arXiv Daily Digest - 2026-02-20

Diffusion Language Models (DLMs) incur high inference cost due to iterative denoising, motivating efficient pruning. Existing pruning heuristics largely inherited from autoregressive (AR) LLMs, typically preserve attention sink tokens because AR sinks serve as stable global anchors. We show that this assumption does not hold for DLMs: the attention-sink position exhibits substantially higher variance over the full generation trajectory (measured by how the dominant sink locations shift across timesteps), indicating that sinks are often transient and less structurally essential than in AR models. Based on this observation, we propose ${\bf \texttt{Sink-Aware Pruning}}$, which automatically identifies and prunes unstable sinks in DLMs (prior studies usually keep sinks for AR LLMs). Without retraining, our method achieves a better quality-efficiency trade-off and outperforms strong prior pruning baselines under matched compute. Our code is available at https://github.com/VILA-Lab/Sink-Aware-Pruning.

HIPE-2026 is a CLEF evaluation lab dedicated to person-place relation extraction from noisy, multilingual historical texts. Building on the HIPE-2020 and HIPE-2022 campaigns, it extends the series toward semantic relation extraction by targeting the task of identifying person--place associations in multiple languages and time periods. Systems are asked to classify relations of two types - $at$ ("Has the person ever been at this place?") and $isAt$ ("Is the person located at this place around publication time?") - requiring reasoning over temporal and geographical cues. The lab introduces a three-fold evaluation profile that jointly assesses accuracy, computational efficiency, and domain generalization. By linking relation extraction to large-scale historical data processing, HIPE-2026 aims to support downstream applications in knowledge-graph construction, historical biography reconstruction, and spatial analysis in digital humanities.

Reward modeling is a core component of modern alignment pipelines including RLHF and RLAIF, underpinning policy optimization methods including PPO and TRPO. However, training reliable reward models relies heavily on human-labeled preference data, which is costly and limited, motivating the use of data augmentation. Existing augmentation approaches typically operate at the representation or semantic level and remain agnostic to the reward model's estimation difficulty. In this paper, we propose MARS, an adaptive, margin-aware augmentation and sampling strategy that explicitly targets ambiguous and failure modes of the reward model. Our proposed framework, MARS, concentrates augmentation on low-margin (ambiguous) preference pairs where the reward model is most uncertain, and iteratively refines the training distribution via hard-sample augmentation. We provide theoretical guarantees showing that this strategy increases the average curvature of the loss function hence enhance information and improves conditioning, along with empirical results demonstrating consistent gains over uniform augmentation for robust reward modeling.

Language Identification (LID) is an important component of many multilingual natural language processing pipelines, where it facilitates corpus curation, training data analysis, and cross-lingual evaluation of large language models. Despite near-perfect performance on high-resource languages, existing systems remain brittle in low-resource and closely related language settings. We introduce UniLID, a simple and efficient LID method based on the UnigramLM tokenization algorithm, leveraging its probabilistic framing, parameter estimation technique and inference strategy. In short, we learn language-conditional unigram distributions over a shared tokenizer vocabulary but treat segmentation as a language-specific phenomenon. Our formulation is data- and compute-efficient, supports incremental addition of new languages without retraining existing models, and can naturally be integrated into existing language model tokenization pipelines. Empirical evaluations against widely used baselines, including fastText, GlotLID, and CLD3, show that UniLID achieves competitive performance on standard benchmarks, substantially improves sample efficiency in low-resource settings - surpassing 70% accuracy with as few as five labeled samples per language - and delivers large gains on fine-grained dialect identification.

We propose a two-stage "Mine and Refine" contrastive training framework for semantic text embeddings to enhance multi-category e-commerce search retrieval. Large scale e-commerce search demands embeddings that generalize to long tail, noisy queries while adhering to scalable supervision compatible with product and policy constraints. A practical challenge is that relevance is often graded: users accept substitutes or complements beyond exact matches, and production systems benefit from clear separation of similarity scores across these relevance strata for stable hybrid blending and thresholding. To obtain scalable policy consistent supervision, we fine-tune a lightweight LLM on human annotations under a three-level relevance guideline and further reduce residual noise via engagement driven auditing. In Stage 1, we train a multilingual Siamese two-tower retriever with a label aware supervised contrastive objective that shapes a robust global semantic space. In Stage 2, we mine hard samples via ANN and re-annotate them with the policy aligned LLM, and introduce a multi-class extension of circle loss that explicitly sharpens similarity boundaries between relevance levels, to further refine and enrich the embedding space. Robustness is additionally improved through additive spelling augmentation and synthetic query generation. Extensive offline evaluations and production A/B tests show that our framework improves retrieval relevance and delivers statistically significant gains in engagement and business impact.

Recent work has shown that language models (LMs) trained on synthetic corpora can exhibit typological preferences that resemble cross-linguistic regularities in human languages, particularly for syntactic phenomena such as word order. In this paper, we extend this paradigm to differential argument marking (DAM), a semantic licensing system in which morphological marking depends on semantic prominence. Using a controlled synthetic learning method, we train GPT-2 models on 18 corpora implementing distinct DAM systems and evaluate their generalization using minimal pairs. Our results reveal a dissociation between two typological dimensions of DAM. Models reliably exhibit human-like preferences for natural markedness direction, favoring systems in which overt marking targets semantically atypical arguments. In contrast, models do not reproduce the strong object preference in human languages, in which overt marking in DAM more often targets objects rather than subjects. These findings suggest that different typological tendencies may arise from distinct underlying sources.

As humans increasingly rely on multiround conversational AI for high stakes decisions, principled frameworks are needed to ensure such interactions reliably improve decision quality. We adopt a human centric view governed by two principles: counterfactual harm, ensuring the AI does not undermine human strengths, and complementarity, ensuring it adds value where the human is prone to err. We formalize these concepts via user defined rules, allowing users to specify exactly what harm and complementarity mean for their specific task. We then introduce an online, distribution free algorithm with finite sample guarantees that enforces the user-specified constraints over the collaboration dynamics. We evaluate our framework across two interactive settings: LLM simulated collaboration on a medical diagnostic task and a human crowdsourcing study on a pictorial reasoning task. We show that our online procedure maintains prescribed counterfactual harm and complementarity violation rates even under nonstationary interaction dynamics. Moreover, tightening or loosening these constraints produces predictable shifts in downstream human accuracy, confirming that the two principles serve as practical levers for steering multi-round collaboration toward better decision quality without the need to model or constrain human behavior.

Black-box adversarial attacks on Large Vision-Language Models (LVLMs) are challenging due to missing gradients and complex multimodal boundaries. While prior state-of-the-art transfer-based approaches like M-Attack perform well using local crop-level matching between source and target images, we find this induces high-variance, nearly orthogonal gradients across iterations, violating coherent local alignment and destabilizing optimization. We attribute this to (i) ViT translation sensitivity that yields spike-like gradients and (ii) structural asymmetry between source and target crops. We reformulate local matching as an asymmetric expectation over source transformations and target semantics, and build a gradient-denoising upgrade to M-Attack. On the source side, Multi-Crop Alignment (MCA) averages gradients from multiple independently sampled local views per iteration to reduce variance. On the target side, Auxiliary Target Alignment (ATA) replaces aggressive target augmentation with a small auxiliary set from a semantically correlated distribution, producing a smoother, lower-variance target manifold. We further reinterpret momentum as Patch Momentum, replaying historical crop gradients; combined with a refined patch-size ensemble (PE+), this strengthens transferable directions. Together these modules form M-Attack-V2, a simple, modular enhancement over M-Attack that substantially improves transfer-based black-box attacks on frontier LVLMs: boosting success rates on Claude-4.0 from 8% to 30%, Gemini-2.5-Pro from 83% to 97%, and GPT-5 from 98% to 100%, outperforming prior black-box LVLM attacks. Code and data are publicly available at: https://github.com/vila-lab/M-Attack-V2.

Traditional electronic recycling processes suffer from significant resource loss due to inadequate material separation and identification capabilities, limiting material recovery. We present A.R.I.S. (Automated Recycling Identification System), a low-cost, portable sorter for shredded e-waste that addresses this efficiency gap. The system employs a YOLOx model to classify metals, plastics, and circuit boards in real time, achieving low inference latency with high detection accuracy. Experimental evaluation yielded 90% overall precision, 82.2% mean average precision (mAP), and 84% sortation purity. By integrating deep learning with established sorting methods, A.R.I.S. enhances material recovery efficiency and lowers barriers to advanced recycling adoption. This work complements broader initiatives in extending product life cycles, supporting trade-in and recycling programs, and reducing environmental impact across the supply chain.

Feature engineering remains a critical yet challenging bottleneck in machine learning, particularly for tabular data, as identifying optimal features from an exponentially large feature space traditionally demands substantial domain expertise. To address this challenge, we introduce FAMOSE (Feature AugMentation and Optimal Selection agEnt), a novel framework that leverages the ReAct paradigm to autonomously explore, generate, and refine features while integrating feature selection and evaluation tools within an agent architecture. To our knowledge, FAMOSE represents the first application of an agentic ReAct framework to automated feature engineering, especially for both regression and classification tasks. Extensive experiments demonstrate that FAMOSE is at or near the state-of-the-art on classification tasks (especially tasks with more than 10K instances, where ROC-AUC increases 0.23% on average), and achieves the state-of-the-art for regression tasks by reducing RMSE by 2.0% on average, while remaining more robust to errors than other algorithms. We hypothesize that FAMOSE's strong performance is because ReAct allows the LLM context window to record (via iterative feature discovery and evaluation steps) what features did or did not work. This is similar to a few-shot prompt and guides the LLM to invent better, more innovative features. Our work offers evidence that AI agents are remarkably effective in solving problems that require highly inventive solutions, such as feature engineering.

Market area models, such as the Huff model and its extensions, are widely used to estimate regional market shares and customer flows of retail and service locations. Another, now very common, area of application is the analysis of catchment areas, supply structures and the accessibility of healthcare locations. The huff Python package provides a complete workflow for market area analysis, including data import, construction of origin-destination interaction matrices, basic model analysis, parameter estimation from empirical data, calculation of distance or travel time indicators, and map visualization. Additionally, the package provides several methods of spatial accessibility analysis. The package is modular and object-oriented. It is intended for researchers in economic geography, regional economics, spatial planning, marketing, geoinformation science, and health geography. The software is openly available via the [Python Package Index (PyPI)](https://pypi.org/project/huff/); its development and version history are managed in a public [GitHub Repository](https://github.com/geowieland/huff_official) and archived at [Zenodo](https://doi.org/10.5281/zenodo.18639559).

Learning time series foundation models has been shown to be a promising approach for zero-shot time series forecasting across diverse time series domains. Insofar as scaling has been a critical driver of performance of foundation models in other modalities such as language and vision, much recent work on time series foundation modeling has focused on scaling. This has resulted in time series foundation models with hundreds of millions of parameters that are, while performant, inefficient and expensive to use in practice. This paper describes a simple recipe for learning efficient foundation models for zero-shot time series forecasting that are orders of magnitude smaller. We show that large-scale transformers are not necessary: small hybrid models that interleave long convolution and linear RNN layers (in particular DeltaNet layers) can match the performance of larger transformer-based models while being more than a hundred times smaller. We also describe several data augmentation and inference strategies that further improve performance. This recipe results in Reverso, a family of efficient time series foundation models for zero-shot forecasting that significantly push the performance-efficiency Pareto frontier.

Reasoning with LLMs increasingly unfolds inside a broader verification loop. Internally, systems use cheap checks, such as self-consistency or proxy rewards, which we call weak verification. Externally, users inspect outputs and steer the model through feedback until results are trustworthy, which we call strong verification. These signals differ sharply in cost and reliability: strong verification can establish trust but is resource-intensive, while weak verification is fast and scalable but noisy and imperfect. We formalize this tension through weak--strong verification policies, which decide when to accept or reject based on weak verification and when to defer to strong verification. We introduce metrics capturing incorrect acceptance, incorrect rejection, and strong-verification frequency. Over population, we show that optimal policies admit a two-threshold structure and that calibration and sharpness govern the value of weak verifiers. Building on this, we develop an online algorithm that provably controls acceptance and rejection errors without assumptions on the query stream, the language model, or the weak verifier.

Modern offline Reinforcement Learning (RL) methods find performant actor-critics, however, fine-tuning these actor-critics online with value-based RL algorithms typically causes immediate drops in performance. We provide evidence consistent with the hypothesis that, in the loss landscape, offline maxima for prior algorithms and online maxima are separated by low-performance valleys that gradient-based fine-tuning traverses. Following this, we present Score Matched Actor-Critic (SMAC), an offline RL method designed to learn actor-critics that transition to online value-based RL algorithms with no drop in performance. SMAC avoids valleys between offline and online maxima by regularizing the Q-function during the offline phase to respect a first-order derivative equality between the score of the policy and action-gradient of the Q-function. We experimentally demonstrate that SMAC converges to offline maxima that are connected to better online maxima via paths with monotonically increasing reward found by first-order optimization. SMAC achieves smooth transfer to Soft Actor-Critic and TD3 in 6/6 D4RL tasks. In 4/6 environments, it reduces regret by 34-58% over the best baseline.

Modern big-data systems generate massive, heterogeneous, and geographically dispersed streams that are large-scale and privacy-sensitive, making centralization challenging. While federated learning (FL) provides a privacy-enhancing training mechanism, it assumes a static data flow and learns a collaborative model over multiple rounds, making learning with \textit{incremental} data challenging in limited-communication scenarios. This paper presents One-Shot Incremental Federated Learning (OSI-FL), the first FL framework that addresses the dual challenges of communication overhead and catastrophic forgetting. OSI-FL communicates category-specific embeddings, devised by a frozen vision-language model (VLM) from each client in a single communication round, which a pre-trained diffusion model at the server uses to synthesize new data similar to the client's data distribution. The synthesized samples are used on the server for training. However, two challenges still persist: i) tasks arriving incrementally need to retrain the global model, and ii) as future tasks arrive, retraining the model introduces catastrophic forgetting. To this end, we augment training with Selective Sample Retention (SSR), which identifies and retains the top-p most informative samples per category and task pair based on sample loss. SSR bounds forgetting by ensuring that representative retained samples are incorporated into training in further iterations. The experimental results indicate that OSI-FL outperforms baselines, including traditional and one-shot FL approaches, in both class-incremental and domain-incremental scenarios across three benchmark datasets.

While emerging Persian NLP benchmarks have expanded into pragmatics and politeness, they rarely distinguish between memorized cultural facts and the ability to reason about implicit social norms. We introduce DivanBench, a diagnostic benchmark focused on superstitions and customs, arbitrary, context-dependent rules that resist simple logical deduction. Through 315 questions across three task types (factual retrieval, paired scenario verification, and situational reasoning), we evaluate seven Persian LLMs and reveal three critical failures: most models exhibit severe acquiescence bias, correctly identifying appropriate behaviors but failing to reject clear violations; continuous Persian pretraining amplifies this bias rather than improving reasoning, often degrading the model's ability to discern contradictions; and all models show a 21\% performance gap between retrieving factual knowledge and applying it in scenarios. These findings demonstrate that cultural competence requires more than scaling monolingual data, as current models learn to mimic cultural patterns without internalizing the underlying schemas.

LLM-based agents show promise for automating penetration testing, yet reported performance varies widely across systems and benchmarks. We analyze 28 LLM-based penetration testing systems and evaluate five representative implementations across three benchmarks of increasing complexity. Our analysis reveals two distinct failure modes: Type A failures stem from capability gaps (missing tools, inadequate prompts) that engineering readily addresses, while Type B failures persist regardless of tooling due to planning and state management limitations. We show that Type B failures share a root cause that is largely invariant to the underlying LLM: agents lack real-time task difficulty estimation. As a result, agents misallocate effort, over-commit to low-value branches, and exhaust context before completing attack chains. Based on this insight, we present Excalibur, a penetration testing agent that couples strong tooling with difficulty-aware planning. A Tool and Skill Layer eliminates Type A failures through typed interfaces and retrieval-augmented knowledge. A Task Difficulty Assessment (TDA) mechanism addresses Type B failures by estimating tractability through four measurable dimensions (horizon estimation, evidence confidence, context load, and historical success) and uses these estimates to guide exploration-exploitation decisions within an Evidence-Guided Attack Tree Search (EGATS) framework. Excalibur achieves up to 91% task completion on CTF benchmarks with frontier models (39 to 49% relative improvement over baselines) and compromises 4 of 5 hosts on the GOAD Active Directory environment versus 2 by prior systems. These results show that difficulty-aware planning yields consistent end-to-end gains across models and addresses a limitation that model scaling alone does not eliminate.

Reinforcement learning (RL) is widely used to improve large language models on reasoning tasks, and asynchronous RL training is attractive because it increases end-to-end throughput. However, for widely adopted critic-free policy-gradient methods such as REINFORCE and GRPO, high asynchrony makes the policy-gradient estimator markedly $\textbf{higher variance}$: training on stale rollouts creates heavy-tailed importance ratios, causing a small fraction of samples to dominate updates. This amplification makes gradients noisy and learning unstable relative to matched on-policy training. Across math and general reasoning benchmarks, we find collapse is reliably predicted by effective sample size (ESS) and unstable gradient norms. Motivated by this diagnosis, we propose $\textbf{V}$ariance $\textbf{C}$ontrolled $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{VCPO}$), a general stabilization method for REINFORCE/GRPO-style algorithms that (i) scales learning rate based on effective sample size to dampen unreliable updates, and (ii) applies a closed-form minimum-variance baseline for the off-policy setting, avoiding an auxiliary value model and adding minimal overhead. Empirically, VCPO substantially improves robustness for asynchronous training across math, general reasoning, and tool-use tasks, outperforming a broad suite of baselines spanning masking/clipping stabilizers and algorithmic variants. This reduces long-context, multi-turn training time by 2.5$\times$ while matching synchronous performance, demonstrating that explicit control of policy-gradient variance is key for reliable asynchronous RL at scale.

Big data scenarios, where massive, heterogeneous datasets are distributed across clients, demand scalable, privacy-preserving learning methods. Federated learning (FL) enables decentralized training of machine learning (ML) models across clients without data centralization. Decentralized training, however, introduces a computational burden on client devices. U-shaped federated split learning (UFSL) offloads a fraction of the client computation to the server while keeping both data and labels on the clients' side. However, the intermediate representations (i.e., smashed data) shared by clients with the server are prone to exposing clients' private data. To reduce exposure of client data through intermediate data representations, this work proposes k-anonymous differentially private UFSL (KD-UFSL), which leverages privacy-enhancing techniques such as microaggregation and differential privacy to minimize data leakage from the smashed data transferred to the server. We first demonstrate that an adversary can access private client data from intermediate representations via a data-reconstruction attack, and then present a privacy-enhancing solution, KD-UFSL, to mitigate this risk. Our experiments indicate that, alongside increasing the mean squared error between the actual and reconstructed images by up to 50% in some cases, KD-UFSL also decreases the structural similarity between them by up to 40% on four benchmarking datasets. More importantly, KD-UFSL improves privacy while preserving the utility of the global model. This highlights its suitability for large-scale big data applications where privacy and utility must be balanced.

Driven by scientific and industry ambition, HPC and AI applications such as operational Numerical Weather Prediction (NWP) require processing and storing ever-increasing data volumes as fast as possible. Whilst POSIX distributed file systems and NVMe SSDs are currently a common HPC storage configuration providing I/O to applications, new storage solutions have proliferated or gained traction over the last decade with potential to address performance limitations POSIX file systems manifest at scale for certain I/O workloads. This work has primarily aimed to assess the suitability and performance of two object storage systems -namely DAOS and Ceph- for the ECMWF's operational NWP as well as for HPC and AI applications in general. New software-level adapters have been developed which enable the ECMWF's NWP to leverage these systems, and extensive I/O benchmarking has been conducted on a few computer systems, comparing the performance delivered by the evaluated object stores to that of equivalent Lustre file system deployments on the same hardware. Challenges of porting to object storage and its benefits with respect to the traditional POSIX I/O approach have been discussed and, where possible, domain-agnostic performance analysis has been conducted, leading to insight also of relevance to I/O practitioners and the broader HPC community. DAOS and Ceph have both demonstrated excellent performance, but DAOS stood out relative to Ceph and Lustre, providing superior scalability and flexibility for applications to perform I/O at scale as desired. This sets a promising outlook for DAOS and object storage, which might see greater adoption at HPC centres in the years to come, although not necessarily implying a shift away from POSIX-like I/O.

Regulatory limits on Maximum Permissible Exposure (MPE) require handheld devices to reduce transmit power when operated near the user's body. Current proximity sensors provide only binary detection, triggering conservative power back-off that degrades link quality. If the device could measure its distance from the body, transmit power could be adjusted proportionally, improving throughput while maintaining compliance. This paper develops a device-centric integrated sensing and communication (ISAC) method for the device to measure this distance. The uplink communication waveform is exploited for sensing, and the natural motion of the user's hand creates a virtual aperture that provides the angular resolution necessary for localization. Virtual aperture processing requires precise knowledge of the device trajectory, which in this scenario is opportunistic and unknown. One can exploit onboard inertial sensors to estimate the device trajectory; however, the inertial sensors accuracy is not sufficient. To address this, we develop an autofocus algorithm based on extended Kalman filtering that jointly tracks the trajectory and compensates residual errors using phase observations from strong scatterers. The Bayesian Cramér-Rao bound for localization is derived under correlated inertial errors. Numerical results at 28GHz demonstrate centimeter-level accuracy with realistic sensor parameters.

The proliferation of Large Language Models (LLMs) necessitates efficient mechanisms to distinguish machine-generated content from human text. While statistical watermarking has emerged as a promising solution, existing methods suffer from two critical limitations: the lack of a principled approach for selecting sampling distributions and the reliance on fixed-horizon hypothesis testing, which precludes valid early stopping. In this paper, we bridge this gap by developing the first e-value-based watermarking framework, Anchored E-Watermarking, that unifies optimal sampling with anytime-valid inference. Unlike traditional approaches where optional stopping invalidates Type-I error guarantees, our framework enables valid, anytime-inference by constructing a test supermartingale for the detection process. By leveraging an anchor distribution to approximate the target model, we characterize the optimal e-value with respect to the worst-case log-growth rate and derive the optimal expected stopping time. Our theoretical claims are substantiated by simulations and evaluations on established benchmarks, showing that our framework can significantly enhance sample efficiency, reducing the average token budget required for detection by 13-15% relative to state-of-the-art baselines.

