, with $β$ the Dyson random-matrix index. The second approach, on the other hand, allows for a unified treatment for any $β$, thanks to the mapping of the Fokker-Planck evolution of eigenvalues onto the Calogero-Sutherland integrable Hamiltonian diagonalized in terms of Jack polynomials. We provide explicit expressions for the universal decrease of Rényi entropies. We show that, approaching the universal scaling limit, numerical simulations of different models agree with each other and with our theoretical predictions. Our results clarify the existence of different classes of universality for the purification process in hybrid quantum systems, accessible in random circuit architectures and weak measurement protocols.

In this letter, we present fast readout of Pauli spin blockade phenomena and interdot coupling tunability in a silicon double quantum dot (DQD) fabricated using industry-compatible processes. The interdot couplings are tuned with a second self-aligned gate layer. The charge sensing and spin readout are performed by using gate-based reflectometry techniques. The results pave the way for scalable fast readout of large-scale industry-standard manufactured Si spin qubit arrays.

Quantum many-body systems can exhibit distinct regimes where dynamics is either ergodic, dynamically exploring an extensive region of available state-space, or non-ergodic, where the dynamics may be restricted. An example is the many-body localization (MBL) transition, where disorder induces non-ergodic behaviour. Most measures of ergodicity notably rely on global quantities, such as level spacing statistics. We explore the ability for a subsystem to probe the ergodicity of dynamics via measurement of local observables, and use expected results from random matrix theory (RMT) as a benchmark for the ergodic regime. We exploit two predictions from RMT as ergodicity is broken: the time evolution of the quantum Fisher information, and a fluctuation-dissipation relation. These are investigated in three different ergodicity breaking mechanisms, namely, as a consequence of transition to integrability, MBL, and Quantum Many-Body Scars (QMBS). We show that the predicted behaviour from RMT can be used as a potential witness for transition to non-ergodic behaviour from the measurement of local observables alone.

This paper investigates stochastic solenoidal magnetohydrodynamics within the field-theoretic Martin-Siggia-Rose-De Dominicis-Janssen formalism, with a specific focus on the stability of the system when spatial mirror (parity) symmetry is explicitly broken. Under helical forcing, the one-particle-irreducible magnetic response function already at one loop contains a curl-type contribution that dominates the bare resistive term in the infrared limit, leading to exponential instability of the trivial state $\langle \mathbf{b} \rangle = \mathbf{0}$. We re-examine a stabilization mechanism proposed in [L. T. Adzhemyan, et al., Theor. Math. Phys. 72, 940-950 (1987)], in which the system evolves into a phase with a dynamically spontaneously broken rotational symmetry and a generated mean magnetic field $\langle \mathbf{b} \rangle = \mathbf{B}_0$. By deriving a self-consistency condition for $ \mathbf{B}_0$, we show that for any physically admissible (infrared) form of the pumping function, the model admits only a singular solution. We illustrate this with the standard power-law and "massive" pumping functions. We further show that previous claims of a finite $ \mathbf{B}_0$ arose from an inconsistent truncation of asymptotic expansions. We argue that a consistent physical resolution requires including a bare curl term in the stochastic induction equation, which naturally arises from a parity-violating modification of Ohm's law. With this modification, stabilization of the system by spontaneous symmetry breaking becomes a viable field-theoretic description of large-scale mean-field generation (turbulent dynamo) in helical turbulent magnetohydrodynamics.

We investigate a gauge transformation based on the successive elimination of time-dependent onsite potentials at individual sites in finite or infinite systems. Our analysis shows that this transformation renormalizes the inward hoppings by a phase factor $e^{i φ(t)}$ and the outward hoppings by $e^{-i φ(t)}$. We further demonstrate how this procedure facilitates the reduction and simulation of pulse propagation in scattering systems, while significantly simplifying the time-ordered integrals involved in the time evolution operator for time-dependent Schrodinger equation.

Quantum fluctuation relations provide a microscopic formulation of thermodynamics beyond equilibrium, but experimentally accessing many-body quantum work statistics remains an outstanding challenge. The quantum piston constitutes a canonical model of boundary-driven nonequilibrium dynamics, where finite-time deformation of a confining potential generates non-adiabatic transitions, dissipation and irreversibility. Here we experimentally simulate the nonequilibrium dynamics of a two-boson quantum piston on a programmable photonic quantum computer. Using two indistinguishable photons, we encode a truncated piston propagator through a quasi-unitary embedding, with an ancilla mode representing leakage into higher-energy states outside the resolved manifold. This architecture enables direct reconstruction of thermodynamic transition statistics for both expansion and compression protocols as functions of driving speed and final trap length. We observe the crossover from quasi-adiabatic to strongly non-adiabatic evolution and show that bosonic interference restructures the resulting two-particle Fock-state populations and work distributions. The measured statistics are in close agreement with theoretical predictions and satisfy the Jarzynski equality across expansion and compression protocols for cyclic driving we further quantify irreversibility through dissipated work and state overlap. Our work identifies programmable photonic quantum hardware as a powerful platform for simulating nonequilibrium quantum thermodynamics and for experimentally resolving how indistinguishability and many-body interference shape quantum work, dissipation and entropy production.

Nonlinear responses in transport phenomena have attracted significant attention because they can arise even when linear responses are forbidden by symmetry, with the quantum geometry of Bloch wave functions playing an essential role. While such effects have been extensively studied in electric transport, similar quantum-geometric mechanisms are also expected to govern nonlinear thermal transport. In particular, thermal responses are crucial in bosonic systems such as magnons and phonons, which are charge-neutral quasiparticles. However, a consistent theoretical description of nonlinear thermal transport remains challenging because of the difficulty in the treatment of energy magnetization in higher-order responses with Luttinger's gravitational potential method. Here, we formulate the intrinsic nonlinear thermal conductivity of bosonic systems using a quantum kinetic equation approach that avoids Luttinger's method and naturally incorporates contributions from energy magnetization. We identify three distinct contributions to the nonlinear thermal conductivity: two expressed in terms of quantum-geometric quantities, namely the quantum metric and the thermal Berry-connection polarizability (TBCP), and a third determined solely by the band dispersions. Applying our formalism to a specific quantum spin model within linear spin-wave theory, we show that the TBCP term dominates the nonlinear thermal Hall effect in the absence of threefold symmetry. Our results differ quantitatively from those obtained using semiclassical theory, thereby highlighting the importance of quantum corrections beyond the semiclassical picture. These findings establish a general framework for intrinsic nonlinear thermal responses in bosonic systems and reveal quantum-geometric mechanisms underlying thermal transport beyond linear response theory.

We investigate the collective behavior of motile rods immersed in a monolayer of apolar rods confined between vertically vibrating plates using numerical simulations. We uncover an antidiffusive instability whereby motile rods segregate from the apolar medium and form flocks whose size increases with the medium concentration. Remarkably, enhanced segregation leads to a reduction of the global polar order. The flock structure is strongly influenced by the anisotropy of the medium rods. For small aspect ratios, the flocks are elongated perpendicular to the mean direction of motion, whereas for larger aspect ratios, they elongate along the direction of motility. We rationalize the emergence of segregation-induced disorder using a minimal mean-field model.

We develop a parabolic-cylinder approach to valley-polarized conductance in tilted anisotropic Dirac-Weyl systems, showing that the smooth-interface scattering problem can be reduced analytically to the Weber equation, which belongs to the same differential-equation class as the quantum harmonic oscillator. This reduction yields closed-form expressions for the angular transmission envelope and clarifies the distinct roles of the tilt components: the perpendicular tilt renormalizes the tunneling-envelope width, while the parallel tilt shifts the Fabry-Perot resonance structure differently in opposite valleys. Combined with the nonlinear mapping between the fixed device frame and the rotated barrier frame, this analytical structure provides a direct route from valley-dependent interface tunneling to net valley-polarized conductance. We apply the formalism to rotated electrostatic barriers and construct phase diagrams over barrier angle, tilt strength, width, height, and Fermi energy. The results reveal a robust optimum near t = 0.2 over the parameter range studied, identify the crossover from oscillatory to monotonic polarization regimes, and delineate practical operating windows for candidate materials including 8-Pmmn borophene and WTe2.

We theoretically investigate how information flows when two particles interact with each other. Understanding the physical mechanisms of directional information flow is crucial for advancing information thermodynamics and stochastic computing. However, the fundamental connection between mechanical motion and causal information transfer remains elusive. To focus only on essential effects of physical dynamics, we examine two interacting Brownian particles confined in a one-dimensional potential. By simulating their Langevin dynamics, we quantify the causal information exchange using transfer entropy. We demonstrate that a mass asymmetry inherently breaks the symmetry of information flow, inducing a net directional transfer from the heavier to the lighter particle. Physically, the heavier particle, possessing larger inertia and higher active information storage, retains the memory of its trajectory longer against thermal fluctuations, thereby acting as a source of information. We analytically clarify that this net transfer is governed by a competition between the difference in memory capacity and the predictability of the particle trajectories. Furthermore, we reveal that the net information flow scales logarithmically with the mass ratio. These findings provide essential insights into the physical significance of transfer entropy and the nature of information flow in general physical systems.

Although two-dimensional periodic structures have functioned as the primary platform for exploring topological phenomena, recent advances have substantially expanded this research boundary to include more intricate, aperiodic structures: quasicrystals, fractals, non-Euclidean lattices, and disordered materials. A network-based perspective not only offers a unified framework for classifying these diverse platforms based on their network connectivity but also unveils unexplored regimes of topological phenomena in complex networks. Here, we implement topological heavy-tailed networks, as an example of high-degree complex networks exhibiting topological phases. By developing a tight-binding model for the Apollonian network and a deterministic algorithm to assign nontrivial gauge fields to this aperiodic geometry, we compute the magnetic-flux-dependent energy spectrum: the Apollonian butterfly. Using spectral localizers, we characterize the topological features of the Apollonian butterfly, whose sensitivity is governed by lower-degree nodes, analogous to the controllability of complex networks. Our framework bridges topological physics and network science, introducing a connectivity-driven paradigm for the control of topological waves.

When single-shot qubit readout protocols are adapted for multilevel systems, theoretical $T_1$ lifetime calculations often fall short of capturing the experimental lifetime trends. We identify {\it extrinsic} population dynamics as the fundamental origin of this disparity, establishing that the lifetime estimates can, in certain operating regions, be distinct from the intrinsic $T_1$ time. We clarify these aspects with an integrated theory to address recent measurements [Nat. Nano, 20, 494, (2025)] on spin-valley states in bilayer graphene. While confirming that phonon and Johnson noise are the dominant intrinsic sources, we show that the inclusion of extrinsic factors provide the critical match to the experimental estimates. The extrinsic factors also effectuate violations of generalized Mathiessen's rules also. With an improved handle on the design space, a revised readout protocol to estimate the $T_1$ lifetime of the valley qubit is proposed.

Symmetry and topology are fundamental concepts deeply intertwined in various fields of physics, especially in the studies of quantum phases of matter. The critical role that Coulomb interactions play in symmetry breaking during topological transitions is a fundamental problem that has not been fully understood. Utilizing gated indium arsenide-gallium antimonide bilayers, we demonstrate that Coulomb interactions play a critical role in symmetry breaking and topological transitions. Whereas the quantum spin Hall insulator (QSHI) dominates the high-density regime, gating the system into the dilute regime enhances interlayer Coulomb interactions and leads to an emergent excitonic topological order (ETO) with spontaneous time-reversal-symmetry breaking. Moreover, applying a magnetic field drives a transition from the QSHI to the ETO accompanied by Coulomb-induced spin-rotation-symmetry breaking, which selects triplet electron-hole pairing in the lowest Landau levels. These results underscore an intricate interplay between symmetry and topology under Coulomb interactions in electron-hole bilayers.

Ferromagnetic domain walls -transitional regions between magnetic domains- are an essential ingredient for racetrack memory, a device concept that promises to deliver faster and more compact memory storage compared to other non-volatile memory devices. Motivated by recent experiments that have found inertial effects in spin dynamics, we explore its consequences on domain wall motion. We find that the inertial dynamics of the individual magnetic moments induce massive dynamics of the domain wall. We investigate these massive dynamics driven by a magnetic field, spin-transfer torque, and spin-orbit torque. We show that, in the absence of Gilbert damping, the domain wall dynamics become chaotic, resembling that of an electron in a two-dimensional crystal. For finite damping, field-like driving of the inertial domain wall significantly increases its velocity compared to conventional massless dynamics, potentially enabling faster racetrack operations. Additionally, in the limit of low driving, we observe that the domain wall width contracts due to the spin inertia of the ferromagnet.

We present a general theoretical framework to capture light-matter interactions beyond the electric-dipole approximation (EDA), applicable to extended nano- and microscale materials interacting with spatially structured electric fields without truncation at finite multipolar order. The approach is based on the Power-Zienau-Woolley (PZW) Hamiltonian for light-matter interactions and a representation of the material's Hamiltonian in a basis of maximally localized Wannier functions (MLWFs), obtainable from first-principles calculations. We utilize this approach to clarify the limitations of the ubiquitous dipole approximation. We consider electric fields with both uniform and non-uniform intensities and a range of ratios of system size to the wavelength of light. Through this analysis, we identify the conditions under which the EDA breaks down, leading to significant errors in the light-induced dynamics. Contrary to conventional belief, we find that the EDA is remarkably robust for uniformly illuminated 1-D or 2-D materials when light propagates perpendicular to the material. For 3-D materials or non-perpendicular illumination of lower-dimensional materials, conventional wisdom holds and the EDA begins to break down when the wavelength becomes comparable to the system size. Furthermore, the EDA fails when the material is illuminated partially or non-uniformly. For slowly varying field intensities this failure can be corrected by finite-order multipolar corrections. However, for fields that vary substantially, correcting via multipolar terms becomes computationally impractical. In contrast, our approach captures beyond-dipole light-matter interactions at the computational cost of a standard dipole calculation. This efficiency enables accurate first-principles simulations of spatially structured light-matter dynamics in nanoscale devices, quantum materials, and interfaces.

Curvature fundamentally alters the collective properties of soft, active, and biological materials. Here we study motility-induced phase separation (MIPS), a canonical non-equilibrium transition, and demonstrate that even weak and slowly varying curvature provides robust geometric control over the dense MIPS phase. This includes dictating both the location and morphology of the MIPS cluster, even in regimes where the effect on the overall phase boundaries is minimal. Focusing on active Brownian particles confined to the surface of a torus, we show that varying the aspect ratio drives a structural transition of the dense cluster from a disk localized at the outer equator to a band wrapping the minor circumference. We then discuss how the curved geometry provides a platform for comparing different theoretical frameworks for the MIPS phase: by analyzing the geometries of the cluster boundaries, we compare the structures predicted by thermodynamic and kinetic pictures. Our results establish curved space not only as a tool to shape and guide non-equilibrium dynamics, but as a uniquely sensitive arena for probing the fundamental mechanisms of active matter.

We develop a mathematical model to describe the persistence of rule-breaking behaviors in societies, such as traffic violations, disregard for legal restrictions and other forms of noncompliance. Using a replicator-type dynamics with utility functions incorporating individual benefits, institutional punishment and social sanctions, we first built a general formulation of the system. Within this framework, we analyze two distinct models differing in the nature of social feedback. In the presence of positive feedback, the system exhibits bistability, with widespread compliance and widespread violation as stable equilibria, and the transition between these states occurs discontinuously once a critical threshold is crossed, resembling a first-order phase transition. By contrast, when negative feedback is present, the population undergoes a continuous phase transition between compliant and noncompliant collective states, driven by an increasing collective cost of rule-breaking. Numerical simulations and analytical results illustrate how changes in enforcement, social tolerance or perceived benefits can shift the system across tipping points. The results provide a theoretical explanation for the fragility of social order under weak institutions and highlight possible pathways to promote compliance.

Superconducting diodes, devices that allow Cooper-pair currents to flow more easily in one direction than the other, are set to become key building blocks for dissipationless electronics. Existing realizations, however, rely on magnetic fields, ferromagnets, or complex heterostructures that hinder integration and scalability. Here we demonstrate a diode effect for Cooper-pairs that arises solely from electron-electron interactions in nanoscale superconducting lead islands. When these islands are driven into the Coulomb blockade regime, Cooper-pair transport occurs through resonant charge states. By tuning the island's electrostatic environment, we controllably break particle-hole symmetry and induce nonreciprocal supercurrents, thereby achieving a gate-switchable superconducting diode without any external magnetic field. Our approach enables robust rectification of superconducting currents and microwave photoresponse, providing a scalable strategy to superconducting logic devices.

The modeling of realistic magnetic materials requires the inclusion of defects. Based on the pseudospectral Landau-Lifshitz description of magnetisation dynamics, we propose a statistical model that takes into account defects, specifically vacancies. This statistical model can be integrated with deep learning techniques that correlate defect thresholds with relevant physical observables. We develop a convolutional neural network and a physics-informed neural network combined with theory of functional connections to predict the dispersion relation given defect parameters and physical constraints. A two-branch convolutional neural network is developed to predict domain-wall widths depending on defects threshold, taking into account the spatial profile and domain-wall width separately to achieve a prediction. The proposed physics-informed approaches leverage deep-learning and achieve statistical predictions measured in physical units. This is a stepping stone towards the discovery of new materials and the determination of minimal defect thresholds required for desired dynamics, states, or topological textures.

We show that there are two distinct mechanisms that can cause the symmetry-restoration Mpemba effect in spin chains with \textit{weakly broken} integrability, such that the asymptotic equilibration is diffusive, but the lifetime of anomalously fast spin hydrodynamics at low temperature is parametrically large. In particular, we consider isotropic spin chains quenched out of equilibrium by suppressing the $z$-components, without inducing any net magnetisation. Initially, the restoration of isotropy is faster in hotter systems -- because they have more phase space available to scramble their initial conditions -- which may cause the equilibration curves to cross at early times in both integrable and non-integrable systems. At later times, however, the equilibration is effectively hydrodynamic, and the \textit{colder} systems start to equilibrate faster as the lifetime over which they evince superdiffusive spin hydrodynamics is parametrically larger -- but only in \textit{non}-integrable models. Depending on the details of the temperatures and the extent of the initial symmetry-breaking, two isotropy-restoration curves may have a crossing at early time, late time, neither, or both.

We study fixed-length bridge paths -- half-space excursions that start and end at a planar boundary -- for three-dimensional random walks with Henyey-Greenstein scattering angles and exponentially distributed step lengths, using Monte Carlo simulation over asymmetry parameter g from 0 to 0.95 and path lengths from 4 to 200 steps. The key structural feature is that the walk evolves on a two-dimensional Markovian state space (depth, direction cosine) rather than the scalar depth coordinate alone. Four anomalies with respect to classical Brownian-excursion theory are reported. The mean amplitude scales super-diffusively, as path length to a power of 0.57--0.58 for isotropic scattering, nine standard deviations above the Brownian prediction of 0.5, with no sign of convergence out to 200 steps. The diffusion coefficient scales as the transport mean free path to the power 0.415 rather than the predicted 1.0. The midpoint depth distribution is Rayleigh rather than half-normal, consistent with a two-dimensional Bessel process. The bridge-conditioned mean direction cosine converges to minus two-thirds at the final step, independently of the asymmetry parameter and initial direction -- the classical Milne result anchored by the H-function moment identity. All anomalies are attributed to the two-dimensional state-space structure. The two anomalous exponents sum to approximately unity, suggesting a common geometric origin. Whether this constitutes a permanent universality-class shift or an anomalously slow crossover to Brownian-excursion behaviour remains the primary open question.

Employing the Kubo linear response formalism to calculate the elasticity of anisotropic systems has been shown to yield odd elastic moduli. For Hamiltonian systems, this result seems to be contradictory as it would violate energy conservation. To resolve this discrepancy, we examine the predictions of quantum linear response in the context of our expectation from classical elasticity theory. Our framework reveals that the geometry of the space of strain perturbations introduces correction factors to the correspondence between the Kubo formula and the elastic moduli which resolves the contradiction. We use a two-dimensional gas of electrons in a magnetic field as a pedagogical example. We use generalized f-sum rules to demonstrate how contact terms may reveal themselves in experimental measurements. Finally, we discuss the implications of our results for interpreting more general linear response functions.

Incorporating oxygen into metallic atomic-scale junctions modifies the interatomic bonding and may even promote the formation of monoatomic chains. In the specific case of copper oxide, first-principles studies have predicted the emergence of ferromagnetic ground states, attributing certain atomic configurations with spin filtering capabilities. By means of low-temperature transport measurements, we provide a series of experimental evidence indicating the presence of local magnetism in air oxidized mechanically controllable copper break junctions. Our investigations on ultimately small contacts range from magnetotransport measurements to the analysis of anomalous shot noise in the presence of strong zero-bias anomalies. The analysis of the latter allows to determine the spin polarization (SP) of the current and that is interpreted with the Kondo physics picture.

We present a geometric formulation of optical, thermoelectric, and thermal linear response in clean, zero temperature band insulators based on a single object: a generalized time-dependent quantum geometric tensor (g-tQGT) built from correlations of projected particle and heat polarization operators. Within this framework, the AC transport tensors admit compact expressions that make their geometric content explicit. The response splits into a Berry curvature contribution that remains finite in the DC limit and a frequency correction governed by the quantum metric, implying geometry driven effects even in topologically trivial insulators. At equal times, the g-tQGT recovers the usual integrated QGT and yields energy-weighted thermal analogs whose antisymmetric parts are fixed by orbital and heat magnetization. Importantly, in the thermal channel, a thermal quantum geometric tensor is obtained. Casting the theory in a Hilbert-Schmidt inner product form yields a bound on the trace of the thermal QGT, an uncertainty relation on the projected polarization operators and a purely geometric upper bound on the finite-time accumulated response. The latter is used in the optical channel to derive a geometric upper bound on the electric current. Finally, time derivatives of the g-tQGT are used to generate a hierarchy of generalized thermoelectric and thermal sum rules, and bounds on these sum rules are obtained. These bounds are used to find inequalities between different physical objects such as the optical mass, susceptibility functions and magnetizations.

Motivated by experiments on spin-orbit coupled magnets with Kitaev exchange in magnetic fields, we present an analysis of the classical Kitaev honeycomb model in the presence of a magnetic field. We show that there is a spin liquid regime that exists within a finite window of fields from zero up to a finite threshold before transitioning into the polarized paramagnet. We uncover constraints that spins need to satisfy in the ground state and show that they determine the exact limiting zero temperature behavior of the heat capacity and magnetic susceptibility within the spin liquid as a function of field. When the field is finite, both the two-point spin and the quadrupolar correlations are short-ranged, in contrast to the zero-field case. We rationalize an effective mass for the quadrupolar correlations in terms of a coarse-grained theory with fluctuating effective charge degrees of freedom. Finally, we show that weak site-dilution does not change the magnetization within the spin liquid -- a kind of "perfect" compensation of the site dilution.

Motivated by the recent twisted MoTe$_2$ experiment [arXiv:2601.18508], we develop a disordered interacting edge theory of a fractional topological insulator at $ν_{\text{tot}}=4/3$, consisting of two time-reversal-conjugated $ν=2/3$ fractional quantum Hall states. For an $S_z$-conserving edge, we uncover three distinct phases with two possible conductance values per edge in the long-edge limit: $\frac{2}{3}\frac{e^2}{h}$ and $\frac{4}{3}\frac{e^2}{h}$. In the presence of $S_z$-changing perturbations (e.g., Rashba spin-orbit coupling), an interaction-induced insulating edge state can emerge without breaking time-reversal or charge-conservation symmetry, corresponding to the absence of a topologically protected edge state. We further provide an exact mapping to a noninteracting fermionic theory exhibiting Anderson localization. Our results showcase an explicit, experimentally relevant example that the edge-state two-terminal transport is insufficient to identify the $ν_{\text{tot}}=4/3$ fractional topological insulators.

Collective modes in Fermi liquids are usually regarded as dissipation channels that relax electronic excitations through Landau damping. Whether such modes can instead mediate the formation of correlated electronic states under non-equilibrium conditions remains an open question. Here we show that, under optical photo-doping, a bulk plasmon can drive correlated inter-band transfer within a transient electronic continuum. Using time- and angle-resolved photoemission spectroscopy (Tr-ARPES) on EuCd$_2$As$_2$ supported by electronic structure calculations, we observe that at high excitation density, plasmons transfer energy from a weakly dispersing bulk band into unoccupied surface states. This bulk-to-surface redistribution stabilizes a long-lived, energy-localized spectral feature consistent with a Mahan exciton. Our results uncover a non-equilibrium regime of Fermi-liquid physics in which collective modes do not merely dissipate energy, but also stabilize correlated bound states.

We study band modulations and topological transitions in a one-dimensional periodic bead-on-string chain. Using an exact transfer-matrix formulation of the wave equation with periodically modulated mass density, combined with numerical spectral searches and tabletop experiments, we characterize band gaps and localized midgap states. We interpret these states by mapping the system to the Su-Schrieffer-Heeger (SSH) model and its low-energy (1+1)-dimensional Dirac theory. This framework reveals that the robust states are topological solitons bound to boundaries or engineered domain walls in the Dirac mass. Through this mapping, we provide an intuitive account of how band structure controls topological phase changes in mechanically realizable lattices.

Collective motion in animals and cells often exhibits rapid reorientations and scale-free velocity correlations. This allows information to spread rapidly through the group, allowing an adequate collective response to environmental changes and threats such as predators. To explain this phenomenon, we introduce a simple, biologically plausible mechanism: a minority-triggered reorientation rule. When local order is high, agents sometimes follow a strongly deviating neighbor instead of the majority. This rule qualitatively changes the macroscopic system behavior compared to traditional flocking models, as it generates heavy-tailed cascades of reorientations over broad parameter ranges. Our mechanism preserves cohesion while markedly enhancing collective responsiveness because localized directional cues elicit amplified group-level reorientation. Our results provide a parsimonious, biologically interpretable route to critical-like fluctuations and high responsiveness during flocking.

We investigate the interplay between phase lag and adaptive learning rules in a network of identical pendulum oscillators, where the coupling strengths evolve dynamically in response to the oscillators' states. Specifically, we examine two biologically inspired adaptation mechanisms, Hebbian and spike-timing-dependent plasticity (STDP), and their influence on the emergence of collective dynamical patterns. Under Hebbian adaptation, the network exhibits a wide range of organized behaviors, including two-cluster, solitary, multi-antipodal, and chimera states. In contrast, STDP coupling induces splay, splay-cluster, and splay-chimera configurations. Importantly, we find that the solitary state arises spontaneously in this adaptive network without requiring delays, nonlocal coupling, or external perturbations; instead, it is induced purely by variations in the phase-lag parameter. To the best of our knowledge, such delay-free and symmetry-preserving emergence of solitary behavior has not been reported previously in adaptive oscillator systems. To systematically characterize the resulting dynamical transitions, we employ two complementary incoherence measures based on the local standard deviation of (i) time-averaged frequencies and (ii) instantaneous phases across spatial bins, enabling the construction of detailed two-parameter phase diagrams. Analytical stability analysis of the two-cluster state shows strong agreement with numerical simulations, revealing regions of pronounced multistability. These findings establish adaptive pendulum networks as a minimal yet powerful framework for studying self-organized synchronization, chimera formation, and multistable transitions driven by diverse adaptation mechanisms.

We investigate collective dynamics in a pulse-coupled adaptive Winfree network under the influence of a frustration (phase-lag) parameter. The coupling strengths coevolve according to a Hebbian adaptation rule and self-organize to support a wide variety of collective states. We observe frequency-clustered states, entrainment, bump states, bump--frequency cluster states, antipodal and multi-antipodal cluster states, chimera states, and incoherent dynamics. Notably, we report for the first time the spontaneous emergence of entrainment, bump, and bump--frequency cluster states in an adaptive network {\it without} any external forcing. To systematically characterize these regimes, we introduce three complementary measures of incoherence based on (i) time-averaged frequencies, (ii) instantaneous phases, and (iii) mean frequencies per bin. These measures enable the construction of one- and two-parameter phase diagrams that clearly delineate transitions between distinct dynamical states. Furthermore, we analytically derive the stability condition for the frequency-entrained state, which shows excellent agreement with numerical simulations. Our results highlight the crucial role of frustration-mediated plasticity in shaping rich self-organized dynamics in pulse-coupled adaptive networks.

Exceptional points (EPs), indicative of parity-time (PT) symmetry breaking, play a central role in non-Hermitian physics, yet most studies begin from deliberately engineered effective Hamiltonians whose parameters are tuned to exhibit exceptional behavior. In realistic open quantum systems, however, dynamics are governed by Lindblad superoperators whose spectral structure is high-dimensional, symmetry-constrained, and not obviously reducible to minimal non-Hermitian models. A general framework for discovering exceptional dynamics directly from microscopic dissipative models has been lacking. Here we introduce a symmetry-resolved approach for identifying and characterizing exceptional points directly from the full Liouvillian generator. Correlated dissipation induces graph symmetries that decompose Liouville space into low-dimensional invariant sectors, within which minimal non-Hermitian blocks govern the onset of EPs and PT-breaking behavior. We further introduce a numerical diagnostic - the exceptional-point strength $\mathcal{E}$ - based on eigenvector conditioning, which quantifies proximity to defective dynamics without requiring analytic reduction. Applied to tight-binding models with correlated dephasing and relaxation, the method reproduces analytically predicted exceptional seams and reveals universal scaling of $\mathcal{E}$ near EP manifolds. More broadly, the framework enables systematic discovery of hidden exceptional structure in complex or high-dimensional open systems and is naturally compatible with matrix-free and tensor-network implementations for scalable many-body applications.