PDEs are central to scientific and engineering modeling, yet designing accurate numerical solvers typically requires substantial mathematical expertise and manual tuning. Recent neural network-based approaches improve flexibility but often demand high computational cost and suffer from limited interpretability. We introduce \texttt{AutoNumerics}, a multi-agent framework that autonomously designs, implements, debugs, and verifies numerical solvers for general PDEs directly from natural language descriptions. Unlike black-box neural solvers, our framework generates transparent solvers grounded in classical numerical analysis. We introduce a coarse-to-fine execution strategy and a residual-based self-verification mechanism. Experiments on 24 canonical and real-world PDE problems demonstrate that \texttt{AutoNumerics} achieves competitive or superior accuracy compared to existing neural and LLM-based baselines, and correctly selects numerical schemes based on PDE structural properties, suggesting its viability as an accessible paradigm for automated PDE solving.

In many real-world settings, such as environmental monitoring, disaster response, or public health, with costly and difficult data collection and dynamic environments, strategically sampling from unobserved regions is essential for efficiently uncovering hidden targets under tight resource constraints. Yet, sparse and biased geospatial ground truth limits the applicability of existing learning-based methods, such as reinforcement learning. To address this, we propose a unified geospatial discovery framework that integrates active learning, online meta-learning, and concept-guided reasoning. Our approach introduces two key innovations built on a shared notion of *concept relevance*, which captures how domain-specific factors influence target presence: a *concept-weighted uncertainty sampling strategy*, where uncertainty is modulated by learned relevance based on readily-available domain-specific concepts (e.g., land cover, source proximity); and a *relevance-aware meta-batch formation strategy* that promotes semantic diversity during online-meta updates, improving generalization in dynamic environments. Our experiments include testing on a real-world dataset of cancer-causing PFAS (Per- and polyfluoroalkyl substances) contamination, showcasing our method's reliability at uncovering targets with limited data and a varying environment.

Molecular generation with diffusion models has emerged as a promising direction for AI-driven drug discovery and materials science. While graph diffusion models have been widely adopted due to the discrete nature of 2D molecular graphs, existing models suffer from low chemical validity and struggle to meet the desired properties compared to 1D modeling. In this work, we introduce MolHIT, a powerful molecular graph generation framework that overcomes long-standing performance limitations in existing methods. MolHIT is based on the Hierarchical Discrete Diffusion Model, which generalizes discrete diffusion to additional categories that encode chemical priors, and decoupled atom encoding that splits the atom types according to their chemical roles. Overall, MolHIT achieves new state-of-the-art performance on the MOSES dataset with near-perfect validity for the first time in graph diffusion, surpassing strong 1D baselines across multiple metrics. We further demonstrate strong performance in downstream tasks, including multi-property guided generation and scaffold extension.

Current speech LLMs largely perform implicit ASR: on tasks solvable from a transcript, they are behaviorally and mechanistically equivalent to simple Whisper$\to$LLM cascades. We show this through matched-backbone testing across four speech LLMs and six tasks, controlling for the LLM backbone for the first time. Ultravox is statistically indistinguishable from its matched cascade ($κ{=}0.93$); logit lens reveals literal text emerging in hidden states; LEACE concept erasure confirms text representations are causally necessary in both architectures tested, collapsing accuracy to near-zero. Qwen2-Audio genuinely diverges, revealing cascade equivalence is architecture-dependent, not universal. For most deployed use cases, current speech LLMs are expensive cascades, and under noise, they are worse ones, with clean-condition advantages reversing by up to 7.6% at 0 dB.

We study the topology of the loss landscape of one-hidden-layer ReLU networks under overparameterization. On the theory side, we (i) prove that for convex $L$-Lipschitz losses with an $\ell_1$-regularized second layer, every pair of models at the same loss level can be connected by a continuous path within an arbitrarily small loss increase $ε$ (extending a known result for the quadratic loss); (ii) obtain an asymptotic upper bound on the energy gap $ε$ between local and global minima that vanishes as the width $m$ grows, implying that the landscape flattens and sublevel sets become connected in the limit. Empirically, on a synthetic Moons dataset and on the Wisconsin Breast Cancer dataset, we measure pairwise energy gaps via Dynamic String Sampling (DSS) and find that wider networks exhibit smaller gaps; in particular, a permutation test on the maximum gap yields $p_{perm}=0$, indicating a clear reduction in the barrier height.

Rigorously evaluating machine intelligence against the broad spectrum of human general intelligence has become increasingly important and challenging in this era of rapid technological advance. Conventional AI benchmarks typically assess only narrow capabilities in a limited range of human activity. Most are also static, quickly saturating as developers explicitly or implicitly optimize for them. We propose that a more promising way to evaluate human-like general intelligence in AI systems is through a particularly strong form of general game playing: studying how and how well they play and learn to play \textbf{all conceivable human games}, in comparison to human players with the same level of experience, time, or other resources. We define a "human game" to be a game designed by humans for humans, and argue for the evaluative suitability of this space of all such games people can imagine and enjoy -- the "Multiverse of Human Games". Taking a first step towards this vision, we introduce the AI GameStore, a scalable and open-ended platform that uses LLMs with humans-in-the-loop to synthesize new representative human games, by automatically sourcing and adapting standardized and containerized variants of game environments from popular human digital gaming platforms. As a proof of concept, we generated 100 such games based on the top charts of Apple App Store and Steam, and evaluated seven frontier vision-language models (VLMs) on short episodes of play. The best models achieved less than 10\% of the human average score on the majority of the games, and especially struggled with games that challenge world-model learning, memory and planning. We conclude with a set of next steps for building out the AI GameStore as a practical way to measure and drive progress toward human-like general intelligence in machines.

We present BMC4TimeSec, an end-to-end tool for verifying Timed Security Protocols (TSP) based on SMT-based bounded model checking and multi-agent modelling in the form of Timed Interpreted Systems (TIS) and Timed Interleaved Interpreted Systems (TIIS). In BMC4TimeSec, TSP executions implement the TIS/TIIS environment (join actions, interleaving, delays, lifetimes), and knowledge automata implement the agents (evolution of participant knowledge, including the intruder). The code is publicly available on \href{https://github.com/agazbrzezny/BMC4TimeSec}{GitHub}, as is a \href{https://youtu.be/aNybKz6HwdA}{video} demonstration.

Despite rapid progress in autonomous web agents, human involvement remains essential for shaping preferences and correcting agent behavior as tasks unfold. However, current agentic systems lack a principled understanding of when and why humans intervene, often proceeding autonomously past critical decision points or requesting unnecessary confirmation. In this work, we introduce the task of modeling human intervention to support collaborative web task execution. We collect CowCorpus, a dataset of 400 real-user web navigation trajectories containing over 4,200 interleaved human and agent actions. We identify four distinct patterns of user interaction with agents -- hands-off supervision, hands-on oversight, collaborative task-solving, and full user takeover. Leveraging these insights, we train language models (LMs) to anticipate when users are likely to intervene based on their interaction styles, yielding a 61.4-63.4% improvement in intervention prediction accuracy over base LMs. Finally, we deploy these intervention-aware models in live web navigation agents and evaluate them in a user study, finding a 26.5% increase in user-rated agent usefulness. Together, our results show structured modeling of human intervention leads to more adaptive, collaborative agents.

We study one-sided and $α$-correct sequential hypothesis testing for data generated by an ergodic Markov chain. The null hypothesis is that the unknown transition matrix belongs to a prescribed set $P$ of stochastic matrices, and the alternative corresponds to a disjoint set $Q$. We establish a tight non-asymptotic instance-dependent lower bound on the expected stopping time of any valid sequential test under the alternative. Our novel analysis improves the existing lower bounds, which are either asymptotic or provably sub-optimal in this setting. Our lower bound incorporates both the stationary distribution and the transition structure induced by the unknown Markov chain. We further propose an optimal test whose expected stopping time matches this lower bound asymptotically as $α\to 0$. We illustrate the usefulness of our framework through applications to sequential detection of model misspecification in Markov Chain Monte Carlo and to testing structural properties, such as the linearity of transition dynamics, in Markov decision processes. Our findings yield a sharp and general characterization of optimal sequential testing procedures under Markovian dependence.

Safety validation for Level 4 autonomous vehicles (AVs) is currently bottlenecked by the inability to scale the detection of rare, high-risk long-tail scenarios using traditional rule-based heuristics. We present Deep-Flow, an unsupervised framework for safety-critical anomaly detection that utilizes Optimal Transport Conditional Flow Matching (OT-CFM) to characterize the continuous probability density of expert human driving behavior. Unlike standard generative approaches that operate in unstable, high-dimensional coordinate spaces, Deep-Flow constrains the generative process to a low-rank spectral manifold via a Principal Component Analysis (PCA) bottleneck. This ensures kinematic smoothness by design and enables the computation of the exact Jacobian trace for numerically stable, deterministic log-likelihood estimation. To resolve multi-modal ambiguity at complex junctions, we utilize an Early Fusion Transformer encoder with lane-aware goal conditioning, featuring a direct skip-connection to the flow head to maintain intent-integrity throughout the network. We introduce a kinematic complexity weighting scheme that prioritizes high-energy maneuvers (quantified via path tortuosity and jerk) during the simulation-free training process. Evaluated on the Waymo Open Motion Dataset (WOMD), our framework achieves an AUC-ROC of 0.766 against a heuristic golden set of safety-critical events. More significantly, our analysis reveals a fundamental distinction between kinematic danger and semantic non-compliance. Deep-Flow identifies a critical predictability gap by surfacing out-of-distribution behaviors, such as lane-boundary violations and non-normative junction maneuvers, that traditional safety filters overlook. This work provides a mathematically rigorous foundation for defining statistical safety gates, enabling objective, data-driven validation for the safe deployment of autonomous fleets.

As models and data scale, independently trained networks often induce analogous notions of similarity. But, matching similarities is weaker than establishing an explicit correspondence between the representation spaces, especially for multimodal models, where consistency must hold not only within each modality, but also for the learned image-text coupling. We therefore ask: given two independently trained multimodal contrastive models (with encoders $(f, g)$ and $(\widetilde{f},\widetilde{g})$) -- trained on different distributions and with different architectures -- does a systematic geometric relationship exist between their embedding spaces? If so, what form does it take, and does it hold uniformly across modalities? In this work, we show that across model families such as CLIP, SigLIP, and FLAVA, this geometric relationship is well approximated by an orthogonal map (up to a global mean shift), i.e., there exists an orthogonal map $Q$ where $Q^\top Q = I$ such that $\widetilde{f}(x)\approx Q f(x)$ for paired images $x$. Strikingly, the same $Q$ simultaneously aligns the text encoders i.e., $\widetilde{g}(y)\approx Q g(y)$ for texts $y$. Theoretically, we prove that if the multimodal kernel agrees across models on a small anchor set i.e. $\langle f(x), g(y)\rangle \approx \langle \widetilde{f}(x), \widetilde{g}(y)\rangle$, then the two models must be related by a single orthogonal map $Q$ and the same $Q$ maps images and text across models. More broadly, this finding enables backward-compatible model upgrades, avoiding costly re-embedding, and has implications for the privacy of learned representations. Our project page: https://canonical-multimodal.github.io/

Omniprediction is a learning problem that requires suboptimality bounds for each of a family of losses $\mathcal{L}$ against a family of comparator predictors $\mathcal{C}$. We initiate the study of omniprediction in a multiclass setting, where the comparator family $\mathcal{C}$ may be infinite. Our main result is an extension of the recent binary omniprediction algorithm of [OKK25] to the multiclass setting, with sample complexity (in statistical settings) or regret horizon (in online settings) $\approx \varepsilon^{-(k+1)}$, for $\varepsilon$-omniprediction in a $k$-class prediction problem. En route to proving this result, we design a framework of potential broader interest for solving Blackwell approachability problems where multiple sets must simultaneously be approached via coupled actions.

Normalization and scaling are fundamental preprocessing steps in time series modeling, yet their role in Transformer-based models remains underexplored from a theoretical perspective. In this work, we present the first formal analysis of how different normalization strategies, specifically instance-based and global scaling, impact the expressivity of Transformer-based architectures for time series representation learning. We propose a novel expressivity framework tailored to time series, which quantifies a model's ability to distinguish between similar and dissimilar inputs in the representation space. Using this framework, we derive theoretical bounds for two widely used normalization methods: Standard and Min-Max scaling. Our analysis reveals that the choice of normalization strategy can significantly influence the model's representational capacity, depending on the task and data characteristics. We complement our theory with empirical validation on classification and forecasting benchmarks using multiple Transformer-based models. Our results show that no single normalization method consistently outperforms others, and in some cases, omitting normalization entirely leads to superior performance. These findings highlight the critical role of preprocessing in time series learning and motivate the need for more principled normalization strategies tailored to specific tasks and datasets.

The significant advancements in computational power cre- ate a vast opportunity for using Artificial Intelligence in different ap- plications of healthcare and medical science. A Hybrid FL-Enabled Ensemble Approach For Lung Disease Diagnosis Leveraging a Combination of SWIN Transformer and CNN is the combination of cutting-edge technology of AI and Federated Learning. Since, medi- cal specialists and hospitals will have shared data space, based on that data, with the help of Artificial Intelligence and integration of federated learning, we can introduce a secure and distributed system for medical data processing and create an efficient and reliable system. The proposed hybrid model enables the detection of COVID-19 and Pneumonia based on x-ray reports. We will use advanced and the latest available tech- nology offered by Tensorflow and Keras along with Microsoft-developed Vision Transformer, that can help to fight against the pandemic that the world has to fight together as a united. We focused on using the latest available CNN models (DenseNet201, Inception V3, VGG 19) and the Transformer model SWIN Transformer in order to prepare our hy- brid model that can provide a reliable solution as a helping hand for the physician in the medical field. In this research, we will discuss how the Federated learning-based Hybrid AI model can improve the accuracy of disease diagnosis and severity prediction of a patient using the real-time continual learning approach and how the integration of federated learn- ing can ensure hybrid model security and keep the authenticity of the information.

Self-distillation (SD) is the process of retraining a student on a mixture of ground-truth labels and the teacher's own predictions using the same architecture and training data. Although SD has been empirically shown to often improve generalization, its formal guarantees remain limited. We study SD for ridge regression in unconstrained setting in which the mixing weight $ξ$ may be outside the unit interval. Conditioned on the training data and without any distributional assumptions, we prove that for any squared prediction risk (including out-of-distribution), the optimally mixed student strictly improves upon the ridge teacher for every regularization level $λ> 0$ at which the teacher ridge risk $R(λ)$ is nonstationary (i.e., $R'(λ) \neq 0$). We obtain a closed-form expression for the optimal mixing weight $ξ^\star(λ)$ for any value of $λ$ and show that it obeys the sign rule: $\operatorname{sign}(ξ^\star(λ))=-\operatorname{sign}(R'(λ))$. In particular, $ξ^\star(λ)$ can be negative, which is the case in over-regularized regimes. To quantify the risk improvement due to SD, we derive exact deterministic equivalents for the optimal SD risk in the proportional asymptotics regime (where the sample and feature sizes $n$ and $p$ both diverge but their aspect ratio $p/n$ converges) under general anisotropic covariance and deterministic signals. Our asymptotic analysis extends standard second-order ridge deterministic equivalents to their fourth-order analogs using block linearization, which may be of independent interest. From a practical standpoint, we propose a consistent one-shot tuning method to estimate $ξ^\star$ without grid search, sample splitting, or refitting. Experiments on real-world datasets and pretrained neural network features support our theory and the one-shot tuning method.

Activation steering, or representation engineering, offers a lightweight approach to align large language models (LLMs) by manipulating their internal activations at inference time. However, current methods suffer from two key limitations: \textit{(i)} the lack of a unified theoretical framework for guiding the design of steering directions, and \textit{(ii)} an over-reliance on \textit{one-step steering} that fail to capture complex patterns of activation distributions. In this work, we propose a unified ordinary differential equations (ODEs)-based \textit{theoretical} framework for activation steering in LLM alignment. We show that conventional activation addition can be interpreted as a first-order approximation to the solution of an ODE. Based on this ODE perspective, identifying a steering direction becomes equivalent to designing a \textit{barrier function} from control theory. Derived from this framework, we introduce ODESteer, a kind of ODE-based steering guided by barrier functions, which shows \textit{empirical} advancement in LLM alignment. ODESteer identifies steering directions by defining the barrier function as the log-density ratio between positive and negative activations, and employs it to construct an ODE for \textit{multi-step and adaptive} steering. Compared to state-of-the-art activation steering methods, ODESteer achieves consistent empirical improvements on diverse LLM alignment benchmarks, a notable $5.7\%$ improvement over TruthfulQA, $2.5\%$ over UltraFeedback, and $2.4\%$ over RealToxicityPrompts. Our work establishes a principled new view of activation steering in LLM alignment by unifying its theoretical foundations via ODEs, and validating it empirically through the proposed ODESteer method.

Continual Learning (CL) with large-scale pre-trained models (PTMs) has recently gained wide attention, shifting the focus from training from scratch to continually adapting PTMs. This has given rise to a promising paradigm: parameter-efficient continual learning (PECL), where task interference is typically mitigated by assigning a task-specific module during training, such as low-rank adapters. However, weight regularization techniques, such as Elastic Weight Consolidation (EWC)-a key strategy in CL-remain underexplored in this new paradigm. In this paper, we revisit weight regularization in low-rank CL as a new perspective for mitigating task interference in PECL. Unlike existing low-rank CL methods, we mitigate task interference by regularizing a shared low-rank update through EWC, thereby keeping the storage requirement and inference costs constant regardless of the number of tasks. Our proposed method EWC-LoRA leverages a low-rank representation to estimate parameter importance over the full-dimensional space. This design offers a practical, computational- and memory-efficient solution for CL with PTMs, and provides insights that may inform the broader application of regularization techniques within PECL. Extensive experiments on various benchmarks demonstrate the effectiveness of EWC-LoRA, achieving a stability-plasticity trade-off superior to existing low-rank CL approaches. These results indicate that, even under low-rank parameterizations, weight regularization remains an effective mechanism for mitigating task interference. Code is available at: https://github.com/yaoyz96/low-rank-cl.

Alzheimer's disease (AD) and Lewy body dementia (LBD) present overlapping clinical features yet require distinct diagnostic strategies. While neuroimaging-based brain network analysis is promising, atlas-based representations may obscure individualized anatomy. Gyral folding-based networks using three-hinge gyri provide a biologically grounded alternative, but inter-individual variability in cortical folding results in inconsistent landmark correspondence and highly irregular network sizes, violating the fixed-topology and node-alignment assumptions of most existing graph learning methods, particularly in clinical datasets where pathological changes further amplify anatomical heterogeneity. We therefore propose a probability-invariant random-walk-based framework that classifies individualized gyral folding networks without explicit node alignment. Cortical similarity networks are built from local morphometric features and represented by distributions of anonymized random walks, with an anatomy-aware encoding that preserves permutation invariance. Experiments on a large clinical cohort of AD and LBD subjects show consistent improvements over existing gyral folding and atlas-based models, demonstrating robustness and potential for dementia diagnosis.

Training large-scale generative models is resource-intensive and relies heavily on heuristic dataset weighting. We address two fundamental questions: Can we train Large Language Models (LLMs) modularly-combining small, domain-specific experts to match monolithic performance-and can we do so robustly for any data mixture, eliminating heuristic tuning? We present a theoretical framework for modular generative modeling where a set of pre-trained experts are combined via a gating mechanism. We define the space of normalized gating functions, $G_{1}$, and formulate the problem as a minimax game to find a single robust gate that minimizes divergence to the worst-case data mixture. We prove the existence of such a robust gate using Kakutani's fixed-point theorem and show that modularity acts as a strong regularizer, with generalization bounds scaling with the lightweight gate's complexity. Furthermore, we prove that this modular approach can theoretically outperform models retrained on aggregate data, with the gap characterized by the Jensen-Shannon Divergence. Finally, we introduce a scalable Stochastic Primal-Dual algorithm and a Structural Distillation method for efficient inference. Empirical results on synthetic and real-world datasets confirm that our modular architecture effectively mitigates gradient conflict and can robustly outperform monolithic baselines.

Error-bounded lossy compression is essential for managing the massive data volumes produced by large-scale HPC simulations. While state-of-the-art compressors such as SZ and ZFP provide strong numerical error guarantees, they often fail to preserve topological structures (example, minima, maxima, and saddle points) that are critical for scientific analysis. Existing topology-aware compressors address this limitation but incur substantial computational overhead. We present TopoSZp, a lightweight, topology-aware, error-controlled lossy compressor that preserves critical points and their relationships while maintaining high compression and decompression performance. Built on the high-throughput SZp compressor, TopoSZp integrates efficient critical point detection, local ordering preservation, and targeted saddle point refinement, all within a relaxed but strictly enforced error bound. Experimental results on real-world scientific datasets show that TopoSZp achieves 3 to 100 times fewer non-preserved critical points, introduces no false positives or incorrect critical point types, and delivers 100 to 10000 times faster compression and 10 to 500 times faster decompression compared to existing topology-aware compressors, while maintaining competitive compression ratios.

Existing Reinforcement Learning with Verifiable Rewards (RLVR) algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanisms that are fundamentally misaligned with the complex optimization dynamics of Large Language Models (LLMs). In this paper, we identify three critical challenges in these methods: (1) inefficient gradient utilization caused by the binary cutoff of hard clipping, (2) insensitive probability mass arising from uniform ratio constraints that ignore the token distribution, and (3) asymmetric signal reliability stemming from the disparate credit assignment ambiguity between positive and negative samples. To bridge these gaps, we propose Mass-Adaptive Soft Policy Optimization (MASPO), a unified framework designed to harmonize these three dimensions. MASPO integrates a differentiable soft Gaussian gating to maximize gradient utility, a mass-adaptive limiter to balance exploration across the probability spectrum, and an asymmetric risk controller to align update magnitudes with signal confidence. Extensive evaluations demonstrate that MASPO serves as a robust, all-in-one RLVR solution, significantly outperforming strong baselines. Our code is available at: https://anonymous.4open.science/r/ma1/README.md.

This paper introduces KLong, an open-source LLM agent trained to solve extremely long-horizon tasks. The principle is to first cold-start the model via trajectory-splitting SFT, then scale it via progressive RL training. Specifically, we first activate basic agentic abilities of a base model with a comprehensive SFT recipe. Then, we introduce Research-Factory, an automated pipeline that generates high-quality training data by collecting research papers and constructing evaluation rubrics. Using this pipeline, we build thousands of long-horizon trajectories distilled from Claude 4.5 Sonnet (Thinking). To train with these extremely long trajectories, we propose a new trajectory-splitting SFT, which preserves early context, progressively truncates later context, and maintains overlap between sub-trajectories. In addition, to further improve long-horizon task-solving capability, we propose a novel progressive RL, which schedules training into multiple stages with progressively extended timeouts. Experiments demonstrate the superiority and generalization of KLong, as shown in Figure 1. Notably, our proposed KLong (106B) surpasses Kimi K2 Thinking (1T) by 11.28% on PaperBench, and the performance improvement generalizes to other coding benchmarks like SWE-bench Verified and MLE-bench.

Instruction-following language models are trained to be helpful and safe, yet their safety behavior can deteriorate under benign fine-tuning and worsen under adversarial updates. Existing defenses often offer limited protection or force a trade-off between safety and utility. We introduce a training framework that adapts regularization in response to safety risk, enabling models to remain aligned throughout fine-tuning. To estimate safety risk at training time, we explore two distinct approaches: a judge-based Safety Critic that assigns high-level harm scores to training batches, and an activation-based risk predictor built with a lightweight classifier trained on intermediate model activations to estimate harmful intent. Each approach provides a risk signal that is used to constrain updates deemed higher risk to remain close to a safe reference policy, while lower-risk updates proceed with standard training. We empirically verify that harmful intent signals are predictable from pre-generation activations and that judge scores provide effective high-recall safety guidance. Across multiple model families and attack scenarios, adaptive regularization with either risk estimation approach consistently lowers attack success rate compared to standard fine-tuning, preserves downstream performance, and adds no inference-time cost. This work demonstrates a principled mechanism for maintaining safety without sacrificing utility.

The paper studies decentralized optimization over networks, where agents minimize a sum of {\it locally} smooth (strongly) convex losses and plus a nonsmooth convex extended value term. We propose decentralized methods wherein agents {\it adaptively} adjust their stepsize via local backtracking procedures coupled with lightweight min-consensus protocols. Our design stems from a three-operator splitting factorization applied to an equivalent reformulation of the problem. The reformulation is endowed with a new BCV preconditioning metric (Bertsekas-O'Connor-Vandenberghe), which enables efficient decentralized implementation and local stepsize adjustments. We establish robust convergence guarantees. Under mere convexity, the proposed methods converge with a sublinear rate. Under strong convexity of the sum-function, and assuming the nonsmooth component is partly smooth, we further prove linear convergence. Numerical experiments corroborate the theory and highlight the effectiveness of the proposed adaptive stepsize strategy.

In multi-agent IR pipelines for tasks such as search and ranking, LLM-based agents exchange intermediate reasoning in terms of Chain-of-Thought (CoT) with each other. Current CoT evaluation narrowly focuses on target task accuracy. However, this metric fails to assess the quality or utility of the reasoning process itself. To address this limitation, we introduce two novel measures: reusability and verifiability. We decouple CoT generation from execution using a Thinker-Executor framework. Reusability measures how easily an Executor can reuse the Thinker's CoT. Verifiability measures how frequently an Executor can match the Thinker's answer using the CoT. We evaluated four Thinker models against a committee of ten Executor models across five benchmarks. Our results reveal that reusability and verifiability do not correlate with standard accuracy, exposing a blind spot in current accuracy-based leaderboards for reasoning capability. Surprisingly, we find that CoTs from specialized reasoning models are not consistently more reusable or verifiable than those from general-purpose LLMs like Llama and Gemma.

Efficient estimation of causal and structural parameters can be automated using the Riesz representation theorem and debiased machine learning (DML). We present genriesz, an open-source Python package that implements automatic DML and generalized Riesz regression, a unified framework for estimating Riesz representers by minimizing empirical Bregman divergences. This framework includes covariate balancing, nearest-neighbor matching, calibrated estimation, and density ratio estimation as special cases. A key design principle of the package is automatic regressor balancing (ARB): given a Bregman generator $g$ and a representer model class, genriesz} automatically constructs a compatible link function so that the generalized Riesz regression estimator satisfies balancing (moment-matching) optimality conditions in a user-chosen basis. The package provides a modulr interface for specifying (i) the target linear functional via a black-box evaluation oracle, (ii) the representer model via basis functions (polynomial, RKHS approximations, random forest leaf encodings, neural embeddings, and a nearest-neighbor catchment basis), and (iii) the Bregman generator, with optional user-supplied derivatives. It returns regression adjustment (RA), Riesz weighting (RW), augmented Riesz weighting (ARW), and TMLE-style estimators with cross-fitting, confidence intervals, and $p$-values. We highlight representative workflows for estimation problems such as the average treatment effect (ATE), ATE on treated (ATT), and average marginal effect estimation. The Python package is available at https://github.com/MasaKat0/genriesz and on PyPI.

Fine-grained skill representations, commonly referred to as knowledge components (KCs), are fundamental to many approaches in student modeling and learning analytics. However, KC-level correctness labels are rarely available in real-world datasets, especially for open-ended programming tasks where solutions typically involve multiple KCs simultaneously. Simply propagating problem-level correctness to all associated KCs obscures partial mastery and often leads to poorly fitted learning curves. To address this challenge, we propose an automated framework that leverages large language models (LLMs) to label KC-level correctness directly from student-written code. Our method assesses whether each KC is correctly applied and further introduces a temporal context-aware Code-KC mapping mechanism to better align KCs with individual student code. We evaluate the resulting KC-level correctness labels in terms of learning curve fit and predictive performance using the power law of practice and the Additive Factors Model. Experimental results show that our framework leads to learning curves that are more consistent with cognitive theory and improves predictive performance, compared to baselines. Human evaluation further demonstrates substantial agreement between LLM and expert annotations.

We study the self-stabilizing leader election problem in anonymous $n$-nodes networks. Achieving self-stabilization with low space memory complexity is particularly challenging, and designing space-optimal leader election algorithms remains an open problem for general graphs. In deterministic settings, it is known that $Ω(\log \log n)$ bits of memory per node are necessary [Blin et al., Disc. Math. \& Theor. Comput. Sci., 2023], while in probabilistic settings the same lower bound holds for some values of $n$, but only for an unfair scheduler [Beauquier et al., PODC 1999]. Several deterministic and probabilistic protocols have been proposed in models ranging from the state model to the population protocols. However, to the best of our knowledge, existing solutions either require $Ω(\log n)$ bits of memory per node for general worst case graphs, or achieve low state complexity only under restricted network topologies such as rings, trees, or bounded-degree graphs. In this paper, we present a probabilistic self-stabilizing leader election algorithm for arbitrary anonymous networks that uses $O(\log \log n)$ bits of memory per node. Our algorithm operates in the state model under a synchronous scheduler and assumes knowledge of a global parameter $N = Θ(\log n)$. We show that, under our protocol, the system converges almost surely to a stable configuration with a unique leader and stabilizes within $O(\mathrm{poly}(n))$ rounds with high probability. To achieve $O(\log \log n)$ bits of memory, our algorithm keeps transmitting information after convergence, i.e. it does not verify the silence property. Moreover, like most works in the field, our algorithm does not provide explicit termination detection (i.e., nodes do not detect when the algorithm has converged).

Robotic camera systems enable dynamic, repeatable motion beyond human capabilities, yet their adoption remains limited by the high cost and operational complexity of industrial-grade platforms. We present the Intelligent Robotic Imaging System (IRIS), a task-specific 6-DOF manipulator designed for autonomous, learning-driven cinematic motion control. IRIS integrates a lightweight, fully 3D-printed hardware design with a goal-conditioned visuomotor imitation learning framework based on Action Chunking with Transformers (ACT). The system learns object-aware and perceptually smooth camera trajectories directly from human demonstrations, eliminating the need for explicit geometric programming. The complete platform costs under $1,000 USD, supports a 1.5 kg payload, and achieves approximately 1 mm repeatability. Real-world experiments demonstrate accurate trajectory tracking, reliable autonomous execution, and generalization across diverse cinematic motions.

This position paper presents a vision for self-driving particle accelerators that operate autonomously with minimal human intervention. We propose that future facilities be designed through artificial intelligence (AI) co-design, where AI jointly optimizes the accelerator lattice, diagnostics, and science application from inception to maximize performance while enabling autonomous operation. Rather than retrofitting AI onto human-centric systems, we envision facilities designed from the ground up as AI-native platforms. We outline nine critical research thrusts spanning agentic control architectures, knowledge integration, adaptive learning, digital twins, health monitoring, safety frameworks, modular hardware design, multimodal data fusion, and cross-domain collaboration. This roadmap aims to guide the accelerator community toward a future where AI-driven design and operation deliver unprecedented science output and reliability.

We present a systematic evaluation framework - thirty-seven analyses, 153 statistical tests, four cell types, two perturbation modalities - for assessing mechanistic interpretability in single-cell foundation models. Applying this framework to scGPT and Geneformer, we find that attention patterns encode structured biological information with layer-specific organisation - protein-protein interactions in early layers, transcriptional regulation in late layers - but this structure provides no incremental value for perturbation prediction: trivial gene-level baselines outperform both attention and correlation edges (AUROC 0.81-0.88 versus 0.70), pairwise edge scores add zero predictive contribution, and causal ablation of regulatory heads produces no degradation. These findings generalise from K562 to RPE1 cells; the attention-correlation relationship is context-dependent, but gene-level dominance is universal. Cell-State Stratified Interpretability (CSSI) addresses an attention-specific scaling failure, improving GRN recovery up to 1.85x. The framework establishes reusable quality-control standards for the field.

This position paper argues that current benchmarking practice in 12-lead ECG representation learning must be fixed to ensure progress is reliable and aligned with clinically meaningful objectives. The field has largely converged on three public multi-label benchmarks (PTB-XL, CPSC2018, CSN) dominated by arrhythmia and waveform-morphology labels, even though the ECG is known to encode substantially broader clinical information. We argue that downstream evaluation should expand to include an assessment of structural heart disease and patient-level forecasting, in addition to other evolving ECG-related endpoints, as relevant clinical targets. Next, we outline evaluation best practices for multi-label, imbalanced settings, and show that when they are applied, the literature's current conclusion about which representations perform best is altered. Furthermore, we demonstrate the surprising result that a randomly initialized encoder with linear evaluation matches state-of-the-art pre-training on many tasks. This motivates the use of a random encoder as a reasonable baseline model. We substantiate our observations with an empirical evaluation of three representative ECG pre-training approaches across six evaluation settings: the three standard benchmarks, a structural disease dataset, hemodynamic inference, and patient forecasting.

Despite significant progress in post-hoc explanation methods for neural networks, many remain heuristic and lack provable guarantees. A key approach for obtaining explanations with provable guarantees is by identifying a cardinally-minimal subset of input features which by itself is provably sufficient to determine the prediction. However, for standard neural networks, this task is often computationally infeasible, as it demands a worst-case exponential number of verification queries in the number of input features, each of which is NP-hard. In this work, we show that for Neural Additive Models (NAMs), a recent and more interpretable neural network family, we can efficiently generate explanations with such guarantees. We present a new model-specific algorithm for NAMs that generates provably cardinally-minimal explanations using only a logarithmic number of verification queries in the number of input features, after a parallelized preprocessing step with logarithmic runtime in the required precision is applied to each small univariate NAM component. Our algorithm not only makes the task of obtaining cardinally-minimal explanations feasible, but even outperforms existing algorithms designed to find the relaxed variant of subset-minimal explanations - which may be larger and less informative but easier to compute - despite our algorithm solving a much more difficult task. Our experiments demonstrate that, compared to previous algorithms, our approach provides provably smaller explanations than existing works and substantially reduces the computation time. Moreover, we show that our generated provable explanations offer benefits that are unattainable by standard sampling-based techniques typically used to interpret NAMs.

Large language models (LLMs) have shown strong potential across a variety of tasks, but their application in the telecom field remains challenging due to domain complexity, evolving standards, and specialized terminology. Therefore, general-domain LLMs may struggle to provide accurate and reliable outputs in this context, leading to increased hallucinations and reduced utility in telecom operations.To address these limitations, this work introduces KG-RAG-a novel framework that integrates knowledge graphs (KGs) with retrieval-augmented generation (RAG) to enhance LLMs for telecom-specific tasks. In particular, the KG provides a structured representation of domain knowledge derived from telecom standards and technical documents, while RAG enables dynamic retrieval of relevant facts to ground the model's outputs. Such a combination improves factual accuracy, reduces hallucination, and ensures compliance with telecom specifications.Experimental results across benchmark datasets demonstrate that KG-RAG outperforms both LLM-only and standard RAG baselines, e.g., KG-RAG achieves an average accuracy improvement of 14.3% over RAG and 21.6% over LLM-only models. These results highlight KG-RAG's effectiveness in producing accurate, reliable, and explainable outputs in complex telecom scenarios.

Some transformer attention heads appear to function as membership testers, dedicating themselves to answering the question "has this token appeared before in the context?" We identify these heads across four language models (GPT-2 small, medium, and large; Pythia-160M) and show that they form a spectrum of membership-testing strategies. Two heads (L0H1 and L0H5 in GPT-2 small) function as high-precision membership filters with false positive rates of 0-4\% even at 180 unique context tokens -- well above the $d_\text{head} = 64$ bit capacity of a classical Bloom filter. A third head (L1H11) shows the classic Bloom filter capacity curve: its false positive rate follows the theoretical formula $p \approx (1 - e^{-kn/m})^k$ with $R^2 = 1.0$ and fitted capacity $m \approx 5$ bits, saturating by $n \approx 20$ unique tokens. A fourth head initially identified as a Bloom filter (L3H0) was reclassified as a general prefix-attention head after confound controls revealed its apparent capacity curve was a sequence-length artifact. Together, the three genuine membership-testing heads form a multi-resolution system concentrated in early layers (0-1), taxonomically distinct from induction and previous-token heads, with false positive rates that decay monotonically with embedding distance -- consistent with distance-sensitive Bloom filters. These heads generalize broadly: they respond to any repeated token type, not just repeated names, with 43\% higher generalization than duplicate-token-only heads. Ablation reveals these heads contribute to both repeated and novel token processing, indicating that membership testing coexists with broader computational roles. The reclassification of L3H0 through confound controls strengthens rather than weakens the case: the surviving heads withstand the scrutiny that eliminated a false positive in our own analysis.

In variational inference (VI), the practitioner approximates a high-dimensional distribution $π$ with a simple surrogate one, often a (product) Gaussian distribution. However, in many cases of practical interest, Gaussian distributions might not capture the correct radial profile of $π$, resulting in poor coverage. In this work, we approach the VI problem from the perspective of optimizing over these radial profiles. Our algorithm radVI is a cheap, effective add-on to many existing VI schemes, such as Gaussian (mean-field) VI and Laplace approximation. We provide theoretical convergence guarantees for our algorithm, owing to recent developments in optimization over the Wasserstein space--the space of probability distributions endowed with the Wasserstein distance--and new regularity properties of radial transport maps in the style of Caffarelli (2000).

Large language models (LLMs) demonstrate strong performance on standard digital logic and Boolean reasoning tasks, yet their reliability under locally redefined semantics remains poorly understood. In many formal settings, such as circuit specifications, examinations, and hardware documentation, operators and components are explicitly redefined within narrow scope. Correct reasoning in these contexts requires models to temporarily suppress globally learned conventions in favor of prompt-local definitions. In this work, we study a systematic failure mode we term semantic override, in which an LLM reverts to its pretrained default interpretation of operators or gate behavior despite explicit redefinition in the prompt. We also identify a related class of errors, assumption injection, where models commit to unstated hardware semantics when critical details are underspecified, rather than requesting clarification. We introduce a compact micro-benchmark of 30 logic and digital-circuit reasoning tasks designed as verifier-style traps, spanning Boolean algebra, operator overloading, redefined gates, and circuit-level semantics. Evaluating three frontier LLMs, we observe persistent noncompliance with local specifications, confident but incompatible assumptions, and dropped constraints even in elementary settings. Our findings highlight a gap between surface-level correctness and specification-faithful reasoning, motivating evaluation protocols that explicitly test local unlearning and semantic compliance in formal domains.

Clinical free-text notes contain vital patient information. They are structured into labelled sections; recognizing these sections has been shown to support clinical decision-making and downstream NLP tasks. In this paper, we advance clinical section segmentation through three key contributions. First, we curate a new de-identified, section-labeled obstetrics notes dataset, to supplement the medical domains covered in public corpora such as MIMIC-III, on which most existing segmentation approaches are trained. Second, we systematically evaluate transformer-based supervised models for section segmentation on a curated subset of MIMIC-III (in-domain), and on the new obstetrics dataset (out-of-domain). Third, we conduct the first head-to-head comparison of supervised models for medical section segmentation with zero-shot large language models. Our results show that while supervised models perform strongly in-domain, their performance drops substantially out-of-domain. In contrast, zero-shot models demonstrate robust out-of-domain adaptability once hallucinated section headers are corrected. These findings underscore the importance of developing domain-specific clinical resources and highlight zero-shot segmentation as a promising direction for applying healthcare NLP beyond well-studied corpora, as long as hallucinations are appropriately managed.

We introduce CRAFT (Cross-layer Rank Adaptation via Frozen Tucker), a parameter-efficient fine-tuning (PEFT) method that applies Tucker tensor decomposition to pre-trained attention weight matrices stacked across transformer layers and trains only small square adaptation matrices on the resulting frozen Tucker factors. Existing tensor-based PEFT methods decompose gradient updates: LoTR applies Tucker decomposition with shared factor matrices, while SuperLoRA groups and reshapes $ΔW$ across layers before applying Tucker decomposition. Separately, methods like PiSSA apply SVD to pre-trained weights but operate independently per layer. CRAFT bridges these two lines of work: it performs full Tucker decomposition via Higher-Order SVD (HOSVD) directly on pre-trained weights organized as cross-layer 3D tensors, freezes all resulting factors, and adapts the model through lightweight trainable transformations applied to each factor matrix. Experiments on the GLUE benchmark using RoBERTa-base and RoBERTa-large demonstrate that CRAFT achieves competitive performance with existing methods while requiring only 41K Tucker adaptation parameters--a count independent of model dimension and depth at fixed Tucker ranks.

This work presents a practical benchmarking framework for optimizing artificial intelligence (AI) models on ARM Cortex processors (M0+, M4, M7), focusing on energy efficiency, accuracy, and resource utilization in embedded systems. Through the design of an automated test bench, we provide a systematic approach to evaluate across key performance indicators (KPIs) and identify optimal combinations of processor and AI model. The research highlights a nearlinear correlation between floating-point operations (FLOPs) and inference time, offering a reliable metric for estimating computational demands. Using Pareto analysis, we demonstrate how to balance trade-offs between energy consumption and model accuracy, ensuring that AI applications meet performance requirements without compromising sustainability. Key findings indicate that the M7 processor is ideal for short inference cycles, while the M4 processor offers better energy efficiency for longer inference tasks. The M0+ processor, while less efficient for complex AI models, remains suitable for simpler tasks. This work provides insights for developers, guiding them to design energy-efficient AI systems that deliver high performance in realworld applications.

Natural Language Processing (NLP) tools support requirements engineering (RE) tasks like requirements elicitation, classification, and validation. However, they are often developed from scratch despite functional overlaps, and abandoned after publication. This lack of interoperability and maintenance incurs unnecessary development effort, impedes tool comparison and benchmarking, complicates documentation, and diminishes the long-term sustainability of NLP4RE tools. To address these issues, we postulate a vision to transition from monolithic NLP4RE tools to an ecosystem of reusable, interoperable modules. We outline a research roadmap towards a software reference architecture (SRA) to realize this vision, elaborated following a standard methodological framework for SRA development. As an initial step, we conducted a stakeholder-driven focus group session to elicit generic system requirements for NLP4RE tools. This activity resulted in 36 key system requirements, further motivating the need for a dedicated SRA. Overall, the proposed vision, roadmap, and initial contribution pave the way towards improved development, reuse, and long-term maintenance of NLP4RE tools.

Learning from self-sampled data and sparse environmental feedback remains a fundamental challenge in training self-evolving agents. Temporal credit assignment mitigates this issue by transforming sparse feedback into dense supervision signals. However, previous approaches typically depend on learning task-specific value functions for credit assignment, which suffer from poor sample efficiency and limited generalization. In this work, we propose to leverage pretrained knowledge from large language models (LLMs) to transform sparse rewards into dense training signals (i.e., the advantage function) through retrospective in-context learning (RICL). We further propose an online learning framework, RICOL, which iteratively refines the policy based on the credit assignment results from RICL. We empirically demonstrate that RICL can accurately estimate the advantage function with limited samples and effectively identify critical states in the environment for temporal credit assignment. Extended evaluation on four BabyAI scenarios show that RICOL achieves comparable convergent performance with traditional online RL algorithms with significantly higher sample efficiency. Our findings highlight the potential of leveraging LLMs for temporal credit assignment, paving the way for more sample-efficient and generalizable RL paradigms.

We develop a method for training neural networks on Boolean data in which the values at all nodes are strictly $\pm 1$, and the resulting models are typically equivalent to networks whose nonzero weights are also $\pm 1$. The method replaces loss minimization with a nonconvex constraint formulation. Each node implements a Boolean threshold function (BTF), and training is expressed through a divide-and-concur decomposition into two complementary constraints: one enforces local BTF consistency between inputs, weights, and output; the other imposes architectural concurrence, equating neuron outputs with downstream inputs and enforcing weight equality across training-data instantiations of the network. The reflect-reflect-relax (RRR) projection algorithm is used to reconcile these constraints. Each BTF constraint includes a lower bound on the margin. When this bound is sufficiently large, the learned representations are provably sparse and equivalent to networks composed of simple logical gates with $\pm 1$ weights. Across a range of tasks -- including multiplier-circuit discovery, binary autoencoding, logic-network inference, and cellular automata learning -- the method achieves exact solutions or strong generalization in regimes where standard gradient-based methods struggle. These results demonstrate that projection-based constraint satisfaction provides a viable and conceptually distinct foundation for learning in discrete neural systems, with implications for interpretability and efficient inference.