Synchrony patterns describe network states in which nodes of a coupled dynamical system are grouped into clusters based on synchronization between nodes. Beyond simple synchrony, synchronized clusters may also exhibit active or inactive states, and the collection of all such clusters constitutes an activity pattern. Although these patterns may arise naturally in networks with permutation symmetries, the requirement of symmetries imposes a restrictive and often unrealistic assumption, as many real-world networks lack such symmetries. In this work, we present synchrony patterns of coexisting active-inactive clusters that cannot be identified through symmetries. Considering dynamical systems in which intrinsic dynamics and coupling functions are odd functions in phase space, we identify all possible patterns a network can exhibit through symmetry breaking of identically synchronized clusters. The symmetry breaking of invariant clusters generates antisynchronized clusters, allowing active-inactive clusters to coexist. We show that while active clusters are external equitable partitions, inactive clusters can be purely dynamics-induced. Starting with a symmetry-broken state, we show that the existence of different invariant patterns is a function of coupling strength and intercluster weights. Finally, by combining synchronization manifolds with the Laplacian eigenvectors, we identify transversal perturbations for these patterns and present a stability analysis.

In this paper, we investigate the geometric, algebraic and analytic properties of the hyperelliptic $\mathrm{al}_{ab}$ functions of a hyperelliptic curve $X$ genus $g$ as the $\mathrm{al}_{ab}$ functions together with the $\mathrm{al}_a$ functions are a generalization of the Jacobi sn, cn, dn functions. We then demonstrate that the $\mathrm{al}_{ab}$ function is a potential hyperelliptic solution to the nonlinear Schrödinger and complex modified Korteweg-de Vries equations as a natural extension of the hyperelliptic solutions of the modified Korteweg-de Vries equation in terms of the $\mathrm{al}_a$ function.

A fundamental open problem in the theory of the multidimensional compressible Navier-Stokes equations is whether smooth solutions can develop singularities in finite time. For constant viscosity coefficients, recent remarkable results show that there exist smooth initial data for which the corresponding smooth solutions of the barotropic flow undergo finite-time implosion at the origin, with the density blowing up to infinity. In contrast, when the viscosity coefficients depend linearly on the density (as in the shallow water case), it has been established that, for general large spherically symmetric initial data, the solutions remain globally regular. These results indicate that the qualitative behavior of multidimensional solutions is sensitive to the structure of the viscosity coefficients. In this paper, we investigate the case of nonlinear viscosity coefficients with power-law density dependence. We identify a threshold value, depending on the adiabatic exponent, such that, for any power below this threshold, there exists a class of smooth initial data with strictly positive density for which the corresponding smooth solutions implode in finite time at the origin. The key issue is to show that, in this regime, the degenerate viscous terms are not sufficiently strong to suppress the convective mechanism driving the implosion. Establishing this rigorously is highly nontrivial due to the degenerate structure of the Navier-Stokes equations. To overcome this difficulty, we first derive a pointwise estimate for the density and then obtain spatial decay estimates for the velocity gradient via carefully constructed weighted high-order energy estimates and interpolation inequalities. The resulting decay rate is sufficiently rapid to compensate for the singular growth of the density, leading to uniform-in-time control of the viscous terms and ultimately to the formation of implosion.

For a system of partial differential equations admitting point, contact, or higher symmetries, the framework of invariant reduction systematically computes how invariant geometric structures, such as conservation laws, presymplectic structures, variational principles, and Poisson brackets, are inherited by the systems governing symmetry-invariant solutions. We extend this mechanism to geometric structures that are not invariant but are $\textit{rescaled}$ by a symmetry. Specifically, if $X$ is the symmetry used for reduction, $X_s$ is a symmetry satisfying $[X_s,X]=aX$, and the Lie derivative $\mathcal{L}_{X_s}$ acts on an $X$-invariant element of the Vinogradov $\mathcal{C}$-spectral sequence as multiplication by $b$, then the restricted symmetry $X_s|_{\mathcal{E}_X}$ acts on the corresponding reduction as multiplication by $a+b$. This shift rule gives rise to two phenomena: the $\textit{emergence of invariance}$, where reductions acquire an invariance that was not present at the level of the original structure, and the $\textit{loss of invariance}$, where reductions of invariant structures are no longer invariant. As an application, we describe a class of exact solutions to systems possessing sufficiently many symmetries and conservation laws subject to certain compatibility conditions. These solutions are invariant under pairs of symmetries and are completely determined by explicitly constructed functions that are constant on them; the description is geometric and does not require any integrability-related structures such as Lax pairs. The framework is illustrated by two examples: the Lin--Reissner--Tsien equation of potential nonstationary transonic gas flows, for which closed-form exact solutions are obtained and validated numerically, and the potential Boussinesq system, for which the inherited Poisson bracket is employed to describe solutions determined by algebraic equations.

In this work, the influence of inertial effects on the formation and evolution of vegetation patterns on sloped arid terrains is investigated from the onset of instability to far-from-equilibrium. Analyses are carried out in a hyperbolic extension of the one-dimensional Klausmeier model, where autotoxicity effects are also taken into account. As the system moves away from the wave bifurcation threshold, two classes of solutions arise: small-amplitude periodic migrating bands near onset and large-amplitude travelling pulses in far-from-equilibrium conditions. For the first class, results of LSA reveal that inertia has a twofold role at onset: it acts as a destabilising mechanism, thereby enlarging the parameter region in which uphill migrating vegetation bands can emerge, and it reduces the pattern migration speed. Its role also manifests itself close to onset, as proved by the Stuart-Landau equation for the pattern amplitude deduced via multiple-scale WNA. Indeed, it is shown that inertial effects may reverse the dynamical regime, from supercritical to subcritical, thus leading to hysteresis. For the second class of solutions, the travelling vegetation pulses are first captured via numerical simulations and then investigated via Geometric Singular Perturbation Theory (GSPT). In far-from-equilibrium conditions, inertia is shown to increase pulse speed while preserving the intrinsic multiscale structure of the solution, in full agreement with the numerical findings. Overall, the proposed combined analytical-numerical investigations have depicted several ecological scenarios as a function of the distance from the instability threshold, elucidating that inertia does not exclusively act as a time lag.

Motile cilia drive biological fluid transport through whip-like beating motions that synchronize into metachronal waves. The lengths of these cilia span three orders of magnitude, from microns in human airways to millimeters in ctenophores. While recent studies have considered ciliary flows at intermediate Reynolds numbers, the effect of inertia on coordinated particle transport remains unexplored. Here, we address this gap using "Pufflets," the inertial counterparts of Stokeslets. These Pufflets describe rapidly accelerating flows generated by short-lived impulses, encoded by spatiotemporally singular momentum injections. To produce such rapid impulses experimentally, we designed an Atwood machine that generates long-lived Pufflet flows, which we capture with high-speed PIV measurements that agree well with analytical theory and simulations. Moreover, we find that pairs of equal and opposite Pufflets can drive net particle displacements and mixing due to time reversal symmetry breaking, which would be impossible in Stokes flow. Finally, we consider metachronal waves of Pufflets. Remarkably, we discover that particles can surf on these waves by coasting inertially from one cilium to the next, leading to highly efficient particle transport. This work paves the way toward understanding rapidly accelerating flows and collective transport driven by biological and artificial cilia.

According to the literature, mobile devices as experimental tools (MDET) can offer educational benefits by creating authentic, real-life contexts for physics learning, enhancing student motivation through the use of familiar technology, and supporting cognitive processes by providing multiple representations of phenomena. However, concerns have been raised about potential distractions and cognitive overload. Regarding these conflicting perspectives, few empirical studies on the impact of MDET in real classroom settings of regular, full-length physics courses are available, focusing on a non-specialized high-school target group. We present a study of a mechanics course in such a new setting, addressing the tight curricular, material, and practical constraints inherent to it. A quasi experimental pre post design comparing a treatment group using MDET and a control group without (same content, lesson plan, and teachers) was used. The 19-week teaching sequence focused on conceptual learning and motivational outcomes, controlled by several predictor variables. Findings reveal substantial pre post learning gains for both groups (Cohen d = 0.9) and small gains for perceived relation to reality (d = 0.29). But no significant differences between treatments were found, indicating that MDET do not outperform conventional teaching under the given constraints. Moreover, no evidence of negative effects such as distraction or cognitive overload was observed, and little to no interactions with predictors such as gender or prior knowledge were found. In conclusion, MDET show considerable potential as an effective option for integrating technology into teaching, offering learning outcomes comparable to those of successful conventional teaching, but not better.

Current research on three-dimensional metamaterial has largely focused on conventional strut, plate, and shell-based lattice designs. Although these designs offer several advantages, they possess inherent limitations that can restrict their performance in certain applications, motivating the exploration of alternative structural topologies. Here, we present a large-scale, symmetry guided framework for the generation and analysis of architected metamaterials based on all 36 cubic space groups. Using a voxel-based representation, we construct a database of approximately 1.95 million periodic unit cells spanning a broad range of relative densities and topological complexity. This dataset reveals a rich elastic property landscape shaped by crystallographic symmetry, including rare pentamode designs with high bulk to shear ratios such as $K/G \approx 166$ , isotropic-auxetic architectures with Poisson's ratio as low as $ν=-0.76$, and structures achieving up to 93% of the Hashin-Shtrikman upper bound on stiffness. Complementing the dataset, we develop a three-dimensional convolutional neural network surrogate model trained and evaluated on the full database to predict strain-energy density values under uniaxial, shear, and hydrostatic loading. Together, this work establishes a comprehensive atlas of cubic symmetric metamaterials and provides a pre-trained model for the accelerated discovery and design of 3D architected materials with extreme mechanical properties.

Light Fidelity (LiFi) has emerged as a promising wireless technology that exploits the vast unlicensed optical spectrum to complement radio frequency networks. Recent advances in laser-based transmitters, particularly vertical-cavity surface-emitting laser (VCSEL) arrays, enable LiFi systems with multi-gigabit data rates, fine-grained spatial multiplexing, and high energy efficiency. However, the highly directional nature of laser beams introduces new challenges related to user mobility, alignment, and dynamic environments. This article introduces VCSEL-enabled holographic communication as a system-level paradigm that addresses these challenges by tightly integrating communication, sensing, and positioning within a single LiFi architecture. The proposed approach leverages individually addressable VCSEL arrays to form a dense grid of controllable beams, while a real-time digital twin of the environment enables adaptive beam management, environmental mapping through sensing, and user localization through positioning, including non-line-of-sight operation. By tightly integrating high-speed data transmission with environmental perception and user tracking, the LiFi access point evolves from a static transmitter into an intelligent environmental hub. The article also provides a tutorial overview of the underlying hardware, system architecture, and operational principles of holographic LiFi, and discusses key applications, open challenges, and future research directions toward next-generation intelligent optical wireless networks.

The synergy of ferroicity with altermagnetism offers a novel platform for designing multifunctional altermagnetic-spintronic device technology. In this work, we propose a mechanism to achieve nonvolatile electrical manipulation of spin and layer degrees of freedom in an altermagnetic bilayer via sliding ferroelectricity. Using first-principles calculations, we show that an interlayer translation can induce a switchable out-of-plane ferroelectric polarization in bilayer CuF2, which directly couples to and reverses the d-wave altermagnetic spin splitting. Notably, the altermangetic spin splitting is layer-locked, the sliding ferroelectricity-driven switching thus embodying a nonvolatile spin-layertronics functionality that couples spin-polarized transport and layer degree of freedom in a single platform. We show that in quadrilayer CuF2, four polarization states are identified which may offer multi-state logic device applications. These findings establish sliding ferroelectricity as a versatile tool for designing voltage-controlled, high-speed and energy-efficient spin-layertronic devices based on altermagnets.

The high directionality of hyperbolic phonon polaritons (HPhPs) has opened radically new ways to route and steer the flow of energy at the nanoscale. However, launching HPhPs requires fabricating efficient and precisely aligned polariton launching structures, which remains time-consuming and expensive with conventional nanofabrication approaches. Recently, using optical laser pulses, polariton launching structures have been programmed into the plasmonic phase-change material In3SbTe2. Here, we leverage this approach to reconfigure HPhPs by programming a variety of launching and confining nanostructures through α-MoO3 flakes deposited onto In3SbTe2. Importantly, optical programming after flake deposition enables alignment of launching stripes to the [001]-axis of the flake, essential to control the directional polariton propagation. We showcase these capabilities in a variety of structures: i) an optically programmed disk, showing similar tuning ranges and confinement as focusing by gold disks; and ii) a cavity for in-plane HPhPs created by reconfiguring the single disk to a double disk structure, tailoring the confinement by simply reprogramming the disk distance. Our fabrication scheme offers fast turn-around times, flexible alignment and the opportunity to reconfigure the structures. Thus, it is a fast, efficient and versatile way to tailor propagation and confinement of highly directional polaritons on demand.

Understanding ion-matter interactions at the atomistic level is key to advancing materials for the semiconductor industry, space systems, and nuclear fusion technologies. However, most atomistic frameworks still rely on simplified descriptions of how ions transfer energy to the electronic subsystem, overlooking the sensitivity of this process to the actual ion path. Existing electron-ion interaction models, such as the tensorial unified two-temperature model, were developed to study self-irradiation scenarios, but their suitability for light-ion irradiation remains unexplored. Here, we propose that for light projectiles, stepping back from the tensorial formulation toward a simpler, local model of electronic stopping provides a more efficient and physically transparent trajectory-dependent description. We parameterize and validate both models for hydrogen and helium in tungsten using ab initio electronic stopping data and large-scale ion range simulations, benchmarked against existing experimental data. This provides a consistent framework for including nonadiabatic electronic stopping in atomistic simulations of light-ion energy dissipation.

A reduced local field model derived from full-wave electromagnetic simulations shows that photonic nanojet formation corresponds to an emergent mesoscopic funnel of propagating power flux sustained by an effective free-space transverse mode structure. This interpretation moves beyond purely geometric-optics or interference-based explanations by identifying a self-consistent redistribution of phase gradients and effective longitudinal wavenumber near the nanojet waist. The model quantitatively captures characteristic nanojet morphology, including the formation and local structure of the jet waist. It also yields a geometry-independent lower bound on the nanojet waist, linking transverse confinement to the effective axial wavenumber through an explicit trade-off. The model establishes a direct connection between full-wave Maxwell fields and a reduced free-space oscillator description, yielding new physical insight into nanojet confinement and suggesting design principles for nanojet-assisted imaging, lithography, and subwavelength field localization.

We develop a quantum statistical framework for passive optical surface metrology. Modelling a surface as an incoherent ensemble of point emitters imaged through a diffraction-limited system, we employ techniques from quantum parameter estimation and hypothesis testing to derive ultimate bounds for jointly estimating geometrical features and for deciding the presence or absence of surface defects, and we identify optimal measurements from the geometry of the point-spread-function manifold. As a representative application, we analyse a minimal surface crack model based on three point sources and show that spatial mode sorting can simultaneously enable near-quantum-limited estimation of crack width and depth and markedly enhanced detectability of the crack, compared with direct imaging. Our results pave the way towards enhanced optical inspection and characterisation of sub-diffraction surface features by probing a limited number of spatial modes without any illumination control.

Magnon-photon coupling in cavity magnonic systems offers a promising route toward integrated wave-based information-processing devices. However, in ultrathin magnetic films the weak magnon response is easily buried beneath photon-dominated spectra. We show that a derivative-divide analysis of the microwave transmission parameter in a planar split-ring-resonator cavity isolates the magnetic contribution and resolves clear anticrossings in yttrium iron garnet and CoFeB films, yielding measurable coupling down to thicknesses of 60 nm and 5 nm, respectively. These results establish derivative-divide method as a simple and sensitive probe of magnon-photon coupling in ultrathin insulating and metallic films, and as a practical tool for characterizing miniaturized cavity-magnonic devices.

Social media platforms have become critical infrastructures for public communication, where large-scale interaction can both support socially beneficial collective pressure and amplify polarization and conflict. While opinion-dynamics research has long modeled how beliefs evolve through interpersonal influence, the central challenge for healthier online environments increasingly lies in algorithmic interventions: mechanisms that steer collective opinion toward desirable outcomes or dampen harmful dynamics. This survey offers a structured synthesis of this fast-growing, interdisciplinary literature. We organize prior work by the objective optimized -- overall opinion (e.g., consensus or mean opinion), polarization and disagreement, and other quantities -- and review the associated optimization formulations and representative algorithms with mathematical rigor. We also compile intervention-relevant theoretical and empirical findings. Finally, we outline concrete future directions that emerge from this survey.

Nonreciprocal structures play an important role in optical physics and applications. Conventional approaches for designing nonreciprocal optical structures rely heavily on extensive numerical simulation and parameter tuning, leading to high computational cost and low efficiency. Here, we apply deep learning to the design of nonreciprocal multilayer photonic structures. Three neural-network models-a forward neural network (FNN), an inverse design network (IDN), and a variational autoencoder (VAE)-are employed to learn the complex mapping between structural/material parameters and nonreciprocal spectral characteristics. We show that the FNN can rapidly and accurately predict the nonreciprocal electromagnetic response of a given structure, while the IDN can directly generate suitable structural parameters for target spectral responses. Both approaches substantially reduce computational cost and design time while improving nonreciprocal performance. Furthermore, the VAE can generate band-limited inverse design under practical performance constraints, facilitating efficient exploration of multiple feasible structures that meet different threshold requirements within specified frequency bands. Our work highlights the potential of deep learning for the advanced design of nonreciprocal optical structures and devices.

Brasília offers a rare test of how urban form shapes experienced segregation. Built almost at once around modernist neighbourhood units, then expanded through planned satellites and informal peripheries, it lets us ask whether urban form turns mobility into mixing or into a more efficient engine of separation. We combine data on human mobility with urban morphometrics, amenities, road networks, along with enclosures and tessellations that capture segregation at the scales where access is structured: districts, neighbourhoods, blocks, and street-and-building cells. We find that segregation intensifies as resolution sharpens, from 0.282 at the district scale to 0.545 at the block scale, indicating that Brasília looks most integrated at coarse units and most segregated where everyday encounters are actually organised. Mobility softens home segregation for most users, but not symmetrically: poorer groups travel farther, while affluent groups remain the most selectively exposed. civic cores and mid-rise, mixed-use areas are the least segregated morphotypes, yet they occupy only a sliver of the metropolis. Elsewhere, rich lakefront suburbs and dense poor settlements reach similarly high segregation through opposite spatial logics. Amenities predict lower segregation, while barriers and enclosed residential interiors predict higher segregation. Built form explains more of this pattern than visit volume alone in the segregation models: integration is less a property of residential design than of shared destinations and porous connections. Planned capitals can build order without building isolation if they distribute mixing space rather than sequestering it.

Breakthrough technologies increasingly shape social institutions, economic systems, and political futures. Yet models of research excellence associated with such technologies often prioritize technical performance, scalability, and short-term innovation metrics while treating ethical, social, and cultural dimensions as secondary considerations. This perspective article argues that such separation is no longer tenable. We propose a broader understanding of excellence that combines technical rigor with ethical robustness, social intelligibility, and long-term relevance. The rapid emergence of generative and agentic artificial intelligence further underscores this argument. As technological systems increasingly operate through language, interpretation, and normative alignment, expertise traditionally cultivated in the humanities and social sciences becomes integral to the design, governance, and responsible deployment of such systems. Drawing on historical examples and contemporary research practices, this article examines five interconnected domains where the humanities and social sciences, treated as integrated dimensions of research practice, can strengthen technological development: (1) ethical, legal, and social integration in agenda-setting and research design; (2) plural and reflexive foresight practices that shape technological futures; (3) graduate education as a leverage point for cross-disciplinary literacy; (4) visualization and communication as epistemic and civic practices; and (5) institutional frameworks that move beyond rigid distinctions between basic and applied research. Across these dimensions, we propose practical strategies for embedding interdisciplinary collaboration structurally rather than symbolically.

We report laser cooling of Yb3+-doped KY3F10 (Yb:KY3F10) driven by a 100-W, 1020-nm pump source. Despite pumping at a non-optimal wavelength, high-quality KY3F10 crystals doped with 3% and 7% Yb were cooled to 145 K and 151 K, respectively, in a double-pass pump configuration. These results establish Yb:KY3F10 as an attractive laser-cooling medium competitive with Yb:YLF, the state-of-the-art laser-cooling material for optical cryocoolers. The observed cooling performance and spectroscopic characteristics suggest that lower cryogenic temperatures may be achieved through pump-wavelength optimization, enhanced pump absorption, and reduced radiative heating.

Orbital angular momentum (OAM) entanglement gives access to multiple qubit and high dimensional Hilbert spaces, but is unfortunately susceptible to disturbance, decaying in real-world noisy channels. Here, we show there is an underlying topology arising from OAM entanglement that is robust to such channels, which we demonstrate using atmospheric turbulence -- exemplary of stochastic or chaotic media. Using a quantum channel with various turbulence strengths, we find the OAM topological observable preserved even though the OAM itself is shown to be highly sensitive to the turbulence. We show this is true for mixed states too, with the OAM topology intact even as the purity of the state decreases due to decoherence. Our work offers a new perspective on OAM entanglement preservation, and may easily be extended to other spatial bases, degrees of freedom, as well as complex channels, whether static or dynamic.

Exceptional points (EPs) exhibit strongly enhanced spectral responses and are therefore promising candidates for sensing applications. Whether these non-Hermitian degeneracies provide a genuine advantage in the quantum regime has been the subject of ongoing debate. Here, we address this issue within a scattering-matrix formalism for sensing with coherent light, which allows the quantum Fisher information (QFI) to be evaluated directly from experimentally accessible scattering data without introducing additional noise channels beyond those inherent to the scattering process. We analyze both nondegenerate and degenerate scattering-matrix poles, including EPs of arbitrary order, and show that the QFI per incoming photon flux is governed by three key factors: the decay rate of the resonant mode, the strength of the spectral response associated with non-normality, and the adjustment between the scattering states and the information source. For spatially localized perturbations, this implies that the Fisher information is fully determined by the local density of states at the perturbation site. Within this framework, we demonstrate that EPs can enhance the QFI compared to isolated modes or diabolic points with identical decay rates, and that the QFI can be further increased by moving away from the EP toward parameter regimes where non- Hermitian linewidth splitting reduces the decay rate of one mode. We further show that sufficiently small additional internal losses do not alter this overall picture, thereby providing a unified and experimentally relevant perspective on the design of quantum-limited non-Hermitian sensors.

Skyrmions are a particle-like topology with a quantised skyrmion number, realised across condensed matter and photonic platforms alike. In quantum photonics, they constitute an emerging resource, promising robust quantum information encoding, so far realised as single photon and bi-photon entangled states. Here we report the first visualisation of tripartite entanglement dynamics through topological structure using spin-skyrmion entangled states, where the topology of a single photon is remotely controlled through the spin of its entangled partner. We visualise our tripartite state theoretically by introducing the notion of a topological Bloch sphere that completely captures the entanglement and topolological features of the state. By leveraging this state, we realise the first quantum multiskyrmions, comprising multiple localised skyrmions within a single structure, that emulate signatures of their magnetic counterparts. We verify this experimentally and show that traversing our topological sphere reveals entanglement-driven particle-like motion of the localised topological structures. These dynamics unveil a physical manifestation of tripartite entanglement correlations which we illustrate by example of GHZ-like states, enabling a visualisation of multiple Bell states encoded within our system. Our work opens exciting possibilities for quantum sensing by mapping complex quantum channel features onto topological observables of multipartite states and offers a promising avenue for harnessing quantum topologies for multi-level encoding quantum communication schemes.

The universal prevalence of cooperation is puzzling, as defection typically yields higher payoffs than cooperation, motivating searches for hidden pathways to cooperation. Here we study a game-theoretic model on a lattice structured population in which interaction payoffs are scaled by the minimum of participants' accumulated wealth, reflecting real-world heterogeneity and incorporating the influence of past strategic choices. This wealth scaling allows frequent cooperators to surpass defectors in payoffs through their greater wealth even at high cooperation costs where defection would otherwise dominate. At the elevated critical cost-benefit ratio, the wealth gradient at the cooperator-defector boundary in one dimension exhibits a discontinuous transition. We show that slowing and effective stalling of the boundary trigger an explosive buildup of the wealth gradient, driving the dominance of cooperation below the critical ratio. Remarkably, this promotion of cooperation is stronger at higher temperatures, revealing a constructive role of fluctuations.

The generation of arbitrary polarization states of light is essential for optical communication and photonic information processing. Photonic crystal and metasurface platforms supporting bound states in the continuum (BICs) provide a powerful route for polarization engineering through tailoring the radiation from the resonant modes. However, existing approaches typically rely on static structural symmetry breaking or off-normal radiation, which limits continuous polarization tuning of vertical radiation. Here, we demonstrate a magnetooptical metasurface that generates arbitrary polarization states of light at normal radiation. By applying an external magnetic field with variable rientation, a symmetry-protected BIC is transformed into a quasi-BIC whose radiation polarization can be continuously tuned. The magneto-optical perturbation drives the controlled migration of polarization singularities in momentum space, allowing the emitted states to continuously span the entire Poincaré sphere without structural modification. This approach establishes a compact platform for actively tunable polarization sources and polarizationencoded photonic devices.

Two Fresnel zone plates zooming optics have been successfully developed and installed at the AR-NE1A beamline of the Photon Factory at the high energy accelerator research organization (KEK) in Japan. To ensure the reliable and versatile operation of this optical instrumentation, a dedicated control architecture has been implemented based on the simple transmission and retrieval system (STARS) framework, incorporating the newly proposed STARS common commands for detector control (CCDC) -- a detector-specific data acquisition (DAQ) state and command system. This system serves as both a practical control system for zooming optics and a demonstration model for modular extensibility using the STARS framework and inter-operability among detector systems enabled by the CCDC command set. The system has been commissioned, and its performance has been verified at the AR NE1A beamline. The control architecture affords enhanced configurational flexibility for optical components and provides an interface appropriate for both routine users and advanced experimental protocols.

Magic angle twisted bilayer graphene (MATBG) hosts narrow moiré bands with meV-scale energy splittings, making its correlated phases sensitive to both material parameters and modeling choices in low-energy downfolding. We develop an ab initio quantum-embedding workflow that derives interacting flat-band Hamiltonians from Kohn-Sham density functional theory (KS-DFT) of a relaxed, unstrained structure. Our model combines constrained random phase approximation (cRPA) screening, controlled double-counting subtraction, and an automated gauge-fixing procedure based on the selected columns of the density matrix (SCDM) that is compatible with symmetry-resolved many-body calculations. Solving the resulting models using Hartree-Fock (HF) and coupled cluster singles and doubles (CCSD), we recover robust insulating Kramers intervalley coherent (KIVC) states at charge neutrality ($ν=0$) and at electron doping ($ν=+2$). The main new physical effect appears on the hole-doped side: at $ν=-2$ we observe a fragile semimetal with a weak $\sqrt{3}\times\sqrt{3}$ Kekulé modulation and enhanced intervalley-scattering peaks in the Fourier-transformed local density of states. Although the underlying KS-DFT band structure is nearly particle-hole symmetric, the effective interacting Hamiltonian exhibits a pronounced particle-hole asymmetry at $ν=\pm 2$ that we trace to momentum-dependent single-particle renormalizations generated by subtraction terms constructed from reference densities consistent with the KS-DFT filling. Our work provides a first-principles route for connecting microscopic electronic structure, screened interactions, subtraction choices, and scanning tunneling microscopy signatures in MATBG.

Propagating uncertainties introduced by chemical reaction rate parameters to high-fidelity numerical simulations of complex combustion devices is necessary to ascertain impact on computational predictions. However, the high cost of detailed computations combined with the need to conduct multiple simulations to propagate uncertainty makes such an estimation computationally challenging. In order to reduce the computational cost, a two-step framework for quantifying uncertainty introduced by detailed chemical kinetics model parameters using reduced chemistry models is developed here. First, reduced-manifold states are uniquely reconstructed in full-composition space by following trajectories at an unburnt mixing state and integrating forward to a prescribed progress variable constraint. Second, parametric uncertainty is propagated by sampling perturbed rate coefficients from mechanism covariance matrices and integrating each realization to the target state, yielding uncertainty maps for reduced-space quantities. The method is applied in two configurations: a subsonic multi-tube combustor with interacting jet flames and recirculation, and a three-dimensional reacting high-speed flowpath. Uncertainty-instrumented estimated are reported for a trajectory time (time for the reconstructed unreacted mixture to reach the local target state) and for the time to equilibrium, revealing order-of-magnitude spatial variations driven by mixing, stratification, and residence-time effects. The largest relative variability occurs in low-to-intermediate temperature regimes associated with induction and the onset of heat release, where branching-related chemistry amplifies sensitivity, particularly away from stoichiometric conditions. The method provides a scalable route to spatially resolved, physically interpretable chemistry-UQ for practical reacting-flow simulations.

The inverse reconstruction of material properties from optical extinction efficiency (Qext) is constrained by the high-dimensional nature of Mie scattering. We demonstrate that the Qext manifold possesses an intrinsic, physics-governed sparsity universal across dielectric materials. By analyzing the spectral topology of a diverse polymer library, we identify a critical Information Bottleneck at the onset of the Mie transition (r approx 0.1 um), where a peak in spectral entropy signifies a fundamental limit on signal compressibility. While the Fast Fourier Transform (FFT) is conventionally used for spectral analysis, we show it is physically mismatched for this domain; its periodic boundary assumptions induce spectral leakage that forces a massive basis expansion to resolve Mie ripples. Conversely, the Discrete Cosine Transform (DCT) mirrors the non-periodic geometry of extinction profiles, uncovering inherent compressibility by capturing over 90% of signal energy using fewer than 10 harmonic modes. Even at the Mie bottleneck, the DCT maintains a 12-fold compression advantage over the FFT at a 99% energy threshold. Notably, while both bases converge to identical error floors for a fixed energy threshold, the DCT achieves this fidelity with significantly lower hardware overhead. Stress-testing under 10% additive Gaussian noise confirms the Information Bottleneck is spatially and structurally invariant, proving this complexity peak is a fundamental physical constant of the manifold. By mapping this sparsity onto a compressed sensing architecture, we resolve a 2.5-20 um spectral range using between 22 and 170 sensors: enabling a 51%-94% reduction in hardware complexity that breaks the traditional Nyquist sampling limit (350 sensors) for high-fidelity clinical and remote sensing applications.