Feedback delays are inevitable in real-world multi-agent learning. They are known to severely degrade performance, and the convergence rate under delayed feedback is still unclear, even for bilinear games. This paper derives the rate of linear convergence of Weighted Optimistic Gradient Descent-Ascent (WOGDA), which predicts future rewards with extra optimism, in unconstrained bilinear games. To analyze the algorithm, we interpret it as an approximation of the Extra Proximal Point (EPP), which is updated based on farther future rewards than the classical Proximal Point (PP). Our theorems show that standard optimism (predicting the next-step reward) achieves linear convergence to the equilibrium at a rate $\exp(-Θ(t/m^{5}))$ after $t$ iterations for delay $m$. Moreover, employing extra optimism (predicting farther future reward) tolerates a larger step size and significantly accelerates the rate to $\exp(-Θ(t/(m^{2}\log m)))$. Our experiments also show accelerated convergence driven by the extra optimism and are qualitatively consistent with our theorems. In summary, this paper validates that extra optimism is a promising countermeasure against performance degradation caused by feedback delays.

Image Copy Detection (ICD) aims to identify manipulated content between image pairs through robust feature representation learning. While self-supervised learning (SSL) has advanced ICD systems, existing view-level contrastive methods struggle with sophisticated edits due to insufficient fine-grained correspondence learning. We address this limitation by exploiting the inherent geometric traceability in edited content through two key innovations. First, we propose PixTrace - a pixel coordinate tracking module that maintains explicit spatial mappings across editing transformations. Second, we introduce CopyNCE, a geometrically-guided contrastive loss that regularizes patch affinity using overlap ratios derived from PixTrace's verified mappings. Our method bridges pixel-level traceability with patch-level similarity learning, suppressing supervision noise in SSL training. Extensive experiments demonstrate not only state-of-the-art performance (88.7% uAP / 83.9% RP90 for matcher, 72.6% uAP / 68.4% RP90 for descriptor on DISC21 dataset) but also better interpretability over existing methods.

Large language models (LLMs), and conversational agents based on them, are exposed to personal data (PD) during pre-training and during user interactions. Prior work shows that PD can resurface, yet users lack insight into how strongly models associate specific information to their identity. We audit PD across eight LLMs (3 open-source; 5 API-based, including GPT-4o), introduce LMP2 (Language Model Privacy Probe), a human-centered, privacy-preserving audit tool refined through two formative studies (N=20), and run two studies with EU residents to capture (i) intuitions about LLM-generated PD (N1=155) and (ii) reactions to tool output (N2=303). We show empirically that models confidently generate multiple PD categories for well-known individuals. For everyday users, GPT-4o generates 11 features with 60% or more accuracy (e.g., gender, hair color, languages). Finally, 72% of participants sought control over model-generated associations with their name, raising questions about what counts as PD and whether data privacy rights should extend to LLMs.

Deep generative models such as flow matching and diffusion models have shown great potential in learning complex distributions and dynamical systems, but often act as black-boxes, neglecting underlying physics. In contrast, physics-based simulation models described by ODEs/PDEs remain interpretable, but may have missing or unknown terms, unable to fully describe real-world observations. We bridge this gap with a novel grey-box method that integrates incomplete physics models directly into generative models. Our approach learns dynamics from observational trajectories alone, without ground-truth physics parameters, in a simulation-free manner that avoids scalability and stability issues of Neural ODEs. The core of our method lies in modelling a structured variational distribution within the flow matching framework, by using two latent encodings: one to model the missing stochasticity and multi-modal velocity, and a second to encode physics parameters as a latent variable with a physics-informed prior. Furthermore, we present an adaptation of the framework to handle second-order dynamics. Our experiments on representative ODE/PDE problems show that our method performs on par with or superior to fully data-driven approaches and previous grey-box baselines, while preserving the interpretability of the physics model. Our code is available at https://github.com/DMML-Geneva/VGB-DM.

Large Language Models (LLMs) consistently excel in diverse medical Natural Language Processing (NLP) tasks, yet their substantial computational requirements often limit deployment in real-world healthcare settings. In this work, we investigate whether "small" LLMs (around one billion parameters) can effectively perform medical tasks while maintaining competitive accuracy. We evaluate models from three major families-Llama-3, Gemma-3, and Qwen3-across 20 clinical NLP tasks among Named Entity Recognition, Relation Extraction, Case Report Form Filling, Question Answering, and Argument Mining. We systematically compare a range of adaptation strategies, both at inference time (few-shot prompting, constraint decoding) and at training time (supervised fine-tuning, continual pretraining). Fine-tuning emerges as the most effective approach, while the combination of few-shot prompting and constraint decoding offers strong lower-resource alternatives. Our results show that small LLMs can match or even surpass larger baselines, with our best configuration based on Qwen3-1.7B achieving an average score +9.2 points higher than Qwen3-32B. We release a comprehensive collection of all the publicly available Italian medical datasets for NLP tasks, together with our top-performing models. Furthermore, we release an Italian dataset of 126M words from the Emergency Department of an Italian Hospital, and 175M words from various sources that we used for continual pre-training.

The core theme of bidirectional alignment is ensuring that AI systems accurately understand human intent and that humans can trust AI behavior. However, this loop fractures significantly across language barriers. Our research addresses Cross-Lingual Sentiment Misalignment between Bengali and English by benchmarking four transformer architectures. We reveal severe safety and representational failures in current alignment paradigms. We demonstrate that compressed model (mDistilBERT) exhibits 28.7% "Sentiment Inversion Rate," fundamentally misinterpreting positive user intent as negative (or vice versa). Furthermore, we identify systemic nuances affecting human-AI trust, including "Asymmetric Empathy" where some models systematically dampen and others amplify the affective weight of Bengali text relative to its English counterpart. Finally, we reveal a "Modern Bias" in the regional model (IndicBERT), which shows a 57% increase in alignment error when processing formal (Sadhu) Bengali. We argue that equitable human-AI co-evolution requires pluralistic, culturally grounded alignment that respects language and dialectal diversity over universal compression, which fails to preserve the emotional fidelity required for reciprocal human-AI trust. We recommend that alignment benchmarks incorporate "Affective Stability" metrics that explicitly penalize polarity inversions in low-resource and dialectal contexts.

The increasing volume of hate speech on online platforms poses significant societal challenges. While the Natural Language Processing community has developed effective methods to automatically detect the presence of hate speech, responses to it, called counter-speech, are still an open challenge. We present PEACE 2.0, a novel tool that, besides analysing and explaining why a message is considered hateful or not, also generates a response to it. More specifically, PEACE 2.0 has three main new functionalities: leveraging a Retrieval-Augmented Generation (RAG) pipeline i) to ground HS explanations into evidence and facts, ii) to automatically generate evidence-grounded counter-speech, and iii) exploring the characteristics of counter-speech replies. By integrating these capabilities, PEACE 2.0 enables in-depth analysis and response generation for both explicit and implicit hateful messages.

Modern language models (LMs) increasingly require two critical resources: computational resources and data resources. Data selection techniques can effectively reduce the amount of training data required for fine-tuning LMs. However, their effectiveness is closely related to computational resources, which always require a high compute budget. Owing to the resource limitations in practical fine-tuning scenario, we systematically reveal the relationship between data selection and uncertainty estimation of selected data. Although large language models (LLMs) exhibit exceptional capabilities in language understanding and generation, which provide new ways to alleviate data scarcity, evaluating data usability remains a challenging task. This makes efficient data selection indispensable. To mitigate these issues, we propose Entropy-Based Unsupervised Data Selection (EUDS) framework. Empirical experiments on sentiment analysis (SA), topic classification (Topic-CLS), and question answering (Q&A) tasks validate its effectiveness. EUDS establishes a computationally efficient data-filtering mechanism. Theoretical analysis and experimental results confirm the effectiveness of our approach. EUDS significantly reduces computational costs and improves training time efficiency with less data requirement. This provides an innovative solution for the efficient fine-tuning of LMs in the compute-constrained scenarios.

We present Jolt Atlas, a zero-knowledge machine learning (zkML) framework that extends the Jolt proving system to model inference. Unlike zkVMs (zero-knowledge virtual machines), which emulate CPU instruction execution, Jolt Atlas adapts Jolt's lookup-centric approach and applies it directly to ONNX tensor operations. The ONNX computational model eliminates the need for CPU registers and simplifies memory consistency verification. In addition, ONNX is an open-source, portable format, which makes it easy to share and deploy models across different frameworks, hardware platforms, and runtime environments without requiring framework-specific conversions. Our lookup arguments, which use sumcheck protocol, are well-suited for non-linear functions -- key building blocks in modern ML. We apply optimisations such as neural teleportation to reduce the size of lookup tables while preserving model accuracy, as well as several tensor-level verification optimisations detailed in this paper. We demonstrate that Jolt Atlas can prove model inference in memory-constrained environments -- a prover property commonly referred to as \textit{streaming}. Furthermore, we discuss how Jolt Atlas achieves zero-knowledge through the BlindFold technique, as introduced in Vega. In contrast to existing zkML frameworks, we show practical proving times for classification, embedding, automated reasoning, and small language models. Jolt Atlas enables cryptographic verification that can be run on-device, without specialised hardware. The resulting proofs are succinctly verifiable. This makes Jolt Atlas well-suited for privacy-centric and adversarial environments. In a companion work, we outline various use cases of Jolt Atlas, including how it serves as guardrails in agentic commerce and for trustless AI context (often referred to as \textit{AI memory}).

Web research and practices have evolved significantly over time, offering users diverse and accessible solutions across a wide range of tasks. While advanced concepts such as Web 4.0 have emerged from mature technologies, the introduction of large language models (LLMs) has profoundly influenced both the field and its applications. This wave of LLMs has permeated science and technology so deeply that no area remains untouched. Consequently, LLMs are reshaping web research and development, transforming traditional pipelines into generative solutions for tasks like information retrieval, question answering, recommendation systems, and web analytics. They have also enabled new applications such as web-based summarization and educational tools. This survey explores recent advances in the impact of LLMs-particularly through the use of retrieval-augmented generation (RAG)-on web research and industry. It discusses key developments, open challenges, and future directions for enhancing web solutions with LLMs.

Multiple-choice question (MCQ) benchmarks have been a standard evaluation practice for measuring LLMs' ability to reason and answer knowledge-based questions. Through a synthetic NonsenseQA benchmark, we observe that different LLMs exhibit varying degrees of label-position-few-shot-prompt bias, where the model either uses the answer position, the label in front of the answer, the distributions of correct answers present in the few-shot prompt, or a combination of all to answer each MCQ question. We propose a simple bias-reduced evaluation protocol that replaces the labels of each question with uniform, unordered labels and prompts the LLM to use the whole answer presented. With a simple sentence similarity model, we demonstrate improved robustness and lower standard deviation between different permutations of answers with a minimal drop in LLM's performance, exposing the LLM's capabilities under reduced evaluation artifacts, without any help from the prompt examples or the option labels. Across multiple benchmarks and models, this protocol substantially improves the robustness to answer permutations, reducing mean accuracy variance $3\times$ with only a minimal decrease in the mean model's performance. Through ablation studies on various embedding models and similarity functions, we show that the method is more robust than the standard ones.

Evaluating the strategic reasoning capabilities of Large Language Models (LLMs) requires moving beyond static benchmarks to dynamic, multi-turn interactions. We introduce AIDG (Adversarial Information Deduction Game), a game-theoretic framework that probes the asymmetry between information extraction (active deduction) and information containment (state maintenance) in dialogue. We propose two complementary tasks: AIDG-I, measuring pragmatic strategy in social deduction, and AIDG-II, measuring constraint satisfaction in a structured "20 Questions" setting. Across 439 games with six frontier LLMs, we observe a clear capability asymmetry: models perform substantially better at containment than deduction, with a 350 ELO advantage on defense;(Cohen's d = 5.47). We identify two bottlenecks driving this gap: (1) Information Dynamics, where confirmation strategies are 7.75x more effective than blind deduction (p < 0.00001), and (2) Constraint Adherence, where instruction-following degrades under conversational load, accounting for 41.3% of deductive failures. These findings suggest that while LLMs excel at local defensive coherence, they struggle with the global state tracking required for strategic inquiry.

Innovation in Recommender Systems is currently impeded by a fractured ecosystem, where researchers must choose between the ease of in-memory experimentation and the costly, complex rewriting required for distributed industrial engines. To bridge this gap, we present WarpRec, a high-performance framework that eliminates this trade-off through a novel, backend-agnostic architecture. It includes 50+ state-of-the-art algorithms, 40 metrics, and 19 filtering and splitting strategies that seamlessly transition from local execution to distributed training and optimization. The framework enforces ecological responsibility by integrating CodeCarbon for real-time energy tracking, showing that scalability need not come at the cost of scientific integrity or sustainability. Furthermore, WarpRec anticipates the shift toward Agentic AI, leading Recommender Systems to evolve from static ranking engines into interactive tools within the Generative AI ecosystem. In summary, WarpRec not only bridges the gap between academia and industry but also can serve as the architectural backbone for the next generation of sustainable, agent-ready Recommender Systems. Code is available at https://github.com/sisinflab/warprec/

Multi-agent planning under Signal Temporal Logic (STL) is often hindered by collaborative tasks that lead to computational challenges due to the inherent high-dimensionality of the problem, preventing scalable synthesis with satisfaction guarantees. To address this, we formulate STL planning as an optimization program under arbitrary multi-agent constraints and introduce a penalty-based unconstrained relaxation that can be efficiently solved via a Block-Coordinate Gradient Descent (BCGD) method, where each block corresponds to a single agent's decision variables, thereby mitigating complexity. By utilizing a quadratic penalty function defined via smooth STL semantics, we show that BCGD iterations converge to a stationary point of the penalized problem under standard regularity assumptions. To enforce feasibility, the BCGD solver is embedded within a two-layer optimization scheme: inner BCGD updates are performed for a fixed penalty parameter, which is then increased in an outer loop to progressively improve multi-agent STL robustness. The proposed framework enables scalable computations and is validated through various complex multi-robot planning scenarios.

Uncertainty quantification has emerged as an effective approach to closed-book hallucination detection for LLMs, but existing methods are largely designed for short-form outputs and do not generalize well to long-form generation. We introduce a taxonomy for fine-grained uncertainty quantification in long-form LLM outputs that distinguishes methods by design choices at three stages: response decomposition, unit-level scoring, and response-level aggregation. We formalize several families of consistency-based black-box scorers, providing generalizations and extensions of existing methods. In our experiments across multiple LLMs and datasets, we find 1) claim-response entailment consistently performs better or on par with more complex claim-level scorers, 2) claim-level scoring generally yields better results than sentence-level scoring, and 3) uncertainty-aware decoding is highly effective for improving the factuality of long-form outputs. Our framework clarifies relationships between prior methods, enables apples-to-apples comparisons, and provides practical guidance for selecting components for fine-grained UQ.

Self-adaptive systems increasingly operate in close interaction with humans, often sharing the same physical or virtual environments and making decisions with ethical implications at runtime. Current approaches typically encode ethics as fixed, rule-based constraints or as a single chosen ethical theory embedded at design time. This overlooks a fundamental property of human-system interaction settings: ethical preferences vary across individuals and groups, evolve with context, and may conflict, while still needing to remain within a legally and regulatorily defined hard-ethics envelope (e.g., safety and compliance constraints). This paper advocates a shift from static ethical rules to runtime ethical reasoning for self-adaptive systems, where ethical preferences are treated as runtime requirements that must be elicited, represented, and continuously revised as stakeholders and situations change. We argue that satisfying such requirements demands explicit ethics-based negotiation to manage ethical trade-offs among multiple humans who interact with, are represented by, or are affected by a system. We identify key challenges, ethical uncertainty, conflicts among ethical values (including human, societal, and environmental drivers), and multi-dimensional/multi-party/multi-driver negotiation, and outline research directions and questions toward ethically self-adaptive systems.

Evaluating machine translation (MT) quality in extremely low-resource language (ELRL) scenarios poses unique challenges, as widely used metrics such as BLEU, effective in high-resource settings, often misrepresent quality in data-scarce contexts. This work presents a comparative analysis of BLEU, an n-gram-based metric, and ChrF++, a character-based metric, for MT evaluation in ELRL settings. We examine how each metric responds to translation artifacts, including hallucinations, repetition, source-text copying, and diacritic (\textit{matra}) variations across three ELRLs: Magahi, Bhojpuri, and Chhattisgarhi, with a focus on outputs from large language models (LLMs) and neural MT (NMT) systems. While recent work often relies solely on ChrF++, our findings show that BLEU, despite its lower absolute scores, provides complementary lexical-precision insights that improve interpretability.

Cross-document coreference resolution (CDCR) identifies and links mentions of the same entities and events across related documents, enabling content analysis that aggregates information at the level of discourse participants. However, existing datasets primarily focus on event resolution and employ a narrow definition of coreference, which limits their effectiveness in analyzing diverse and polarized news coverage where wording varies widely. This paper proposes a revised CDCR annotation scheme of the NewsWCL50 dataset, treating coreference chains as discourse elements (DEs) and conceptual units of analysis. The approach accommodates both identity and near-identity relations, e.g., by linking "the caravan" - "asylum seekers" - "those contemplating illegal entry", allowing models to capture lexical diversity and framing variation in media discourse, while maintaining the fine-grained annotation of DEs. We reannotate the NewsWCL50 and a subset of ECB+ using a unified codebook and evaluate the new datasets through lexical diversity metrics and a same-head-lemma baseline. The results show that the reannotated datasets align closely, falling between the original ECB+ and NewsWCL50, thereby supporting balanced and discourse-aware CDCR research in the news domain.

We investigate the convergence guarantee of two-layer neural network training with Gaussian randomly masked inputs. This scenario corresponds to Gaussian dropout at the input level, or noisy input training common in sensor networks, privacy-preserving training, and federated learning, where each user may have access to partial or corrupted features. Using a Neural Tangent Kernel (NTK) analysis, we demonstrate that training a two-layer ReLU network with Gaussian randomly masked inputs achieves linear convergence up to an error region proportional to the mask's variance. A key technical contribution is resolving the randomness within the non-linear activation, a problem of independent interest.

Children are increasingly using technologies powered by Artificial Intelligence (AI). However, there are growing concerns about privacy risks, particularly for children. Although existing privacy regulations require companies and organizations to implement protections, doing so can be challenging in practice. To address this challenge, this article proposes a framework based on Privacy-by-Design (PbD), which guides designers and developers to take on a proactive and risk-averse approach to technology design. Our framework includes principles from several privacy regulations, such as the General Data Protection Regulation (GDPR) from the European Union, the Personal Information Protection and Electronic Documents Act (PIPEDA) from Canada, and the Children's Online Privacy Protection Act (COPPA) from the United States. We map these principles to various stages of applications that use Large Language Models (LLMs), including data collection, model training, operational monitoring, and ongoing validation. For each stage, we discuss the operational controls found in the recent academic literature to help AI service providers and developers reduce privacy risks while meeting legal standards. In addition, the framework includes design guidelines for children, drawing from the United Nations Convention on the Rights of the Child (UNCRC), the UK's Age-Appropriate Design Code (AADC), and recent academic research. To demonstrate how this framework can be applied in practice, we present a case study of an LLM-based educational tutor for children under 13. Through our analysis and the case study, we show that by using data protection strategies such as technical and organizational controls and making age-appropriate design decisions throughout the LLM life cycle, we can support the development of AI applications for children that provide privacy protections and comply with legal requirements.

In current inter-organizational data spaces, usage policies are enforced mainly at the asset level: a whole document or dataset is either shared or withheld. When only parts of a document are sensitive, providers who want to avoid leaking protected information typically must manually redact documents before sharing them, which is costly, coarse-grained, and hard to maintain as policies or partners change. We present DAVE, a usage policy-enforcing LLM spokesperson that answers questions over private documents on behalf of a data provider. Instead of releasing documents, the provider exposes a natural language interface whose responses are constrained by machine-readable usage policies. We formalize policy-violating information disclosure in this setting, drawing on usage control and information flow security, and introduce virtual redaction: suppressing sensitive information at query time without modifying source documents. We describe an architecture for integrating such a spokesperson with Eclipse Dataspace Components and ODRL-style policies, and outline an initial provider-side integration prototype in which QA requests are routed through a spokesperson service instead of triggering raw document transfer. Our contribution is primarily architectural: we do not yet implement or empirically evaluate the full enforcement pipeline. We therefore outline an evaluation methodology to assess security, utility, and performance trade-offs under benign and adversarial querying as a basis for future empirical work on systematically governed LLM access to multi-party data spaces.

Large language models (LLMs) have shown great promise in recommender systems, where supervised fine-tuning (SFT) is commonly used for adaptation. Subsequent studies further introduce preference learning to incorporate negative samples into the training process. However, existing methods rely on sequence-level, offline-generated negatives, making them less discriminative and informative when adapting LLMs to recommendation tasks with large negative item spaces. To address these challenges, we propose ILRec, a novel preference fine-tuning framework for LLM-based recommendation, leveraging self-hard negative signals extracted from intermediate layers to improve preference learning. Specifically, we identify self-hard negative tokens from intermediate layers as fine-grained negative supervision that dynamically reflects the model's preference learning process. To effectively integrate these signals into training, we design a two-stage framework comprising cross-layer preference optimization and cross-layer preference distillation, enabling the model to jointly discriminate informative negatives and enhance the quality of negative signals from intermediate layers. In addition, we introduce a lightweight collaborative filtering model to assign token-level rewards for negative signals, mitigating the risk of over-penalizing false negatives. Extensive experiments on three datasets demonstrate ILRec's effectiveness in enhancing the performance of LLM-based recommender systems.