Accurate measurement of the pulse duration of ultrashort, ultra-intense laser pulses at focus is essential for strong-field science. Most existing diagnostics, however, cannot allow direct in situ measurement in the focal region because of damage-threshold limits and unavoidable spatial averaging. We present a direct single-shot far-field diagnostic based on a plasma grating. In this method, the pulse duration is encoded in the axial length of an interference-written plasma grating and retrieved from the corresponding Bragg-diffraction signal. Comparison with near-field (pre-focus) autocorrelator measurements and far-field (at-focus) scanning measurements confirms single-shot pulse-duration retrieval in the focal region over 35-130 fs, and the method remains effective at a peak intensity of $\sim 10^{16}{\rm W/cm^2}$. In principle, the measurable range can be extended to 15-300 fs and to higher peak intensities. The method is insensitive to the laser central wavelength and offers a practical approach to far-field diagnostics in high-power laser systems.

We present a systematic methodology for designing slow-light photonic integrated circuits with a frozen mode based on a special kind of exceptional point of degeneracy (EPD) of order three named stationary inflection points (SIPs). This is realized through three-way coupled waveguides with lateral gratings operating at telecommunication wavelengths. We provide two designs and analyze sensitivity to geometric perturbations. We have fabricated a periodic waveguide with integrated taper loads and demonstrate reasonable agreement with full-wave simulations. These findings confirm the feasibility of integrating SIP-based delay functionalities in standard silicon photonic platforms.

Sensors based on optical microcavities enhance light-matter interactions within an ultraconfined volume, enabling high-sensitivity detection across a wide range of sensing applications. In these systems, environmental perturbations modify the intrinsic resonance properties of the cavity, typically manifested as frequency shifts, linewidth broadening, or mode splitting. However, the minimum resolvable change in these spectral properties fundamentally limits the overall sensor sensitivity. Here, we propose a new avalanche sensing scheme enabled by Kerr nonlinearity. Instead of relying on the detection of frequency shifts, our approach exploits abrupt state transitions in a Kerr frequency comb to amplify weak perturbations. We provide a theoretical analysis of the underlying mechanism of this scheme and validate the concept through both coupled-mode theory (CMT) modeling and full-wave electromagnetic simulations.

Solid-state semiconductor lasers underpin technologies from telecommunications and data storage to sensing, medical diagnostics, and emerging quantum communication. Polaritons-hybrid exciton-photon states have further extended this reach, enabling room-temperature quantum effects such as low-threshold lasing and single-photon nonlinearities. Organic semiconductors are ideal for polaritonics due to their large exciton binding energy, strong optical nonlinearities, and straightforward processing, making them attractive for both classical and quantum photonics. While solution-processed organic films have been widely explored, their optical cavities have almost always been fabricated using vacuum deposition, limiting the realization of truly scalable and low-cost devices. Here, we report the first organic laser microcavities fabricated entirely by solution processing, which operate in the strong coupling regimeThe resulting platform can be driven reliably to high excitation densities, where we observe a reversible, interaction-driven redistribution of the polariton condensate, revealing a distinct polariton lasing behaviour in organic microcavities. Together, the fabrication approach and the observed lasing dynamics establish a route toward scalable polaritonic and quantum photonic technologies and provide new opportunities for studying nonlinear polariton physics in organic systems.

For decades, scientists have studied the intricate charge migration dynamics, where after ionization a localized charge distribution ("hole") migrates across the molecule on a femtosecond timescale. This has the potential for controlling electrons in molecules, yet a comprehensive understanding of the many aspects of charge migration is still missing. In this work, we analyze charge migration using an extension of localized intrinsic bond orbitals (IBOs). These orbitals lead to a compact representation of the dynamics and map the complex, correlated many-electron charge migration to chemical concepts such as curly arrows and orbital-orbital interactions. By analyzing multiple challenging scenarios, we show how IBOs enable us to identify key mechanisms in charge migration. For example, we show that different mechanisms are responsible for converting a $π$-shaped hole to a $σ$-shaped hole and vice versa. We explain these in terms of hyperconjugation interactions and configurations that couple orbitals with different symmetries. We further demonstrate how IBOs can be used to find molecules with high charge migration efficiency. We carry out all simulations using an efficient set up of the time-dependent density matrix renormalization group (TDDMRG), correlating as many as 45 electrons in 50 orbitals. We believe that our results will be useful to design future experiments. The proposed IBO analysis is applicable to other types of real-time electron dynamics and spectroscopy.

Dark matter (DM) could be subject to non-gravitational self-interactions which is relevant to resolve potential problems of cold DM on small scales. Their impact on astrophysical objects such as galaxies and galaxy clusters allows for constraining the strength of this scattering and eventually further properties of the cross-section. To model self-interacting dark matter (SIDM), N-body simulations are a crucial tool widely employed by the SIDM community. In this paper, we describe the SIDM implementation in the cosmological hydrodynamical N-body code OpenGadget3 and release it to the public. It is capable of simulating elastic scattering for various differential cross-sections, including strongly anisotropic cross-sections. Beyond single-species models, the code also allows simulating a two-species model with cross-species interactions. In addition to describing the numerical schemes for modelling various flavours of SIDM, we discuss the technical challenges of implementing them. Moreover, we demonstrate through several test problems that OpenGadget3 can accurately simulate DM self-interactions. Furthermore, we assess the performance of the code and provide scaling tests. Lastly, we highlight remaining challenges in the context of SIDM and describe directions for improving the current state of the art.

Indonesia's nickel ore export ban has driven rapid expansion of smelting and hydrometallurgical processing capacity at the Indonesia Morowali Industrial Park (IMIP), now the world's largest integrated nickel processing complex, on the coast of Central Sulawesi. Whether this industrialization has degraded the adjacent marine environment remains unquantified. We apply Bayesian structural time-series (BSTS) causal inference to a multi-decadal, multi-sensor satellite ocean color record of the diffuse attenuation coefficient at 490 nm, $K_d(490)$, to test for a causal link between IMIP expansion and nearshore turbidity change. A consensus structural breakpoint, a significant posterior causal effect estimated against a Banda Sea counterfactual, and a distribution-free placebo rank test collectively establish that coastal water clarity deteriorated after the transition from initial nickel pig iron production to hyper-expansion of high-pressure acid leaching facilities for battery-grade nickel. Satellite-derived land cover analysis independently corroborates this timing, showing substantial built-area growth and concurrent tree cover loss within the IMIP footprint. The resulting euphotic zone shoaling occurs in oligotrophic waters supporting high marine biodiversity, where even moderate optical degradation may impair coral photosynthesis and compress depth-dependent reef habitat. These findings quantify a marine environmental cost absent from Indonesia's mineral downstreaming policy discourse and demonstrate a transferable, satellite-based quasi-experimental framework for causal impact assessment at coastal industrial sites in data-limited tropical settings.

How do competing populations convert a spatial advantage into macroscopic dominance? We introduce a stochastic model for resource competition that decouples the transient discovery phase from monopolization. Initial symmetry breaking is governed by extreme value statistics of first-passage times: a linear spatial disadvantage requires an exponentially larger population to overcome. However, transient superiority cannot stabilize dominance. A non-reciprocal interaction bias is strictly necessary to arrest local fluctuations and drive the system into a robust absorbing state.

Trigger waves are self-regenerating propagating fronts that emerge from the coupling of nonlinear reaction kinetics and diffusion. In cells, trigger waves coordinate large-scale processes such as mitotic entry and stress responses. Although the roles of circuit topology and feedback architecture in generating bistability are well established, how nonequilibrium energetic driving shapes wave propagation is less well understood. Here, we employ a thermodynamically consistent reaction--diffusion framework to investigate trigger-wave dynamics in ATP-dependent phosphorylation--dephosphorylation systems. We first recapitulate general expressions for trigger-wave speed in the bistable regime and analyze curvature-induced corrections that determine the minimum critical nucleus required for sustained propagation in higher dimensions. We then apply this framework to two representative systems, treating ATP concentration and the nonequilibrium parameter $γ= [ATP]/(K_{\mathrm{eq}}[ADP][P_i])$ as independent control variables to examine how energetic driving regulates wave propagation. Our results show that ATP and $γ$ not only modulate wave speed, but can also reverse the direction of propagation and reshape the parameter regime supporting trigger waves. The critical excitation radius also depends on both ATP concentration and phosphorylation free energy. These findings identify the intracellular energetic state as a regulator of trigger-wave behavior, linking metabolic conditions to the spatial dynamics of wave propagation. More broadly, this framework connects classical reaction--diffusion theory with ATP-driven biochemical regulation and provides a general perspective on related energy-dependent cellular decision-making processes.

The COVID-19 pandemic exposed young people to a prolonged and evolving societal stressor, yet it remains unclear whether symptom networks were reorganized or whether control was redistributed across a conserved modular scaffold. Here we analysed repeated cross-sectional data on 47 self-reported mental-health symptoms from 14,181 U.S. young adults aged 18-24 years across five COVID-19 phases between 2020 and 2023. For each phase, we estimated Gaussian graphical models, identified symptom communities, and characterized minimum-dominating-set-based module control. Symptom networks showed broadly conserved community organization across phases, indicating a stable mesoscale scaffold despite marked temporal variation. By contrast, intermodule control shifted from an early configuration centered on stress-related symptoms to a later, more distributed pattern spanning emotional, cognitive and social domains. Resampling analyses showed high stability for node strength and moderate stability for module-to-module control, whereas average within-module control was less robust. These findings suggest that prolonged crisis may preserve the modular architecture of youth psychopathology while redistributing control across symptom domains, and they identify intermodule control as a comparatively robust mesoscale feature for cross-phase comparison.

We revisit the controversial "abscopal" effect in the context of Personalized Ultra-Fractionated Stereotactic Adaptive Radiotherapy (PULSAR). By allowing long interval between fractions, PULSAR may enhance systemic immune activation and increase the likelihood of abscopal responses compared with conventional daily fractionation. To quantify treatment-induced effects, we introduce an interaction-picture transformation adapted from quantum mechanics, which separates intrinsic tumor growth from radiation and immune-mediated perturbations. In this preliminary study, we tested this method to two preclinical bilateral tumor models (4T1 and MC38). Our model provides a quantitative measure of the interaction strength between primary and secondary tumors at the individual level, capturing dynamics over time rather than relying solely on cohort averages. This approach frames the abscopal effect as a continuous, stochastic phenomenon rather than a binary response. The framework is flexible for future studies, particularly in concurrent radiation and immunotherapy with PULSAR, where different radiation doses and fractionation schedules can be compared, and immune checkpoint inhibitors (ICIs) can be incorporated to further enhance systemic anti-tumor immunity. The framework can also help us make cross-study comparison of abscopal effects and standardizes the reporting of abscopal magnitude beyond simple statistical significance.

This letter investigates phase noise (PN) mitigation in generalized receiver spatial modulation (GRSM) massive MIMO systems at mmWave under a common local oscillator (CLO). Under CLO, the received energy remains invariant relative to the no-PN scenario, enabling reliable energy-based spatial detection using the no-PN threshold. PN-sensitivity and geometry-based metrics are introduced to design compact, PN-resilient MQAM symbol pools with low detection complexity. PN robustness is further improved through an enhanced PN-aware GRSM-MQAM system that exploits spatial modulation (SM) to recover part of the MQAM bits and strategically maps spatial-pattern Hamming weights to reduce the effective PN impact. In addition, a practical single-stage PN estimation/compensation architecture is proposed, while a benchmark double-stage compensation is adopted to quantify the upper bound achievable via separate Tx/Rx PN mitigation. Results show that under PN, the overall BER is mainly dominated by MQAM symbol detection errors, especially for denser constellations, whereas spatial detection remains robust. The proposed single-stage compensation improves PN resilience, while the benchmark double-stage compensation approaches near PN-free performance.

Monostatic ISAC sensing differs from communication because the transmitter can monitor its distorted transmit waveform. Thus, the aggregate $κ$ distortion model, which treats impairments as unknown noise, is appropriate for communication but pessimistic for monostatic sensing. We derive PA-aware sensing Cramér--Rao bounds (CRBs) and a PN-aware CRB that reveals an irreducible velocity-error floor, and quantify when $κ$-based bounds overestimate sensing degradation. Simulations validate the analysis and show robustness to practical DPD template errors (less than 1~dB overhead at a typical $-25$~dB NMSE).

We analyse the level crossing rate (LCR) of an uplink single-user (SU) reconfigurable intelligent surface (RIS) aided system. It is assumed that the RIS to base station (RIS-BS) channel is deployed as line-of-sight (LoS), and the user (UE)-RIS and UE-BS channels are correlated Rayleigh. For the optimal RIS reflection matrix, we derive a novel and exact analytical LCR expression for when the direct (UE-BS) channel is blocked, i.e. the RIS-only channel. Also, the existing exact expression for the direct-only channel (equivalent to classical maximal-ratio-combining (MRC)) suffers from extreme numerical precision problems when the BS has many elements. Therefore, we propose a new stable and accurate approximation to the LCR of the direct channel. The approximation is based on replacing any small similar eigenvalues of the channel correlation matrix by their average. We show that increasing the number of elements at the RIS or BS and decreasing channel correlation makes the LCR drop more rapidly for thresholds away from the mean SNR. Crucially, we find that RIS systems do not significantly amplify temporal variations in the channel. This is particularly beneficial for RIS systems considering the difficulty in acquiring channel state information (CSI).

In this paper, we analyse the performance of a reconfigurable intelligent surface (RIS) aided system where the RIS is divided into subsurfaces. Each subsurface is designed specifically for one user, who is served on their own frequency band. The other subsurfaces (those not designed for this user) provide additional uncontrolled scattering. A new subsurface RIS design is developed based on the optimal single-user design for a pure line-of-sight (LoS) base station (BS) to RIS channel. This is also extended to arbitrary BS-RIS channels. For our method, exact closed form solutions for the mean SNR and a mean rate upper bound are derived for the BS-RIS LoS scenario. For each user, the designed subsurface performs optimally in LoS conditions and is remarkably robust to non-LoS conditions. The system design drives down complexity to extremely low levels, reducing RIS design and receiver processing complexity and reducing the channel estimation requirements. We also quantify the complexity-performance trade-off for the new design relative to multi-user approaches.

Cell switching is a promising approach for improving energy efficiency in wireless networks; however, existing studies largely rely on simplified models and energy-centric formulations that overlook key performance-limiting factors. This paper revisits the cell switching concept by redefining its modeling assumptions and mathematical formulation, explicitly incorporating realistic propagation effects such as building entry loss (BEL) and atmospheric losses relevant to non-terrestrial networks (NTN), particularly high-altitude platform station (HAPS). Beyond proposing a new cell switching strategy, the conventional energy-focused problem is reformulated as a multi-objective optimization framework that jointly minimizes power consumption, unconnected users, and data rate degradation. Through this reformulation, the proposed methods ensure that energy-efficient operation is achieved without compromising user connectivity and data rate performance, thereby inherently supporting sustainability objectives for sixth-generation (6G) networks. To solve this reformulated problem, two complementary approaches are employed: the weighted sum method (WSM), which enables flexible and adaptive weighting mechanism, and the {ε-constraint-inspired method (εCM), which converts connectivity and rate-related objectives into constraints within the conventional energy-focused problem. Moreover, unlike prior work relying only on simulations, this study combines system-level simulations with Sionna-OpenAirInterface (OAI) based emulation on a smaller network to validate the proposed cell switching concept under realistic conditions. The results show that, compared to the conventional approach, WSM reduces rate degradation for up to 70% for high-loss indoor users and eliminates the 44% drop for low-loss indoor users.

Over-the-air (OTA) emulation of diverse sensing target characteristics in a controlled laboratory environment is pivotal for advancing integrated sensing and communication (ISAC) technology, as it facilitates the non-invasive performance evaluation of ISAC base stations (BSs) across complex scenarios. In this work, a flexible multi-target OTA emulation framework based on a wireless cable method is proposed to evaluate the sensing performance of large-scale ISAC BSs. The core concept leverages an amplitude and phase modulation (APM) network to simultaneously establish wireless cables and simulate target spatial characteristics without consuming additional resources on costly radar target emulators. For the wireless cable method, the condition number increases as the number of antennas scales up, which affects the performance of the wireless cable. Although the wireless cable concept has been established for devices-under-test (DUTs) with a limited number of antenna ports, establishing wireless cables for large-scale DUTs remains an open question in the community. We address this problem by optimizing the OTA probe array configuration based on the theoretical properties of strictly diagonally dominant matrices. Experimental results validate the proposed framework, demonstrating high-isolation wireless cables for a 32-element DUT and an extremely low condition number for a 128-element synthetic array. Furthermore, the OTA emulation of a dynamic dual-drone scenario confirms the method's effectiveness and practicality in reproducing complex sensing environments.

Accurate estimation of the sound speed profile (SSP) is essential for underwater acoustic communication, sonar performance, and navigation, as the acoustic wave propagation depends strongly on the SSP. This work considers SSP estimation in a region of interest using an autonomous underwater vehicle (AUV) equipped with a conductivity-temperature-depth (CTD) sensor and an acoustic receiver measuring transmission loss (TL) from a sonar transmitter. The SSP is modeled using a linear basis-function expansion and is sequentially estimated with an unscented Kalman filter that fuses local CTD measurements with TL measurements. A receding-horizon path planning scheme is also employed to select future AUV positions by minimizing the predicted total sound speed variance. Simulations using the Bellhop acoustic wave propagation solver show that CTD measurements provide accurate local SSP estimates, whereas TL measurements are seen to capture the global characteristics of the SSP, with their joint use improving the reconstruction of both local variations and large-scale SSP behavior. The results also indicate that the proposed path planning strategy reduces the estimation uncertainty compared to constant-velocity motion, thereby enabling improved environmental characterization for underwater acoustic systems.

High-frequency acoustic wave transducers, vibrating at gigahertz (GHz), favored for their compact size, are not only dominating the front-end of mobile handsets but are also expanding into various interdisciplinary fields, including quantum acoustics, acoustic-optics, acoustic-fluids, acoustoelectric, and sustainable power conversion systems. However, like strong vibration can "shake off" substances and produce heat, a long-standing bottleneck has been the ability to harness acoustics under high-power vibration loads, while simultaneously suppressing temperature rise, especially for IDT-based surface acoustic wave (SAW) systems. Here, we proposed a layered acoustic wave (LAW) platform, utilizing a quasi-infinite multifunctional top layer, that redefines mechanical and thermal boundary conditions to overcome three fundamental challenges in high-power acoustic wave vibration: self-heating, thermal instability, and acoustomigration. By simply leveraging a simplified, thick single-material overlayer to achieve electro-thermo-mechanical co-design, this acoustic platform moves beyond prior substrate-focused thermal management in SAW technology. It demonstrates, for the first time from the top boundary, simultaneous redistribution of the von Mises stress field and the creation of an efficient vertical thermal dissipation path. The LAW transducer, vibrating at over 2 GHz, achieves a 70% reduction in temperature rise under identical power loads, a first-order temperature coefficient of frequency (TCF) of -13 ppm/C with minimal dispersion, and an unprecedented threshold power density of 45.61 dBm/mm2 - over one order-of-magnitude higher than that of state-of-the-art thin-film surface acoustic wave (TF-SAW) counterparts at the same wavelength.

Intelligent reflecting surfaces (IRSs) are poised to revolutionize next-generation wireless communication systems by enhancing channel quality and spectrum efficiency through advanced wave manipulation. However, extremely large-scale IRS {(XL-IRS)} deployments face significant challenges in channel estimation due to multiplicative path loss and near-field (NF) effects, where spherical wavefronts couple distance and angle parameters. Existing polar-domain codebook-based compressive sensing methods for NF channel estimation suffer from low accuracy and high complexity, caused by the need for high-resolution grids of both distance and angle parameters. To address this, we propose a harmonic processing-inspired channel estimation framework for NF {XL-IRS} systems by leveraging tensor modalization to decouple channel parameters. Drawing an analogy to musical harmonic analysis, our approach decomposes the high-dimensional NF channel tensor into independent factor matrices, modeled as ``chords," representing distance and angle parameters. Through harmonic analysis-inspired distance parameter decoupling, we design a compact, distance-dependent codebook that enables high-resolution NF channel parameter estimation. This approach significantly reduces the codebook size compared to polar-domain methods. {Then, we} derive the Cramér-Rao lower bound (CRLB) to evaluate the estimators. Finally, simulation results show an 8.5 dB improvement in normalized mean square error (NMSE) compared to conventional methods, underscoring its low complexity and high accuracy.

Radio Dynamic Zones (RDZs) are geographically defined areas specifically allocated for testing new wireless technologies. It is essential to safeguard the regular spectrum users outside the zones from the interference caused by the deployed equipment within this zone. Previous works have utilized sparse reference signal received power (RSRP) measurements collected by unmanned aerial vehicles (UAVs) to construct a dense 3D radio map through ordinary Kriging. In this work, we illustrate that matrix completion can outperform ordinary Kriging. We partitioned a 2D area of interest into small square grids where each grid corresponds to a single entry of a matrix. The matrix completion algorithm learns the global structure of the radio environment map by leveraging the low-rank property of propagation maps. Additionally, we illustrate that the simple Kriging and trans-Gaussian Kriging yield better results when the density of known measurements is lower. Earlier works of RSRP prediction involved a training dataset at a single altitude. In this work, we also show that performance can be improved by utilizing a combined dataset from multiple altitudes.

This paper presents a novel wireless sensing system where a movable antenna (MA) continuously moves and receives sensing signals within a three-dimensional (3-D) region to enhance sensing performance compared with conventional fixed-position antenna (FPA)-based sensing. We show that the performance of direction vector estimation for a target is fundamentally related to the 3-D MA trajectory in terms of the mean square angular error lower-bound (MSAEB), which is adopted as a coordinate-invariant performance metric. In particular, the closed-form expression of the MSAEB is derived as a function of the trajectory covariance matrix. Theoretical analysis shows that two-dimensional (2-D) antenna movement suffers from performance divergence for target direction close to the endfire direction of the 2-D MA plane, whereas 3-D movement can achieve isotropic sensing performance over the entire angular region. To achieve robust sensing performance, we formulate a min-max optimization problem to minimize the maximum (worst-case) MSAEB over a given continuous angular region wherein the target is located. An efficient successive convex approximation (SCA) algorithm is developed to optimize the 3-D MA trajectory and obtain a locally optimal solution. Numerical results demonstrate that the proposed 3-D MA sensing scheme is able to significantly reduce the worst-case mean square angular error (MSAE) compared with conventional arrays with FPAs and MA systems with 2-D movement only, thus achieving more accurate and robust direction estimation over the entire angular region.

In this paper, we introduce Spyglass, a spectrum sensor designed to address the challenges of effective spectrum usage in dense wireless environments. Spyglass is capable of observing a frequency band and accurately estimating the Angle of Arrival (AoA) of any signal during a single transmission. This includes additional signal context such as center frequency, bandwidth, and I/Q samples. We overcome challenges such as the clutter of fleeting transmissions in common bands, the high cost of array processing for AoA estimation, and the difficulty of detecting and estimating channels for unknown signals. Our first contribution is the development of Searchlite, a protocol-agnostic signal detection and separation algorithm. We use a switched array to reduce cost and processing complexity, and we develop SSFP, a signal processing technique using Fourier transforms that is synchronized to switching boundaries. Spyglass performs multi-channel blind AoA estimation synchronized with the array. Implemented using commercially available hardware, Spyglass demonstrates a median AoA accuracy of 1.4$^\circ$ and the ability to separate simultaneous signals from multiple devices in an unconstrained RF environment, providing valuable tools for large-scale RF data collection and analysis.

Movable antennas (MAs) have emerged as a promising technology for wireless sensing by reconfiguring antenna positions to exploit additional spatial degrees of freedom (DoFs). This paper investigates a robust movable antenna placement strategy for near-field wireless sensing to minimize the worst-case squared position error bound (SPEB). By temporarily relaxing the minimum inter-element spacing constraint, we first establish the optimality of centro-symmetric antenna position distribution, which simplifies the identification of the worst-case source, locating it at the array broadside on the Rayleigh boundary. Moreover, by leveraging moment-based analysis with the Richter-Tchakaloff theorem, we derive a closed-form optimal solution with three points supported on the center and two edges of the array. Guided by this structural insight, we finally develop an efficient three-point discrete deployment strategy to ensure the minimum inter-element spacing. Simulations demonstrate that the proposed design consistently outperforms conventional fixed antenna arrays and matches the exhaustive search benchmark at negligible computational complexity.

Spectrum sensing and the generation of 3D Radio Environment Maps (REMs) are essential for enabling spectrum sharing within cognitive radio networks. While Uncrewed Aerial Vehicles (UAVs) offer high-mobility 3D sensing, REM accuracy is challenged by dynamic flight behaviors, where fluctuations in UAV speed and direction introduce measurement inconsistencies. Furthermore, the structural influence of the airframe itself impacts the onboard antenna's radiation characteristics. In this paper, we present a comprehensive analysis of REM reconstruction at various altitudes, using real-world data from a fixed base station tower and a ground-vehicle source. We evaluate diverse reconstruction methodologies, including Kriging (simple, ordinary, and trans-Gaussian), matrix completion, and Gaussian process regression (GPR) for recovery from sparse samples. Our results indicate that simple Kriging and GPR remain more robust under extreme sample sparsity. We also propose a framework to enhance reconstruction accuracy in deep-shadowed regions by decomposing the REM into distinct smooth and deep-shadowed spatial components. We further investigate how REM reconstruction performance is influenced by physical and UAV-related external parameters. First, we demonstrate that the impact of UAV altitude on accuracy follows a tri-phasic trend: an initial performance gain up to $h_1$, a performance dip between $h_1$ and $h_2$, and a final stage of increasing accuracy. Additionally, we show that performance improves with increased spectrum bandwidth. Second, our analysis of UAV trajectories reveals that the variance of shadow fading exhibits a non-monotonic trend, peaking at both very low and mid-high elevation angles. Finally, we demonstrate that antenna pattern calibration from in-field measurements significantly enhances REM reconstruction accuracy by accounting for shadowing induced by the UAV airframe.

To meet the evolving demands of sixth-generation (6G) wireless channel modeling, such as precise prediction capability, extension capabilities, and system participation capability, multi-modal intelligent channel modeling (MMICM) has been proposed based on Synesthesia of Machines (SoM) which explores the mapping relationship between multi-modal sensing in physical environment and channel characteristics in electromagnetic space. Furthermore, for integrating heterogeneous sensing, reasoning across scales, and generalizing to complex air-space-ground-sea communication environments, two new paradigms of MMICM are explored, including fine-tuned large language models (LLMs) for Channel Modeling (LLM4CM) and Wireless Channel Foundation Model (WiCo). LLM4CM leverages pre-trained LLMs on channel representations for cross-modal alignment and lightweight adaptation, enabling flexible channel modeling for 6G multi-band and multi-scenario communication systems. WiCo, which pre-trained on physically valid channel realizations and their associated environmental and modal observations, embeds electromagnetic equations for physical interpretability and uses parameterized adapters for scalability. This article details the architectures and features of LLM4CM and WiCo, laying a foundation for artificial intelligence (AI)-native 6G wireless communication systems. Then, we conducts a comparative analysis of the two emerging paradigms, focusing on their distinct characteristics, relative advantages, inherent limitations, and performance attributes. Finally, we discuss the future research directions.

Adaptive ansatz construction has emerged as a powerful technique for reducing circuit depth and improving optimization efficiency in variational quantum eigensolvers. However, existing adaptive methods, including ADAPT-VQE, rely solely on first-order gradients and therefore ignore the underlying geometry of the quantum state space, limiting both convergence behavior and operator-selection efficiency. We introduce Geo-ADAPT-VQE, a geometry-aware adaptive VQE algorithm that selects operators from a pool using the natural gradient rule. The geometric operator-selection rule enables the ansatz to grow along directions aligned with the underlying quantum-state geometry, thereby improving convergence and reducing the algorithm's susceptibility to shallow local minima and saddle-point regions. We further provide an asymptotic convergence result. We present numerical simulations involving five molecules, which demonstrate that Geo-ADAPT-VQE achieves faster and more stable convergence compared to existing methods, while producing significantly shorter ansatz. In particular, Geo-ADAPT achieves up to 100-fold reduction in energy error compared to existing methods.