Predicting survival outcomes for non-small cell lung cancer (NSCLC) patients is challenging due to the different individual prognostic features. This task can benefit from the integration of whole-slide images, bulk transcriptomics, and DNA methylation, which offer complementary views of the patient's condition at diagnosis. However, real-world clinical datasets are often incomplete, with entire modalities missing for a significant fraction of patients. State-of-the-art models rely on available data to create patient-level representations or use generative models to infer missing modalities, but they lack robustness in cases of severe missingness. We propose a Multimodal Contrastive Variational AutoEncoder (MCVAE) to address this issue: modality-specific variational encoders capture the uncertainty in each data source, and a fusion bottleneck with learned gating mechanisms is introduced to normalize the contributions from present modalities. We propose a multi-task objective that combines survival loss and reconstruction loss to regularize patient representations, along with a cross-modal contrastive loss that enforces cross-modal alignment in the latent space. During training, we apply stochastic modality masking to improve the robustness to arbitrary missingness patterns. Extensive evaluations on the TCGA-LUAD (n=475) and TCGA-LUSC (n=446) datasets demonstrate the efficacy of our approach in predicting disease-specific survival (DSS) and its robustness to severe missingness scenarios compared to two state-of-the-art models. Finally, we bring some clarifications on multimodal integration by testing our model on all subsets of modalities, finding that integration is not always beneficial to the task.

In recent years, significant advances in computer vision have also propelled progress in remote sensing. Concurrently, the use of drones has expanded, with many organizations incorporating them into their operations. Most drones are equipped by default with RGB cameras, which are both robust and among the easiest sensors to use and interpret. The body of literature on optical remote sensing is vast, encompassing diverse tasks, capabilities, and methodologies. Each task or methodology could warrant a dedicated survey. This work provides a comprehensive overview of the capabilities of the field, while also presenting key information, such as datasets and insights. It aims to serve as a guide for researchers entering the field, offering high-level insights and helping them focus on areas most relevant to their interests. To the best of our knowledge, no existing survey addresses this holistic perspective.

Generalized Category Discovery (GCD) aims to identify novel categories in unlabeled data while leveraging a small labeled subset of known classes. Training a parametric classifier solely on image features often leads to overfitting to old classes, and recent multimodal approaches improve performance by incorporating textual information. However, they treat modalities independently and incur high computational cost. We propose SpectralGCD, an efficient and effective multimodal approach to GCD that uses CLIP cross-modal image-concept similarities as a unified cross-modal representation. Each image is expressed as a mixture over semantic concepts from a large task-agnostic dictionary, which anchors learning to explicit semantics and reduces reliance on spurious visual cues. To maintain the semantic quality of representations learned by an efficient student, we introduce Spectral Filtering which exploits a cross-modal covariance matrix over the softmaxed similarities measured by a strong teacher model to automatically retain only relevant concepts from the dictionary. Forward and reverse knowledge distillation from the same teacher ensures that the cross-modal representations of the student remain both semantically sufficient and well-aligned. Across six benchmarks, SpectralGCD delivers accuracy comparable to or significantly superior to state-of-the-art methods at a fraction of the computational cost. The code is publicly available at: https://github.com/miccunifi/SpectralGCD.

Unmanned Aerial Vehicle (UAV)-assisted networks are increasingly foreseen as a promising approach for emergency response, providing rapid, flexible, and resilient communications in environments where terrestrial infrastructure is degraded or unavailable. In such scenarios, voice radio communications remain essential for first responders due to their robustness; however, their unstructured nature prevents direct integration with automated UAV-assisted network management. This paper proposes SIREN, an AI-driven framework that enables voice-driven perception for UAV-assisted networks. By integrating Automatic Speech Recognition (ASR) with Large Language Model (LLM)-based semantic extraction and Natural Language Processing (NLP) validation, SIREN converts emergency voice traffic into structured, machine-readable information, including responding units, location references, emergency severity, and Quality-of-Service (QoS) requirements. SIREN is evaluated using synthetic emergency scenarios with controlled variations in language, speaker count, background noise, and message complexity. The results demonstrate robust transcription and reliable semantic extraction across diverse operating conditions, while highlighting speaker diarization and geographic ambiguity as the main limiting factors. These findings establish the feasibility of voice-driven situational awareness for UAV-assisted networks and show a practical foundation for human-in-the-loop decision support and adaptive network management in emergency response operations.

Information retrieval lies at the foundation of the modern digital industry. While natural language search has seen dramatic progress in recent years largely driven by embedding-based models and large-scale pretraining, the field still faces significant challenges. Specifically, queries that involve complex relationships, object compositions, or precise constraints such as identities, counts and proportions often remain unresolved or unreliable within current frameworks. In this paper, we propose a novel framework that integrates formal verification into deep learning-based image retrieval through a synergistic combination of graph-based verification methods and neural code generation. Our approach aims to support open-vocabulary natural language queries while producing results that are both trustworthy and verifiable. By grounding retrieval results in a system of formal reasoning, we move beyond the ambiguity and approximation that often characterize vector representations. Instead of accepting uncertainty as a given, our framework explicitly verifies each atomic truth in the user query against the retrieved content. This allows us to not only return matching results, but also to identify and mark which specific constraints are satisfied and which remain unmet, thereby offering a more transparent and accountable retrieval process while boosting the results of the most popular embedding-based approaches.

Task Arithmetic yields a modular, scalable way to adapt foundation models. Combining multiple task vectors, however, can lead to cross-task interference, causing representation drift and degraded performance. Representation drift regularization provides a natural remedy to disentangle task vectors; however, existing approaches typically require external task data, conflicting with modularity and data availability constraints (e.g., privacy requirements). We propose a dataless approach by framing regularization against representation drift as a curvature matrix approximation problem. This allows us to leverage well-established techniques; in particular, we adopt Kronecker-Factored Approximate Curvature and obtain a practical regularizer that achieves state-of-the-art results in task addition and negation. Our method has constant complexity in the number of tasks and promotes robustness to task vector rescaling, eliminating the need for held-out tuning.

When students are unsure of the correct answer to a multiple-choice question (MCQ), guessing is common practice. The availability heuristic, proposed by A. Tversky and D. Kahneman in 1973, suggests that the ease with which relevant instances come to mind, typically operationalised by the mere frequency of exposure, can offer a mental shortcut for problems in which the test-taker does not know the exact answer. Is simply choosing the option that comes most readily to mind a good strategy for answering MCQs? We propose a computational method of assessing the cognitive availability of MCQ options operationalised by concepts' prevalence in large corpora. The key finding, across three large question sets, is that correct answers, independently of the question stem, are significantly more available than incorrect MCQ options. Specifically, using Wikipedia as the retrieval corpus, we find that always selecting the most available option leads to scores 13.5% to 32.9% above the random-guess baseline. We further find that LLM-generated MCQ options show similar patterns of availability compared to expert-created options, despite the LLMs' frequentist nature and their training on large collections of textual data. Our findings suggest that availability should be considered in current and future work when computationally modelling student behaviour.

We cast episodic Markov decision process (MDP) planning as Bayesian inference over _policies_. A policy is treated as the latent variable and is assigned an unnormalized probability of optimality that is monotone in its expected return, yielding a posterior distribution whose modes coincide with return-maximizing solutions while posterior dispersion represents uncertainty over optimal behavior. To approximate this posterior in discrete domains, we adapt variational sequential Monte Carlo (VSMC) to inference over deterministic policies under stochastic dynamics, introducing a sweep that enforces policy consistency across revisited states and couples transition randomness across particles to avoid confounding from simulator noise. Acting is performed by posterior predictive sampling, which induces a stochastic control policy through a Thompson-sampling interpretation rather than entropy regularization. Across grid worlds, Blackjack, Triangle Tireworld, and Academic Advising, we analyze the structure of inferred policy distributions and compare the resulting behavior to discrete Soft Actor-Critic, highlighting qualitative and statistical differences that arise from policy-level uncertainty.

Long-tail question answering presents significant challenges for large language models (LLMs) due to their limited ability to acquire and accurately recall less common knowledge. Retrieval-augmented generation (RAG) systems have shown great promise in mitigating this limitation by integrating external retrieval mechanisms. However, dense retrieval models often face the same difficulties when generalizing to rare or niche knowledge. In this study, we introduce RPDR, a novel data augmentation framework that selects high-quality easy-to-learn training data, to enhance dense retrievers. Our approach is built around three core components: synthetic data generation, data selection with Round-Trip prediction to identify easy-to-learn instances, and retriever training with these instances. We evaluate RPDR on two long-tail retrieval benchmarks, PopQA and EntityQuestion, demonstrating substantial improvements over existing retrievers like BM25 and Contriver, especially on extremely long-tail categories. We identify the strengths and limitations of RPDR through detailed human analysis and propose a dynamic routing mechanism to dynamically route queries to specialized retrieval modules to further improve retrieval performance.

Agents operating in complex software environments benefit from reasoning about the consequences of their actions, as even a single incorrect user interface (UI) operation can derail long, artifact-preserving workflows. This challenge is particularly acute for computer-using scenarios, where real execution does not support counterfactual exploration, making large-scale trial-and-error learning and planning impractical despite the environment being fully digital and deterministic. We introduce the Computer-Using World Model (CUWM), a world model for desktop software that predicts the next UI state given the current state and a candidate action. CUWM adopts a two-stage factorization of UI dynamics: it first predicts a textual description of agent-relevant state changes, and then realizes these changes visually to synthesize the next screenshot. CUWM is trained on offline UI transitions collected from agents interacting with real Microsoft Office applications, and further refined with a lightweight reinforcement learning stage that aligns textual transition predictions with the structural requirements of computer-using environments. We evaluate CUWM via test-time action search, where a frozen agent uses the world model to simulate and compare candidate actions before execution. Across a range of Office tasks, world-model-guided test-time scaling improves decision quality and execution robustness.

Machine learning models are increasingly applied to biomedical data, yet their adoption in high stakes domains remains limited by poor robustness, limited interpretability, and instability of learned features under realistic data perturbations, such as missingness. In particular, models that achieve high predictive performance may still fail to inspire trust if their key features fluctuate when data completeness changes, undermining reproducibility and downstream decision-making. Here, we present CACTUS (Comprehensive Abstraction and Classification Tool for Uncovering Structures), an explainable machine learning framework explicitly designed to address these challenges in small, heterogeneous, and incomplete clinical datasets. CACTUS integrates feature abstraction, interpretable classification, and systematic feature stability analysis to quantify how consistently informative features are preserved as data quality degrades. Using a real-world haematuria cohort comprising 568 patients evaluated for bladder cancer, we benchmark CACTUS against widely used machine learning approaches, including random forests and gradient boosting methods, under controlled levels of randomly introduced missing data. We demonstrate that CACTUS achieves competitive or superior predictive performance while maintaining markedly higher stability of top-ranked features as missingness increases, including in sex-stratified analyses. Our results show that feature stability provides information complementary to conventional performance metrics and is essential for assessing the trustworthiness of machine learning models applied to biomedical data. By explicitly quantifying robustness to missing data and prioritising interpretable, stable features, CACTUS offers a generalizable framework for trustworthy data-driven decision support.

Linear attention transformers have become a strong alternative to softmax attention due to their efficiency. However, linear attention tends to be less expressive and results in reduced accuracy compared to softmax attention. To bridge the accuracy gap between softmax attention and linear attention, we manipulate Mamba-2, a very strong linear attention variant. We first simplify Mamba-2 down to its most fundamental and important components, evaluating which specific choices make it most accurate. From this simplified Mamba variant (Mamba-2S), we improve the A-mask and increase the order of the hidden state, resulting in a method, which we call 2Mamba, that is nearly as accurate as softmax attention, yet much more memory efficient for long context lengths. We also investigate elements to Mamba-2 that help surpass softmax attention accuracy. Code is provided for all our experiments

Classifying the topology of closed curves is a central problem in low dimensional topology with applications beyond mathematics spanning protein folding, polymer physics and even magnetohydrodynamics. The central problem is how to determine whether two embeddings of a closed arc are equivalent under ambient isotopy. Given the striking ability of neural networks to solve complex classification tasks, it is therefore natural to ask if the knot classification problem can be tackled using Machine Learning (ML). In this paper, we investigate generic shortcut methods employed by ML to solve the knot classification challenge and specifically discover hidden non-topological features in training data generated through Molecular Dynamics simulations of polygonal knots that are used by ML to arrive to positive classifications results. We then provide a rigorous foundation for future attempts to tackle the knot classification challenge using ML by developing a publicly-available (i) dataset, that aims to remove the potential of non-topological feature classification and (ii) code, that can generate knot embeddings that faithfully explore chosen geometric state space with fixed knot topology. We expect that our work will accelerate the development of ML models that can solve complex geometric knot classification challenges.

Preordering is a generalization of clustering and partial ordering with applications in bioinformatics and social network analysis. Given a finite set $V$ and a value $c_{ab} \in \mathbb{R}$ for every ordered pair $ab$ of elements of $V$, the preordering problem asks for a preorder $\lesssim$ on $V$ that maximizes the sum of the values of those pairs $ab$ for which $a \lesssim b$. Building on the state of the art in solving this NP-hard problem partially, we contribute new partial optimality conditions and efficient algorithms for deciding these conditions. In experiments with real and synthetic data, these new conditions increase, in particular, the fraction of pairs $ab$ for which it is decided efficiently that $a \not\lesssim b$ in an optimal preorder.

Embodied AI systems (e.g., autonomous vehicles, service robots, and LLM-driven interactive agents) are rapidly transitioning from controlled environments to safety critical real-world deployments. Unlike disembodied AI, failures in embodied intelligence lead to irreversible physical consequences, raising fundamental questions about security, safety, and reliability. While existing research predominantly analyzes embodied AI through the lenses of Large Language Model (LLM) vulnerabilities or classical Cyber-Physical System (CPS) failures, this survey argues that these perspectives are individually insufficient to explain many observed breakdowns in modern embodied systems. We posit that a significant class of failures arises from embodiment-induced system-level mismatches, rather than from isolated model flaws or traditional CPS attacks. Specifically, we identify four core insights that explain why embodied AI is fundamentally harder to secure: (i) semantic correctness does not imply physical safety, as language-level reasoning abstracts away geometry, dynamics, and contact constraints; (ii) identical actions can lead to drastically different outcomes across physical states due to nonlinear dynamics and state uncertainty; (iii) small errors propagate and amplify across tightly coupled perception-decision-action loops; and (iv) safety is not compositional across time or system layers, enabling locally safe decisions to accumulate into globally unsafe behavior. These insights suggest that securing embodied AI requires moving beyond component-level defenses toward system-level reasoning about physical risk, uncertainty, and failure propagation.

Graph Out-of-Distribution (OOD) detection aims to identify whether a test graph deviates from the distribution of graphs observed during training, which is critical for ensuring the reliability of Graph Neural Networks (GNNs) when deployed in open-world scenarios. Recent advances in graph OOD detection have focused on test-time training techniques that facilitate OOD detection without accessing potential supervisory information (e.g., training data). However, most of these methods employ a one-pass inference paradigm, which prevents them from progressively correcting erroneous predictions to amplify OOD signals. To this end, we propose a \textbf{S}elf-\textbf{I}mproving \textbf{G}raph \textbf{O}ut-\textbf{o}f-\textbf{D}istribution detector (SIGOOD), which is an unsupervised framework that integrates continuous self-learning with test-time training for effective graph OOD detection. Specifically, SIGOOD generates a prompt to construct a prompt-enhanced graph that amplifies potential OOD signals. To optimize prompts, SIGOOD introduces an Energy Preference Optimization (EPO) loss, which leverages energy variations between the original test graph and the prompt-enhanced graph. By iteratively optimizing the prompt by involving it into the detection model in a self-improving loop, the resulting optimal prompt-enhanced graph is ultimately used for OOD detection. Comprehensive evaluations on 21 real-world datasets confirm the effectiveness and outperformance of our SIGOOD method. The code is at https://github.com/Ee1s/SIGOOD.

Parquet is the de facto columnar file format in modern analytical systems, yet its configuration guidelines have largely been shaped by CPU-centric execution models. As GPU-accelerated data processing becomes increasingly prevalent, Parquet files generated with CPU-oriented defaults can severely underutilize GPU parallelism, turning GPU scans into a performance bottleneck. In this work, we systematically study how Parquet configurations affect GPU scan performance. We show that Parquet's poor GPU performance is not inherent to the format itself but rather a consequence of suboptimal configuration choices. By applying GPU-aware configurations, we increase effective read bandwidth up to 125 GB/s without modifying the Parquet specification.

Comparative analysis of adaptive immune repertoires at population scale is hampered by two practical bottlenecks: the near-quadratic cost of pairwise affinity evaluations and dataset imbalances that obscure clinically important minority clonotypes. We introduce SubQuad, an end-to-end pipeline that addresses these challenges by combining antigen-aware, near-subquadratic retrieval with GPU-accelerated affinity kernels, learned multimodal fusion, and fairness-constrained clustering. The system employs compact MinHash prefiltering to sharply reduce candidate comparisons, a differentiable gating module that adaptively weights complementary alignment and embedding channels on a per-pair basis, and an automated calibration routine that enforces proportional representation of rare antigen-specific subgroups. On large viral and tumor repertoires SubQuad achieves measured gains in throughput and peak memory usage while preserving or improving recall@k, cluster purity, and subgroup equity. By co-designing indexing, similarity fusion, and equity-aware objectives, SubQuad offers a scalable, bias-aware platform for repertoire mining and downstream translational tasks such as vaccine target prioritization and biomarker discovery.

We introduce WebFAQ 2.0, a new version of the WebFAQ dataset, containing 198 million FAQ-based natural question-answer pairs across 108 languages. Compared to the previous version, it significantly expands multilingual coverage and the number of bilingual aligned QA pairs to over 14.3M, making it the largest FAQ-based resource. Unlike the original release, WebFAQ 2.0 uses a novel data collection strategy that directly crawls and extracts relevant web content, resulting in a substantially more diverse and multilingual dataset with richer context through page titles and descriptions. In response to community feedback, we also release a hard negatives dataset for training dense retrievers, with 1.25M queries across 20 languages. These hard negatives were mined using a two-stage retrieval pipeline and include cross-encoder scores for 200 negatives per query. We further show how this resource enables two primary fine-tuning strategies for dense retrievers: Contrastive Learning with MultipleNegativesRanking loss, and Knowledge Distillation with MarginMSE loss. WebFAQ 2.0 is not a static resource but part of a long-term effort. Since late 2025, structured FAQs are being regularly released through the Open Web Index, enabling continuous expansion and refinement. We publish the datasets and training scripts to facilitate further research in multilingual and cross-lingual IR. The dataset itself and all related resources are publicly available on GitHub and HuggingFace.

Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, yet early risk detection is often limited by available diagnostics. Carotid ultrasound, a non-invasive and widely accessible modality, encodes rich structural and hemodynamic information that is largely untapped. Here, we present a machine learning (ML) framework that extracts clinically meaningful representations of vascular damage (VD) from carotid ultrasound videos, using hypertension as a weak proxy label. The model learns robust features that are biologically plausible, interpretable, and strongly associated with established cardiovascular risk factors, comorbidities, and laboratory measures. High VD stratifies individuals for myocardial infarction, cardiac death, and all-cause mortality, matching or outperforming conventional risk models such as SCORE2. Explainable AI analyses reveal that the model relies on vessel morphology and perivascular tissue characteristics, uncovering novel functional and anatomical signatures of vascular damage. This work demonstrates that routine carotid ultrasound contains far more prognostic information than previously recognized. Our approach provides a scalable, non-invasive, and cost-effective tool for population-wide cardiovascular risk assessment, enabling earlier and more personalized prevention strategies without reliance on laboratory tests or complex clinical inputs.

Quantum computing has gained significant attention due to its potential to solve computational problems beyond the capabilities of classical computers. With major corporations and academic institutions investing in quantum hardware and software, there has been a rise in the development of quantum-enabled systems, particularly within open-source communities. However, despite the promising nature of quantum technologies, these communities face critical socio-technical challenges, including the emergence of socio-technical anti-patterns known as community smells. These anti-patterns, prevalent in open-source environments, have the potential to negatively impact both product quality and community health by introducing technical debt and amplifying architectural and code smells. Despite the importance of these socio-technical factors, there remains a scarcity of research investigating their influence within quantum open-source communities. This work aims to address this gap by providing a first step in analyzing the socio-technical well-being of quantum communities through a cross-sectional study. By understanding the socio-technical dynamics at play, it is expected that foundational knowledge can be established to mitigate the risks associated with community smells and ensure the long-term sustainability of open-source quantum initiatives.

Optimizing resource utilization in high-performance computing (HPC) clusters is essential for maximizing both system efficiency and user satisfaction. However, traditional rigid job scheduling often results in underutilized resources and increased job waiting times. This work evaluates the benefits of resource elasticity, where the job scheduler dynamically adjusts the resource allocation of malleable jobs at runtime. Using real workload traces from the Cori, Eagle, and Theta supercomputers, we simulate varying proportions (0-100%) of malleable jobs with the ElastiSim software. We evaluate five job scheduling strategies, including a novel one that maintains malleable jobs at their preferred resource allocation when possible. Results show that, compared to fully rigid workloads, malleable jobs yield significant improvements across all key metrics. Considering the best-performing scheduling strategy for each supercomputer, job turnaround times decrease by 37-67%, job makespan by 16-65%, job wait times by 73-99%, and node utilization improves by 5-52%. Although improvements vary, gains remain substantial even at 20% malleable jobs. This work highlights important correlations between workload characteristics (e.g., job runtimes and node requirements), malleability proportions, and scheduling strategies. These findings confirm the potential of malleability to address inefficiencies in current HPC practices and demonstrate that even limited adoption can provide substantial advantages, encouraging its integration into HPC resource management.

The rapid advancement of Large Language Models (LLMs) has established standardized evaluation benchmarks as the primary instrument for model comparison. Yet, their reliability is increasingly questioned due to sensitivity to shallow variations in input prompts. This paper examines how controlled, truth-conditionally equivalent lexical and syntactic perturbations affect the absolute performance and relative ranking of 23 contemporary LLMs across three benchmarks: MMLU, SQuAD, and AMEGA. We employ two linguistically principled pipelines to generate meaning-preserving variations: one performing synonym substitution for lexical changes, and another using dependency parsing to determine applicable syntactic transformations. Results show that lexical perturbations consistently induce substantial, statistically significant performance degradation across nearly all models and tasks, while syntactic perturbations have more heterogeneous effects, occasionally improving results. Both perturbation types destabilize model leaderboards on complex tasks. Furthermore, model robustness did not consistently scale with model size, revealing strong task dependence. Overall, the findings suggest that LLMs rely more on surface-level lexical patterns than on abstract linguistic competence, underscoring the need for robustness testing as a standard component of LLM evaluation.

We introduce Flickering Multi-Armed Bandits (FMAB), a new MAB framework where the set of available arms (or actions) can change at each round, and the available set at any time may depend on the agent's previously selected arm. We model this constrained, evolving availability using random graph processes, where arms are nodes and the agent's movement is restricted to its local neighborhood. We analyze this problem under two random graph models: an i.i.d. Erdős--Rényi (ER) process and an Edge-Markovian process. We propose and analyze a two-phase algorithm that employs a lazy random walk for exploration to efficiently identify the optimal arm, followed by a navigation and commitment phase for exploitation. We establish high-probability and expected sublinear regret bounds for both graph settings. We show that the exploration cost of our algorithm is near-optimal by establishing a matching information-theoretic lower bound for this problem class, highlighting the fundamental cost of exploration under local-move constraints. We complement our theoretical guarantees with numerical simulations, including a scenario of a robotic ground vehicle scouting a disaster-affected region.

Open datasets play a crucial role in three research domains that intersect data science and education: learning analytics, educational data mining, and artificial intelligence in education. Researchers in these domains apply computational methods to analyze data from educational contexts, aiming to better understand and improve teaching and learning. Providing open datasets alongside research papers supports reproducibility, collaboration, and trust in research findings. It also provides individual benefits for authors, such as greater visibility, credibility, and citation potential. Despite these advantages, the availability of open datasets and the associated practices within the learning analytics research communities, especially at their flagship conference venues, remain unclear. We surveyed available datasets published alongside research papers in learning analytics. We manually examined 1,125 papers from three flagship conferences (LAK, EDM, and AIED) over the past five years. We discovered, categorized, and analyzed 172 datasets used in 204 publications. Our study presents the most comprehensive collection and analysis of open educational datasets to date, along with the most detailed categorization. Of the 172 datasets identified, 143 were not captured in any prior survey of open data in learning analytics. We provide insights into the datasets' context, analytical methods, use, and other properties. Based on this survey, we summarize the current gaps in the field. Furthermore, we list practical recommendations, advice, and 8-item guidelines under the acronym PRACTICE with a checklist to help researchers publish their data. Lastly, we share our original dataset: an annotated inventory detailing the discovered datasets and the corresponding publications. We hope these findings will support further adoption of open data practices in learning analytics communities and beyond.

Offline safe reinforcement learning (RL) is increasingly important for cyber-physical systems (CPS), where safety violations during training are unacceptable and only pre-collected data are available. Existing offline safe RL methods typically balance reward-safety tradeoffs through constraint relaxation or joint optimization, but they often lack structural mechanisms to prevent safety drift. We propose LexiSafe, a lexicographic offline RL framework designed to preserve safety-aligned behavior. We first develop LexiSafe-SC, a single-cost formulation for standard offline safe RL, and derive safety-violation and performance-suboptimality bounds that together yield sample-complexity guarantees. We then extend the framework to hierarchical safety requirements with LexiSafe-MC, which supports multiple safety costs and admits its own sample-complexity analysis. Empirically, LexiSafe demonstrates reduced safety violations and improved task performance compared to constrained offline baselines. By unifying lexicographic prioritization with structural bias, LexiSafe offers a practical and theoretically grounded approach for safety-critical CPS decision-making.

Large language models (LLMs) are increasingly used for diagnostic tasks in medicine. In clinical practice, the correct diagnosis can rarely be immediately inferred from the initial patient presentation alone. Rather, reaching a diagnosis often involves systematic history taking, during which clinicians reason over multiple potential conditions through iterative questioning to resolve uncertainty. This process requires considering differential diagnoses and actively excluding emergencies that demand immediate intervention. Yet, the ability of medical LLMs to generate informative follow-up questions and thus reason over differential diagnoses remains underexplored. Here, we introduce MedClarify, an AI agent for information-seeking that can generate follow-up questions for iterative reasoning to support diagnostic decision-making. Specifically, MedClarify computes a list of candidate diagnoses analogous to a differential diagnosis, and then proactively generates follow-up questions aimed at reducing diagnostic uncertainty. By selecting the question with the highest expected information gain, MedClarify enables targeted, uncertainty-aware reasoning to improve diagnostic performance. In our experiments, we first demonstrate the limitations of current LLMs in medical reasoning, which often yield multiple, similarly likely diagnoses, especially when patient cases are incomplete or relevant information for diagnosis is missing. We then show that our information-theoretic reasoning approach can generate effective follow-up questioning and thereby reduces diagnostic errors by ~27 percentage points (p.p.) compared to a standard single-shot LLM baseline. Altogether, MedClarify offers a path to improve medical LLMs through agentic information-seeking and to thus promote effective dialogues with medical LLMs that reflect the iterative and uncertain nature of real-world clinical reasoning.

While frontier large language models demonstrate strong reasoning and mathematical capabilities, the practical process of training domain-specialized scientific language models from raw sources remains under-documented. In this work, we present a detailed case study of training a 1.36B-parameter scientific language model directly from raw arXiv LaTeX sources spanning mathematics, computer science, and theoretical physics. We describe an end-to-end pipeline covering metadata filtering, archive validation, LaTeX extraction, text normalization, domain-aware tokenization, and dense transformer training under constrained compute (2xA100 GPUs). Through 24 experimental runs, we analyze training stability, scaling behavior, data yield losses, and infrastructure bottlenecks. Our findings highlight how preprocessing decisions significantly affect usable token volume, how tokenization impacts symbolic stability, and how storage and I/O constraints can rival compute as limiting factors. We further analyze convergence dynamics and show stable training behavior in a data-rich regime (52B pretraining tokens). Rather than proposing a novel architecture, this work provides an engineering-grounded, transparent account of training a small scientific language model from scratch. We hope these insights support researchers operating under moderate compute budgets who seek to build domain-specialized models.

Modern neural translation models based on the Transformer architecture are known for their high performance, particularly when trained on high-resource datasets. A standard next-token prediction training strategy, while widely adopted in practice, may lead to overlooked artifacts such as representation collapse. Previous works have shown that this problem is especially pronounced in the representation of the deeper Transformer layers, where it often fails to efficiently utilize the geometric space. Representation collapse is even more evident in end-to-end training of continuous-output neural machine translation, where the trivial solution would be to set all vectors to the same value. In this work, we analyze the dynamics of representation collapse at different levels of discrete and continuous NMT transformers throughout training. We incorporate an existing regularization method based on angular dispersion and demonstrate empirically that it not only mitigates collapse but also improves translation quality. Furthermore, we show that quantized models exhibit similar collapse behavior and that the benefits of regularization are preserved even after quantization.

We consider the privacy amplification properties of a sampling scheme in which a user's data is used in $k$ steps chosen randomly and uniformly from a sequence (or set) of $t$ steps. This sampling scheme has been recently applied in the context of differentially private optimization (Chua et al., 2024a; Choquette-Choo et al., 2025) and communication-efficient high-dimensional private aggregation (Asi et al., 2025), where it was shown to have utility advantages over the standard Poisson sampling. Theoretical analyses of this sampling scheme (Feldman & Shenfeld, 2025; Dong et al., 2025) lead to bounds that are close to those of Poisson sampling, yet still have two significant shortcomings. First, in many practical settings, the resulting privacy parameters are not tight due to the approximation steps in the analysis. Second, the computed parameters are either the hockey stick or Renyi divergence, both of which introduce overheads when used in privacy loss accounting. In this work, we demonstrate that the privacy loss distribution (PLD) of random allocation applied to any differentially private algorithm can be computed efficiently. When applied to the Gaussian mechanism, our results demonstrate that the privacy-utility trade-off for random allocation is at least as good as that of Poisson subsampling. In particular, random allocation is better suited for training via DP-SGD. To support these computations, our work develops new tools for general privacy loss accounting based on a notion of PLD realization. This notion allows us to extend accurate privacy loss accounting to subsampling which previously required manual noise-mechanism-specific analysis.

While large language models (LLMs) have become pivotal to content safety, current evaluation paradigms primarily focus on detecting explicit harms (e.g., violence or hate speech), neglecting the subtler value dimensions conveyed in digital content. To bridge this gap, we introduce X-Value, a novel Cross-lingual Values Assessment Benchmark designed to evaluate LLMs' ability to assess deep-level values of content from a global perspective. X-Value consists of more than 5,000 QA pairs across 18 languages, systematically organized into 7 core domains grounded in Schwartz's Theory of Basic Human Values and categorized into easy and hard levels for discriminative evaluation. We further propose a unique two-stage annotation framework that first identifies whether an issue falls under global consensus (e.g., human rights) or pluralism (e.g., religion), and subsequently conducts a multi-party evaluation of the latent values embedded within the content. Systematic evaluations on X-Value reveal that current SOTA LLMs exhibit deficiencies in cross-lingual values assessment ($Acc < 77\%$), with significant performance disparities across different languages ($ΔAcc > 20\%$). This work highlights the urgent need to improve the nuanced, values-aware content assessment capability of LLMs. Our X-Value is available at: https://huggingface.co/datasets/Whitolf/X-Value.

Processing sensory data close to the data source, often involving Edge devices, promises low latency for pervasive applications, like smart cities. This commonly involves a multitude of processing services, executed with limited resources; this setup faces three problems: first, the application demand and the resource availability fluctuate, so the service execution must scale dynamically to sustain processing requirements (e.g., latency); second, each service permits different actions to adjust its operation, so they require individual scaling policies; third, without a higher-level mediator, services would cannibalize any resources of services co-located on the same device. This demo first presents a platform for context-aware autoscaling of stream processing services that allows developers to monitor and adjust the service execution across multiple service-specific parameters. We then connect a scaling agent to these interfaces that gradually builds an understanding of the processing environment by exploring each service's action space; the agent then optimizes the service execution according to this knowledge. Participants can revisit the demo contents as video summary and introductory poster, or build a custom agent by extending the artifact repository.

Quantum generative modeling, where the Born rule naturally defines probability distributions through measurement of parameterized quantum states, is a promising near-term application of quantum computing. We propose a Quantum Scrambling Born Machine in which a fixed entangling unitary -- acting as a scrambling reservoir -- provides multi-qubit entanglement, while only single-qubit rotations are optimized. We consider three entangling unitaries -- a Haar random unitary and two physically realizable approximations, a finite-depth brickwork random circuit and analog time evolution under nearest-neighbor spin-chain Hamiltonians -- and show that, for the benchmark distributions and system sizes considered, once the entangler produces near-Haar-typical entanglement the model learns the target distribution with weak sensitivity to the scrambler's microscopic origin. Finally, promoting the Hamiltonian couplings to trainable parameters casts the generative task as a variational Hamiltonian problem, with performance competitive with representative classical generative models at matched parameter count.

Reinforcement learning (RL) is a subfield of machine learning that focuses on developing models that can autonomously learn optimal decision-making strategies over time. In a recent pioneering paper, Wagner demonstrated how the Deep Cross-Entropy RL method can be applied to tackle various problems from extremal graph theory by reformulating them as combinatorial optimization problems. Subsequently, many researchers became interested in refining and extending the framework introduced by Wagner, thereby creating various RL environments specialized for graph theory. Moreover, a number of problems from extremal graph theory were solved through the use of RL. In particular, several inequalities concerning the Laplacian spectral radius of graphs were refuted, new lower bounds were obtained for certain Ramsey numbers, and contributions were made to the Turán-type extremal problem in which the forbidden structures are cycles of length three and four. Here, we present Reinforcement Learning for Graph Theory (RLGT), a novel RL framework that systematizes the previous work and provides support for both undirected and directed graphs, with or without loops, and with an arbitrary number of edge colors. The framework efficiently represents graphs and aims to facilitate future RL-based research in extremal graph theory through optimized computational performance and a clean and modular design.

Semantic communication aims to convey meaning for effective task execution, but differing latent representations in AI-native devices can cause semantic mismatches that hinder mutual understanding. This paper introduces a novel approach to mitigating latent space misalignment in multi-agent AI- native semantic communications. In a downlink scenario, we consider an access point (AP) communicating with multiple users to accomplish a specific AI-driven task. Our method implements a protocol that shares a semantic pre-equalizer at the AP and local semantic equalizers at user devices, fostering mutual understanding and task-oriented communication while considering power and complexity constraints. To achieve this, we employ a federated optimization for the decentralized training of the semantic equalizers at the AP and user sides. Numerical results validate the proposed approach in goal-oriented semantic communication, revealing key trade-offs among accuracy, com- munication overhead, complexity, and the semantic proximity of AI-native communication devices.

We present Unified Latents (UL), a framework for learning latent representations that are jointly regularized by a diffusion prior and decoded by a diffusion model. By linking the encoder's output noise to the prior's minimum noise level, we obtain a simple training objective that provides a tight upper bound on the latent bitrate. On ImageNet-512, our approach achieves competitive FID of 1.4, with high reconstruction quality (PSNR) while requiring fewer training FLOPs than models trained on Stable Diffusion latents. On Kinetics-600, we set a new state-of-the-art FVD of 1.3.

Controlling the longitudinal laser pulse shape in photoinjectors of Free-Electron Lasers is a powerful lever for optimizing electron beam quality, but systematic exploration of the vast design space is limited by the cost of brute-force pulse propagation simulations. We present a generative modeling framework based on Wasserstein Autoencoders to learn a differentiable latent interface between pulse shaping and downstream beam dynamics. Our empirical findings show that the learned latent space is continuous and interpretable while maintaining high-fidelity reconstructions. Pulse families such as higher-order Gaussians trace coherent trajectories, while standardizing the temporal pulse lengths shows a latent organization correlated with pulse energy. Analysis via principal components and Gaussian Mixture Models reveals a well behaved latent geometry, enabling smooth transitions between distinct pulse types via linear interpolation. The model generalizes from simulated data to real experimental pulse measurements, accurately reconstructing pulses and embedding them consistently into the learned manifold. Overall, the approach reduces reliance on expensive pulse-propagation simulations and facilitates downstream beam dynamics simulation and analysis.

Human self-report questionnaires are increasingly used in NLP to benchmark and audit large language models (LLMs), from persona consistency to safety and bias assessments. Yet these instruments presume honest responding; in evaluative contexts, LLMs can instead gravitate toward socially preferred answers-a form of socially desirable responding (SDR)-biasing questionnaire-derived scores and downstream conclusions. We propose a psychometric framework to quantify and mitigate SDR in questionnaire-based evaluation of LLMs. To quantify SDR, the same inventory is administered under HONEST versus FAKE-GOOD instructions, and SDR is computed as a direction-corrected standardized effect size from item response theory (IRT)-estimated latent scores. This enables comparisons across constructs and response formats, as well as against human instructed-faking benchmarks. For mitigation, we construct a graded forced-choice (GFC) Big Five inventory by selecting 30 cross-domain pairs from an item pool via constrained optimization to match desirability. Across nine instruction-tuned LLMs evaluated on synthetic personas with known target profiles, Likert-style questionnaires show consistently large SDR, whereas desirability-matched GFC substantially attenuates SDR while largely preserving the recovery of the intended persona profiles. These results highlight a model-dependent SDR-recovery trade-off and motivate SDR-aware reporting practices for questionnaire-based benchmarking and auditing of LLMs.

AllReduce is a fundamental collective operation in distributed computing and a key performance bottleneck for large-scale training and inference. Its completion time is determined by the number of communication steps, which dominates latency-sensitive workloads, and the communication distance affecting both latency- and bandwidth-bound regimes. Direct-connect topologies, such as torus networks used in Google's TPUv4, are particularly prone to large communication distances due to limited bisection bandwidth. Latency-optimal algorithms such as Bruck's complete AllReduce in $\log_3 n$ steps on a bidirectional ring, but incur large communication distances that result in substantial congestion. In contrast, recent approaches such as Swing reduce communication distance and congestion, but are inherently required to perform $\log_2 n$ steps to complete AllReduce, sacrificing latency-optimality. In this paper, we present Trivance, a novel AllReduce algorithm that completes within $\log_3 n$ steps, while reducing congestion compared to Bruck's algorithm by a factor of three and preserving bandwidth-optimality. Trivance exploits both transmission ports of a bidirectional ring within each step to triple the communication distance along both directions simultaneously. Furthermore, by performing joint reductions, Trivance improves both the number of steps and network congestion. We further show that Trivance extends naturally to multidimensional torus networks, retaining its latency advantage while achieving performance comparable to bandwidth-optimal algorithms for large messages. Our empirical evaluation shows that Trivance improves state-of-the-art approaches by 5-30% for message sizes up to 8\,MiB, in high-bandwidth settings up to 32MiB and for 3D tori up to 128MiB. Throughout the evaluation, Trivance remains the best-performing latency-optimal algorithm.

Electroencephalography (EEG) foundation models (EFMs) have achieved strong performance under full fine-tuning but exhibit poor generalization when subject-level supervision is limited, a common constraint in real-world clinical settings. We show that this failure stems not merely from limited supervision, but from a structural mismatch between noisy, limited supervision and the highly plastic parameter space of EFMs. To address this challenge, we propose SCOPE, a Structured COnfidence-aware Prototype-guided adaptation framework for EFM fine-tuning. SCOPE follows a two-stage pipeline. In the first stage, we construct reliable external supervision by learning geometry-regularized task priors, constructing balanced class-level prototypes over the resulting embeddings, and producing confidence-aware pseudo-labels from their agreement to filter unreliable signals on unlabeled data. In the second stage, we introduce ProAdapter, which adapts frozen EEG foundation models via a lightweight adapter conditioned on the structured prototypes. Experiments across three EEG tasks and five foundation model backbones demonstrate that SCOPE consistently achieves strong performance and efficiency under label-limited cross-subject settings.

The Web is evolving from a medium that humans browse to an environment where software agents act on behalf of users. Advances in large language models (LLMs) make natural language a practical interface for goal-directed tasks, yet most current web agents operate on low-level primitives such as clicks and keystrokes. These operations are brittle, inefficient, and difficult to verify. Complementing content-oriented efforts such as NLWeb's semantic layer for retrieval, we argue that the agentic web also requires a semantic layer for web actions. We propose \textbf{Web Verbs}, a web-scale set of typed, semantically documented functions that expose site capabilities through a uniform interface, whether implemented through APIs or robust client-side workflows. These verbs serve as stable and composable units that agents can discover, select, and synthesize into concise programs. This abstraction unifies API-based and browser-based paradigms, enabling LLMs to synthesize reliable and auditable workflows with explicit control and data flow. Verbs can carry preconditions, postconditions, policy tags, and logging support, which improves \textbf{reliability} by providing stable interfaces, \textbf{efficiency} by reducing dozens of steps into a few function calls, and \textbf{verifiability} through typed contracts and checkable traces. We present our vision, a proof-of-concept implementation, and representative case studies that demonstrate concise and robust execution compared to existing agents. Finally, we outline a roadmap for standardization to make verbs deployable and trustworthy at web scale.

Counterfactual explanations (CFs) provide human-interpretable insights into model's predictions by identifying minimal changes to input features that would alter the model's output. However, existing methods struggle to generate multiple high-quality explanations that (1) affect only a small portion of the features, (2) can be applied to tabular data with heterogeneous features, and (3) are consistent with the user-defined constraints. We propose CounterFlowNet, a generative approach that formulates CF generation as sequential feature modification using conditional Generative Flow Networks (GFlowNet). CounterFlowNet is trained to sample CFs proportionally to a user-specified reward function that can encode key CF desiderata: validity, sparsity, proximity and plausibility, encouraging high-quality explanations. The sequential formulation yields highly sparse edits, while a unified action space seamlessly supports continuous and categorical features. Moreover, actionability constraints, such as immutability and monotonicity of features, can be enforced at inference time via action masking, without retraining. Experiments on eight datasets under two evaluation protocols demonstrate that CounterFlowNet achieves superior trade-offs between validity, sparsity, plausibility, and diversity with full satisfaction of the given constraints.

We introduce \emph{TAPO-Structured Description Logic} (TAPO--DL), a formal extension of classical description logic designed to model \emph{information behavior} as a structured, dynamic process. TAPO--DL extends the standard T--Box/A--Box architecture with two additional layers: a \emph{Procedural Box} (P--Box), which supports concept-driven, imperative-style programs such as conditional and iterative actions, and an \emph{Oracle Box} (O--Box), which formalizes controlled interaction with external information sources. While the terminological and assertional components capture static conceptual and factual knowledge, the procedural and oracle-based components enable the explicit representation of information-generating actions and external validation. We provide a unified semantic framework for TAPO--DL based on a co-generative, sheaf-theoretic interpretation, in which local informational states are modeled as sections and informational stability corresponds to the existence of coherent global structures. Within this setting, informational truth is characterized as stability under repeated agentive interaction rather than correspondence to a fixed global state. By integrating description logic with procedural dynamics, oracle-based reasoning, and sheaf-theoretic semantics, TAPO--DL offers a principled formal framework for analyzing information behavior in contexts involving interaction, uncertainty, and contextuality.

We introduce a compositional approach to model-based test generation in Behavior-Driven Development (BDD). BDD is an agile methodology in which system behavior is specified through textual scenarios that, in our approach, are translated into transition systems used for model-based testing. This paper formally defines disjunction composition, to combine BDD transition systems that represent alternative system behaviors. Disjunction composition allows for modeling and testing the integrated behavior while ensuring that the testing power of the original set of scenarios is preserved. This is proved using a symbolic semantics for BDD transition systems, with the property that the symbolic equivalence of two BDD transition systems guarantees that they fail the same test cases. Also, we demonstrate the potential of disjunction composition by applying the composition in an industrial case study.