Timing degradation in SRAM-based FPGAs arises from multiple physical mechanisms that manifest differently in the routing fabric, most notably power-distribution-network (PDN) marginality and configuration-induced routing perturbations. Existing in-situ timing monitors provide limited insight into the physical origin, spatial structure, or statistical characteristics of the degradation. This paper presents a scalable in-situ timing diagnosis architecture that enables fine-grained, routing-aware characterization of timing behavior directly within the FPGA fabric during normal operation. The proposed approach combines non-intrusive delay taps placed at routing switch-matrix boundaries with distributed phase-swept delay monitoring elements and centralized statistical analysis. By extracting probabilistic delay distributions rather than binary timing margins, the framework captures both mean delay shifts and timing variability across spatially distributed routing locations. Experimental results obtained on a modern SRAM-based FPGA show that PDN-induced timing degradation produces globally correlated delay shifts with minimal change in variance, whereas routing-induced perturbations exhibit localized, topology-dependent delay growth and increased timing dispersion. Spatial correlation analysis and two-dimensional correlation heatmaps further reveal distinct signatures that enable systematic differentiation between these mechanisms. The presented architecture operates concurrently with an active user design and does not require external instrumentation, radiation sources, or design modification. These results establish a practical foundation for in-situ timing diagnosis, reliability assessment, and architecture-aware timing management in large FPGA-based systems.

The microscopic origin of high-temperature superconductivity in cuprates remains one of the central open questions in condensed matter physics. Growing experimental and theoretical evidence suggests that the bare single-band Fermi-Hubbard model may not fully capture properties of cuprates such as superconductivity, motivating us to revisit the canonical three-band model of the copper-oxide planes - the Emery model - from which the single-band counterpart was originally derived. Here, we propose and analyze a quantum simulation scheme for realizing the Emery model in regimes relevant to cuprates and infinite-layer nickelates with today's ultracold atom quantum simulation platforms, enabling the exploration of the three-band physics on system sizes that are challenging for current numerical methods. Specifically, we show that a two-dimensional optical lattice with a superimposed pattern of repulsive potentials can be designed to study low-temperature properties for variable parameter regimes of the Emery model relevant to cuprates as well as infinite-layer nickelates. Our results pave the way for real material simulations with ultracold atom quantum simulators and a better understanding of the physics of unconventional superconductors.

We study quantum-to-classical correspondence of the Krylov space for evolutions driven by unitary maps with a classical limit. This entails a proper definition of corresponding quantum and classical operators, inner products and initial states. We prove that with these definitions the purely classical Krylov space is indeed obtained as the asymptotic $\hbar\to 0$ expansion of the quantum Krylov space, and provide several examples of such correspondence. We use these examples to analyze some general aspects about the evolution of the Krylov complexity as they relate to the phase-space representation for the Krylov states. Additionally, we discuss alternative definitions to obtain the correspondence and why they fail. This paper constitutes a first step in understanding complexity and ergodicity of unitary evolution through the Krylov perspective as they relate to classical dynamical notions.

Single-qubit gates on superconducting quantum processors are typically implemented using microwave pulses applied through dedicated control lines. However, these microwave pulses may also drive other qubits due to crosstalk arising from capacitive coupling and wavefunction overlap in systems with closely spaced transition frequencies. Crosstalk and frequency crowding increase errors during simultaneous single-qubit operations relative to isolated gates, thus forming a major bottleneck for scaling superconducting quantum processors. In this work, we combine model-based qubit frequency optimization with pulse shaping to demonstrate crosstalk error mitigation in single-qubit gates on a 49-qubit superconducting quantum processor. We introduce and experimentally verify an analytical model of simultaneous single-qubit gate error caused by microwave crosstalk that depends on a given pulse shape. By employing a model-based optimization strategy of qubit frequencies, we minimize the crosstalk-induced error across the processor and achieve a mean simultaneous single-qubit gate fidelity of 99.96% for a 16-ns gate duration, approaching the mean individual gate fidelity. To further reduce the simultaneous error and required qubit frequency bandwidth on high-crosstalk qubit pairs, we introduce a crosstalk transition suppression (CTS) pulse shaping technique that minimizes the spectral energy around transitions inducing leakage and crosstalk errors. Finally, we combine CTS with model-based frequency optimization across the device and experimentally show a systematic reduction in the required qubit frequency bandwidth for high-fidelity simultaneous gates, supported by simulations of systems with up to 1000 qubits. By alleviating constraints on qubit frequency bandwidth for parallel single-qubit operations, this work represents an important step for scaling towards larger quantum processors.

Recent work on the instantaneous quantum polynomial-time (IQP) quantum-circuit Born machine (QCBM) highlights a promising paradigm for generative modeling: train classically, deploy quantumly. In this setting, the training objective can be evaluated efficiently on a classical computer, while sampling from the resulting model may still be classically intractable. Furthermore, in the IQP-QCBM framework, extending the model family with ancillary qubits has been proven to yield universality. This paper asks whether similar results hold for linear-optical generative models. To this end, we introduce the Boson Sampling Born Machine (BSBM). Our analysis retraces analogous steps as were found for IQP-QCBMs with twists. Using recent results that enable classical approximation of broad classes of expectation values in linear optics, we show that BSBMs can be trained classically for wide families of loss functions. Next, we argue that "basic" BSBMs are not universal generative models, and that universality can be achieved by expanding the model while preserving efficient classical training and sampling hardness. In our approach, we introduce and analyze the role of constant-function postprocessing, generalizing the construction for IQP-QCBMs, which under suitable conditions can lead to universality while preserving the hardness of classically simulating the models. We showcase a family of BSBMs, characterized by a single hyperparameter, that allows for a monotonic increase in expressivity toward universality while retaining the capacity to represent ostensibly hard distributions. Furthermore, we discuss the possible modalities for the efficient classical training, in the sense of efficient estimation of gradients of the loss function.

Thermal correlators in holographic conformal field theories are known to exhibit singularities in complex time, sometimes referred to as ``bouncing singularities", which are believed to be related to bulk geodesics probing the black hole interior. These singularities correspond to exponentially suppressed contributions in the high-frequency limit of the thermal correlators. We revisit in detail the calculation of retarded two-point functions of local operators dual to bulk scalar fields in the planar AdS black hole background. We confirm that these correlators develop bouncing singularities, and highlight the agreement of two independent methods: a large frequency WKB analysis with infalling boundary conditions at the horizon; and an asymptotic OPE analysis that relies only on the near-boundary expansion, without any direct input from the black hole interior. We then extend these calculations to the case of the retarded two-point function of displacement operators on a Wilson line in the finite temperature gauge theory. This is computed holographically by solving the wave equation for the transverse fluctuations of the dual string worldsheet in the planar AdS black hole background. We find that these defect correlators also exhibit bouncing singularities, and again observe exact agreement between the WKB analysis sensitive to the black hole interior and the asymptotic OPE analysis. This agreement suggests that the bouncing singularities and the corresponding OPE data encode a universal high-frequency structure of the retarded correlators, and we propose a factorization formula that encodes the deviations from this universality.

The definition of the phase space of field theories in presence of boundaries of Cauchy surfaces requires a choice of boundary conditions or decay behaviour of those fields. Often these conditions are motivated in part by the decay behaviour of the initial data of known exact solutions. In the case of gauge field theories the initial data are not free but are subject to initial value constraints. Still, the decay behaviour is commonly specified for the kinematical, i.e. unconstrained phase space. This can lead to the following practical problem: The constraints are preferably solved for field variables on which they depend only algebraically, i.e. not involving derivatives, as otherwise one would need to solve partial differential equations. However, the specified decay behaviour may prevent from doing that. On the other hand, a precise specification of decay for all kinematical fields appears unnecessary because the decay of gauge degrees of freedom is not observable. Yet, knowledge of their decay is required as one needs to compute Poisson brackets on the kinematical phase space in order to define what gauge invariance means. Thus the interplay between the constraint structure and the decay properties of the kinematical phase space is complex. In this contribution we develop a reduced phase space induced approach to the decay problem. Upon specifying gauge conditions tailored to the algebraic structure of the constraints, these define a split of the kinematical phase space into gauge and true degrees of freedom. Then the decay conditions of the kinematical phase space is systematically parametrised by a choice of decay for just the true degrees of freedom (i.e. the reduced phase space), the decay of the gauge degrees of freedom then follows unambiguously from solving both the constraints and the gauge conditions.

No-boundary instantons and Euclidean "wineglass" wormholes have both been proposed as providing suitable initial conditions for the current expanding phase of our universe, and in particular for providing conditions that are favorable to an inflationary phase. These finite action solutions have generally been regarded as unrelated, and enacting different scenarios - in one case the creation of spacetime from nothing, and in the other up-tunneling from a Euclidean Anti-de Sitter vacuum. By studying explicit solutions of both axionic and magnetic wineglass wormholes, we find that in the zero-charge limit the throat of the wormholes pinches off, leaving a no-boundary instanton that disconnects from the asymptotic Anti-de Sitter region. Thus wormholes and no-boundary instantons are part of a common family of Euclidean solutions. Along the way, we resolve the long-known puzzle that the action of wineglass wormholes can become negative. Moreover, small-charge wormholes lead to a longer inflationary phase than large-charge solutions, while no-boundary instantons dominate the probability distribution overall.

We investigate Permutation-Invariant (PI) quantum error-correcting codes encoding a logical qudit of dimension $\mathrm{d}_\mathrm{L}$ in PI states using physical qudits of dimension $\mathrm{d}_\mathrm{P}$. We extend the Knill--Laflamme (KL) conditions for $d-1$ deletion errors from qubits to qudits and investigate numerically both qubit ($\mathrm{d}_\mathrm{L} = \mathrm{d}_\mathrm{P} = 2$) and qudit ($\mathrm{d}_\mathrm{L} > 2$ or $\mathrm{d}_\mathrm{P} > 2$) PI codes. We analyze the scaling of the block length $n$ in terms of the code distance $d$, and compare to existing families of PI codes due to Ouyang, Aydin--Alekseyev--Barg (AAB) and Pollatsek--Ruskai (PR). Our three main findings are: (i) We conjecture that qubit PI codes correcting up to $d-1$ deletion errors have block length $n(d) \geq (3d^2 + 1) / 4$, which implies an upper bound $d \leq \sqrt{12n-3}/3$ on their code distance, and that PR codes can saturate this bound. (ii) For qudit PI codes encoding a single qudit we numerically observe that increasing $\mathrm{d}_\mathrm{P}$ results in $n$ monotonically decreasing and approaching the quantum Singleton bound $n(d) \geq 2d-1$. (iii) We propose a semi-analytic extension of the qubit AAB construction to qudits that finds explicit solutions by solving a linear program. Our results therefore provide key insights into lower bounds on the block length scaling of both qubit and qudit PI codes, and demonstrate the benefit of increased physical local dimension in the context of PI codes.

We formulate an extended holographic dark energy scenario based on a recently proposed two-parameter generalized entropic functional. Unlike constructions that phenomenologically impose modified entropy-area relations at the horizon level, the present framework is rooted in a microscopic entropy functional and the corresponding microstate counting. For bounded systems, the entropy acquires a generalized holographic scaling with two independent area contributions, recovering the Bekenstein-Hawking entropy in the appropriate limits. Implementing this entropy within the holographic principle, we derive a generalized dark energy density containing two distinct holographic sectors, naturally embedding standard holographic dark energy and $Λ$CDM as limiting cases. We analyze the cosmological evolution for both Hubble and future event horizon cutoffs and show that the model successfully reproduces the matter-to-dark-energy transition. The two entropic exponents enrich the dynamics, allowing for quintessence-like behavior or phantom regimes, while remaining compatible with the standard thermal history of the Universe.

In quantum computations of gauge theories at finite temperature and finite density, it is challenging to enforce Gauss's law for all states contributing to the thermal ensemble. While various techniques for implementing gauge constraints have been proposed, they often involve practical trade-offs. In this work, we adopt the Quantum Minimally Entangled Typical Thermal States (QMETTS) algorithm for $\mathbb{Z}_2$ gauge-constrained systems, which allows us to capture thermal equilibrium states with chemical potential while mitigating these trade-offs. To ensure that gauge invariance is preserved throughout the procedure while maintaining computational efficiency, we derive the specific measurement bases within the algorithm. Furthermore, since the estimation of expectation values on quantum hardware is inherently noisy, we rigorously account for shot noise in estimating expectation values, and propose a sampling method that is more efficient than those in previous works. We validate our approach numerically by studying a (1+1)-dimensional $\mathbb{Z}_2$ gauge theory coupled to staggered fermions. Our proposed algorithm reproduces the correct equilibrium states at finite temperature and finite density.

Time-slicing has emerged as a strategy for incorporating semiclassical propagation into real-time path integral formulation and recovering full quantum mechanical dynamics. A central step is the decomposition of a time-evolved wave function into a superposition of Gaussian wave packets. Here we introduce a quadrature-free variational framework for Gaussian wave packet decomposition, reformulating it as an optimization problem in which the parameters of Gaussian wave packets are chosen to maximize the overlap with the time-evolving wave function. An autoencoder-decoder neural network is used for this optimization, with the representation being adaptively reoptimized during propagation. Each wave packet in this decomposition represents a localized patch of the underlying semiclassical manifold, while retaining full correlations between all degrees of freedom. This variational adaptive Gaussian decomposition (VAGD) approach yields a compact Gaussian expansion, providing a scalable route to time-sliced semiclassical quantum dynamics. While general, applying VAGD to facilitate time-slicing of thawed Gaussian approximation (TGA) simulation allows a route to improving the semiclassical result to the full quantum mechanical result in a systematic manner.

The influence of chiral molecular potentials on phase-coherent transport in superconducting Josephson junctions is investigated. Within a Bogoliubov-de Gennes tight-binding framework, an SNS junction functionalized by adsorbed chiral molecules is modeled, where electrostatic gradients generated by the molecules induce spin-orbit coupling in the normal region. The equilibrium charge current-phase relation is found to remain largely insensitive to molecular chirality in symmetric, zero-field configurations. In contrast, the spin supercurrent exhibits a pronounced chirality-dependent response, with opposite enantiomers producing distinct and anisotropic spin-polarized Josephson currents. The resulting handedness contrast can be enhanced through control parameters such as molecular orientation and the strength of the induced spin-orbit coupling. The temperature dependence of these currents further shows that the chirality-dependent signatures persist across a range of temperatures well below the superconducting critical temperature. These results establish Josephson interferometry as a phase-sensitive and accessible platform for detecting molecular chirality and highlight spin-polarized superconducting transport as a promising route toward integrating chiral molecular functionality into superconducting spintronic devices.

Quantum hypergraph states extend the well-studied class of graph states by taking into account multi-qubit interactions through hyperedges. They provide a powerful framework to represent a family of quantum states with genuine multipartite entanglement. In this review, we provide a compact overview of the formal structure, entanglement characteristics, and operational relevance of hypergraph states in quantum information theory. We begin by introducing their mathematical foundations and generalizations of the stabilizer formalism. A central focus is placed on their entanglement properties, including the classification under local unitary (LU) and stochastic local operations with classical communication (SLOCC), the quantification of multipartite entanglement, and detection techniques via entanglement witnesses. We also explore other nonclassical features of hypergraph states, such as contextuality and genuine multipartite nonlocality, derived from stabilizer-based Bell-type inequalities. Additional attention is given to the role of hypergraph states in error correction, and as a computational resource in measurement-based quantum computation (MBQC), and to their non-stabilizer character - quantified via resource-theoretic measures of quantum magic. Finally we review their generalization to higher dimensions, i.e. to qudits and continuous variables.

In the most general formulation of gravity, the metric and connection are independent degrees of freedom, and the connection may include torsion and non-metricity (or distortion, collectively) degrees of freedom, resulting in a huge number of possible dynamical fields. However, the most fields are either non-dynamical or extremely heavy and the general relativity is recovered at low energy. We find a unique naturally light vector- or scalar-like distortion field, which can be dynamical and have phenomenological implications. In particular, a light scalar particle that mixes with the Higgs boson naturally appears.

Quantum telepathy is the concept of using quantum entanglement to solve real-world problems involving decision coordination between parties with restricted communication. One possible reason for this restriction is a latency constraint: some pairs of parties do not have enough time to communicate with each other before they have to produce their outputs. Example scenarios include high frequency trading and distributed systems. Another reason is isolation: for some pairs of parties, there is an obstacle to communication. Example scenarios include locating a stray traveler by a rescue team and coordination within a network where nodes are owned by competing firms. In this paper we give a concise overview of the different application areas of quantum telepathy. We find that these real-world problems can be modeled as nonlocal games or its generalizations. We also discuss possible physical implementations. Quantum telepathy guarantees a quantum advantage via Bell's theorem and can directly solve real-world problems, such as reducing risk in high frequency trading or balancing data loads efficiently in ad hoc networks. Moreover, this quantum advantage can be physically realized with existing NISQ hardware.

This project is aimed at studying the first-order phase transitions, that is presumed to have ensued in the early universe, and its consequences on the primordial gravitational waves. The effects of bubble nucleation, growth, and coalescence are reviewed. The resulting first-order phase transition is taken as the source of the gravitational waves that were produced, in order to determine the energy density, amplitude, and frequency spectra of the relic gravitational wave background. This is accomplished by modelling the first-order phase transition as a turbulent fluid and employing relativistic hydrodynamic equations to estimate the required physical quantities. Two models are majorly studied for all the analysis done in this project. Both models compute the necessary gravitational wave spectra using the exponential Kraichnan function as the temporal decorrelation function. Also, both models contemplate the turbulence in the flow of plasma to be stationary, obeying the conditions dictated by the Kolmogorov turbulence. However, the first model uses a de-coherence function that depends on the wavenumber and time, while, the second model uses the top hat correlation, to compute the anisotropic stress. The new model introduced here adheres to the freely decaying turbulence model, but employs the time dependent de-coherence function in its computations.

We investigate massive scalar perturbations and several characteristics of particle motion in the spacetime of regular black holes arising in four-dimensional quasi-topological gravity. Quasinormal modes are computed using high-order WKB approximations with Padé resummation and verified through time-domain integration. For moderate values of the scalar-field mass, the time-domain profiles confirm the WKB results with excellent accuracy. As the mass increases, the damping rate decreases substantially, indicating the approach to the quasi-resonant regime of long-lived modes. For sufficiently large masses, the late-time signal becomes dominated by oscillatory power-law tails, which mask the quasi-resonant mode in the time-domain profile. In addition, we analyze photon motion and circular geodesics, including the photon-sphere radius, shadow size, Lyapunov exponent, and ISCO characteristics. These quantities exhibit only moderate deviations from their Schwarzschild values, unlike the Hawking temperature of the black hole.

Efficient entanglement distillation is a central task in quantum information science and future quantum networks. At the core of distillation protocols are the quantum error correction and detection schemes which enhance the fidelity of entangled pairs. Conventional protocols focus on digital systems, which typically require complicated compiled circuits, high-fidelity multi-qubit operations and delicate pulse-level control that impose high demands on near-term hardware. Crucially, the leading physical platforms for quantum networks, trapped ions and neutral atoms, are governed by native many-body Hamiltonians inherently suited for analog, continuous-time evolution. Adopting these natural dynamics is simpler than engineering digital logic via delicate pulse-level control. Motivated by this experimental reality, we seek to leverage the intrinsic analog capabilities for efficient entanglement distillation. In this work, we introduce the Hamiltonian entanglement distillation protocol, which exploits the intrinsic information scrambling generated by random time evolution under native Hamiltonians. We establish a quantitative connection between output fidelity and Out-of-Time-Order Correlators, showing that efficient scrambling directly implies good distillation performance. Since generic Hamiltonians are naturally efficient scramblers, the capability for distillation is ubiquitous: almost all Hamiltonians in the Hilbert space suffice for high-fidelity distillation. Numerical simulations of representative Rydberg-atom and trapped-ion systems further confirm that robust performance could be achieved using only short-range interactions and evolution times feasible in current experiments. By avoiding the complexity of digital circuit control, our approach substantially relaxes experimental requirements, providing a scalable route to entanglement engineering on current analog quantum platforms.

We present a scheme for implementing a high-fidelity non-linear phase shift on a photonic state. The scheme is based on repeated scattering off a two-level quantum emitter embedded in a chiral or one-sided waveguide. The waveguide is equipped with elements inducing second-order dispersion and temporal phase shifts, which effectively form a harmonic trap and confine the photon pulses to a Gaussian shape. The same quantum emitter can be used for each scattering, and thus, only one quantum emitter is needed in this scheme. To illustrate the application of our scheme for photonic quantum computing and quantum communication, we analyze the implementation of a control-Z gate and a deterministic Bell-state analyzer for photonic qubits. Through numerical optimization, we show that we can reach a control-Z gate fidelity of $\mathcal{F} \sim 99.2\%$ ($\mathcal{F} \sim 96\%$) and a success probability of $P_s \sim 99.6 \%$ ($P_s\sim 98 \%$) for a Bell-state measurement with $N=17$ ($N=5$) scatterings.

The recent release of high-precision cosmological data, particularly the small-scale cosmic microwave background (CMB) measurements from ACT and baryon acoustic oscillation (BAO) data from DESI, has opened a new landscape for probing the neutrino mass. In this work, we present updated constraints on the total neutrino mass, $\sum m_ν$, and its hierarchy within the $Λ$CDM, $w$CDM, holographic dark energy (HDE), and $w_0w_a$CDM models, using the latest ACT DR6, DESI DR2, and DESY5 datasets. We find that the upper limits on $\sum m_ν$ are critically governed by the evolutionary behavior of the dark energy equation of state. Specifically, models exhibiting early-time quintessence features (e.g., HDE) yield the most stringent constraints, whereas those allowing for early-time phantom behavior (e.g., $w_0w_a$CDM) result in significantly looser bounds. Despite these model-dependent variations, we observe a robust hierarchy dependence across all scenarios, where the inverted hierarchy consistently yields weaker constraints and the degenerate hierarchy consistently yields tightest constraints. Our analysis demonstrates that the improved small-scale CMB information from ACT, combined with high-precision BAO data, systematically tightens the limits on $\sum m_ν$, providing a crucial benchmark for future neutrino mass measurement.

The differential scalar polarizability $Δα_0(ω)$ of the Ba$^+$ S$_{1/2}$-to-D$_{5/2}$ clock transition has a zero crossing near 481nm, which is measured to be 623.603\,13(17)\,THz. From this measurement, we infer a ratio of reduced matrix elements $\langle P_{3/2}\|r\|S_{1/2}\rangle/\langle P_{1/2}\|r\|S_{1/2}\rangle=1.411\,81(13)$, which provides a stringent test of atomic structure calculations and experimental determination of matrix elements. Additionally, it enables the construction of an accurate approximation to $Δα_0(ω)$, valid for frequencies up to 450\,THz, with only one reduced matrix element, $\langle P_{1/2}\|r\|S_{1/2}\rangle$, appearing in the model's parameterization. We discuss the achievable accuracy of the model, the application to the assessment of blackbody radiation (BBR) shifts in ion-based clocks, and the applicability of the approach to other alkaline-earth ions.

Energy extraction from compact objects has been a central topic in general relativity since the introduction of the Penrose process. In this work we present a unified analysis of rotational and electromagnetic energy extraction in Kerr, Reissner-Nordstrom, and Kerr-Newman spacetimes. Using particle energetics and the irreducible mass formalism, we compare the efficiencies of these mechanisms and examine their consequences for horizonless objects. While purely rotational extraction in Kerr spacetime is fundamentally limited by geometric constraints, electromagnetic interactions enlarge the region of negative energy orbits through an effective ergoregion, allowing significantly higher efficiencies. In Kerr-Newman geometry, the combined effect of rotation and charge further enhances the extractable energy. We then study the long-term evolution of over-extremal cases under continuous extraction. By deriving coupled evolution equations for the mass, spin, and charge parameters, we show that continuous extraction can gradually drive a naked singularity toward the extremal bound. For astrophysically realistic luminosities, the characteristic evolution timescale is of order about 10^9 years. These results suggest that energy extraction provides an energetic indication of instability in Reissner-Nordstrom, Kerr, and Kerr-Newman naked singularities and may lead to horizon formation as a long-term stabilizing outcome.

We propose a hybrid photonic quantum reservoir computing (QRC) framework for swaption surface prediction. The pipeline compresses 224-dimensional surfaces to a 20-dimensional latent space via a sparse denoising autoencoder, extracts 1,215 Fock-basis features from an ensemble of three fixed photonic reservoirs, concatenates them with a 120-dimensional classical context, and maps the resulting 1,335-dimensional feature vector to predictions with Ridge regression. We benchmark against 10 classical and quantum baselines on six held-out trading days. Our approach achieves the lowest surface RMSE of~$0.0425$ while maintaining sub-millisecond inference. The quantum layer has zero trainable parameters, sidestepping barren plateaus entirely. Variational quantum methods (VQC, Quantum LSTM) yield negative $R^{2}$ on test data, confirming that fixed quantum feature extractors paired with regularised readouts are more viable for low-data financial applications.

Increasing connectivity and decreasing qubit-state delocalization without compromising the speed and accuracy of elementary gate operations are topical challenges in the development of large-scale superconducting quantum computers. In this theoretical work, we study a special honeycomb qubit lattice where each qubit inside a unit cell is coupled to every other one via two dedicated tunable couplers and a common central element. This results in an effective multi-mode interaction enabling tunable, on-demand, all-to-all connectivity between each qubit pair within the unit cell. We provide a thorough analysis of the unit cell, including a proposal for a novel and efficient conditional-Z gate scheme which takes advantage of the effective multi-mode coupling. We develop an experimentally viable pulse protocol for a single-step gate implementation which considerably improves the gate speed compared to the previous two-qubit-gate realizations suggested for architectures utilizing a center mode. We also show numerical results on how the presence of spectator qubits affects the average two-qubit-gate fidelity, and analyse how the multi-mode coupling structure mitigates the delocalization-induced crosstalk during simultaneous single-qubit gates within the unit cell. We also provide analytical estimates for the errors caused by relaxation and dephasing during a two-qubit-gate operation, including noise terms for the multi-mode coupling structure. Our multi-mode coupling architecture results in a good balance between increased connectivity and available parallelism, especially when several interacting unit cells form a quantum processing unit. We anticipate that the obtained results pave the way towards high-connectivity quantum processors with efficient and low-overhead quantum algorithms.

Efficient optomechanical cooling typically requires high photon occupancy to maximize cooling power, a constraint that generally limits the degree of coherent quantum control available in the few-photon regime. Here, we investigate this trade-off by considering a strongly nonlinear cavity operated as a discrete quantum system. In the weak-coupling limit, we derive a general connection between the optomechanical damping rate and the cavity's dressed-state manifold. This framework reveals that the damping rate (determined by the population imbalance across dressed states) is directly tunable via the coherent manipulation tools which are standard in circuit quantum electrodynamics. We illustrate this framework using a Josephson photonics architecture, where a dc-biased junction induces a photon blockade that truncates the cavity to an $N$-level system. By sacrificing raw cooling (or heating) power, this platform enables full quantum mechanical control over optomechanical properties, offering a versatile avenue for the quantum manipulation of mechanical modes.

Self-testing--the attractive possibility to infer the underlying physics of a quantum device in a black-box scenario--has gained increased traction in recent years, with applications to device-independent quantum information processing. Thus far, self-testing has been done under the assumption that the settings for the requisite Bell test are chosen freely and independently of the device tested in the experiment. That is, the random number generator used to generate the settings has been assumed to have no correlations with the device tested. Here, we extend self-testing protocols beyond the independence assumption. Surprisingly, we show that all pure bipartite partially entangled states can be self-tested provided that the random number generator obeys a residual randomness constraint strictly weaker than the independence assumption. This in itself provides a semi-device-independent certification of independence between the randomness source and the device.

We introduce a new class of interacting dark sector models that couple the intrinsic entropy of dark matter to scalar field dark energy. Using the Lagrangian formulation for relativistic perfect fluids, we construct consistent covariant actions that incorporate algebraic and derivative entropy couplings. These interactions leave the expansion history unchanged, rendering the background cosmology indistinguishable from $Λ$CDM or uncoupled quintessence. At the level of cosmological perturbations, the entropy couplings generate scale-dependent modifications to the dark matter Euler equation, while the continuity equation remains unaltered at linear order. The resulting interactions correspond to a pure-momentum exchange within the dark sector. We show that intrinsic entropy perturbations can carry primordial scale dependence, and non-minimal couplings can lead to a scale-dependent suppression or enhancement of structure growth. Finally, we demonstrate that these models are generically compatible with current Cosmic Microwave Background observations, while inducing distinctive signatures in large-scale structure. The framework provides a theoretically well-motivated and observationally viable extension to the standard cosmological model, opening new directions to explore novel interactions in the dark sector.

In this paper I argue against Carcassi, Calderon, and Aidala's recent claim that the Hilbert spaces are unphysical and should be replaced with the Schwartz spaces in quantum mechanics, since Hilbert spaces include states with infinite expectation values for certain observables. I also review and discuss issues regarding unbounded operators in quantum mechanics raised by Streater and Wightman, Heathcote, and Lemos. I argue that the existence of infinite expectation values does not cause problems in quantum mechanics. On the other hand, replacing the Hilbert spaces with the Schwartz spaces would cause more issues, as it would exclude a class of meaningful Hamiltonian evolutions. I also discuss the question in literature whether reformulating quantum mechanics with essentially self-adjoint operators instead of self-adjoint operators may cause problems. I further analyse the hierarchies of the notions of "physicality" and possibility in fundamental physics, and suggest that "physicality" is a vague concept. Finally, I connect the problem raised by Carcassi, Calderon, and Aidala with the problem of the Hadamard condition in quantum field theory.

We develop a quantum teleportation protocol of an unknown optical single rail qubit using a hybrid quantum channel composed of continuous variable (CV) states of certain parity. The quantum channel is characterized by two parameters: a squeezing parameter of single-mode squeezed vacuum (SMSV) state and the beam splitter (BS) parameter used to implement it. The CV part of the hybrid state belongs to Alice, while discrete variable (DV) half is controlled by Bob. The third parameter of the protocol is a parameter of the beam splitter, used to mix the CV components of the hybrid quantum state with unknown optical single-rail qubit. Even though the number of measurement results Alice sends may increase, Bob can obtain the original qubit half the time with an appropriate choice of parameter values. In almost half the remaining cases, Bob obtains the original qubit with distorted amplitudes, and both participants know the value of the distortion factors. This means that as the amount of classical information transmitted by Alice increases, they both gain greater access to partial information about the unitary transformations that the teleported qubits undergo, allowing Bob to continue using them or attempt to recover them to improve the protocol's efficiency. The proposed method is a generalization of quantum teleportation with a nonlocal photon used as a quantum channel and unknown single-rail optical qubit.

We propose two schemes to achieve remote entanglement distribution between two mechanical nodes with a significant frequency mismatch, based on optomechanical systems. The first scheme utilizes the physical mechanism to redistribute the quantum entanglement initially established in a dispersively-coupled optomechanical system with a megahertz mechanical resonance to a distant optomechanical system which embodies the tripleresonant interaction induced by the scattering of gigahertz mechanical phonon. We also provide a fast optical pulse protocol to realize the long-distance entanglement distribution from the optomechanical system supporting the gigahertz mechanical mode to the megahertz mechanical mode included in a distant optomechanical system. Specifically, these two schemes respectively demonstrate the megahertz-to-gigahertz and gigahertz-tomegahertz entanglement distribution in the quantum network of optical photons and phonons. This work may facilitate the application of various mechanical systems in hybrid quantum network-based quantum technologies and fundamental physical research.

Electric-dipole forbidden spectroscopic transitions in atoms form the basis of many advanced implementations of quantum computers, atomic clocks and quantum sensors. Coherently addressing such transitions in molecules which are among the most ubiquitous and versatile quantum objects has remained a long-standing challenge owing to their complex energy-level structure. Here, we report the search, observation and coherent manipulation of electric-quadrupole rotational-vibrational transitions in single trapped molecules using a quantum-logic-spectroscopy protocol. We identified individual hyperfine-Zeeman-rotational components of the fundamental vibrational transition of the nitrogen molecular ion, N$_2^+$, and performed coherent population transfer between energy levels. Our work opens up new perspectives for precision molecular spectroscopy, for high-fidelity qubits encoded in the rotational-vibrational motion of molecules, for precise infrared molecular clocks and for searches for new physics

Electronegativity is a cornerstone of chemical intuition, essential for rationalizing bonding, reactivity, and material properties. However, prevailing scales remain empirically derived, often relying on parameterized models or composite physical quantities. In this work, we introduce a universal electronegativity scale founded on the atomic mean inner potential (AMIP), also known as the average Coulomb potential, a fundamental, quantum-mechanical property accessible through both first-principles computation and electron-scattering experiments. Our scale, denoted $χ_{\mathrm{AMIP},p}$, is an analytic function of just three ground-state atomic descriptors and carries explicit physical units. It demonstrates excellent agreement with established scales and successfully classifies bonding types across 358 compounds, including adherence to the metalloid ``Si rule". Beyond replicating known trends, $χ_{\mathrm{AMIP,1/2}}$ proves to be a powerful predictive tool, accurately determining Lewis acid strengths for over 14,000 coordination environments ($R^2=0.93$) and $γ$-ray annihilation spectral widths for 36 elements ($R^2=0.97$), outperforming previous methods. By linking electronegativity directly to a measurable quantum property, this work provides a unified and predictive descriptor for electronic structure and chemical behavior across the periodic table.

Airfoil shape optimization presents a challenge where classical solvers frequently struggle with computational efficiency and local minima. In the promising paradigm of quantum computing, the coherent Ising machine (CIM), a specialized physical solver, offers acceleration capabilities. However, its native discrete binary architecture restricts the application in aerodynamic design. To bridge this gap, we propose a comprehensive framework that translates airfoil shape optimization into hardware-compliant quadratic unconstrained binary optimization formulations. We integrate high-order response surface models via the Rosenberg order reduction, enabling the CIM to capture strong nonlinearities in the aerodynamic performance response. Furthermore, we introduce a block-diagonal scalarization strategy that compose trade-off scenarios into a single optimization. Validated on the NACA 4-digit airfoil series using CIM hardware with 615 spins, the framework successfully locates the global optimum with a computational speedup of three orders of magnitude compared to the classical simulated annealing. The parallel embedding capacity allows for the extraction of an entire optimal Pareto front in a single hardware execution. This work demonstrates a viable, quantum-enhanced paradigm for engineering optimization.

Quantum communication between distant superconducting qubits on separate chips using itinerant microwave photons has been studied to realize distributed quantum information processing. To enhance information capacity and fault tolerance in quantum networks, it is beneficial to encode a large quantity of quantum information using auxiliary degrees of freedom of these photons. In this work, we experimentally investigate the use of temporal modes of photon wave packets. Through the photon-shaping technique with a fixed-frequency transmon qubit, we generate single microwave photons in four orthogonal temporal modes propagating along a waveguide. We demonstrate mode-selective absorption across orthogonal modes via the time-reversed process of emission, achieving absorption efficiencies exceeding 0.89 for mode-matched cases, while remaining below 0.13 for orthogonal modes. Photons rejected by a given receiver mode can remain mutually orthogonal, enabling selective absorption at subsequent receivers in future multi-node architectures. These results highlight the feasibility of temporal-mode engineering for constructing a higher-dimensional orthogonal basis for multiplexed quantum networks.

Continuous-variable quantum key distribution (CV-QKD) is a promising quantum-safe alternative to classical asymmetric cryptography that enables two authenticated parties to establish a shared secret over a potentially eavesdropped quantum channel. A key step in CV-QKD post-processing is information reconciliation, which leverages forward error correction (FEC) techniques to extract identical bit strings from noisy correlated data. In this work, we analyze the strict limitations on operating distance that are imposed by constant-rate FEC, severely limiting the practicability of CV-QKD systems in deployed optical networks. To overcome the distance limitations, we evaluate three strategies: (i) tuning modulation variance, (ii) adding controlled amounts of trusted detector loss, and (iii) the use of rate-adaptive FEC. All approaches are validated experimentally, compared in terms of performance, and we discuss implementation aspects. Our results show that while methods (i) and (ii) extend the operational distance of constant-rate FEC without the need for additional hardware components, they incur a significant penalty in secret key rate (SKR). In contrast, rate-adaptive FEC enables CV-QKD operation with performance close to the asymptotic SKR over a wide range of distances, provided that the reconciliation efficiency is chosen appropriately.

The Gauss-Bonnet (GB) scalarization for charged quantum Oppenheimer-Snyder (cqOS)-black holes is investigated in the Einstein-Gauss-Bonnet-scalar theory with the nonlinear electrodynamics (NED) term. Here, the scalar coupling function to GB term is given by $f(φ)=2λφ^2$ with a coupling constant $λ$. Three parameters of mass ($M$), action parameter ($α$), and magnetic charge ($P$) are necessary to describe the cqOS-black hole, and it may become the qOS-black hole when $P=M$. The GB scalarization of cqOS-black holes comes into two cases GB$^\pm$, depending on the sign of GB term which triggers the different phenomena. For $α=0$ and $λ>0$, GB$^+$ scalarization is allowed, while for $α\not=0$ and $λ<0$, GB$^-$ scalarization appears for a narrow band of $3.5653\le α\le 4.6875$. After discussing the onset GB$^-$ scalarization, we construct scalarized qcOS-black holes which belong to the single branch. The scalar field is nonmonotonic near the horizon while it asymptotes to a finite value at infinity, indicating a distinct scalarization mechanism for negative coupling $λ$. Stability analysis shows these scalarized black holes are linearly stable under scalar perturbations.

We investigate, how finite temperature influences quantum coherence in multipartite open systems by analyzing a tripartite spin boson model subjected to non-Markovian dephasing. Two distinct environmental configurations are considered viz. independent local reservoir and a common structured reservoir characterized by an Ohmic spectral density. In this framework, temperature enters explicitly through the time dependent dephasing rates, enabling a systematic exploration of thermal effects on coherence dynamics. Using the relative entropy of coherence, we examine representative pure states belonging to inequivalent entanglement classes along with physically relevant mixed states constructed from them. Under local non-Markovian dephasing, all states exhibit monotonic coherence decay, with temperature acting as a universal accelerator of decoherence. In contrast, the common reservoir scenario reveals a strikingly non-universal behaviour. While $GHZ$ and $Star$ type states undergo temperature enhanced degradation, $W$ class states and certain Werner type mixtures display robust stationary coherence that remains largely insensitive to thermal fluctuations. These results demonstrate that the thermal susceptibility of coherence is governed not only by environmental configuration but also by the internal architecture of multipartite quantum states. The interplay between reservoir structure and state geometry leads to qualitatively distinct dynamical regimes ranging from rapid thermal fragility to temperature resilient coherence preservation. Our findings identify coherence dynamics as a sensitive probe of structured finite temperature environments and suggest a pathway toward coherence based quantum thermometry and nanoscale calorimetry using engineered multipartite states.

In this work, we scrutinize the thermodynamic properties of the Simpson-Visser (SV) spacetime. Working within Einstein gravity coupled to nonlinear electrodynamics (NLED) and a scalar field with negative kinetic energy, we rederive the solution in a formulation where the integration constants do not explicitly appear in the action, allowing them to vary consistently in the thermodynamic analysis. Using the Euclidean method, we show that the regular spacetime structure modifies the boundary contributions to the conserved charge associated with time translations, allowing the NLED sector to cancel the mass term and yielding a black hole with vanishing thermodynamic mass. Nevertheless, the spacetime admits a conserved magnetic charge and describes a regular black hole with a single horizon, finite temperature, and entropy, while the first law of thermodynamics holds in a modified form. We further compare this solution with the corresponding scalar-free singular black hole obtained when the regular parameter vanishes. Placing the two configurations in the same heat bath with identical temperature and magnetic chemical potential, we find that the SV regular black hole always has a larger free energy, indicating that the scalar-free singular configuration is thermodynamically preferred.

We derive and analytically prove a tight quantum speed limit (QSL) for ergotropy charging in the $N$-qubit Dicke quantum battery: the first-passage time to normalised ergotropy $ε$ satisfies $τ^{*}(ε) \geq \sqrt{Nε}/(2λ\sqrt{\bar{n}})$, where $λ$ is the coupling and $\bar{n}$ is the mean charger photon number. The bound follows from an exact perturbative identity $ε(t) = Aλ^2\bar{n}t^2 + \mathcal{O}((λt)^4)$, where $A=4/N$ is the short-time ergotropy coefficient, combined with a global upper bound proved analytically for all $N$. The composite parameter $Γ_N = 2λ\sqrt{\bar{n}/N}$ is the unique figure of merit for charging speed; all protocols collapse onto $Γ_N τ^{*} \geq \sqrtε$, with the bound saturated to within 1% at small $ε$.

We study a multi-field model in Loop Quantum Cosmology for a maximally symmetric spacetime governed by the Einstein--Hilbert action minimally coupled to scalar fields. Using a Legendre transformation, we formulate the Hamiltonian dynamics in canonically equivalent geometrodynamical and Yang--Mills--type representations, incorporating nontrivial couplings through a geometric structure on the multi-field configuration space. Implementing the $\barμ$-scheme polymerization, we obtain the loop-quantum-corrected Friedmann equations. By focusing on the two-field model as an example, we analyze the effective dynamics for specific potentials. The \textit{quantum bouncing, transition, and slow-roll inflationary} phases are investigated numerically, and viability of the models is assessed by evaluating the number of e-folds during the inflationary phase for certain given initial conditions. The global behavior of the background evolution is further examined through linear stability and dynamical-systems analyses.

In this paper, we give a complete description of the black hole threshold, locally near the Reissner-Nordström family, in the infinite-dimensional moduli space $\mathfrak M$ of dynamical spherically symmetric solutions to the Einstein-Maxwell-neutral scalar field system. In a neighborhood of the full Reissner-Nordström family in $\mathfrak M$, we prove the following: (i) Any solution that forms a black hole eventually decays to a Reissner-Nordström black hole. (ii) Any solution that fails to collapse into a black hole eventually becomes superextremal along null infinity and exists globally in the domain of dependence of the bifurcate characteristic initial data. (iii) The subset of this neighborhood consisting of black hole solutions admits a $C^1$ foliation by hypersurfaces of constant final charge-to-mass ratio, up to and including extremality. (iv) The mutual boundary between the set of black hole solutions and noncollapsing solutions, i.e., the black hole threshold, is the extremal leaf of the foliation. Black holes which are not on the threshold are asymptotically subextremal. Our quantitative control of near-threshold solutions allows us to prove "universal" scaling laws for the location of the event horizon and its final area and temperature (surface gravity), with critical exponent $\frac 12$. Moreover, we show that the celebrated Aretakis instability is activated for an open and dense set of threshold solutions and that generic near-threshold subextremal black holes experience a transient horizon instability on the timescale of their inverse final temperature.

Understanding the emergence of the thermodynamic arrow of time in microscopic systems is of fundamental importance, particularly given that unitary evolution preserves time-reversal symmetry. While projective measurements introduce temporal irreversibility, identifying this asymmetry from single evolution trajectories in the presence of stochastic fluctuations presents a considerable challenge. Here, we harness machine learning to identify the arrow of time from individual trajectories generated by a programmable ten-qubit quantum processor based on a nitrogen-vacancy center in diamond. We implement quantum circuits that realize unitary evolutions where heat flows from hotter to colder subsystems and their time-reversed counterparts. Projective measurements inserted in these processes induce entropy production, and their outcomes constitute the evolution trajectory. We demonstrate that an unsupervised clustering algorithm autonomously divides the experimental trajectories into two distinct groups without prior knowledge, while a convolutional neural network identifies the temporal direction of these trajectories with approximately 92% accuracy. In addition, we show that a diffusion-based generative model reproduces essential signatures of directional energy flow and entropy production. Our results establish machine learning as a powerful tool for uncovering underlying physical processes from complex experimental data, advancing the interface between quantum thermodynamics and artificial intelligence.

We study the quantum dynamics of the Schwarzschild interior in the Ashtekar-Barbero formulation, focusing on the fate of the classical singularity and the annihilation-to-nothing scenario. Using minisuperspace Wheeler-DeWitt quantization, we first analyze the standard Schrödinger representation and show that the annihilation-to-nothing behavior appears only for a specific choice of factor ordering and is not generic. We then introduce a generalized uncertainty principle (GUP), which induces minimal-length effects through a deformation of the canonical algebra. Solving the modified Wheeler-DeWitt equation and constructing Gaussian wave packets localized at the horizon, we find that the annihilation-to-nothing behavior is suppressed once the GUP corrections are included. Our results indicate that minimal-length effects qualitatively alter the quantum interior dynamics and challenge the robustness of this scenario as a mechanism for singularity resolution.

We resolve a conjecture of Hanotel and Loubenets concerning CHSH inequality in bipartite qutrits. It states that nonseparable pure states of two qutrits do not violate the CHSH inequality when each party is restricted to spin-1 observables. We prove a stronger result that \emph{all} bipartite states on $\mathbb{C}^3 \otimes \mathbb{C}^3$ satisfy the CHSH inequality under spin-1 measurements, regardless of whether the state is pure or mixed.

We construct the Gamow Golden Rule of multichannel scattering, and use it to obtain the decay distributions, the partial decay constants, the partial decay widths, and the branching fractions of a resonance that has several decay modes. We exemplify the results using two coupled-channel square well potentials.

We present a simple, malleable and low-overhead approach for improving generic biased quantum error mitigation (QEM) methods, achieving up to 15% fidelity improvements over standard QEM on 100-qubit circuits with up to 2000 entangling gates. We do so by constructing verifiable benchmark circuits which mirror the application circuit's native-gate structure and thus noise profile. These circuits can be used to benchmark and mitigate the bias of the underlying error mitigation method, requiring only the application circuit and hardware native gate set. We present two methods for generating benchmark circuits; one is agnostic to the target hardware at the expense of a small overhead of single-qubit gates, while the other is specific to the IBM superconducting hardware and has no gate overhead. As a corollary, we introduce benchmarked-noise zero-noise extrapolation (bnZNE) as a simple adaptation of zero-noise extrapolation (ZNE), one of the most popular error mitigation methods. We consider as an example the bias-mitigated ZNE and bnZNE of Trotterized Hamiltonian simulations, observing that our approaches outperform standard ZNE using both small-scale classical simulations and 100-qubit utility-scale experiments on the IBM superconducting hardware. We consider the measurement of both single-site observables as well as two-site correlations along a one-dimensional qubit chain. We also provide a software package for implementing the error mitigation techniques used in this research.

Tidal disruption events (TDEs) occur when stars are destroyed by supermassive black holes and are among the brightest nuclear transients. It has been thought that strong relativistic effects rapidly dissipate orbital energy and produce prompt disk formation when the stellar pericenter is smaller than $\sim 10$ gravitational radii. Using a general relativistic hydrodynamic simulation of a strongly relativistic TDE around a $10^{6}\,M_{\odot}$ black hole, we find instead that the overall evolution is similar to weakly relativistic TDEs: the debris remains highly eccentric, with most of the returned mass residing near the orbital apocenter ($\sim 250\times$ the initial pericenter distance), and shocks, rather than accretion, power the event. The simulation starts from the initial stellar approach and follows the debris evolution up to $35\,\text{days}$ after the peak mass-return time ($\simeq$ $23\,\text{days}$). Although early shocks driven by strong relativistic apsidal precession and pericenter nozzle compression dissipate orbital energy efficiently, they last only about a week ($\sim 0.3$ of the peak mass-return time). Stream self-interactions increase the incoming stream's angular momentum, thereby expanding its pericenter distance, weakening precession and shocks, and reducing dissipation. These results, along with previous work on weakly relativistic TDEs, suggest that circularization may be intrinsically slow regardless of the strength of relativistic effects, and the flow remains highly eccentric and extended during the peak of optical/UV luminosity.

The quantum trajectories in the de Broglie-Bohm formulation of quantum mechanics depend on an additional quantum potential derived from the full wave solution of Schrödinger's equation. The task of supplying collectively all the correct quantum results strongly alters the characteristics of the corresponding classical trajectories, which underlie semiclassical approximations to the evolving wave function. Both classical and quantum trajectories are here considered to be conservative with no influence of an external environment, even though this is the source of eventual classicality in quantum systems, that is, decoherence. The concept of integrability, closely correspondent in classical and quantum mechanics, is not preserved by the quantum trajectories. General systems, in which classical chaotic motion participates, are much harder to treat semiclassically, but quantum trajectories can be chaotic even for integrable systems. This discrepancy between the character of classical and quantum trajectories in the de Broglie-Bohm interpretation does not clarify the singular classical-quantum transition.

Demonstrating quantum heuristics that outperform strong classical solvers on large-scale optimization remains an open challenge. Here we introduce Regularized Warm-Started QAOA (RWS-QAOA), which initializes qubits by minimizing expected energy with a regularizer that penalizes near-bitstring states, preventing QAOA from stalling. We further propose a protocol that yields fixed, instance-independent parameters, enabling RWS-QAOA to operate as a non-variational algorithm in which the quantum circuit parameters are fixed and only a classical warm starting step is instance-dependent. We evaluate RWS-QAOA on the Max-Cut problem for random regular graphs, where this protocol yields a constant-depth quantum circuit, across three complementary settings. First, on Quantinuum's trapped-ion processor, RWS-QAOA outperforms the classical algorithms with the best provable guarantees for Max-Cut on $3$-regular graphs, namely Goemans-Williamson and Halperin-Livnat-Zwick, on $96$-node instances. Second, tensor-network simulations on graphs with up to $N{=}10{,}000$ nodes show that depth-$6$ RWS-QAOA, achieving an average cut fraction of $0.9167$, surpasses the best classical heuristics under matched restrictions (no local-search post-processing and no iterative refinement). Third, we remove these restrictions and benchmark against the strongest unrestricted classical heuristics, including an optimized parallel Burer-Monteiro solver that improves upon the MQLib implementation. Even against this stronger baseline, we project that surface-code RWS-QAOA reaches a quantum-classical runtime crossover below $0.2$ seconds on $3{,}000$-node graphs with fewer than $1.3$ million physical qubits. Our results show that constant-depth quantum circuits combined with a classical warm start have a credible potential to surpass classical solvers on the Max-Cut problem when executed on future quantum computers.

In globular clusters, hierarchical mergers are among the most promising pathways to forming massive black holes such as GW231123. A key factor determining whether a merger-remnant black hole will be retained in these environments and thus participate in subsequent hierarchical mergers is the recoil kick velocity. Analytic models for the recoil velocity are currently employed in nearly all population-synthesis frameworks. We instead use a state-of-the-art recoil-kick model gwModel_flow_prec developed from a combination of numerical-relativity and black-hole perturbation-theory data, together with data-driven techniques such as normalizing flows and the post-Newtonian structure of the kick. Employing both back-of-the-envelope estimates and detailed N-body as well as semi-analytical cluster simulations, we show that gwModel_flow_prec leads to a noticeable increase in the retention probability of hierarchical-merger remnants compared to the previously used analytic model and changes the mass and spin distribution of the black holes formed through hierarchical mergers. Additionally, we discuss the implications of our results in the context of massive binaries such as GW231123.

We investigate primordial magnetogenesis by assuming the gauge field is prepared in a thermal state during inflation rather than the standard Bunch-Davies vacuum. The temperature $\mathcal{T}$ introduces a physical scale that breaks conformal invariance at the level of the state while preserving the standard Maxwell action. This modification results in a {\it dissipative boost} that alters the magnetic energy density scaling from $a^{-4}$ to $a^{-3}$, resulting in a present-day magnetic field $B_0$ enhancement that can potentially range from about $10^{8}$ to $10^{16}$ on galactic scales. While this toy model alone does not satisfy observational lower bounds, it demonstrates that thermal initial conditions can significantly mitigate the conformal obstruction. Our results suggest that embedding this mechanism within a fully dynamical warm inflation framework, where dissipation continuously maintains the thermal bath, provides a highly promising path towards successfully realizing a minimal model of inflationary magnetogenesis without the need to invoke non-minimal couplings, anomalous background dynamics or nonlinear extensions of electrodynamics.

We provide a complete study of the factors influencing gravitational-wave signal localization using pulsar timing arrays. We derive analytical expressions for the Cramér-Rao sky localization precision that delineate the impact of the angular proximity, $ξ$, between the pulsar and the gravitational wave source, and the precision, $σ_L$, with which pulsar distances are known. Interference between the Earth and pulsar terms creates rapid angular oscillations for sky-coordinate Fisher matrix elements that aids localization, which is complemented by more broadly varying antenna response gradient information. The relative importance of these factors depends on whether pulsar distances are known precisely [i.e., $σ_L\leqλ_\mathrm{GW}/(1-\cosξ)$] or imprecisely, respectively. If the former, tightening pulsar distance precisions improves signal localization according to $ΔΩ_\mathrm{sky}\proptoσ_L^2$ until the Earth-pulsar system reaches its diffraction limit. If the latter, localization precision is degraded, but more pulsars in close proximity to the source is the best means of improving. With $α$ indexing pulsars, this scales as $ΔΩ_\mathrm{sky}~\propto~(\sum_α\mathrm{SNR}_α^2/ξ_α^2)^{-1}$ in the small-angle limit of the unmarginalized Fisher matrix, and we derive the analytic generalization to any angle between a pulsar and the source. Finally, we study a scenario where pulsar-term phases are treated as nuisance variables that are unconnected to binary or PTA properties. This phase-decoupled scenario, which is how all PTA continuous wave searches are currently conducted, delivers localization performance similar to the antenna-response--driven case, and does not exhibit significant improvement as pulsar distance precisions are tightened.

Preparing algebraically correlated ground states of quantum many-body systems is an important, yet challenging task for quantum simulation. We introduce a protocol that employs local projective measurements and unitary feedback for frustration-free gapless systems. Our approach prepares a priori unknown ground states in time that scales polynomially with system size. We analytically show the performance our protocol for the dynamics of a single-particle; we argue the same mechanism generalizes to many-body systems based on the physics of quasiparticles. Our theory predicts that a transient cooling dynamics directly reveals the system's universal critical properties. In particular, the state preparation time is linear in the inverse of the finite-size gap (up to log correction) when the system's dynamical critical exponent is larger or equal the effective spatial dimension explored by the quasiparticles. We verify these predictions in numerical simulations of ferromagnetic Heisenberg models in one- and two dimensions, a Fredkin spin chain, and a two-dimensional model of resonating valence bond states. Our protocol stabilizes gapless many-body ground states fully digitally without requiring analog rotations, enabling access to high-fidelity states beyond conventional adiabatic approaches in near-term experiments.

We investigate the dynamics of asymmetric binaries in extensions of General Relativity featuring a massless scalar field non-minimally coupled to gravity, focusing on the interplay between eccentricity and inclination in fully generic bound orbits. Building on an effective field theory framework tailored to extreme- and intermediate-mass-ratio inspirals, we compute scalar-field perturbations using a new arbitrary-precision C++ code capable of evolving perturbations along generic Kerr geodesics, STORM. We investigate the complete set of scalar fluxes at infinity and through the horizon across the relevant parameter space and analyze their harmonic structure as a function of orbital geometry and black-hole spin. Our results advance ongoing efforts to construct accurate waveform models for asymmetric binaries beyond GR and lay the groundwork for precision tests of fundamental physics with next-generation gravitational-wave detectors.

Vacuum quasi-topological gravity with infinitely many terms in the action satisfies Markov's limiting curvature hypothesis: the spherically symmetric solutions are regular and all curvature invariants are bounded by solution-independent scales. We study how this picture changes when the theory is coupled to matter. We find that minimally coupled matter spoils the scaling properties of the vacuum equations that lead to the validity of Markov's hypothesis, but the corresponding geometries often remain regular. We make this precise by developing a set of sufficient conditions on general static, spherically symmetric stress-tensors such that the corresponding solutions have bounded curvature. These conditions cover regular matter sectors but also singular matter profiles that are sufficiently singular in a sense we quantify. Our conclusions hold independently of the matter field equations and include configurations in which matter exhibits divergent energy density and pressure at finite radius or at Killing horizons, results that may have implications for mass inflation in these models. We then explore non-minimal couplings, focusing on theories with infinite towers of higher-curvature and electromagnetic terms in the action. In this class, Markov's hypothesis can be restored: we present theories admitting a universal upper bound on curvature, independent of the mass and charge. Overall, our results highlight subtleties in coupling quasi-topological gravity to matter and suggest Markov's hypothesis as a potential selection criterion for resummed gravity-matter effective theories.

We develop a systematic effective field theory calculation of the quasinormal modes of Kerr black holes valid for arbitrary spin, providing model-independent corrections to their ringdown spectrum directly relevant for gravitational-wave observations. Close to extremality, the effective field theory corrections in the grand-canonical ensemble exhibit an oscillatory dependence on $\log τ_{H}$, with $τ_H \equiv T_H/Ω_H$ a dimensionless measure of the black hole temperature. This behaviour signals an underlying discrete-scale-invariant structure.

Horizon scale imaging and precision lensing have turned black holes into quantitative laboratories for strong gravity and for non standard electromagnetic physics. We study the optical appearance and orbital dynamics of a new class of static spherically symmetric black holes sourced by a Palatini inspired nonlinear electrodynamics model, minimally coupled to Einstein-Hilbert gravity. Using a unified geodesic analysis, we identify the key radii that organize the strong field phenomenology. For photons we determine the unstable photon sphere, the associated critical capture threshold, and the resulting shadow size for a distant observer, and we map how these observables respond to the charge and to the nonlinearity index $n$. For massive probes we compute circular orbits and the innermost stable circular orbit, clarifying the departure from the Schwarzschild and Reissner-Nordström cases. We then connect to classical tests by evaluating the light deflection angle and periastron advance, providing additional diagnostics that complement the shadow. Our results furnish a practical reference model for confronting first order nonlinear electrodynamics black holes with current and forthcoming imaging and lensing data.

We develop and apply the Batalin-Fradkin-Vilkovisky (BFV) formalism for quantizing off-diagonal solutions of the Einstein equations in general relativity. In the quasi-classical limit of quantum gravity, such solutions possess specific nonlinear symmetries and encode Hořava - Lifshitz (HL) configurations with anisotropic scaling and effective cosmological constants. The geometric constructions are performed on Lorentz manifolds enabled with nonholonomic 2+2 and 3+1 fibration structures.

Cryogenic sub-Hz cROss torsion-bar detector with quantum NOn-demolition Speed meter(CHRONOS) is a proposed next-generation ground-based gravitational-wave observatory designed to explore the sub-Hz frequency band with unprecedented sensitivity. Utilizing a cryogenic torsion-bar interferometric configuration with quantum non-demolition speed-meter readout, CHRONOS targets a frequency window that bridges space-based missions and current high-frequency ground-based detectors, opening a new frontier in gravitational-wave astronomy. The observatory will enable long-duration tracking of compact binary inspirals well before merger, significantly improving source localization, parameter estimation, and tests of general relativity. In addition to transient signals, CHRONOS is optimized to probe the stochastic gravitational-wave background (SGWB) in the sub-Hz regime, providing powerful constraints on primordial gravitational waves, inflationary tensor spectra with red or blue tilts, first-order phase transitions, cosmic strings, and other relics of high-energy physics. By connecting gravitational-wave measurements across cosmological frequency scales-from cosmic microwave background polarization to pulsar timing arrays and high-frequency interferometers-CHRONOS will contribute to a coherent reconstruction of the gravitational-wave spectrum over more than twenty orders of magnitude. Crossing critical sensitivity thresholds in the sub-Hz band, CHRONOS will establish a new pillar of gravitational-wave astronomy and cosmology, enabling transformative advances in astrophysics and fundamental physics.