To evaluate whether LLMs can accurately predict future events, we need the ability to \textit{backtest} them on events that have already resolved. This requires models to reason only with information available at a specified past date. Yet LLMs may inadvertently leak post-cutoff knowledge encoded during training, undermining the validity of retrospective evaluation. We introduce a claim-level framework for detecting and quantifying this \emph{temporal knowledge leakage}. Our approach decomposes model rationales into atomic claims and categorizes them by temporal verifiability, then applies \textit{Shapley values} to measure each claim's contribution to the prediction. This yields the \textbf{Shapley}-weighted \textbf{D}ecision-\textbf{C}ritical \textbf{L}eakage \textbf{R}ate (\textbf{Shapley-DCLR}), an interpretable metric that captures what fraction of decision-driving reasoning derives from leaked information. Building on this framework, we propose \textbf{Time}-\textbf{S}upervised \textbf{P}rediction with \textbf{E}xtracted \textbf{C}laims (\textbf{TimeSPEC}), which interleaves generation with claim verification and regeneration to proactively filter temporal contamination -- producing predictions where every supporting claim can be traced to sources available before the cutoff date. Experiments on 350 instances spanning U.S. Supreme Court case prediction, NBA salary estimation, and stock return ranking reveal substantial leakage in standard prompting baselines. TimeSPEC reduces Shapley-DCLR while preserving task performance, demonstrating that explicit, interpretable claim-level verification outperforms prompt-based temporal constraints for reliable backtesting.

The black-box nature of Large Language Models necessitates novel evaluation frameworks that transcend surface-level performance metrics. This study investigates the internal neural representations of cognitive complexity using Bloom's Taxonomy as a hierarchical lens. By analyzing high-dimensional activation vectors from different LLMs, we probe whether different cognitive levels, ranging from basic recall (Remember) to abstract synthesis (Create), are linearly separable within the model's residual streams. Our results demonstrate that linear classifiers achieve approximately 95% mean accuracy across all Bloom levels, providing strong evidence that cognitive level is encoded in a linearly accessible subspace of the model's representations. These findings provide evidence that the model resolves the cognitive difficulty of a prompt early in the forward pass, with representations becoming increasingly separable across layers.

As large language models (LLMs) continue to grow in size, fewer users are able to host and run models locally. This has led to increased use of third-party hosting services. However, in this setting, there is a lack of guarantees on the computation performed by the inference provider. For example, a dishonest provider may replace an expensive large model with a cheaper-to-run weaker model and return the results from the weaker model to the user. Existing tools to verify inference typically rely on methods from cryptography such as zero-knowledge proofs (ZKPs), but these add significant computational overhead, and remain infeasible for use for large models. In this work, we develop a new insight -- that given a method for performing private LLM inference, one can obtain forms of verified inference at marginal extra cost. Specifically, we propose two new protocols which leverage privacy-preserving LLM inference in order to provide guarantees over the inference that was carried out. Our approaches are cheap, requiring the addition of a few extra tokens of computation, and have little to no downstream impact. As the fastest privacy-preserving inference methods are typically faster than ZK methods, the proposed protocols also improve verification runtime. Our work provides novel insights into the connections between privacy and verifiability in LLM inference.

Predicting human decision-making in high-stakes environments remains a central challenge for artificial intelligence. While large language models (LLMs) demonstrate strong general reasoning, they often struggle to generate consistent, individual-specific behavior, particularly when accurate prediction depends on complex interactions between psychological traits and situational constraints. Prompting-based approaches can be brittle in this setting, exhibiting identity drift and limited ability to leverage increasingly detailed persona descriptions. To address these limitations, we introduce the Large Behavioral Model (LBM), a behavioral foundation model fine-tuned to predict individual strategic choices with high fidelity. LBM shifts from transient persona prompting to behavioral embedding by conditioning on a structured, high-dimensional trait profile derived from a comprehensive psychometric battery. Trained on a proprietary dataset linking stable dispositions, motivational states, and situational constraints to observed choices, LBM learns to map rich psychological profiles to discrete actions across diverse strategic dilemmas. In a held-out scenario evaluation, LBM fine-tuning improves behavioral prediction relative to the unadapted Llama-3.1-8B-Instruct backbone and performs comparably to frontier baselines when conditioned on Big Five traits. Moreover, we find that while prompting-based baselines exhibit a complexity ceiling, LBM continues to benefit from increasingly dense trait profiles, with performance improving as additional trait dimensions are provided. Together, these results establish LBM as a scalable approach for high-fidelity behavioral simulation, enabling applications in strategic foresight, negotiation analysis, cognitive security, and decision support.

Generative AI is reshaping knowledge work, yet existing research focuses predominantly on software engineering and the natural sciences, with limited methodological exploration for the humanities and social sciences. Positioned as a "methodological experiment," this study proposes an AI Agent-based collaborative research workflow (Agentic Workflow) for humanities and social science research. Taiwan's Claude.ai usage data (N = 7,729 conversations, November 2025) from the Anthropic Economic Index (AEI) serves as the empirical vehicle for validating the feasibility of this methodology. This study operates on two levels: the primary level is the design and validation of a methodological framework - a seven-stage modular workflow grounded in three principles: task modularization, human-AI division of labor, and verifiability, with each stage delineating clear roles for human researchers (research judgment and ethical decisions) and AI Agents (information retrieval and text generation); the secondary level is the empirical analysis of AEI Taiwan data - serving as an operational demonstration of the workflow's application to secondary data research, showcasing both the process and output quality (see Appendix A). This study contributes by proposing a replicable AI collaboration framework for humanities and social science researchers, and identifying three operational modes of human-AI collaboration - direct execution, iterative refinement, and human-led - through reflexive documentation of the operational process. This taxonomy reveals the irreplaceability of human judgment in research question formulation, theoretical interpretation, contextualized reasoning, and ethical reflection. Limitations including single-platform data, cross-sectional design, and AI reliability risks are acknowledged.

Efficiently navigating complex environments requires agents to internalize the underlying logic of their world, yet standard world modelling methods often struggle with sample inefficiency, lack of transparency, and poor scalability. We propose a framework for constructing symbolic causal world models entirely online by integrating continuous model learning and repair into the agent's decision loop, by leveraging the power of Meta-Interpretive Learning and predicate invention to find semantically meaningful and reusable abstractions, allowing an agent to construct a hierarchy of disentangled, high-quality concepts from its observations. We demonstrate that our lifted inference approach scales to domains with complex relational dynamics, where propositional methods suffer from combinatorial explosion, while achieving sample-efficiency orders of magnitude higher than the established PPO neural-network-based baseline.

We present an AI-assisted framework for predicting individual runs of complex quantum experiments, including contextuality and causality (adaptive measurements), within our long-term programme of discovering a local hidden-variable theory that extends quantum theory. In order to circumvent impossibility theorems, we replace the assumption of free choice (measurement independence and parameter independence) with a weaker, compatibilistic version called contingent free choice. Our framework is based on interpreting complex quantum experiments as a Chess-like game between observers and the universe, which is seen as an economic agent minimizing action. The game structures corresponding to generic experiments such as fixed-causal-order process matrices or causal contextuality scenarios, together with a deterministic non-Nashian resolution algorithm that abandons unilateral deviation assumptions (free choice) and assumes Perfect Prediction instead, were described in previous work. In this new research, we learn the reward functions of the game, which contain a hidden variable, using neural networks. The cost function is the Kullback-Leibler divergence between the frequency histograms obtained through many deterministic runs of the game and the predictions of the extended Born rule. Using our framework on the specific case of the EPR 2-2-2 experiment acts as a proof-of-concept and a toy local-realist hidden-variable model that non-Nashian quantum theory is a promising avenue towards a local hidden-variable theory. Our framework constitutes a solid foundation, which can be further expanded in order to fully discover a complete quantum theory.

Generating samples from limited information is a fundamental problem across scientific domains. Classical maximum entropy methods provide principled uncertainty quantification from moment constraints but require sampling via MCMC or Langevin dynamics, which typically exhibit exponential slowdown in high dimensions. In contrast, generative models based on diffusion and flow matching efficiently transport noise to data but offer limited theoretical guarantees and can overfit when data is scarce. We introduce Moment Guided Diffusion (MGD), which combines elements of both approaches. Building on the stochastic interpolant framework, MGD samples maximum entropy distributions by solving a stochastic differential equation that guides moments toward prescribed values in finite time, thereby avoiding slow mixing in equilibrium-based methods. We formally obtain, in the large-volatility limit, convergence of MGD to the maximum entropy distribution and derive a tractable estimator of the resulting entropy computed directly from the dynamics. Applications to financial time series, turbulent flows, and cosmological fields using wavelet scattering moments yield estimates of negentropy for high-dimensional multiscale processes.

We present softdtw-cuda-torch, an open-source PyTorch library for computing Soft Dynamic Time Warping (SoftDTW) on GPUs. Our implementation addresses three key limitations of existing GPU implementations of SoftDTW: a hard sequence-length cap of 1024, numerical instability in the backward pass for small smoothing parameters, and excessive GPU memory consumption from materializing pairwise distance tensors. We introduce (1) tiled anti-diagonal kernel execution that removes the sequence-length constraint, (2) a log-space back-ward pass that prevents floating-point overflow, and (3) a fused distance-computation mode that eliminates the O(BN M ) intermediate distance tensor, achieving up to 98% memory reduction compared to prior work. The library supports arbitrary sequence lengths, full PyTorch autograd integration, and Soft-DTW Barycenter computation. Code is available at https://github.com/BGU-CS-VIL/sdtw-cuda-torch.

The detection limit of astronomical imaging observations is limited by several noise sources. Some of that noise is correlated between neighbouring image pixels and exposures, so in principle could be learned and corrected. We present an astronomical self-supervised transformer-based denoising algorithm (ASTERIS), that integrates spatiotemporal information across multiple exposures. Benchmarking on mock data indicates that ASTERIS improves detection limits by 1.0 magnitude at 90% completeness and purity, while preserving the point spread function and photometric accuracy. Observational validation using data from the James Webb Space Telescope (JWST) and Subaru telescope identifies previously undetectable features, including low-surface-brightness galaxy structures and gravitationally-lensed arcs. Applied to deep JWST images, ASTERIS identifies three times more redshift > 9 galaxy candidates, with rest-frame ultraviolet luminosity 1.0 magnitude fainter, than previous methods.

The threat of algorithmic collusion, and whether it merits regulatory intervention, remains debated, as existing evaluations of its emergence often rely on long learning horizons, assumptions about counterparty rationality in adopting collusive strategies, and symmetry in hyperparameters and economic settings among players. To study collusion risk, we introduce a meta-game design for analyzing algorithmic behavior under test-time constraints. We model agents as possessing pretrained policies with distinct strategic characteristics (e.g., competitive, naively cooperative, robustly collusive), and formulate the problem as selecting a meta-strategy that combines a pretrained, initial policy with an in-game adaptation rule. We seek to examine whether collusion can emerge under rational choices and how agents co-adapt toward cooperation or competition. To this end, we sample normal-form empirical games over meta-strategy profiles, % across random initial game states, compute relevant game statistics (e.g., payoffs against individuals and regret against an equilibrium mixture of opponents), and construct empirical best-response graphs to uncover strategic relationships. We evaluate both reinforcement-learning and LLM-based strategies in repeated pricing games under symmetric and asymmetric cost settings, and present findings on the feasibility of algorithmic collusion and the effectiveness of pricing strategies in practical ``test-time'' environments. The source code and the full paper with appendix are available at: https://github.com/chailab-rutgers/CollusionMetagame.

Text-based telemedicine has become a common mode of care, requiring clinicians to deliver medical advice clearly and effectively in writing. As platforms increasingly rely on patient ratings and feedback, clinicians face growing pressure to maintain satisfaction scores, even though these evaluations often reflect communication quality more than clinical accuracy. We analyse patient satisfaction signals in Romanian text-based telemedicine. Using a sample of 77,334 anonymised patient question--doctor response pairs, we model feedback as a binary outcome, treating thumbs-up responses as positive and grouping negative or absent feedback into the other class. We extract interpretable, predominantly language-agnostic features (e.g., length, structural characteristics, readability proxies), along with Romanian LIWC psycholinguistic features and politeness/hedging markers where available. We train a classifier with a time-based split and perform SHAP-based analyses, which indicate that patient and clinician history features dominate prediction, functioning as strong priors, while characteristics of the response text provide a smaller but, crucially, actionable signal. In subgroup correlation analyses, politeness and hedging are consistently positively associated with patient feedback, whereas lexical diversity shows a negative association.

Since its introduction in the early 90s, the web has become the largest application platform available globally. HyperText Markup Language (HTML) has been an essential part of the web since the beginning, as it allows defining webpages in a tree-like manner, including semantics and content. Although the web was never meant to be an application platform, it evolved as such, especially since the early 2000s, as web application frameworks became available. While the emergence of frameworks made it easier than ever to develop complex applications, it also put HTML on the back burner. As web standards caught up, especially with milestones such as HTML5, the gap between the web platform and frameworks was reduced. HTML First development emphasizes this shift and puts focus on literally using HTML first when possible, while encouraging minimalism familiar from the early days of the web. It seems HTML-oriented web development can provide clear benefits to developers, especially when it is combined with comple- mentary approaches, such as embracing hypermedia and moving a large part of application logic to the server side. In the context of the htmx project, it was observed that moving towards HTML can reduce the size of a codebase greatly while leading to maintenance and development benefits due to the increased conceptual simplicity. Holotype-based comparisons for content-oriented websites show performance benefits, and the same observation was confirmed by a small case study where the Yle website was converted to follow HTML First principles. In short, the HTML First approach seems to have clear advantages for web developers, while there are open questions related to the magnitude of the benefits and the alignment with the recent trend of AI-driven web development.

In this paper we present Texo, a minimalist yet highperformance formula recognition model that contains only 20 million parameters. By attentive design, distillation and transfer of the vocabulary and the tokenizer, Texo achieves comparable performance to state-of-the-art models such as UniMERNet-T and PPFormulaNet-S, while reducing the model size by 80% and 65%, respectively. This enables real-time inference on consumer-grade hardware and even in-browser deployment. We also developed a web application to demonstrate the model capabilities and facilitate its usage for end users.

The problem of domain generalization concerns learning predictive models that are robust to distribution shifts when deployed in new, previously unseen environments. Existing methods typically require labeled data from multiple training environments, limiting their applicability when labeled data are scarce. In this work, we study domain generalization in an anti-causal setting, where the outcome causes the observed covariates. Under this structure, environment perturbations that affect the covariates do not propagate to the outcome, which motivates regularizing the model's sensitivity to these perturbations. Crucially, estimating these perturbation directions does not require labels, enabling us to leverage unlabeled data from multiple environments. We propose two methods that penalize the model's sensitivity to variations in the mean and covariance of the covariates across environments, respectively, and prove that these methods have worst-case optimality guarantees under certain classes of environments. Finally, we demonstrate the empirical performance of our approach on a controlled physical system and a physiological signal dataset.

Large Language Model-powered conversational agents (CAs) are increasingly capable of projecting sophisticated personalities through language, but how these projections affect users is unclear. We thus examine how CA personalities expressed linguistically affect user decisions and perceptions in the context of charitable giving. In a crowdsourced study, 360 participants interacted with one of eight CAs, each projecting a personality composed of three linguistic aspects: attitude (optimistic/pessimistic), authority (authoritative/submissive), and reasoning (emotional/rational). While the CA's composite personality did not affect participants' decisions, it did affect their perceptions and emotional responses. Particularly, participants interacting with pessimistic CAs felt lower emotional state and lower affinity towards the cause, perceived the CA as less trustworthy and less competent, and yet tended to donate more toward the charity. Perceptions of trust, competence, and situational empathy significantly predicted donation decisions. Our findings emphasize the risks CAs pose as instruments of manipulation, subtly influencing user perceptions and decisions.

Large language models (LLMs) increasingly assist software engineering tasks that require reasoning over long code contexts, yet their robustness under varying input conditions remains unclear. We conduct a systematic study of long-context code question answering using controlled ablations that test sensitivity to answer format, distractors, and context scale. Extending LongCodeBench Python dataset with new COBOL and Java question-answer sets, we evaluate state-of-the-art models under three settings: (i) shuffled multiple-choice options, (ii) open-ended questions and (iii) needle-in-a-haystack contexts containing relevant and adversarially irrelevant information. Results show substantial performance drops in both shuffled multiple-choice options and open-ended questions, and brittle behavior in the presence of irrelevant cues. Our findings highlight limitations of current long-context evaluations and provide a broader benchmark for assessing code reasoning in both legacy and modern systems.

Crystal structure prediction (CSP), which aims to predict the three-dimensional atomic arrangement of a crystal from its composition, is central to materials discovery and mechanistic understanding. Existing deep learning models often treat crystallographic symmetry only as a soft heuristic or rely on space group and Wyckoff templates retrieved from known structures, which limits both physical fidelity and the ability to discover genuinely new material structures. In contrast to retrieval-based methods, our approach leverages large language models to encode chemical semantics and directly generate fine-grained Wyckoff patterns from composition, effectively circumventing the limitations inherent to database lookups. Crucially, we incorporate domain knowledge into the generative process through an efficient constrained-optimization search that rigorously enforces algebraic consistency between site multiplicities and atomic stoichiometry. By integrating this symmetry-consistent template into a diffusion backbone, our approach constrains the stochastic generative trajectory to a physically valid geometric manifold. This framework achieves state-of-the-art performance across stability, uniqueness, and novelty (SUN) benchmarks, alongside superior matching performance, thereby establishing a new paradigm for the rigorous exploration of targeted crystallographic space. This framework enables efficient expansion into previously uncharted materials space, eliminating reliance on existing databases or a priori structural knowledge.

Robust control of mechanical systems with multiple uncertainties remains a fundamental challenge, particularly when nonlinear dynamics and operating-condition variations are intricately intertwined. While deep reinforcement learning (DRL) combined with domain randomization has shown promise in mitigating the sim-to-real gap, simultaneously handling all sources of uncertainty often leads to sub-optimal policies and poor learning efficiency. This study formulates a new curriculum-based continual learning framework for robust control problems involving nonlinear dynamical systems in which multiple sources of uncertainty are simultaneously superimposed. The key idea is to decompose a complex control problem with multiple uncertainties into a sequence of continual learning tasks, in which strategies for handling each uncertainty are acquired sequentially. The original system is extended into a finite set of plants whose dynamic uncertainties are gradually expanded and diversified as learning progresses. The policy is stably updated across the entire plant sets associated with tasks defined by different uncertainty configurations without catastrophic forgetting. To ensure learning efficiency, we jointly incorporate a model-based controller (MBC), which guarantees a shared baseline performance across the plant sets, into the learning process to accelerate the convergence. This residual learning scheme facilitates task-specific optimization of the DRL agent for each uncertainty, thereby enhancing sample efficiency. As a practical industrial application, this study applies the proposed method to designing an active vibration controller for automotive powertrains. We verified that the resulting controller is robust against structural nonlinearities and dynamic variations, realizing successful sim-to-real transfer.

Recent work has demonstrated that transformers and linear attention models can perform in-context learning (ICL) on simple function classes, such as linear regression. In this paper, we empirically study how these two attention mechanisms differ in their ICL behavior on the canonical linear-regression task of Garg et al. We evaluate learning quality (MSE), convergence, and generalization behavior of each architecture. We also analyze how increasing model depth affects ICL performance. Our results illustrate both the similarities and limitations of linear attention relative to quadratic attention in this setting.

This paper presents SimulatorCoder, an agent powered by large language models (LLMs), designed to generate and optimize deep neural network (DNN) accelerator simulators based on natural language descriptions. By integrating domain-specific prompt engineering including In-Context Learning (ICL), Chain-of-Thought (CoT) reasoning, and a multi-round feedback-verification flow, SimulatorCoder systematically transforms high-level functional requirements into efficient, executable, and architecture-aligned simulator code. Experiments based on the customized SCALE-Sim benchmark demonstrate that structured prompting and feedback mechanisms substantially improve both code generation accuracy and simulator performance. The resulting simulators not only maintain cycle-level fidelity with less than 1% error compared to manually implemented counterparts, but also consistently achieve lower simulation runtimes, highlighting the effectiveness of LLM-based methods in accelerating simulator development. Our code is available at https://github.com/xiayuhuan/SimulatorCoder.

Genomic Foundation Models (GFMs) have largely relied on Masked Language Modeling (MLM) or Next Token Prediction (NTP) to learn the language of life. While these paradigms excel at capturing local genomic syntax and fine-grained motif patterns, they often fail to capture the broader functional context, resulting in representations that lack a global biological perspective. We introduce JEPA-DNA, a novel pre-training framework that integrates the Joint-Embedding Predictive Architecture (JEPA) with traditional generative objectives. JEPA-DNA introduces latent grounding by coupling token-level recovery with a predictive objective in the latent space by supervising a CLS token. This forces the model to predict the high-level functional embeddings of masked genomic segments rather than focusing solely on individual nucleotides. JEPA-DNA extends both NTP and MLM paradigms and can be deployed either as a standalone from-scratch objective or as a continual pre-training enhancement for existing GFMs. Our evaluations across a diverse suite of genomic benchmarks demonstrate that JEPA-DNA consistently yields superior performance in supervised and zero-shot tasks compared to generative-only baselines. By providing a more robust and biologically grounded representation, JEPA-DNA offers a scalable path toward foundation models that understand not only the genomic alphabet, but also the underlying functional logic of the sequence.

Zeroth-order (ZO) optimization provides a gradient-free alternative to first-order (FO) methods by estimating gradients via finite differences of function evaluations, and has recently emerged as a memory-efficient paradigm for fine-tuning large-scale models by avoiding backpropagation. However, ZO optimization has a fundamental tension between accuracy and query efficiency. In this work, we show that ZO optimization can be substantially improved by unifying two complementary principles: (i) a projection-based subspace view that reduces gradient estimation variance by exploiting the intrinsic low-rank structure of model updates, and (ii) Muon-style spectral optimization that applies gradient orthogonalization to extract informative spectral structure from noisy ZO gradients. These findings form a unified framework of subspace gradient orthogonalization, which we instantiate in a new method, ZO-Muon, admitting a natural interpretation as a low-rank Muon optimizer in the ZO setting. Extensive experiments on large language models (LLMs) and vision transformers (ViTs) demonstrate that ZO-Muon significantly accelerates convergence and achieves a win-win improvement in accuracy and query/runtime efficiency. Notably, compared to the popular MeZO baseline, ZO-Muon requires only 24.7% of the queries to reach the same SST-2 performance for LLM fine-tuning, and improves accuracy by 25.1% on ViT-B fine-tuning on CIFAR-100.