Using Vakarelov's theory of lattice logics with negation, we introduce the (co)quasiintuitionistic logic, and prove its soundness and completeness with respect to the class of (co)quasiintuitionistic algebras. Combining these algebras together, we obtain biquasiintuitionistic algebras and the biquasiintuitionistic logic. Their further extension with the Skolem algebra structure defines Akchurin algebras and the respective logic, which is a product of biquasiintuitionistic and biintuitionistic logics, featuring four distinct negations. Next we generalise the framework of spectral presheaves (which is a main object in the Butterfield--Isham--Döring topos theoretic approach to quantum mechanics) to arbitrary complete orthocomplemented lattices, and show that the orthocomplementation determines two negation operators on the spectral presheaf (one paraconsistent, another paracomplete), equipping the set of all closed-and-open subpresheaves of a spectral presheaf with the structure of a biquasiintuitionistic algebra. Combined with the generic Skolem (i.e. Heyting and Brouwer) algebra structure of this set, this gives a particular instance of an Akchurin algebra. We also show that the underlying orthocomplemented lattice can be reconstructed as an internal object of the spectral presheaf, resulting as the image of a double coquasiintuitionistic (resp., quasiintuitionistic) negation monad (resp., comonad). Finally, we prove a no-go theorem for the claim that the spectral presheaf is a model of a dialectical (or any other) relevance logic.

We present a scalable formal verification methodology for Quantum Phase Estimation (QPE) circuits. Our approach uses a symbolic qubit abstraction based on quantifier-free bit-vector logic, capturing key quantum phenomena, including superposition, rotation, and measurement. The proposed methodology maps quantum circuit functional behaviour from Hilbert space to a bit-vector domain. We develop formal properties aligned with this abstraction to ensure functional correctness of QPE circuits. The method scales efficiently, verifying QPE circuits with up to 6 precision qubits and 1,024 phase qubits using under 3.5~GB of memory.

Massive leptophobic Z' bosons decaying to a pair of charginos are searched for in proton-proton collisions at $\sqrt{s}$ = 13 TeV, using data samples collected by the CMS experiment in 2016, 2017, and 2018, corresponding to a total integrated luminosity of 138 fb$^{-1}$. The Z' bosons originate from an additional U(1)' gauge symmetry extended to the minimal supersymmetric standard model. The final state consists of two oppositely charged leptons and missing transverse momentum. The signal extraction is performed with a parametrized neural network. The measurements are found to be consistent with the standard model expectations. Upper limits are set on the Z' boson production cross sections as a function of the Z' and chargino masses. The analysis excludes Z' boson masses up to about 3.5 TeV for the specific case of Z' bosons decaying exclusively to charginos, with the charginos decaying to W bosons and neutralinos.

The SUPerconduction AXion search experiment (Supax) is a future haloscope-type detector designed to probe axion-like particles (ALPs) as candidates for dark matter and solutions to the strong-CP problem in the mass range between $8\,μ$eV and $30\,μ$eV. In the course of the preparation of Supax, a prototype experiment was built and operated. Using a copper cavity, cooled down to a temperature of 2 K and operated in a magnetic field of 12 T, we probe axion masses around $34\,μ$eV and exclude axion-photon couplings down to $|g_{aγγ}|> 1.6\cdot 10^{-13}$GeV$^{-1}$. The data was also used to exclude dark photons in the same mass range with a kinetic mixing parameter of $χ> 1.4\cdot 10^{-12}$. Details of the experimental setup and the analysis strategy are summarized in this paper.

While hidden, singly, doubly, and fully heavy tetraquark states have been experimentally observed, triply heavy tetraquark states remain unconfirmed. We systematically investigate the spectroscopic and decay properties of four triply heavy-flavor tetraquark systems ($cc\bar{c}\bar{n}$, $cc\bar{c}\bar{s}$, $bb\bar{b}\bar{n}$, $bb\bar{b}\bar{s}$; $n=u,d$) based on the nonrelativistic quark model. Using an effective Hamiltonian, we employ the Gaussian expansion method to solve the four-body Schrödinger equation and incorporate the effect of color-spin configuration mixing. Results show both $cc\bar{c}\bar{q}$ and $bb\bar{b}\bar{q}$ systems have two $J^{P}=0^{+}$, three $J^{P}=1^{+}$, and one $J^{P}=2^{+}$ states, with ground-state masses of 5.2-5.5 GeV and 15.0-15.3 GeV, respectively. Root-mean-square radius analysis supports compact tetraquark configurations. All states are unstable, with rearrangement strong decays dominant and negligible radiative decays. Narrow resonances (e.g., $T_{c^{2}\bar{c}\bar{s}}(5360,0^{+})$, $T_{b^{2}\bar{b}\bar{n}}(15052,0^{+})$) arise from Feynman amplitude cancellation. We propose experimental searches in $J/ψD^{*}_{s}$/ $η_{c}D_{s}$ (5.3-5.4 GeV) and $ΥB^{*}$ (15.0-15.1 GeV) channels, providing key guidance for triply heavy tetraquark identification.

A measurement of the flavour-untagged decay-time-integrated ${C\!P}$ asymmetry in the flavour-specific decay ${B_s^0 \to D_s^-π^+}$, ${\langle A^s_{\rm untagged}\rangle}$, is performed using proton-proton collision data collected by the LHCb experiment between 2016 and 2018 at a center-of-mass energy of ${13\,{\rm TeV}}$, corresponding to a total integrated luminosity of ${5.4\,{\rm fb}^{-1}}$. The ${C\!P}$ asymmetry is measured in two $D_s^-$ meson decay modes, ${D_s^- \to K^-K^+π^-}$ and ${D_s^- \to π^-π^+π^-}$. The combined result, $\langle A^s_{\rm untagged}\rangle = ( -1.4 \pm 5.9\,\rm{(stat)} \pm 1.1\,\rm{(syst)}) \times 10^{-3}$, is consistent with the Standard Model expectation and provides a direct constraint on new physics in tree-level $b$-hadron decays.

The temporal resolution of Low-Gain Avalanche Detectors (LGADs), also known as Ultra-Fast Silicon Detectors (UFSDs), is governed by two contributions: jitter, arising from electronic noise and signal slew rate, and the Landau noise term, arising from the non-uniform energy deposition of minimum ionizing particles (MIPs). We show that a correct simulation of the initial ionization alone significantly overestimates the measured Landau noise. Two additional physical mechanisms are necessary to reproduce the data: space charge effects during electron/hole drift, which smooth the granularity of the initial charge distribution, and gain saturation during multiplication, which preferentially suppresses large-amplitude fluctuations. All steps of the model have been implemented in the fast simulation program Weightfield2 (WF2). The model is validated against several independent experimental observations: the evolution of the measured charge distribution with gain, the temporal resolution of events in the Landau tail, and the thickness dependence of timing performance. We also discuss a data-driven gain measurement method based on gain saturation, and implications for gain layer design.

Heavy-ion experiments at the CERN SPS began in the mid-1980s to study nuclear matter at extreme temperatures and densities. The program started with light ions, such as oxygen and sulphur, at energies of 60A GeV and 200A GeV, later advancing to lead ions at 158A GeV. A series of experiments, employing novel detector technologies, explored various signatures of quark-gluon plasma (QGP) formation. In February 2000, these results led CERN to announce evidence for the QGP formation. Subsequently, an energy scan was conducted with lead ions from 20A GeV to 158A GeV, to locate the threshold of QGP creation.

Courant algebroid relations are used to define notions of relations between Dirac structures and spinors. It is shown under which circumstances a spinor relation gives a Courant algebroid relation and how it descends to a relation between Dirac structures. A converse to this result is proved: a T-duality relation induces a spinor relation that links the Dirac generating operators defining T-dual Courant algebroids, generalising the isomorphism of twisted cohomology associated with topological T-duality. We introduce the notion of relation between generalised complex structures and characterise their reduction. We also define relations between generalised Kähler structures, and rephrase them in terms of bi-Hermitian structures which induce T-duality relations between $\mathcal{N}=(2,2)$ supersymmetric sigma-models. We prove existence results for T-dual structures, and demonstrate compatibility of T-duality relations with Type II supergravity equations.

We report the first searches for charged-lepton-flavor violation in decays of $χ_{bJ}(1P)$ ($J=0, 1,$ and $2$) to a pair of charged leptons using 158 million $Υ(2S)$ decays collected with the Belle detector in $e^+e^-$ collisions at the KEKB collider. No significant signal is observed, and we set upper limits on the branching fractions for $χ_{bJ}(1P)$ decays to $e^\pmμ^\mp$ at the level of $10^{-6}$ and to $e^\pmτ^\mp$ or $μ^\pmτ^\mp$ at the level of $10^{-5}$. Limits on $χ_{b0}(1P)$ decays are translated into bounds on the corresponding Wilson coefficients of scalar operators that mediate charged-lepton-flavor violation.

In this work, we adopt the diquark model, construct the diquark-diquark-antiquark type currents with the light quarks $uus$ in two octets, and study the $uusc\bar{c}$ pentaquark states in the framework of the QCD sum rules systematically. We obtain the mass spectrum of the hidden-charm-singly-strange pentaquark states with the quantum numbers $IJ^{P}=1{\frac{1}{2}}^-$, $1{\frac{3}{2}}^-$ and $1{\frac{5}{2}}^-$. And we can search for those $P_{cs}$ states in the processes $Σ_b^+\to P_{cs}^+φ\to J/ψΣ^+ \,φ$ and $Ξ_b^0\to P_{cs}^+ K^- \to J/ψΣ^+\, K^-$.

This review traces the evolution of precision timing in particle physics experiments, from the first large-scale applications of scintillator and photomultiplier tube (PMT) systems in the 1990s to the picosecond-precision detectors of future colliders. Four technological generations are identified: (i) the era of discrete electronics (NIM, CAMAC, VME) and PMTs, which established the three canonical uses of timing -- particle identification via time-of-flight, background and pile-up rejection, and directionality triggering; (ii) the silicon revolution enabled by Silicon Photomultipliers (SiPMs), Low-Gain Avalanche Diodes (LGADs), and Application-Specific Integrated Circuits (ASICs); (iii) the current transition to ubiquitous four-dimensional (4D) tracking, in which time is a coordinate measured at every point along a particle trajectory. Under-construction systems at the HL-LHC (CMS MTD, ATLAS HGTD, CMS HGCAL) demonstrate the maturity of 30--50\,ps silicon timing at the million-channel scale. The EIC, LHCb VELO Upgrade~II, and ALICE~3 push this technology into new regimes of radiation hardness, material budget, and granularity. (iv) The Far-future facilities such as the Muon Collider and FCC require a further leap to $\sim$10\,ps, setting the agenda for the next decade of detector R\&D.

We give two novel proofs that the path integral and stochastic quantizations of generic scalar Euclidean quantum field theories are equivalent. Our proofs rely on Taylor interpolations indexed by forests, in the fashion of constructive field theory. The first proof works at the level of individual terms in the Feynman expansion, with the forests appearing as spanning forests in Feynman graphs. The second one works at the level of the path integral and avoids the full expansion of the Feynman perturbation series.

We study the spin-dependent cross section in semi-inclusive deep-inelastic lepton-nucleon collisions, $\ell + N^\uparrow\to \ell+h+X$. We focus on the cross section that is integrated over the transverse momentum $P_{h\perp}$ of the detected unpolarized hadron. We analyze this cross section at large virtuality $Q^2$ of the exchanged virtual photon within the framework of collinear twist-3 factorization in perturbative QCD. The two main transverse spin observables, the single nucleon spin asymmetry (SSA) and the double longitudinal lepton-transverse nucleon spin asymmetry (DSA), both receive contributions from chiral-odd twist-3 three-parton fragmentation functions. We study these fragmentation contributions to Next-To-Leading Order (NLO) in perturbative QCD. We explicitly observe that collinear twist-3 factorization holds for these contributions at the one-loop level. We confront our NLO formulae with HERMES data and provide numerical predictions for EIC kinematics.

Conformal blocks in odd spacetime dimensions are not known in closed analytic form. To facilitate efficient computations in the conformal bootstrap, we introduce $\texttt{GoBlocks}$: a novel conformal-block generator implemented in the Go programming language, designed for rapid, on-the-fly, parallel evaluation using recursive relations. The package supports both multi-point and derivative-based bootstrap approaches and allows flexible control over accuracy and performance. We benchmark $\texttt{GoBlocks}$ against the $\texttt{scalar_blocks}$ package, finding significant speed improvements in applications where computational speed and moderate accuracy are critical, but ultra-high precision is not essential. As an illustration, we apply $\texttt{GoBlocks}$ to the mixed-correlator bootstrap of the three-dimensional Ising model, formulated as a non-convex optimisation problem in a suitable truncation scheme. We simultaneously optimise over external scaling dimensions and OPE CFT data. In addition, we discuss how the approach scales as we increase the number of mixed correlators in more general $O(N)$ vector models.

This study explores the heavy-quark contributions to polarized structure functions in deep-inelastic scattering at next-to-leading order. The structure functions $g_1$, $g_4$, $g_5$, $g_6$, and $g_7$ are computed, while $g_2$ and $g_3$ are excluded due to the higher-twist suppression. The calculations are performed within the ACOT renormalization scheme, which ensures theoretical consistency across kinematic regions where heavy quarks transition from being dynamically produced to fully active degrees of freedom. By incorporating heavy-flavor contributions at next-to-leading-order, this work provides deeper insights into their role in polarized structure functions and the spin-dependent dynamics of QCD. Both analytical results and their numerical implementation are presented.

For radially excited heavy quarkonia $ V=ψ(2S) $ and $ Υ(2S) $, the nodal structure of their wave function renders the three-gluon decay $ V\to ggg $ acutely sensitive to relativistic corrections, presenting a longstanding challenge for reliable theoretical predictions. We perform a comprehensive analysis of these decays within the Bethe-Salpeter formalism, constructing analytic harmonic oscillator wave functions that explicitly incorporate the $2S$ nodal structure. Model-independent relations among polarized decay widths are derived from helicity-flip and phase-space symmetries. To obtain physically consistent results beyond $ q^2 $-order relativistic corrections, we introduce a concise phenomenological treatment that effectively incorporates partial higher-order contributions while preserving the correct low-momentum limit. Taking into account both relativistic and QCD radiative corrections, we compute $ Γ(V\to ggg) $, $ Γ(V\to e^+e^-) $, and their ratio $ R_V = Γ(V\to ggg)/Γ(V\to e^+e^-) $, which allows a clean extraction of the harmonic oscillator parameter $ β_V $, with the extracted values lying at the lower end of the ranges from typical phenomenological models. Our predictions for the branching ratios $ \mathcal{B}(V\to ggg) $ and $ \mathcal{B}(V\to e^+e^-) $ are in excellent agreement with experimental data. A striking feature is the distinct convergence behavior of the relativistic expansion: the leptonic width converges rapidly already at $ q^2 $ order, whereas the gluonic width converges far more slowly due to destructive interference induced by the nodal structure in the multi-gluon convolution. These results offer new insights into the dynamics of excited quarkonia and provide valuable constraints on non-perturbative descriptions of heavy-quark bound states.

Scenarios where dark matter is boosted to relativistic velocities provide a promising probe of sub-GeV dark matter. Cosmic-ray upscattered and supernova-produced dark matter generate relativistic fluxes peaked toward the Galactic Centre, an anisotropy that offers a strong directional signature and is not mimicked by any terrestrial or cosmic background. We determine how many directional recoil events are required in a gas time-projection chamber to distinguish various scenarios for the origin of dark matter particles arriving in the solar system, which are otherwise indistinguishable without directionality. We find that standard halo dark matter particles can be distinguished from boosted populations with as few as $\mathcal{O}(20)$ events under reasonable track reconstruction performance and background conditions.

The global spin polarization of hyperons in heavy-ion collisions have been investigated by including spin correlation effects among their constituent quarks. The available data on global spin polarizations of hyperons and spin alignments of vector mesons provide constraints on phase space functions of the spin polarization and correlation. These constraints can lead to inequalities under some approximations, which might provide possible clues for the presence of quark spin correlation inside hyperons at lower collision energies.

Finite-size scaling (FSS) provides a framework for investigating the possible presence of a critical end point (CEP) in the QCD phase diagram using fluctuation observables measured in relativistic heavy-ion collisions. A recent analysis reported a FSS collapse of a susceptibility constructed from net-proton cumulants, which was interpreted as evidence for a CEP near $μ_B \approx 625$ MeV \cite{FSSpaper}. This note examines several aspects of that scaling construction. These include identifying the pseudorapidity acceptance window with the physical system size used in the finite-size scaling relations, assessing the influence of acceptance-driven multiplicity scaling on the susceptibility used for the scaling analysis, and treating thermodynamic scaling fields in the scaling variable. These considerations clarify several issues relevant for the consistent implementation and interpretation of finite-size scaling analyses in heavy-ion collision experiments.

We investigate the phenomenology induced by generic PQ-breaking operators within the axion kinetic misalignment framework. We analyze their impact on the relic density of axion dark matter (DM), the PQ quality problem, axion-mediated fifth-force, as well as Big Bang Nucleosynthesis (BBN) and Cosmic Microwave Background (CMB) constraints. A nonzero initial axion velocity gives rise to brief periods of early matter domination and axion kinetic domination, leading to a nonstandard cosmological evolution. We compute the resulting gravitational wave (GW) signal from global cosmic strings and find that, because these nonstandard epochs are extremely short, the signal is highly suppressed and beyond the reach of existing experiments. Finally, we perform a parameter space scan, identify the regions and benchmark point that are consistent with all experimental constraints.

We investigate the low-energy dynamics of small-amplitude surface oscillations of spherical superfluid droplets in vacuum. Starting from the effective field theory of superfluid phonons, we derive an effective action governing the surface oscillations under a fixed particle-number constraint. The normal-mode eigenfrequencies $ω_{\ell}$ for each angular momentum quantum number $\ell$ are determined and shown to depend on a dimensionless parameter measuring the ratio of surface tension to bulk compressibility energy. We identify a critical value of this parameters at which the breathing mode ($\ell = 0$) becomes mechanically unstable, and show that all multipole surface modes with $\ell \geq 2$ enter the low-energy regime when the surface tension is sufficiently small. Within this regime, we further quantize the surface oscillations, whose quanta correspond to ripplons, allowing the construction of general multi-ripplon states obeying angular-momentum selection rules. We also apply our formalism to a concrete example: a weakly interacting two-component Bose mixture realizing a self-bound superfluid droplet. The resulting description is universal in the sense that it applies to surface dynamics of generic nonrelativistic superfluids with a free interface, independent of microscopic details.

It was recently shown that the decoupling limits leading to matrix (gauge) theories on D-branes give rise to non-Lorentzian target space geometries. Perturbatively, matrix theory describes a quantum gravity theory whose low-energy supergravity description exhibits non-Lorentzian behavior. Focusing on the D-particle case associated with the Banks-Fischler-Shenker-Susskind matrix theory, and using techniques from ambitwistor string theory, we show evidence that the dynamics of this non-Lorentzian gravity should be related to anomalies in the current algebra of the associated fundamental string worldsheet theory. At large N, the D-particle backreaction deforms the non-Lorentzian supergravity to the Lorentzian IIA theory, providing a holographic description of the BFSS matrix theory. At a moderately large N such that the D-particles decouple at the leading order, this non-Lorentzian supergravity maps holographically to the leading-order contribution of weakly coupled bulk gravity. This approximately non-Lorentzian regime is related to the null reduction of eleven-dimensional supergravity. Within the non-Lorentzian supergravity, non-trivial dynamics arises from the backreaction of extended brane objects that form BPS states with the D-particles. Finally, we generalize these results to other D-brane and string soliton holographic constructions.

Nearly thirty years ago, Bern, Dixon and Kosower computed all helicity amplitudes for the annihilation of an electron-positron pair into four QCD partons through an electroweak vector boson. More recently, the leading-color two-loop amplitudes for the same process were obtained. When such amplitudes are expressed in the massless spinor-helicity formalism, they effectively correspond to the decay of a transversely polarized vector boson. However, for several reasons, it is highly desirable to extend these calculations to the case where the polarization of the vector boson is longitudinal. Due to the complexity of such computations, repeating them to obtain the result for the ''missing'' polarization of the electroweak boson is a significant undertaking even at one loop. Besides, when attempting new higher-loop computations, it is beneficial to identify the minimal set of quantities (e.g. form factors) that must be determined to obtain the full amount of physically-relevant information. In this paper, we show that amplitudes involving vector-boson decays to massless leptons-although they appear to project onto the transverse polarization-still encode the full information about all polarization states of the vector boson, including the longitudinal one. This follows from the little-group (spin) covariance of the amplitude, which allows us-largely through simple replacement rules-to rewrite the helicity amplitudes as a matrix with open SU(2) spin indices in the massive spinor-helicity (or spin-spinor) formalism. Therefore, knowledge of an amplitude for any polarization component suffices to reconstruct the full covariant matrix.

We investigate the extent of the critical scaling region of the chiral phase transition at finite chemical potential within the quark-meson (QM) model using the functional renormalization group (fRG) approach. By analyzing the scaling behavior of the chiral order parameter and correlation length with respect to temperature and pion mass near the second-order phase transition, we extract critical exponents from the data and quantify the range over which the scaling relations remain valid. We find that both the leading order and the next-to-leading-order scaling regions systematically shrink as the chemical potential increases. This behavior is observed in both the local potential approximation (LPA) and its extension including anomalous dimensions (LPA'), with qualitatively consistent results, while the scaling region in LPA' is slightly smaller than that in LPA.

We expand on recent work arXiv:2601.17199 demonstrating the existence of a novel entanglement radius $ξ_0$ characterizing flux tube entanglement entropy (FTE$^2$) in (2+1)D Yang-Mills theory. This physical scale corresponds to the intrinsic thickness of the flux tube that must be fully severed by an entangling region for color degrees of freedom in the flux tube to contribute non-zero FTE$^2$. We consider here geometries of the entanglement region $V$ on the lattice where the length of the region cross-cutting the flux tube is of the same magnitude as $ξ_0$. Our results further the conclusions of arXiv:2601.17199 by adding detailed new information on the topological structure of the entanglement radius of color flux tubes.

We study large-spin operators in conformal field theories (CFTs) in spacetime dimensions $d>2$ by placing the theory on appropriate pp-wave backgrounds. We show that these geometries admit Heisenberg-group symmetries, and that these symmetries, combined with locality of quantum fields on such spacetimes, impose strong constraints on the asymptotic spectrum in the large-spin limit. The pp-wave backgrounds probe both the small-twist regime, corresponding to the Regge or light-cone bootstrap, and a strongly coupled regime of large twist. Finally, we demonstrate that causality (or the requirement that the energy be bounded from below) leads to a new unitarity bound in $3+1$ dimensions.

Pulsars possess strong dipole magnetic fields that can source axion fields through the axion-photon interaction. Pulsars may therefore be surrounded by axion field configurations oscillating with the pulsar's rotational period. These axions could be detected by observing their effect on the polarization of the pular's emission. In this paper, we use time resolved observations of the optical polarization of the Crab pulsar to place bounds on the axion-photon coupling, demonstrating the potential of time resolved pulsar birefringence in the search for axions.

We investigate the vacuum expectation value of bosonic current induced by a magnetic flux in a $(2+1)-$dimensional conical spacetime in the presence of a circular boundary, in a Ho{ř}ava-Lifshitz Lorentz violation symmetry scenario. We assume that the circular boundary is concentric with magnetic flux, and the massive scalar quantum field obeys the Robin boundary condition on the boundary. In order to develop this analysis, we calculate the positive frequency Wightman functions for both regions, inside and outside the boundary. Using these functions, we obtain analytical expressions for the vacuum expectation bosonic currents. As we will see, these functions are presented in the form of the sum of boundary-free and boundary-induced parts. As to the boundary-induced currents, some asymptotic behaviors are investigated for specific limiting cases; moreover, in order to provide a better understanding about the behavior of the currents, some plots are given.

We study entanglement in two-dimensional Yang-Mills theory, viewed as a quasi-topological model of emergent space. The most familiar class of states in this theory are states defined by Euclidean path integrals over Riemann surfaces. Bipartite states of this class have thermofield double structure, with entanglement consistently reducing with total area and the number of topological defects, turning separable in the infinite-area limit. In contrast, Wilson lines and loops generate rich non-monotonic behavior of the entanglement entropy. Most notably, we find that for a certain discrete set of configurations, entanglement remains finite at infinite area. The reduced density matrices, in such configurations, take the form of finite-dimensional projectors onto non-trivial vacuum sectors. We also discuss the implications of the large-volume effects for confinement and find that special asymptotic configurations are related to transitions in the confining force.

We compute the graviton-mediated two-to-two scattering amplitude and cross section for scalar particles in asymptotically safe quantum gravity. Specifically, we compute the full momentum dependence of the scalar-graviton three-point scattering vertex for spacelike momenta with the functional renormalisation group. We also discuss the analytic continuation to the Minkowski branch, and in particular its angular dependence. Then, the timelike part of the vertex is reconstructed and used to compute the scattering amplitude and cross-section. We show that the cross-section reduces to that in General Relativity at small energies, and it respects unitarity in the UV.

Continuous-wave Nuclear Magnetic Resonance (CW-NMR) operated in constant-current mode has served as a foundational technique for polarization measurement in solid-state dynamically polarized targets within nuclear and high-energy physics experiments for several decades, and it remains an essential tool. Conventional Q-meter-based phase-sensitive detection is critical for precise real-time determination of target polarization during scattering runs. However, the accuracy and reliability of these measurements are frequently compromised by elevated noise levels, baseline drift, and systematic uncertainties arising from signal isolation and fitting, ultimately degrading the overall experimental figure of merit. In this work, we report the first successful application of neural network architectures to continuous-wave NMR polarization metrology. By leveraging advanced machine learning techniques for signal extraction and denoising, we achieve a substantial reduction of fitting uncertainties under a variety of realistic simulated and experimental conditions. These improvements translate directly into more robust real-time (online) polarization monitoring and significantly higher precision in subsequent offline analysis. The resulting methodology offers an improved figure of merit for scattering experiments employing dynamically polarized targets and establishes a new tools for NMR-based polarimetry in high-energy and nuclear physics.

We study nonperturbative aspects of quantum field theory (QFT) in rigid anti de Sitter (AdS) spacetime using quantum information theoretic methods. While irreversibility of renormalization group (RG) flows is well established in flat space, it is not obvious whether it persists in AdS, where negative curvature and an asymptotic timelike boundary significantly modify infrared dynamics. Using strong subadditivity and AdS invariance, we derive an entropic second-order differential inequality for the difference between the vacuum entanglement entropy of a QFT and that of its ultraviolet fixed point, evaluated for spherical bulk regions. This inequality allows us to define RG charges that measure the relevant number of degrees of freedom, and we prove the irreversibility of the RG in $2,\,3,$ and $4$ spacetime dimensions. We further analyze free scalar and fermion theories in AdS, developing lattice formulations adapted to the geometry and computing entanglement entropies and RG charges. In AdS$_2$, we obtain analytic results for a massive Dirac fermion and compare them with numerical lattice calculations. These examples illustrate the general irreversibility theorem and clarify the distinction between conformal and massive theories in AdS.

We study global vortex production in (numerical) scattering experiments when the scatterer and the vortex degrees of freedom interact only due to intermediary quantum variables. We work in $2+1$ dimensions with a complex scalar field, $φ$, that supports global vortices, a real scalar field, $ψ$, that is the scatterer, and a quantum field, $ρ$, that couples to both $φ$ and $ψ$, acting as a mediator. We simulate the scattering of relativistic Gaussian wavepackets of $ψ$ and scan parameter space for regions where vortex-antivortex pairs are produced. The results show that vortex production is highly sensitive to the initial Gaussian parameters, and the parameter space is chaotic with ``holes" and isolated regions of vortex production.

We study the Stodolsky effect utilizing the most general form of neutrino interactions with electrons below the electroweak scale by considering all possible Lorentz invariant operators respecting SU(3)$\otimes$U(1) symmetry. We perform our calculation for both Dirac and Majorana neutrinos and find that in the most general setting, only the non-standard neutrino interactions and the tensor interaction terms provide a non-zero contribution, apart from the Standard Model contribution. We investigate the implications for the possible detection of the cosmic neutrino background (C$ν$B) by analysing the energy shifts that are characteristic of the Stodolsky effect. We also discuss the implication of considerable asymmetry in the C$ν$B on the present scenario.

We study Krylov (spread) complexity in strongly coupled six-dimensional ${\cal N}=(1,0)$ superconformal field theories with holographic duals in massive type IIA supergravity. Extending recent holographic proposals relating Krylov complexity growth to the proper momentum of an infalling particle, we analyse the dynamics of massive geodesic probes in these geometries. In our setup, the bulk particle is allowed to move along three directions: the radial AdS coordinate, the internal $S^2$ associated with the $SU(2)_R$ symmetry, and the coordinate parametrising the quiver. In the dual field theory these motions encode, respectively, operator growth, the presence of R-symmetry charges, and spreading across different nodes of the quiver. We analyse the geodesic motion both analytically and numerically for representative quiver configurations. The motion along the quiver direction is typically damped and localised at early times, while the late-time behaviour is dominated by the radial AdS motion. As a consequence, the generalised proper momentum grows linearly at late times, consistent with expectations for Krylov complexity in conformal theories. The inclusion of angular momentum ($SU(2)_R$ charge) introduces additional constraints on the allowed motion and modifies the early-time dynamics while leaving the asymptotic behaviour unchanged. These results provide a first exploration of Krylov complexity in higher-dimensional holographic conformal theories and reveal how operator growth can probe both internal symmetries and quiver structure in strongly coupled conformal field theories.

This work presents an analysis of event-by-event multiplicity fluctuations as a sensitive tool for diagnosing the state of matter produced in relativistic heavy-ion collisions. Using a modified version of the HIJING Monte Carlo generator, which integrates various models of partonic energy loss in hot (quark-gluon plasma) and cold media, the connection between fluctuation dynamics and system properties is investigated. It is shown that the nature and magnitude of fluctuations allow for the identification of the created medium type (hot or cold), the testing of the adequacy of energy loss models, and the detection of signatures of a first-order phase transition in different kinematic regions. The results obtained are important for interpreting the data from experiments aimed at mapping the QCD phase diagram and searching for the critical point.

We study type II$_0$ loci in the moduli space of type IIB string theory compactified on Calabi-Yau manifolds. We show that around these infinite distance singular loci the leading order behaviour of the gauge kinetic matrix, and of the prepotential, can always be written in the form of a threshold correction from integrating out a BPS state, but one with an effectively complex charge. In order to understand the physical meaning of this, we carefully identify the splitting in the effective supergravity between the graviphoton direction and matter vector multiplets. Within a specific two-parameter example of a Calabi-Yau, we use this to identify a strongly-coupled matter sector involving both light electric and light magnetic states. We propose that the leading gauge kinetic matrix arises as a threshold correction from integrating out this non-perturbative sector, and that the sector has an effective weakly-coupled infrared description in terms of the complex-charged state. The region in moduli space has a Heterotic string dual microscopic description. The light magnetic state in this description corresponds to a Kaluza-Klein monopole, which becomes lighter than the fundamental Heterotic string, leading to the non-perturbative sector. Assuming this picture is correct, it implies the existence of infrared emergent infinite distance loci in moduli spaces of quantum gravity.

1FGL J0523.5$-$2529 is a Fermi selected redback millisecond pulsar candidate that exhibited luminous optical and X-ray flares in 2020-2021. We obtained a simultaneous X-ray and $U$-band observation with XMM-Newton in 2025, the first to cover the 16.5 hr orbit of 1FGL J0523.5$-$2529. The X-ray luminosity was in an intermediate state with a power-law photon spectral index of $Γ=1.53\pm0.02$. Frequent flares were superposed on a broad, single-peaked modulation, the latter characteristic of intrabinary shock models in which the shock front is wrapped around the pulsar. We speculate that density enhancements in the shocked companion wind cause flares, as well as variable optical recombination lines. The $U$-band light curve was dominated by ellipsoidal modulation of the nearly Roche lobe filling companion star, similar to that seen in ground-based optical photometry. We also used this effect in 10 years of ATLAS monitoring to improve the precision of the orbital period to 0.6881366(19) days. Considering that searches for radio pulsations from 1FGL J0523.5$-$2529 at all orbital phases have been unsuccessful, the shocked wind usually surrounds the pulsar.

Observations of the Cosmic Microwave Background (CMB) B-mode polarisation provide a unique probe of inflationary physics. Extracting a reliable constraint on the tensor-to-scalar ratio $r$ nonetheless demands stringent suppression of diffuse Galactic foregrounds, whose residuals can bias the inferred signal. This work introduces and evaluates two extensions of the Needlet Internal Linear Combination (NILC) framework aimed at reducing foreground-induced biases on $r$. The first extension implements the deprojection of selected foreground moments directly within the component-separation step. The second performs a likelihood-level marginalisation over residual foreground power using a data-driven template. Using Simons Observatory Small Aperture Telescope (SO-SAT) - like simulations, we show that both methods effectively control residual contamination, yielding unbiased estimates of $r$ and a consistent reconstruction of the lensing B-mode amplitude. These results indicate that enhanced foreground-mitigation strategies will be useful for next-generation CMB polarisation analyses seeking a robust detection of primordial B-modes.

Recent work by Chae et al. (2026) reported a gravitational anomaly in 36 wide-binary pairs, finding a gravity boost factor of $γ\equiv G_{\rm eff}/G_{\rm N} \approx 1.60_{-0.14}^{+0.17}$ at low accelerations, consistent with predictions from Modified Newtonian Dynamics (MOND). We reanalyze the same dataset using a hierarchical Bayesian model that infers a global $γ$ across all systems while fitting three-dimensional orbital elements. Our model yields $γ= 1.12^{+0.27}_{-0.22}$, consistent with Newtonian gravity ($γ= 1$) at the $\sim0.4σ$ level. To identify the source of the discrepancy, we perform a test using an approach similar to Chae et al. (2026), replacing the semi-major axis with a geometric de-projection of the observed projected separation. This test yields $γ= 1.56^{+0.21}_{-0.18}$, closely matching the result of Chae et al. (2026). This suggests that the inferred value of $γ$ is sensitive to how the three-dimensional orbital separation is modeled, and including an independent semi-major axis parameter can account for velocity excesses that would otherwise be attributed to non-Newtonian gravity.

Magnetic white dwarfs (MWDs) are key to understanding the origin and evolution of magnetic fields in compact stars. While large spectroscopic surveys such as SDSS have greatly expanded the known sample, the potential of LAMOST has not yet been fully explored. Our aim is to identify and characterize isolated MWDs in the LAMOST DR10 database. We cross-matched LAMOST DR10 spectra with white dwarf candidates from Gaia EDR3 and with recent SDSS-based catalogs of MWDs. Zeeman splitting in Balmer and helium absorption lines was used as the primary diagnostic to identify magnetic fields and to estimate their strengths. Reference objects from SDSS catalogs were used to test the detectability of MWDs in LAMOST low-resolution spectra. We identified 63 isolated MWDs in LAMOST DR10, of which 32 are new discoveries. Surface magnetic field strengths were measured from Zeeman splitting, covering a range from a few MG up to several tens of MG. For previously known SDSS MWDs, our LAMOST-based field measurements show mostly agreement with published values. This work demonstrates the capability of LAMOST low-resolution spectroscopy to identify and characterize isolated MWDs. The newly discovered objects expand the known population and provide valuable targets for future high-resolution spectroscopic and polarimetric follow-up studies. Our results highlight the potential of combining LAMOST with Gaia and other large surveys to build a more complete census of MWDs.

The Shakura-Sunyaev $α$-viscosity coefficient, defined as the ratio of total stress to total pressure, $α= \mathbb{T}/p$, played an important role in the development of the accretion disc theory in the early 1970s. The origin of turbulence that causes the stress $\mathbb{T}$ was unknown at that time. Shakura and Sunyaev assumed $α=$ const. Today we know that this was not quite realistic - the modern general relativistic magneto-hydrodynamic simulations (GRMHD) of black hole accretion discs revealed that $α$ changes by about an order of magnitude within the disc, being smaller far away from the black hole and larger in the plunging region close in. It was found that the behaviour of $α$ reflects some underlying, fundamental properties of the stress $\mathbb{T}$ itself. In particular, as argued by several authors, the stress must be zero at the black hole horizon. We notice that the stress calculated in GRMHD simulations by different authors, including us, has a maximum rather close to the location of the circular photon orbit. We propose a formula that accurately describes this universal behaviour of $α$ in terms of the "gyration radius'', a physical characteristic of rotation well known in Newtonian dynamics and in the black hole case uniquely defined by the Kerr space-time geometry. Analytic and semi-analytic models of black hole accretion discs provide an invaluable insight into fundamental physics, and the GRMHD simulations do not aspire to replace them. Rather, simulations could help to improve analytic models by making them more realistic. For example, our $α$-formula, deduced from the GRMHD simulations, may be handy in the construction of improved versions of thin and slim disc models.

The SPEARHEAD (specification, analysis, and re-calibration of high-energy particle data) project, funded by the European Union Horizon Europe programme, enhances the accuracy and usability of high-energy particle measurements. It investigates particle acceleration, release, and transport during solar eruptions by refining instrument response functions and cross-calibrating datasets. We present a comprehensive catalogue of greater than 100 MeV proton events identified from May 1996 to August 2024 using SOHO-ERNE penetrating-particle rates, together with associated solar phenomena derived from multi-instrument observations. The SEP events were detected through a systematic scan of ERNE-HED counter data and cross-calibrated with SOHO-EPHIN to derive peak fluxes and fluences. Each event was associated with its likely parent eruption using X-ray (XR) (GOES-XRS, RHESSI, SolO-STIX), radio (Wind-WAVES, STEREO-WAVES, ground-based observatories), and gamma-ray (Fermi-LAT) observations, CMEs (SOHO-LASCO), and ground-level enhancements (GLEs) (neutron monitors). Timings and physical properties were systematically compared to investigate the relationships between SEP onset, flare evolution, CME kinematics, and radio signatures. Statistical analyses reveal that most SEP releases closely follow flare and CME onsets, with moderate SEP-XR-CME correlations, and a strong SEP-GLE fluence link. These results indicate that high-energy SEP events are typically associated with strong solar activity signatures, with the observed intensities and timings strongly modulated by magnetic connectivity and coronal conditions. This catalogue provides the most extensive reference to date for high-energy SEP events over solar cycles 23-25, establishing a unified framework for future investigations of extreme particle acceleration.

We aim to identify the physical mechanism responsible for the observed X-ray emission and polarization of Cyg X-1 in the soft state. We perform a detailed spectral analysis of X-ray data obtained by NICER, NuSTAR, and INTEGRAL during observations simultaneous with IXPE on 2023 June 20, supplemented at higher energies with archival CGRO data. We develop a new model, retBB, that describes thermal disk emission and its returning reflection, and apply it together with accurate models of Comptonization and relativistically broadened reflection. Using the resulting spectral solution, we compute the expected polarization signal and compare it with the IXPE measurements. Our spectropolarimetric modeling shows that the observed polarization is produced by Comptonization in a corona undergoing a semi-relativistic outflow with a velocity of ~0.3c. Our spectral solutions admit either low or high black-hole spin values, depending on the adopted model setup. However, the observed polarization strongly favors a low spin. At high spin, the polarization angle would inevitably rotate significantly across the energy band, which is not consistent with the observations. Apart from this rotation of the polarization angle, general relativistic effects do not play a significant role in producing the observed polarization. In particular, we find that, at most, returning disk radiation contributes only a minor amount.

The Peanut cluster (SRGe J023820.8+200556, SRGe CL0238.3+2005, $z_{spec}$ = 0.42) has recently emerged as a candidate for a rare, massive merger, potentially analogous to the Bullet cluster. We present the results of optical identification and spectroscopic redshift measurements for 31 galaxies in the Peanut cluster, including 26 new redshifts obtained with the 6-m telescope BTA (Big Telescope Alt-azimuthal) at SAO RAS between October 2024 and January 2025. The derived distribution of line-of-sight velocities reveals the possible presence of two subclusters with a line-of-sight velocity difference of ~2000 km/s. However, statistical tests and the Dressler-Schectman test show that the hypothesis that the observed velocity distribution can be described by a normal distribution for a single cluster cannot be ruled out, and the evidence for the existence of two gravitationally bound substructures remains ambiguous. Assuming a single cluster with the normal velocity distribution, the estimated galaxy velocity dispersion is $σ_{los} = 1455 \pm 83$ km/s, corresponding to the total cluster mass of $M_{200} = 2 \times 10^{15} M_\odot$ based on the mass-velocity dispersion scaling relation. In either scenario -- a single extremely massive cluster or an ongoing merger -- the Peanut cluster appears to be a very rare and peculiar object, comparable to such extreme systems as the Bullet cluster (1E 0657-56) or El Gordo (ACT-CL J0102-4915).

The recent evidence for a nanohertz gravitational wave background from Pulsar Timing Arrays highlights the urgent need to identify electromagnetic counterparts to supermassive binary black holes. Here, we perform global 3D general relativistic magnetohydrodynamic (GRMHD) simulations of a secondary black hole (mass ratio $q=0.1$) interacting with a Magnetically Arrested Disk around a primary black hole using a time-dependent superposed Kerr-Schild metric and post-processed general relativistic radiation transfer calculations based on thermal electron distribution function (eDF). We explore three orbital configurations: a vertical impact orbit, a coplanar embedded orbit, and a high-spin, eccentric, inclined scenario. Despite clear orbital periodicity and recurrent shock formation, the thermal synchrotron light curves frequently lack expected shock-induced flares. In vertical impacts, shock brightenings are typically sub-dominant to the stochastic MAD variability of the primary black hole, unless viewed at specific alignment phases. Conversely, coplanar orbits produce distinctive, rapid flares driven by gravitational self-lensing. We identify a frequency-dependent emission hierarchy: the primary black hole dominates sub-millimeter flux, while the secondary dominates near-infrared emission due to higher electron temperatures in thermal eDF. Finally, spin-orbit coupling drives Lense-Thirring precession, yielding twisted, wobbling jets reminiscent of OJ~287. Crucially, we demonstrate that intrinsic MAD turbulence may easily mask shock-induced flares at radio frequencies. We strongly advocate coordinated sub-millimeter and near-infrared monitoring to robustly isolate supermassive binary black hole self-lensing signatures.

Pulsar timing of double neutron star (DNS) systems is one of the best methodologies to study the neutron star masses distribution. Here we report the discovery of a double neutron star system PSR J0641+0448 in the Five-hundred-meter Aperture Spherical radio Telescope (FAST) Galactic Plane Pulsar Snapshot (GPPS) survey. This pulsar has a 25.7 ms spin period and moves in a 3.73-days eccentric orbit with an eccentricity of 0.145. Using FAST observations, we obtained its phase-connected timing solution with periastron advance and Shapiro delay detected. Using $χ^2$ analysis based on DDGR model, we constrain the pulsar mass to $1.319^{+0.021}_{-0.035}~M_\odot$, and the companion mass to $1.269^{+0.022}_{-0.016}~M_\odot$ with a 68.3\% confidence level. The low companion mass and mild orbital eccentricity is consistent with the correlation between neutron masses and orbital eccentricities.

We study the properties of galaxy cluster 2-point correlation function covariance matrices estimated using the linear-construction (LC) method, which is computationally up to 20 times faster than the standard sample-covariance method. Our goal is to assess how well the LC method performs in cosmological parameter estimation compared to the sample covariance. We use a set of 1000 mock dark matter halo catalogues to compute both the LC-covariance and the sample-covariance estimates in four redshift shells. These numerical matrices are used to fit a theoretical four-parameter model for the covariance. We then use the two fitted covariance models in a likelihood function to estimate two cosmological parameters - the matter density parameter $Ω_{\rm m}$ and the amplitude of the matter density fluctuations $σ_8$ - from the simulated mock catalogues. The purpose of this is to validate the LC-covariance-based model against the sample-covariance model. The catalogues were simulated assuming the spatially flat $Λ$CDM cosmology, with $Ω_{\rm m} = 0.30711$ and $σ_8=0.8288$. We find that the parameter posteriors obtained using the sample- and LC-covariance models agree well with each other and with the simulation cosmology. The two pairs of marginalized constraints are $Ω_{\rm m} = 0.307 \pm 0.003$ and $σ_8 = 0.826\pm 0.009$ (sample covariance), and $Ω_{\rm m} = 0.308 \pm 0.003$ and $σ_8 = 0.825 \pm 0.009$ (LC covariance). The posterior widths are the same, and the difference in the median values is less than $0.16\,σ$ for both parameters.

Recently, a number of methods have emerged to detect galaxy clusters solely through their weak lensing signal. Using the recently-introduced wavelet multi-scale detection method, we focus here on the potential for the use of tomographic information of the source galaxies to increase the number of weak lensing detections. We apply the $z_{s,\mathrm{min}}$-cut technique, consisting of the combination of weak lensing peak detections emerging from lensing maps obtained using different source redshift bins, to mock data sets of progressively increasing sophistication. The source redshift distribution is chosen to be $Euclid$-like, with a maximum depth of $z_{s,\mathrm{max}}=3$, and overlapping tomographic redshift bins are constructed by progressively increasing the minimum source redshift $z_{s,\mathrm{min}}$. Considering all possible detection combinations from one to four tomographic bins, we find that a single source redshift bin, with $z_{s,\mathrm{min}}=0.4$, performs as well as the combination of multiple redshift bins. By running detections on synthetic clusters of varying complexity -- from isolated Navarro Frenk White haloes to haloes embedded in and formed within N-body cosmological simulations, and considering both true and photometric source redshifts -- we show that while large-scale structure contamination and photometric redshift errors reduce the potential gains of the tomographic approach, the dominant limitation is the accumulation of spurious detections across redshift bins, leading to decreased purity at a fixed detection threshold.

The accretion of metal-poor gas sustains galactic star formation. In the Milky Way, this process is fueled by high-velocity clouds (HVCs), yet their fundamental properties have remained elusive in the absence of stellar tracers. Here we report a binary open cluster within HVC Complex H. With an age of 11.2 +- 0.6 Myr and a subsolar metallicity of 0.05(+0.05-0.02) Zsun, the clusters provide a direct stellar distance anchor to the cloud at 13.8 +- 0.6 kpc. Their proper motions indicate Complex H is on a prograde, south-to-north orbit through the outer Galactic disk. The resulting interaction produces a 'slow-fast-slow' velocity gradient, with the cloud's outer layers decelerating as they merge into the disk. Orbit integration suggests the clusters formed from an internal cloud-cloud collision. This triggering mechanism implies other HVCs could similarly produce high-velocity stars. The scarcity of previous stellar detections in HVCs is explained by the rapid escape of young stars (< 20 Myr), while CO non-detections may stem from weak emission due to low metallicity and gas dispersal. This work reveals that the circumgalactic medium can sustain star formation, offering a tangible laboratory to probe the physical conditions of accreting gas before it merges with the Galactic disk.

Remnant black holes (BHs) of massive stars (``light seeds'') are a potential origin for supermassive black holes (SMBHs). We use magnetohydrodynamic simulations to study the formation and growth of light seeds in star-forming giant molecular clouds (GMCs) with masses $10^5$--$10^9\,M_\odot$, which evolve for $\sim 10$--$30\,\rm Myr$ and form compact star clusters, akin to high-redshift nuclear star clusters. In particular, the simulations resolve very massive stars (VMSs, 100--$300\,M_\odot$), including their radiative and mechanical feedback, and model feedback-regulated accretion onto remnant BHs. We find that, even in compact GMCs capable of forming deep potential wells, the gas reservoir is expelled by sustained stellar feedback and rapidly dispersed after supernova explosions. Remnant BH populations emerge $\sim 3\,\rm Myr$ after the starburst and concentrate at the cluster center (where $ρ_{\rm BH}\sim 10^4$--$10^6\,M_\odot\,{\rm pc}^{-3}$). With our fiducial sub-grid BH accretion/feedback model, in-situ BH accretion is inefficient for forming heavy seeds: some direct-collapse BHs briefly accrete at $\sim$\,(1--10)$\times$ the Eddington rate, but they reach only $\sim 400$--$500\,M_\odot$. A top-heavy initial mass function or natal kicks do not change this conclusion. Runaway accretion is only possible if the sub-grid BH model allows a high fraction of Bondi inflow to reach the BH, in which case a few seeds can grow to $\sim 10^6\,M_\odot$. We also discuss multiple-generation star formation that may be intrinsically correlated with remnant BH accretion.

Extended radio sources present unique challenges for automated detection and classification in wide-field radio surveys. With current surveys such as the Evolutionary Map of the Universe (EMU), robust and scalable methods are essential to identify and catalogue these complex sources. We apply three automatic approaches to detect complex radio emission in EMU observations of the Galaxy And Mass Assembly (GAMA) 09 field (EMU-G09) in order to evaluate their relative strengths and limitations in preparation for large-scale application across future EMU data releases. These include DRAGNHunter, designed to detect likely DRAGNs (Double Radio sources associated with Active Galactic Nuclei) from a component catalogue; coarse-grained complexity, a metric designed to highlight regions of complex emission; and RG-CAT, a machine learning pipeline trained on radio sources identified in the EMU pilot survey. We find that together, the three methods recover nearly all extended sources in EMU-G09 but identify largely distinct, partially-overlapping subsets, with only 375 sources identified by all finders. This demonstrates that a combination of complementary techniques will be required to achieve a complete census of extended radio sources in future large-scale surveys.

Aims. We aim to determine how ion neutral coupling and ambipolar diffusion affect the linear and the nonlinear growth of the RTinstability under astrophysically relevant conditions, and to identify the coupling regimes in which departures from the classical single fluid picture become significant. Methods. We perform high resolution two fluid numerical simulations using the MPI AMRVAC code, spanning a wide range of perturbation wavelengths, coupling strengths, from uncoupled to strongly coupled passing by intermediate or ambipolar diffusion dominated regimes, and magnetic field configurations. The linear theory is revisited using a physically consistent formulation with different ion neutral coupling strengths across the interface and validated against the simulations. We investigate the physics of the instability using morphology based diagnostics of the mixing layer to compare simulations at equivalent nonlinear stages, complemented by spectral, force, and energy budgets analyses. Results. In the linear regime, theoretical growth rates are recovered over a wide range of wavelengths, from the single fluid limit to intermediate bi fluid coupling. In the nonlinear regime, ambipolar diffusion modifies the classical quadratic growth and introduces a coupling dependent evolution. For multi wavelength perturbations, the nonlinear dynamics becomes strongly scale dependent: intermediate coupling enhances fragmentation in hydrodynamic configurations, while magnetised cases exhibit a non monotonic reorganisation of the interface, with the smoothest morphologies occurring at intermediate coupling. Spectral and energetic diagnostics indicate that these behaviours correlate with changes in the relative contributions of ion neutral drift and magnetic stresses during thenonlinear evolution

The diversity of pulsar light-curves and radio polarisation properties originates in the structure of the magnetic field close to the stellar surface. For millisecond pulsars, this complexity is particularly puzzling. Fortunately, some means exist to uncover the magnetic field topology which indeed impacts the emission within the magnetosphere but also on the surface through its hot spot thermal radiation. We aim at deducing a plausible magnetic field geometry for the millisecond pulsar J0437$-$4715 by using combined information from the soft X-ray hot spot geometry deduced from NICER observations by pulse profile modelling and from radio and $γ$-ray pulse profile fitting. We also check the consistency between the geometry obtained and the radio polarisation data. Our $γ$-ray light-curve shapes rely on the striped wind model, whereas the radio polarisation fits rely on the rotating vector model. The magnetosphere structure is obtained from dipolar force-free magnetosphere simulations. We demonstrate that a slightly off-centred dipole augmented by a small scale dipole located on one polar cap explains simultaneously the shape of the hot spot and the radio and $γ$-ray data with a magnetic obliquity of $α\approx (42\pm5) \degr$ and a line-of-sight inclination angle of $ζ\approx (136 \pm5) \degr$. Our simple dipole model reproduces all the radio and $γ$-ray characteristics of PSR~J0437$-$4715, including its radio polarisation data. It shows that the radio emission could be produced in regions where the magnetic field is mainly of dipolar nature.

We aim to investigate the blowout jet-like prominence eruption, which occurred on October 6$^{th}$, 2023, with the help of imaging and spectroscopic observations. Firstly, the prominence rises slowly with a speed of 33 km/s, followed by a fast rise (i.e., 338 km/s). Later, the northern leg breaks completely, and the eruption forms the blowout jet. The jet consists of different plasma threads, which show a range of upflow (i.e., 125 to 593 km/s) and downflow velocities (i.e., 43 to 158 km/s). The jet plasma column exhibits transverse oscillations, and this motion (untwisting motion) propagate at the speed of 267 km/s, are consistent with being Alfev{é}n waves. The transverse motion has the time period, amplitude, and transverse velocity of 1332 s, 26.19 Mm, and 126.18$\pm$7.27 km/s, respectively, and this transverse oscillation decays over time. Interestingly, the different plasma threads within the jet's body exhibit decayless transverse oscillations, and these decayless oscillations are related to the main decaying transverse oscillation. The transverse velocity of these decayless oscillations ranges from 66 to 30 km/s, the amplitudes from 8.52 to 2.74 Mm, and periods from 811 to 406 s. In addition, the spectroscopic analysis reveals Si~{\sc iv} lines are forming in the optically thick conditions in high electron density regions (i.e., near the base of the blowout jet). Lastly, we mention that two weak C-class flares occurred during this event, and further, one CME also occurred, which propagated with the speed of $\sim$250 km/s.

This study presents an investigation of nearly two dozen candidate Luminous Blue Variables (cLBVs) in the galaxies M31 and M33. Eight stars have been studied in detail, while an additional sixteen objects are briefly mentioned. Multi-epoch spectra of confirmed cLBVs from LAMOST and previous literature show broad hydrogen, He I lines, abundant Fe II and [Fe II] emission lines, and discernible spectral variability, consistent with the characteristics of known LBVs. Low outflow velocities inferred from P Cygni profiles are also incorporated into the classification criteria. Moreover, key stellar properties, including temperature and luminosity, are determined using the Spectral Energy Distribution (SED) fitting and spectral modeling. By comparison with stellar evolutionary tracks on the temperature luminosity diagram, the initial masses are estimated to be in the range of approximately 32 to 60 $M_{\odot}$. Except for J013401 and J013411, other stars locate within the typical LBV region between the S Doradus instability strip and their outburst phase. More importantly, our sample, except for the binary system, are all positioned in the LBVs region rather than that of B[e]SGs in the near-infrared color-color diagram. Based on all available information, one of the eight sources is confirmed as an LBV, four stars are designated as high-probability cLBVs, and the remaining three stars await further photometric observations to secure their classification. Given the current scarcity of known cLBVs, our study has the potential to make a significant increase in the number of LBVs in M31 and M33.

Matter ejected from the magnetar crust during giant flares (GFs) may undergo $r$-process nucleosynthesis, producing short-lived optical transients termed "novae breves". Although intrinsically much fainter than kilonovae from compact binary mergers, novae breves may occur within or near the Galaxy, making them promising observational targets. We aim to investigate how the neutron star (NS) equation of state (EOS) and the mass of the central magnetar affect the ejecta properties following GFs and the resulting nova brevis emission. We employ a semi-analytical ejecta model combined with nuclear reaction network calculations to compute nucleosynthesis yields and multi-band light curves for different EOSs and magnetar masses, and assess their detectability with current and future facilities. We find that variations in the EOS and magnetar mass modify the ejecta mass and its density and velocity distributions, etc., leading to observable differences in nova brevis light curves. In particular, both the peak luminosity and the characteristic peak timescale are EOS-dependent. Assuming a fixed Galactic magnetar mass of 1.4 solar mass and taking the $u$ band as an example, we find that the minimum apparent AB magnitudes range from 7 mag (H4 EOS) to 8.5 mag (WFF EOS) with peak timescales of 100-1000 s. A more massive magnetar produces fainter emission with a shorter peak timescale. For a magnetar mass of 1.4 solar mass, novae breves associated with known magnetars may reach peak luminosities of 1e37-1e39 erg/s, enabling targeted searches, particularly following high-energy GF alerts. Moreover, a detection horizon of 10 Mpc or beyond is achievable with current and future facilities, allowing searches for novae breves from previously unknown magnetars in the Local Volume. Although challenging, detection of such rapidly evolving transients is feasible.

Infrared (IR) excess observed around white dwarfs (WDs) is typically attributed to companions or debris disks. These systems are interesting because they offer a unique opportunity to study the late stages of stellar evolution and the interactions between WDs and surrounding material. The 11th data release (DR11) of the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) - one of the largest spectroscopic surveys to date - has recently provided spectra for 3092 WDs, many of which have yet to be systematically investigated for IR excess. In this study, we cross-correlated the LAMOST DR11 WD catalog with optical and IR surveys, including the Sloan Digital Sky Survey (SDSS), Two Micron All-Sky Survey (2MASS), UKIRT Infrared Deep Sky Survey (UKIDSS), and Wide-field Infrared Survey Explorer (WISE). We performed spectral energy distribution fitting using the VOSA tool for 1818 WDs and identified 167 IR excess WD candidates. After excluding 23 sources with potential contamination within 6" and five additional sources identified through WISE ccf flag analysis, we identified 139 objects with candidate IR excess. These include 30 candidate WD + M dwarf binaries (18 new systems), 19 candidate WD + brown dwarf (BD) binaries (eight new systems), 66 candidate WD + dust disks (38 new systems), and 24 candidate either WD + BD or WD + dust disks (19 new systems). Given the limited spatial resolution of WISE, all candidate systems require follow-up IR observations for confirmation, such as high spatial resolution imaging or IR spectroscopy. This will help expand the parameter space of dust disks, allowing us to explore a broader range of possibilities.

We present a multi-wavelength study of the Flat Spectrum Radio Quasar PKS 1222+216, analyzing its long-term variability of radio data obtained in 2013-2020 from the iMOGABA, MOJAVE, and VLBA-BU-BLAZAR programs, along with $γ$-ray data from Fermi-LAT. We found that the radio flux densities at 15, 22, 43, and 86 GHz declined exponentially by 37%-56% over a year following a $γ$-ray flare in November 2014. We estimated jet physical parameters through Gaussian model fitting of VLBA 43 GHz data, identifying 10 jet components. The cooling timescales of the jet emission regions, i.e., newly ejected components C9, C10, and C11, range from 43 to 222 days, with the estimated jet viewing angles of approximately 8 degrees and magnetic field strengths of 77-134 mG in the jet emission regions. Additionally, by determining the magnetic field strength at different frequencies, we found that the magnetic field scales as $B\propto r^{-0.3\pm0.04}$, indicating a non-equipartition condition ($k_\text{r}\gtrsim 1$) or a slow decline in magnetic field strength profile ($m<1$). By analyzing component ejection times, we discovered that the $γ$-ray flare in 2014 coincided with the interaction between the moving component C9 and the stationary feature A1. We estimated that the $γ$-ray emission region is located at $9.2\pm1.0$ pc from the central engine, beyond the BLR and dusty torus, suggesting that the seed photons for inverse Compton scattering originate from the jet itself, external CMB radiation, or a surrounding sheath. Our results favor a scenario where $γ$-ray emission originates further downstream from the central engine through interactions between moving and stationary components. Additionally, our study presents an alternative method for estimating magnetic field strengths in AGNs undergoing long-term synchrotron cooling based on the associated timescale.

China's Tianwen-1 Mars orbiter successfully imaged the third interstellar object, 3I/ATLAS, during its close encounter with Mars using the onboard HiRIC CMOS camera. This is China's first deep-space observation of an astronomical object. These observations constitute the first imaging of this object from a vantage point significantly out of its orbital plane, providing a unique constraint on dust dynamics. Three observing epochs between 2025 September 30 and October 3 reveal clear changes in coma and tail morphology driven by the rapidly evolving viewing geometry. Comparison with Finson-Probstein dust dynamical models indicates that the coma is dominated by large grains with solar radiation pressure parameter $β\approx 10^{-3} $ - $10^{-2}$, corresponding to grain sizes of a few 100s $μ$m. The extent of the sunward coma implies dust ejection velocities of $3$ - $10$ m s$^{-1}$. Despite the morphological evolution, the azimuthally averaged surface brightness profile remains nearly unchanged through the three epochs, transitioning from a radial slope near -1 close to the nucleus to slightly steeper than -1.5 at larger cometocentric distances, consistent with steady-state dust outflow accelerated by solar radiation pressure. Photometry yields an average $Afρ\sim (2.0\pm0.2)\times10^4$ cm and a corresponding dust mass loss rate of $\dot{M} \sim 10^3$ kg s$^{-1}$. The dominance of large grains in both interstellar comets discovered to date, 2I/Borisov and 3I/ATLAS, together with their high supervolatile contents, may indicate that these objects originate from the outer regions of their parent planetary disks.

We propose a pulsar candidate cross matching algorithm to sift radio pulsar search candidates from repeated observations of the same sky location such as globular clusters, high energy sources, or supernova remnants. Our method uses both the candidate spin period ($P$) and dispersion measure (DM) value; if two or more candidates from different observations have similar spin periods to within 1\%, and dispersion measure values within 10\%, they are likely to correspond to the same candidate detection. We have demonstrated the effectiveness of our method through the discovery of the pulsar M12B with the Five-hundred-meter Aperture Spherical radio Telescope (FAST). This pulsar has a spin period of 2.76\,ms and a dispersion measure of $42.70 \pm 0.05\,\mathrm{cm}^{-3}~\mathrm{pc}$. This pulsar has a profile with three peaks, being faint, showing scintillation. It is in an approximately 0.53-day orbit. Our discovery indicates that more pulsars might be effectively discovered if the algorithm is applied to the search results from other archival globular cluster observations.

Ultraluminous X-ray sources represent extreme super-Eddington accretion regimes, and a subset is now known to host highly magnetized neutron stars. However, direct observational probes of their surface magnetic fields remain scarce. In this Letter, we report the detection of a narrow X-ray absorption feature at $3.29\pm0.02$ keV in the XMM$-$Newton/EPIC-pn spectrum of NGC 4656 ULX-1. The source exhibits a hard-ultraluminous state, while our timing analysis reveals a candidate pulsation at $\sim$0.9736 Hz, with a local significance of $5.5σ$ and a pulsed fraction of $\sim11\%$. The feature is robust against changes in continuum modeling and data-selection criteria, retaining a statistical significance of $\gtrsim3σ$ in Monte Carlo simulations. Interpreting the absorption as a proton cyclotron resonant scattering feature implies a local magnetic field of $B\sim(6-7)\times10^{14}$ G in the line-forming region. This value is consistent with strong magnetic fields anchored near the neutron star surface, even if the large-scale dipole is substantially weaker. Although we discuss electron cyclotron features and atomic transitions as possible alternatives, they appear less consistent with the observed phenomenology.

Galaxy clusters are the largest gravitationally bound systems, and they continue their growth through mergers in a hierarchical ΛCDM Universe. Therefore, we can describe the merger stage of a cluster as the dynamical state of clusters. Previous studies have investigated this phenomenon, but several limitations remain, including reliance on dichotomous classifications, constraints on the number of indicators used, absence of reliability, and incompatibility of methods between observation and simulation studies. To overcome this, we developed an enhanced and observation-applicable cluster dynamical state classification method using the Bayesian classifier with the class-conditional Gaussian mixture distribution model using the N-cluster Run simulation data. The Bayesian classifier was designed for two merger stages (merger and relaxed) as well as three merger stages (recent merger, ancient merger, and relaxed) to provide a more detailed interpretation of the merger processes. In the results, using a larger number of indicators yields better results, with their order of importance being: magnitude difference, center offset, sparsity, Kuiper V statistic, and mirror asymmetry. Additionally, our analyses show that a projected classifier (built on the 6D space, but evaluated on lower dimensional projections) consistently produces better outcomes than non-projected classifiers (i.e., classifiers built directly on the corresponding low dimensional spaces), which means limited observation data can be used to classify with enhanced performance. Furthermore, the new classification method outperforms our previous research. This new method can suggest a way of overcoming previous limitations and provides new insights by providing the reliability of dynamical state classification results.

We propose a variable ADAF disk model for X-ray binary systems. In this model, the accretion flow consists of an outer thin disk and an inner optically thick advection-dominated accretion flow (ADAF) torus. The size of the turbulent ADAF is variable. A complete cycle of ADAF contraction, transition to a thin disk, and subsequent re-expansion corresponds to the rapid rise, peak, and decay phases observed in the X-ray outbursts of black hole binaries. This cycle also tracks the canonical evolution through the low-hard, high-soft, and back to the low-hard state in the hardness-intensity diagram. The model unifies the presence of near-ISCO Fe emission lines with the truncated disk paradigm, as observed in the black hole system GX 339-4. It explains the 35-day period in the neutron star system Her X-1 more effectively through variable ADAF sizes than through a precessing disk. This variable ADAF framework may be extended to explain similar phenomena in active galactic nuclei.

In neutron star (NS) magnetospheres, plasma waves propagate as normal modes with distinct propagation dynamics that strongly influence observable signals. This letter presents a unified theory of linear mode conversion between Alfv'en (A), superluminal ordinary (O), and extraordinary (X) modes, incorporating the effect of magnetic-field geometry and local plasma response. Magnetic field-line curvature induces A-X conversion for low frequencies and O-X conversion at high frequencies, whereas plasma gradients alone do not drive X-mode coupling. We show that a single dimensionless parameter controls both conversion channels. The conversion efficiency follows the universal nonadiabatic transition probability of a multilevel quantum system. Efficient conversion occurs within a narrow angular window between the wave vector and magnetic field, localizing potential conversion sites in the NS magnetosphere. This linear mechanism naturally accounts for complex polarization features observed in pulsars and some fast radio bursts.

Cygnus~X-3 is a Galactic X-ray binary with a 4.8-hr orbital period operating in the ultraluminous regime. Although the system is viewed at relatively low inclination ($i\approx28^\circ$), it exhibits a deep orbital modulation. Recent IXPE observations show strong linear polarization orthogonal to the radio jet, indicating that the X-ray emission is dominated by reflection from the inner walls of a supercritical outflow funnel. We propose a Hybrid Roche-lobe overflow (RLOF) scenario in which a massive Wolf-Rayet donor effectively fills its Roche lobe with a focused wind driving a super-Eddington accretion stream. Using a numerical synthesis of the folded light curve, we show that the modulation is reproduced when the central funnel is periodically occulted by a vertically extended, shock-heated Turbulent Wall formed by stream impact on the outer disk rim. This produces a phase lag ($Δφ\approx0.11$) between X-ray minimum and binary conjunction, with extended attenuation by the WR wind defining a broader Suppression Region. This geometry explains the enhanced iron-line equivalent width during X-ray minimum via a coronagraphic effect. The large radial-velocity amplitude of FeXXVI measured by XRISM ($K_{\rm obs}\approx430$ km s$^{-1}$) and its zero-crossing at $φ_X=0.0$ arise naturally in the stream-impact region rather than from orbital motion of the compact object. Finally, we show that the observed secular orbital expansion ($\dot P>0$) follows directly from highly non-conservative mass transfer with inner-disk mass loss, indicating that Cygnus~X-3 is a stable, long-lived system in a supercritical accretion regime.

Radio bridges are diffuse synchrotron emission observed between merging galaxy clusters. Recent radio observations have reported both detections and non-detections of radio bridges between clusters. The detections imply the presence of cosmic rays (CRs) and magnetic fields permeating the cosmic web that produce synchrotron emission observable with current facilities, whereas the non-detections suggest that specific physical conditions are required for their formation. We study the CR reacceleration by solenoidal turbulence in the filament connecting two massive clusters at an early stage of the merger. Our aim is to test whether this mechanism can generate diffuse emission in the inter-cluster region. We perform a cosmological magneto-hydrodynamical (MHD) simulation using the Enzo code. We improved a run-time Lagrangian tracer method implemented in Enzo, and follow the trajectories of baryonic matter using $N=\mathcal{O}(10^7)$ tracer particles. In post-processing, we conduct a parallel computation of the Fokker-Planck (FP) equation for all tracers, with cooling and reacceleration efficiencies evaluated from the local quantities recorded along each tracer trajectory. Our simulation generate a Mpc-sized radio bridge in the early stage of the cluster merger. Within a reasonable parameter range, the reacceleration model produces a broad variety of spectra. In our fiducial model, the simulated bridge matches several properties of the one found between Abell 399 and Abell 401, such as its spectral shape, intensity profile, and pixel-by-pixel correlation between radio and X-ray intensities. The inter-cluster region is filled with turbulence induced by infalling mass clumps and subsequently amplified by the approaching motion of the clusters. The CR reacceleration by the turbulence is a viable mechanism to power a Mpc-sized synchrotron emission observed as radio bridges.

We present a synopsis of the project to establish thirty-two new faint ($ 16.5 \leq V \leq 19.8 $) DA white dwarfs as spectrophotometric standards distributed over the whole sky. Our results validate the use of fully radiative pure hydrogen model fluxes for hot DA white dwarfs to predict the observed broadband fluxes from near ultraviolet through the near infrared to accuracies of a few parts per thousand. After fitting the line of sight reddenings simultaneously with the model spectral energy distributions of these stars against spectroscopic and multi-band photometric observations, we have shown that residuals have an rms of typically 0.4 percent. This indicates that the complications from interstellar dust extinction have been adequately mitigated. Our stars supplement the three brighter DA white dwarfs that define the flux scale of CALSPEC. The consequent photometric accuracy, their all sky coverage, and their brightness range that matches the dynamic range of large telescopes, constitutes an unprecedented ensemble of standard stars for both ground as well as space based use. This paper targets readers who may wish to use these as standard stars, and provides for them the essential content to understand their strengths and limitations, without traversing the technical details of analysis that are already captured in a series of papers since 2016. The narrative here describes the motivation, justification, and evolution of the analysis methods; the input data that constrain the modeling; as well as the stability of our results in the face of future improvements in models.

Accurate modelling of redshift-space distortions (RSD) in galaxy clustering is essential for extracting cosmological information from current and forthcoming large-scale structure surveys. While perturbation theory is reliable on large scales, much of the constraining power lies at intermediate and small separations, where nonlinear dynamics within and between dark matter haloes dominate. We present a halo streaming model for nonlinear galaxy clustering in redshift space that is accurate and physically interpretable. Our framework combines the streaming model for RSD with a halo-model decomposition of the galaxy clustering into central/satellite and one-/two-halo contributions. We build dedicated emulators for the key physical ingredients, trained on a suite of $N$-body simulations: halo mass functions, real-space halo two-point correlation functions, and pairwise velocity moments. By emulating these modular building blocks rather than the final redshift-space observable, this approach preserves physical transparency, enables targeted optimisation for each ingredient, and remains flexible to changes in tracer populations and galaxy-halo connection models. The resulting halo streaming model reproduces the simulated nonlinear anisotropic clustering signal down to highly nonlinear scales, while achieving the computational efficiency required for cosmological parameter inference. This framework is designed to support full-shape RSD analyses for surveys such as DESI and \textit{Euclid}, facilitating precision measurements of structure growth and tests of gravity. All codes and trained emulators are publicly available in the \href{https://github.com/chzruan/freyja}{\texttt{freyja}} repository.

We tested the effectiveness of radio selection to discover heavily obscured AGNs, particularly at high-z, and we measured their abundance for the first time from a radio perspective. We consider the radio sources detected in the J1030 field, which is one of the fields with the deepest combination of 1.4 GHz radio and X-ray observations. We defined a radio excess parameter as the ratio between the star formation rate (SFR) that would correspond to the observed radio luminosity and the one directly derived from the spectral energy distribution (SED) fitting, $\rm REX=SFR_{1.4GHz}/SFR^{corr}_{SED}$. We then select as radio excess AGN those sources with $\rm REX>8.5$, corresponding to a $3σ$ excess above the median value. In this way, we find 145 radio-excess sources falling into the \textit{Chandra} X-ray image footprint but without X-ray detection. From the deep X-ray upper limits, we estimated a lower limit to the obscuration of each radio-excess AGN, finding on average $\log (N_H/\rm{cm^{-2}})>23.7$. A CTK AGN scenario is also supported by the results of the X-ray stacking analysis performed on sources at $z>1.5$, which revealed X-ray luminosities and hardness ratios compatible with very highly obscured AGN. Finally, we computed the number density of these radio-selected CTK AGN. While at $z\sim 2$ the radio number density agrees well with the CTK AGN predictions of different population synthesis models, at $z\sim3$ the radio selection returns a CTK AGN number density $\sim 2-3$ times larger than what is predicted by the X-ray models and observations. This result supports the effectiveness of radio emission in selecting the most obscured sources, unraveling a population of AGN potentially missed by X-rays surveys at $z>3$, paving the way to a synergistic use of the future radio and X-ray facilities such as the \textit{SKAO} and \textit{NewAthena}.

In this paper I review recent results on high-energy neutrino astronomy and what they can reveal about some of the most extreme cosmic accelerators. I discuss recent measurements of the diffuse TeV-PeV cosmic neutrino spectrum by the IceCube observatory and the current flux limits in the ultra-high-energy regime, contextualizing the recent detection of an ultra-high-energy neutrino by the KM3NeT observatory. I review the recent emergence of a TeV signal from nearby Seyfert galaxies such as NGC 1068, the potential of $γ$-ray blazars as neutrino sources above the PeV regime, and the current status of tidal disruption events and other transient classes as possible neutrino sources. For each of these topics, I discuss ongoing developments in source models and their current limitations. I argue for the indispensable role of next-generation multi-messenger facilities, such as IceCube-Gen2, in solidifying current source associations, probing the ultra-high-energy regime, and resolving vast transient populations that remain unidentified with current statistics.

We investigate the origin of the observed X-ray weakness in high z LRDs and other JWST-selected broad line AGN by comparing their X-ray and optical properties with those of a diverse sample of low z AGN, including super-Eddington accreting massive black holes (SEAMBHs), NLS1s, and type I AGN from large surveys. We examine the relations between X-ray luminosity, broad Hα line luminosity, Eddington ratio, bolometric luminosity and X-ray-to-bolometric luminosity correction, and explore whether high z sources may represent analogues of local highly accreting systems. While a few LRDs and JWST-selected AGN are consistent with the SEAMBHs population in the $L_x/L_{Hα}$ versus $λ_{Edd}$ plane, most lie below it, suggesting either more extreme accretion conditions, suppressed coronal emission or heavy obscuration. We identify an anti-correlation between $L_x/L_{Hα}$ and $λ_{Edd}$ in the low z, high accreting subsample, consistent with theoretical expectations of slim-disc accretion. We further show that, for SEAMBHs, $Hα$-based bolometric luminosities underestimate SED-based values even after dust correction. We find that SEAMBHs, LRDs, and JWST-selected AGN occupy a similar high-$κ_{bol,x}$ regime, indicating that the relative deficit of X-ray emission compared to the bolometric output could potentially support the view that suppression of the hot corona emission is a common feature of highly accreting systems across cosmic time. Our results are consistent with the idea that the observed X-ray weakness of LRDs and JWST-selected AGN may be linked to the physics of highly accreting SMBHs. Moreover, observational limitations at high z, including instrumental sensitivity and the steep X-ray spectra expected for highly accreting systems, likely further suppress the detected X-ray signal.

Long Gamma Ray Bursts are thought to originate from the core collapse of massive stars that give rise to energetic broad-lined Type Ic supernovae. The brightest burst ever recorded, GRB 221009A, has been linked to a broad-lined Type Ic supernova through late-time observations by the James Webb Space Telescope. An emission line evolving from $\sim$37 to $\sim$6~MeV is detected during the prompt phase. We propose that this time-evolving line is consistent with Doppler-boosted radioactive decay of nickel synthesized in the associated supernova and entrained in the relativistic jet, corresponding to the boosted 158~keV decay branch. We also report evidence for an additional higher-energy excess near $\sim$24~MeV at 290--300~s, detected at moderate statistical significance and consistent with the boosted 270~keV decay branch. The observed kinematics and flux evolution are compatible with expectations from radioactive decay, providing direct spectroscopic evidence linking prompt emission to supernova nucleosynthesis.

Accreting white dwarfs (AWDs) are among the best natural laboratories for understanding disk accretion. Their proximity, brightness, and purely classical nature make them ideal systems in which to probe the fundamental physics that governs the transport of angular momentum, the generation of outflows, and the coupling between disks, magnetospheres, and accretors. Yet despite decades of study, many critical questions remain unresolved. In this ``unreview'', we therefore focus not on what is known, but on what is unknown. What drives viscosity and sustains accretion in largely neutral disks? How are powerful winds launched, and how do they feed back on the disk and binary evolution? Why do so many systems show persistent retrograde precession, and what drives bursts in magnetic AWDs? By identifying these open problems -- and suggesting ways to resolve them -- we aim to motivate new observational, numerical, and theoretical efforts that will advance our understanding of accretion physics across all mass scales, from white dwarfs to black holes.

Photometric selection of luminous high-redshift ($z\gtrsim4$) quasars is plagued by contamination from numerous low-mass Galactic stars, reddened lower-redshift quasars, as well as compact luminous red galaxies. Confirmation of these rare objects thus requires extensive spectroscopic campaigns on 4 and 8-meter-class telescopes with relatively low success rates. Here we demonstrate the utility of SPHEREx spectrophotometric survey data for quasar confirmation with no ground-based follow-up required, "from the couch," applied to candidates from a purposefully simplistic photometric and astrometric Gaia+WISE selection down to low Galactic latitudes ($|b|\geq8^\circ$). Primarily from the detection of their strong broad H$α$ emission lines, we discover 87 new luminous $4.0 < z < 5.7$ quasars with median $M_\text{1450} = -27.5$, including 19 quasars at $z>5$, and recover 219 previously published quasars at $z>4$. We validate our SPHEREx selection with a 100% confirmation rate in ground-based spectroscopic follow-up of 29 of our new $z>4$ quasars, including 11 unpublished archival spectra. We also discover 203 additional lower-redshift quasars at $0.3 < z < 4$, consisting primarily of relatively rare highly-reddened and strong broad-absorption-line objects that are likely missed by traditional quasar surveys. Finally, we show that the Ly$α$ absorption breaks and H$α$ lines of luminous quasars are already detectable at redshifts $5.7\lesssim z\lesssim6.5$ after the completion of only the first of four all-sky surveys to be performed by SPHEREx during its planned two-year mission.

The tidal field of a black hole can turn a star into a gas stream whose orbit can precess, especially if the a black hole is rapidly spinning. In this work, we investigate the impact of precession on the light curves of tidal disruption events (TDE). To do so, we perform two-dimensional radiation-hydrodynamic simulations of the interaction of the TDE wind and luminosity with the precessed stream wrapped around the black hole. Our results show that in events with black holes of $\sim10^6~\text{M}_{\odot}$ and no orbit-spin inclination, the line of sight has little effect on the light curves, since the stream covers a small fraction of the solid angle as the precession is confined to the orbital plane. In the case of black holes of $\gtrsim10^7~\text{M}_{\odot}$ and high inclination ($i\sim90^{\circ}$), the light curve peaks can be delayed by $\sim$100 days due to presence of the precessed stream blocking the radiation in the early phase of the event. We also discuss our efforts to model self-consistently the hydrodynamic evolution of a tidal stellar stream on curved spacetimes by the presence of a massive black hole.

We study the transport of baryons in magnetically arrested accretion flows and relativistic jets using general relativistic magnetohydrodynamic simulations that incorporate a passive Eulerian tracer. The tracer allows us to reconstruct a proxy for the physical baryon density supplied by the accretion disk while excluding the mass injected numerically to maintain stability in highly magnetized, low-density regions. Applying this method to axisymmetric black hole simulations with varying spin, we show that baryon loading of the jet is intrinsically episodic and regulated by magnetic flux eruption cycles occurring in the inner accretion flow. Each eruption evacuates baryons from the innermost equatorial region, drives reconnection in extended current sheets, and expels moderately magnetized disk material along the funnel wall, establishing a recurrent mass-loading channel. In spinning black holes, shear-driven waves along the jet boundary further enhance baryon entrainment, whereas this mechanism is suppressed in the non-spinning case. For parameters representative of the black hole accretion flow in M87, we map the global structure and time evolution of the Goldreich-Julian screening boundary, defined as the surface separating regions where the plasma density is sufficient to supply the charges required to screen electric fields parallel to the magnetic field from regions that are charge starved. For spinning black holes, we find that the electromagnetic power of the jet is predominantly carried by baryon-poor plasma, with extended time intervals of charge starvation. Our results provide a framework for diagnosing jet composition, charge starvation, and reconnection-driven mass loading in magnetically arrested black hole systems, with direct implications for particle acceleration and non-thermal emission in low-luminosity accretion flows.

The LMC's Globular Clusters (GCs) bring a novel opportunity to understand the LMC's assembly history and dark matter (DM) properties, provided the kinematically outlying GCs can be reliably identified. However, traditional diagnostics like the Energy-Angular Momentum space fail because of large uncertainties on the GC velocities. In this work, we develop a new, robust statistical framework for identifying kinematically outlying LMC GCs, by using their Gaia-DR3 Proper Motions (PMs) combined with previous Line-of-Sight (LoS) velocity measurements. We use the difference between a GC's velocity vector and the average velocity vector of the surrounding red clump stars as a metric for quantifying a GC's kinematic peculiarity. We account for both the velocity measurement uncertainties and the LMC's intrinsic velocity dispersion. We find 5 LMC GCs to be kinematically outlying based on PM differences alone, and additional 6 GCs if LoS velocity information is also used. Majority of the GCs with outlying PMs are clustered at a distance of 3-4 kpc from the LMC center. The inclusion of outlying LMC GCs introduces a bias of upto 30% in the LMC's enclosed mass estimates derived using GCs as dynamical tracers; caution must be exercised in choosing the GC sample for precisely determining the LMC's DM content. We discuss the possibility that the kinematically outlying LMC GCs may have been accreted from external galaxies, and motivate future spectroscopic follow-up of the GC population to better understand the assembly history of massive satellite galaxies of Milky Way like hosts.

We investigate the cosmological constraining power of combined weak galaxy lensing and galaxy clustering probes, i.e. $3\times2$-point analyses, assuming flexible models for redshift uncertainty, and Lagrangian perturbation theory and hybrid effective field theory models for galaxy intrinsic alignments, galaxy bias and baryonic physics. In this context, we provide a detailed accounting of the limiting systematics on $3\times2$-point analyses. Our main finding is that in the presence of current levels of uncertainty on baryonic physics, the information content of weak lensing analyses saturates on quasi-linear scales, allowing the use of source galaxy samples that are significantly less dense, e.g. with number densities of $5\rm \, arcmin^{-2}$, without sacrificing constraining power, provided that redshift distributions can be calibrated at the $σ(\langle z\rangle)=0.005$ level. We show that for sufficiently narrow lens and source redshift distributions, intrinsic alignment contributions can be largely self-calibrated, though sufficient flexibility must be given to the redshift and scale dependence of this signal. The near optimality of such relatively sparse source galaxy samples opens the possibility to directly calibrate the redshift distributions and intrinsic alignment contamination of such a sample using a spectroscopic instrument like DESI, thus mitigating the dominant systematics in weak lensing analyses.

We present the results of a detailed kinematic analysis of a significant fraction of the known population of Galactic star clusters aimed at constraining the physical mechanisms driving the onset and evolution of cluster rotation. Our study reveals for the very first time the presence of rotation in clusters at any age, with about $25\%-30\%$ of systems in the sample showing significant evidence of rotation. This result increases by a factor of $\sim5$ the number of clusters identified as rotators so far and it finally enables an observational reading of cluster rotation as a function of time. Young ($<500$ Myr) clusters show a larger range of rotation velocities than older systems. In addition, at young ages we observe a significantly larger fraction ($50\%-60\%$) of rotating systems than at older ones ($\sim 15\%$). These purely empirical results are compatible with rotation being imprinted during the very early stages of cluster formation and early evolution and then being progressively erased by the long-term effects of dynamical evolution. For the sub-sample of clusters for which we were able to perform a full 3D analysis, we calculated the angle between the internal rotation axis and that of the cluster orbital motion. Interestingly, while for clusters with an age smaller than their orbital period we observe similar fractions of prograde and retrograde systems, more evolved clusters appear to be preferentially prograde. We argue that such a behavior is in qualitative agreement with the expectations for the evolution of systems in which primordial rotation was imprinted by the parent molecular cloud and/or by the following hierarchical cluster assembly processes, and in which internal cluster dynamics and interactions with the Galactic field have induced a torque-driven alignment between cluster rotation and orbital motion.

We report on a pair of X-ray Multi-Mirror Mission (XMM-Newton) observations of the Supergiant X-ray binary 4U 1909+07, which were performed on 2021 October 3 and 2021 October 8, respectively. We measure the neutron star rotation period in each observation to be $\sim$602.62 s. This continues a long spin-up trend that has persisted since 2001 where the neutron star spin period was found to be $\sim$604.66 s. In our timing analysis, we observe strong variations in the amplitude of the 1--10 keV pulse profile as a function of time, and for the first time we find a low flux interval extending for a single pulse period in which pulsations are no longer detected. We interpret this low flux interval as a pulse dropout similar to those observed in Vela X-1 and GX 301-2, which were each explained by a low-density cavity in the wind driving the propeller effect. In our time-resolved spectral analysis, we observed the spectral continuum, which can be described as an absorbed power law modified by a high-energy cutoff, to significantly soften during the pulse-dropout phase. No evidence of an increasing absorption column density was found. The observed softening in 4U 1909+07 also supports an interpretation that the observed pulse dropout may be driven by the propeller effect, but the quasi-spherical settling accretion regime cannot be ruled out.

We present new observations from JWST NIRCam that reveal a striking kpc-wide cavity in the stellar distribution of the central galaxy in the cluster Abell402. Supporting data from HST allow us to rule out extinction due to dust as an explanation and, instead, suggest that this is a localized depression in the stellar density field corresponding to ~2x10^9 Msun in missing stars within a volume of 0.5kpc^3. On larger scales, both the JWST and HST data show evidence for a 2.2kpc flattened core in the stellar distribution (on which the smaller-scale cavity is superimposed), which implies the presence of a central ultra-massive black hole with M_BH = 6 +/- 4 x10^10 Msun. We report evidence for a mid-IR-bright point source at one edge of the cavity, suggesting that this black hole is actively accreting. MUSE spectroscopy reveal that this source is a LINER AGN and that there is a second candidate AGN on the opposite side of the cavity with a relative velocity of 370km/s -- if real, this implies the presence of a kpc-separation dual AGN with a total binary mass of 6 +/- 2 x10^10 Msun, which would make this the most massive binary black hole system discovered to date. We propose that this unique stellar cavity is the result of a short-lived dynamical interaction between at least one supermassive black hole and the background stellar density field, caused either by three-body scattering during binary hardening or the induction of a dipole instability in the stellar density field.

Protoclusters represent sites of accelerated galaxy formation and extreme astrophysical activity characteristic of dense environments. Identifying massive protoclusters and mapping their spatial structures are therefore crucial first steps in understanding how the large-scale environment influences galaxy evolution. We combine wide-field Ly$α$ imaging from the ODIN survey with extensive DESI and ancillary spectroscopy across the extended COSMOS and XMM-LSS fields ($\approx$14 deg$^2$) to search for massive protoclusters. We confirm six systems at $z\approx 2.4$ and $z\approx 3.1$, reconstruct their three-dimensional structures, estimate descendant halo masses, and, for one structure at $z\approx 3.12$, demonstrate that overlapping narrowband filters ($NB497$ and $N501$) provide accurate redshift tomography for emission-line galaxies. One protocluster at $z\approx 2.45$ overlaps with one of the LATIS tomographic fields, enabling direct comparison between galaxy and H {\sc i} overdensities traced by Ly$α$ forest absorption. Another at $z\approx 3.12$ hosts a massive quiescent galaxy ($M_{\ast} \approx 1.2 \times 10^{11}M_\odot$), indicating early quenching in a dense environment. By comparing Ly$α$ emission properties across environments, we find that protocluster galaxies exhibit higher median line fluxes and a deficit of faint emitters relative to the field. The effect is strongest when both 2D and 3D density information are combined, indicating that galaxies in the densest protocluster cores are most affected by environmental processes. This effect is stronger at $z\approx3.1$ than at $z\approx2.4$, suggesting possible redshift evolution.