Recent time series modeling faces a sharp divide between numerical generation and semantic understanding, with research showing that generation models often rely on superficial pattern matching, while understanding-oriented models struggle with high-fidelity numerical output. Although unified multimodal models (UMMs) have bridged this gap in vision, their potential for time series remains untapped. We propose TimeOmni-VL, the first vision-centric framework that unifies time series understanding and generation through two key innovations: (1) Fidelity-preserving bidirectional mapping between time series and images (Bi-TSI), which advances Time Series-to-Image (TS2I) and Image-to-Time Series (I2TS) conversions to ensure near-lossless transformations. (2) Understanding-guided generation. We introduce TSUMM-Suite, a novel dataset consists of six understanding tasks rooted in time series analytics that are coupled with two generation tasks. With a calibrated Chain-of-Thought, TimeOmni-VL is the first to leverage time series understanding as an explicit control signal for high-fidelity generation. Experiments confirm that this unified approach significantly improves both semantic understanding and numerical precision, establishing a new frontier for multimodal time series modeling.

As the need for more accurate and powerful Convolutional Neural Networks (CNNs) increases, so too does the size, execution time, memory footprint, and power consumption. To overcome this, solutions such as pruning have been proposed with their own metrics and methodologies, or criteria, for how weights should be removed. These solutions do not share a common implementation and are difficult to implement and compare. In this work, we introduce Combine, a criterion- based pruning solution and demonstrate that it is fast and effective framework for iterative pruning, demonstrate that criterion have differing effects on different models, create a standard language for comparing criterion functions, and propose a few novel criterion functions. We show the capacity of these criterion functions and the framework on VGG inspired models, pruning up to 79\% of filters while retaining or improving accuracy, and reducing the computations needed by the network by up to 68\%.

Learning to Defer (L2D) enables a classifier to abstain from predictions and defer to an expert, and has recently been extended to multi-expert settings. In this work, we show that multi-expert L2D is fundamentally more challenging than the single-expert case. With multiple experts, the classifier's underfitting becomes inherent, which seriously degrades prediction performance, whereas in the single-expert setting it arises only under specific conditions. We theoretically reveal that this stems from an intrinsic expert identifiability issue: learning which expert to trust from a diverse pool, a problem absent in the single-expert case and renders existing underfitting remedies failed. To tackle this issue, we propose PiCCE (Pick the Confident and Correct Expert), a surrogate-based method that adaptively identifies a reliable expert based on empirical evidence. PiCCE effectively reduces multi-expert L2D to a single-expert-like learning problem, thereby resolving multi expert underfitting. We further prove its statistical consistency and ability to recover class probabilities and expert accuracies. Extensive experiments across diverse settings, including real-world expert scenarios, validate our theoretical results and demonstrate improved performance.

Vector Quantized Variational Autoencoders (VQ-VAEs) are fundamental to modern generative modeling, yet they often suffer from training instability and "codebook collapse" due to the inherent coupling of representation learning and discrete codebook optimization. In this paper, we propose VP-VAE (Vector Perturbation VAE), a novel paradigm that decouples representation learning from discretization by eliminating the need for an explicit codebook during training. Our key insight is that, from the neural network's viewpoint, performing quantization primarily manifests as injecting a structured perturbation in latent space. Accordingly, VP-VAE replaces the non-differentiable quantizer with distribution-consistent and scale-adaptive latent perturbations generated via Metropolis--Hastings sampling. This design enables stable training without a codebook while making the model robust to inference-time quantization error. Moreover, under the assumption of approximately uniform latent variables, we derive FSP (Finite Scalar Perturbation), a lightweight variant of VP-VAE that provides a unified theoretical explanation and a practical improvement for FSQ-style fixed quantizers. Extensive experiments on image and audio benchmarks demonstrate that VP-VAE and FSP improve reconstruction fidelity and achieve substantially more balanced token usage, while avoiding the instability inherent to coupled codebook training.

Throughout the history of software, evolution has occurred in cycles of rise and fall driven by competition, and open-source software (OSS) is no exception. This cycle is accelerating, particularly in rapidly evolving domains such as web development and deep learning. However, the impact of competitive relationships among OSS projects on their survival remains unclear, and there are risks of losing a competitive edge to rivals. To address this, this study proposes a new automated method called ``Mutual Impact Analysis of OSS (MIAO)'' to quantify these competitive relationships. The proposed method employs a structural vector autoregressive model and impulse response functions, normally used in macroeconomic analysis, to analyze the interactions among OSS projects. In an empirical analysis involving mining and analyzing 187 OSS project groups, MIAO identified projects that were forced to cease development owing to competitive influences with up to 81\% accuracy, and the resulting features supported predictive experiments that anticipate cessation one year ahead with up to 77\% accuracy. This suggests that MIAO could be a valuable tool for OSS project maintainers to understand the dynamics of OSS ecosystems and predict the rise and fall of OSS projects.

We propose a novel parallel algorithm for decomposing hard CircuitSAT instances. The technique employs specialized constraints to partition an original SAT instance into a family of weakened formulas. Our approach is implemented as a parameterized parallel algorithm, where adjusting the parameters allows efficient identification of high-quality decompositions, guided by hardness estimations computed in parallel. We demonstrate the algorithm's practical efficacy on challenging CircuitSAT instances, including those encoding Logical Equivalence Checking of Boolean circuits and preimage attacks on cryptographic hash functions.

As Large Language Models (LLMs) transition from standalone chat interfaces to foundational reasoning layers in multi-agent systems and recursive evaluation loops (LLM-as-a-judge), the detection of durable, provider-level behavioral signatures becomes a critical requirement for safety and governance. Traditional benchmarks measure transient task accuracy but fail to capture stable, latent response policies -- the ``prevailing mindsets'' embedded during training and alignment that outlive individual model versions. This paper introduces a novel auditing framework that utilizes psychometric measurement theory -- specifically latent trait estimation under ordinal uncertainty -- to quantify these tendencies without relying on ground-truth labels. Utilizing forced-choice ordinal vignettes masked by semantically orthogonal decoys and governed by cryptographic permutation-invariance, the research audits nine leading models across dimensions including Optimization Bias, Sycophancy, and Status-Quo Legitimization. Using Mixed Linear Models (MixedLM) and Intraclass Correlation Coefficient (ICC) analysis, the research identifies that while item-level framing drives high variance, a persistent ``lab signal'' accounts for significant behavioral clustering. These findings demonstrate that in ``locked-in'' provider ecosystems, latent biases are not merely static errors but compounding variables that risk creating recursive ideological echo chambers in multi-layered AI architectures.

3D scene reconstruction and rendering are core tasks in computer vision, with applications spanning industrial monitoring, robotics, and autonomous driving. Recent advances in 3D Gaussian Splatting (GS) and its variants have achieved impressive rendering fidelity while maintaining high computational and memory efficiency. However, conventional vision-based GS pipelines typically rely on a sufficient number of camera views to initialize the Gaussian primitives and train their parameters, typically incurring additional processing cost during initialization while falling short in conditions where visual cues are unreliable, such as adverse weather, low illumination, or partial occlusions. To cope with these challenges, and motivated by the robustness of radio-frequency (RF) signals to weather, lighting, and occlusions, we introduce a multimodal framework that integrates RF sensing, such as automotive radar, with GS-based rendering as a more efficient and robust alternative to vision-only GS rendering. The proposed approach enables efficient depth prediction from only sparse RF-based depth measurements, yielding a high-quality 3D point cloud for initializing Gaussian functions across diverse GS architectures. Numerical tests demonstrate the merits of judiciously incorporating RF sensing into GS pipelines, achieving high-fidelity 3D scene rendering driven by RF-informed structural accuracy.

Augmented reality (AR)-enabled Metaverse is a promising technique to provide immersive service experience for mobile users. However, the limited network resources and unpredictable wireless propagation environments are key design bottlenecks of AR-enabled Metaverse systems. Therefore, this paper presents a resource management framework for simultaneously transmitting and reflecting RIS (STAR-RIS)-assisted AR-enabled Metaverse, where the STAR-RIS is configured to improve the communication efficiency between AR users and the Metaverse server located at the base station (BS). Moreover, we formulate a service latency minimization problem via jointly optimizing the computation resource allocation of the BS, coefficient matrix of the STAR-RIS, central processing unit (CPU) frequency and transmit power of the AR users. To tackle the non-convex problem, we utilize an approximate method to transform it to a tractable form, and decouple the multi-dimensional variables via the alternating optimization method. Particularly, the optimal coefficient matrix is obtained by a penalty function-based method with proved convergence, the CPU frequencies of AR users are derived as the closed-form solution, and the transmit power of AR users and computation resource allocation of the BS are obtained by the Lagrange duality method and convex optimization theory. Finally, simulation results demonstrates that the proposed method achieves remarkable latency reduction than several benchmark methods.

Nonstationary time series forecasting suffers from the distribution shift issue due to the different distributions that produce the training and test data. Existing methods attempt to alleviate the dependence by, e.g., removing low-order moments from each individual sample. These solutions fail to capture the underlying time-evolving structure across samples and do not model the complex time structure. In this paper, we aim to address the distribution shift in the frequency space by considering all possible time structures. To this end, we propose a Time-Invariant Frequency Operator (TIFO), which learns stationarity-aware weights over the frequency spectrum across the entire dataset. The weight representation highlights stationary frequency components while suppressing non-stationary ones, thereby mitigating the distribution shift issue in time series. To justify our method, we show that the Fourier transform of time series data implicitly induces eigen-decomposition in the frequency space. TIFO is a plug-and-play approach that can be seamlessly integrated into various forecasting models. Experiments demonstrate our method achieves 18 top-1 and 6 top-2 results out of 28 forecasting settings. Notably, it yields 33.3% and 55.3% improvements in average MSE on the ETTm2 dataset. In addition, TIFO reduces computational costs by 60% -70% compared to baseline methods, demonstrating strong scalability across diverse forecasting models.

Domain-specific accelerators deliver exceptional performance on their target workloads through fabrication-time orchestrated datapaths. However, such specialized architectures often exhibit performance fragility when exposed to new kernels or irregular input patterns. In contrast, programmable architectures like FPGAs, CGRAs, and GPUs rely on compile-time orchestration to support a broader range of applications; but they are typically less efficient under irregular or sparse data. Pushing the boundaries of programmable architectures requires designs that can achieve efficiency and high-performance on par with specialized accelerators while retaining the agility of general-purpose architectures. We introduce Canon, a parallel architecture that bridges the gap between specialized and general purpose architectures. Canon exploits data-level and instruction-level parallelism through its novel design. First, it employs a novel dynamic data-driven orchestration mechanism using programmable Finite State Machines (FSMs). These FSMs are programmed at compile time to encode high-level dataflow per state and translate incoming meta-information (e.g., sparse coordinates) into control instructions at runtime. Second, Canon introduces a time-lapsed SIMD execution in which instructions are issued across a row of processing elements over several cycles, creating a staggered pipelined execution. These innovations amortize control overhead, allowing dynamic instruction changes while constructing a continuously evolving dataflow that maximizes parallelism. Experimental evaluation shows that Canon delivers high performance across diverse data-agnostic and data-driven kernels while achieving efficiency comparable to specialized accelerators, yet retaining the flexibility of a general-purpose architecture.

Physical simulation predicts future states of objects based on material properties and external loads, enabling blueprints for both Industry and Engineering to conduct risk management. Current 3D reconstruction-based simulators typically rely on explicit, step-wise updates, which are sensitive to step time and suffer from rapid accuracy degradation under complicated scenarios, such as high-stiffness materials or quasi-static movement. To address this, we introduce i-PhysGaussian, a framework that couples 3D Gaussian Splatting (3DGS) with an implicit Material Point Method (MPM) integrator. Unlike explicit methods, our solution obtains an end-of-step state by minimizing a momentum-balance residual through implicit Newton-type optimization with a GMRES solver. This formulation significantly reduces time-step sensitivity and ensures physical consistency. Our results demonstrate that i-PhysGaussian maintains stability at up to 20x larger time steps than explicit baselines, preserving structural coherence and smooth motion even in complex dynamic transitions.

Generative AI presents an unprecedented challenge to our understanding of knowledge and its production. Unlike previous technological transformations, where engineering understanding preceded or accompanied deployment, generative AI operates through mechanisms whose epistemic character remains obscure, and without such understanding, its responsible integration into science, education, and institutional life cannot proceed on a principled basis. This paper argues that the missing account must begin with a paradigmatic break that has not yet received adequate philosophical attention. In the Turing-Shannon-von Neumann tradition, information enters the machine as encoded binary vectors, and semantics remains external to the process. Neural network architectures rupture this regime: symbolic input is instantly projected into a high-dimensional space where coordinates correspond to semantic parameters, transforming binary code into a position in a geometric space of meanings. It is this space that constitutes the active epistemic condition shaping generative production. Drawing on four structural properties of high-dimensional geometry concentration of measure, near-orthogonality, exponential directional capacity, and manifold regularity the paper develops an Indexical Epistemology of High-Dimensional Spaces. Building on Peirce semiotics and Papert constructionism, it reconceptualizes generative models as navigators of learned manifolds and proposes navigational knowledge as a third mode of knowledge production, distinct from both symbolic reasoning and statistical recombination.

Graph neural networks (GNNs) work remarkably well in semi-supervised node regression, yet a rigorous theory explaining when and why they succeed remains lacking. To address this gap, we study an aggregate-and-readout model that encompasses several common message passing architectures: node features are first propagated over the graph then mapped to responses via a nonlinear function. For least-squares estimation over GNNs with linear graph convolutions and a deep ReLU readout, we prove a sharp non-asymptotic risk bound that separates approximation, stochastic, and optimization errors. The bound makes explicit how performance scales with the fraction of labeled nodes and graph-induced dependence. Approximation guarantees are further derived for graph-smoothing followed by smooth nonlinear readouts, yielding convergence rates that recover classical nonparametric behavior under full supervision while characterizing performance when labels are scarce. Numerical experiments validate our theory, providing a systematic framework for understanding GNN performance and limitations.

With the availability of automation machinery and its superiority, are being slothful and inviting many diseases to invade them. The world still has so many places where people lack basic health facilities. Due to early detection and intervention, CDV can be cured to an extreme extent. It heavily reduces travel and associated costs. A remote ECG monitoring system enables community health workers to support and empower patients through telemedicine. However, there remains some financial and logistical burden. Heart disease cannot be taken lightly. These patients require regular health check-ups and the attention of health personnel in a short period if their health deteriorates suddenly and rapidly. Chronic diseases are extremely variable in their symptoms and evolution of treatment. Some, if not treated early, will end the patient's life. The trend of the INTERNET OF THINGS, IoT, is spreading massively. This paper focuses on the three main: the operator, the doctor, and the server over which the data is being sent.

Many OSS projects join foundations such as Apache, Eclipse, and OSGeo, to aid their immediate plans and improve long-term prospects by getting governance advice, incubation support, and community-building mechanisms. But foundations differ in their policies, funding models, and support strategies. Moreover, since projects joining these foundations are diverse, coming at different lifecycle stages and having different needs, it can be challenging to decide on the appropriate project-foundation match and on the project-specific plan for sustainability. Here, we present an empirical study and quantitative analysis of the sustainability of incubator projects in the Apache, Eclipse, and OSGeo foundations, and, additionally, of OSS projects from GitHub outside of foundations. We develop foundation-specific sustainability models and a project triage, based on projects' sociotechnical trace profiles, and demonstrate their effectiveness across the foundations. Our results show that our models with triage can effectively forecast sustainability outcomes not only within but across foundations. In addition, the generalizability of the framework allows us to apply the approach to GitHub projects outside the foundations. We complement our findings with actionable recovery strategies from previous work and apply them to case studies of failed incubator projects. Our study highlights the value of sociotechnical frameworks in characterizing and addressing software project sustainability issues.

Mastering educational concepts requires understanding both their prerequisites (e.g., recursion before merge sort) and sub-concepts (e.g., merge sort as part of sorting algorithms). Capturing these dependencies is critical for identifying students' knowledge gaps and enabling targeted intervention for personalized learning. This is especially challenging in large-scale courses, where instructors cannot feasibly diagnose individual misunderstanding or determine which concepts need reinforcement. While knowledge graphs offer a natural representation for capturing these conceptual relationships at scale, existing approaches are either surface-level (focusing on course-level concepts like "Algorithms" or logistical relationships such as course enrollment), or disregard the rich pedagogical signals embedded in instructional materials. We propose InstructKG, a framework for automatically constructing instructor-aligned knowledge graphs that capture a course's intended learning progression. Given a course's lecture materials (slides, notes, etc.), InstructKG extracts significant concepts as nodes and infers learning dependencies as directed edges (e.g., "part-of" or "depends-on" relationships). The framework synergizes the rich temporal and semantic signals unique to educational materials (e.g., "recursion" is taught before "mergesort"; "recursion" is mentioned in the definition of "merge sort") with the generalizability of large language models. Through experiments on real-world, diverse lecture materials across multiple courses and human-based evaluation, we demonstrate that InstructKG captures rich, instructor-aligned learning progressions.

Thematic Apperception Test (TAT) is a psychometrically grounded, multidimensional assessment framework that systematically differentiates between cognitive-representational and affective-relational components of personality-like functioning. This test is a projective psychological framework designed to uncover unconscious aspects of personality. This study examines whether the personality traits of Large Multimodal Models (LMMs) can be assessed through non-language-based modalities, using the Social Cognition and Object Relations Scale - Global (SCORS-G). LMMs are employed in two distinct roles: as subject models (SMs), which generate stories in response to TAT images, and as evaluator models (EMs), who assess these narratives using the SCORS-G framework. Evaluators demonstrated an excellent ability to understand and analyze TAT responses. Their interpretations are highly consistent with those of human experts. Assessment results highlight that all models understand interpersonal dynamics very well and have a good grasp of the concept of self. However, they consistently fail to perceive and regulate aggression. Performance varied systematically across model families, with larger and more recent models consistently outperforming smaller and earlier ones across SCORS-G dimensions.

Shapley value-based methods have become foundational in explainable artificial intelligence (XAI), offering theoretically grounded feature attributions through cooperative game theory. However, in practice, particularly in vision tasks, the assumption of feature independence breaks down, as features (i.e., pixels) often exhibit strong spatial and semantic dependencies. To address this, modern SHAP implementations now include the Owen value, a hierarchical generalization of the Shapley value that supports group attributions. While the Owen value preserves the foundations of Shapley values, its effectiveness critically depends on how feature groups are defined. We show that commonly used segmentations (e.g., axis-aligned or SLIC) violate key consistency properties, and propose a new segmentation approach that satisfies the $T$-property to ensure semantic alignment across hierarchy levels. This hierarchy enables computational pruning while improving attribution accuracy and interpretability. Experiments on image and tabular datasets demonstrate that O-Shap outperforms baseline SHAP variants in attribution precision, semantic coherence, and runtime efficiency, especially when structure matters.

Sustainability or ESG rating agencies use company disclosures and external data to produce scores or ratings that assess the environmental, social, and governance performance of a company. However, sustainability ratings across agencies for a single company vary widely, limiting their comparability, credibility, and relevance to decision-making. To harmonize the rating results, we propose adopting a universal human-AI collaboration framework to generate trustworthy benchmark datasets for evaluating sustainability rating methodologies. The framework comprises two complementary parts: STRIDE (Sustainability Trust Rating & Integrity Data Equation) provides principled criteria and a scoring system that guide the construction of firm-level benchmark datasets using large language models (LLMs), and SR-Delta, a discrepancy-analysis procedural framework that surfaces insights for potential adjustments. The framework enables scalable and comparable assessment of sustainability rating methodologies. We call on the broader AI community to adopt AI-powered approaches to strengthen and advance sustainability rating methodologies that support and enforce urgent sustainability agendas.

We propose a streamlined spectral algorithm for community detection in the two-community stochastic block model (SBM) under constant edge density assumptions. By reducing algorithmic complexity through the elimination of non-essential preprocessing steps, our method directly leverages the spectral properties of the adjacency matrix. We demonstrate that our algorithm exploits specific characteristics of the second eigenvalue to achieve improved error bounds that approach information-theoretic limits, representing a significant improvement over existing methods. Theoretical analysis establishes that our error rates are tighter than previously reported bounds in the literature. Comprehensive experimental validation confirms our theoretical findings and demonstrates the practical effectiveness of the simplified approach. Our results suggest that algorithmic simplification, rather than increasing complexity, can lead to both computational efficiency and enhanced performance in spectral community detection.

We investigate the recently introduced model of learning with improvements, where agents are allowed to make small changes to their feature values to be warranted a more desirable label. We extensively extend previously published results by providing combinatorial dimensions that characterize online learnability in this model, by analyzing the multiclass setup, learnability in a bandit feedback setup, modeling agents' cost for making improvements and more.

This paper presents a serverless MLOps framework orchestrating the complete ML lifecycle from data ingestion, training, deployment, monitoring, and retraining to using event-driven pipelines and managed services. The architecture is model-agnostic, supporting diverse inference patterns through standardized interfaces, enabling rapid adaptation without infrastructure overhead. We demonstrate practical applicability through an industrial implementation for Harmonized System (HS) code prediction, a compliance-critical task where short, unstructured product descriptions are mapped to standardized codes used by customs authorities in global trade. Frequent updates and ambiguous descriptions make classification challenging, with errors causing shipment delays and financial losses. Our solution uses a custom text embedding encoder and multiple deep learning architectures, with Text-CNN achieving 98 percent accuracy on ground truth data. Beyond accuracy, the pipeline ensures reproducibility, auditability, and SLA adherence under variable loads via auto-scaling. A key feature is automated A/B testing, enabling dynamic model selection and safe promotion in production. Cost-efficiency drives model choice; while transformers may achieve similar accuracy, their long-term operational costs are significantly higher. Deterministic classification with predictable latency and explainability is prioritized, though the architecture remains extensible to transformer variants and LLM-based inference. The paper first introduces the deep learning architectures with simulations and model comparisons, then discusses industrialization through serverless architecture, demonstrating automated retraining, prediction, and validation of HS codes. This work provides a replicable blueprint for operationalizing ML using serverless architecture, enabling enterprises to scale while optimizing performance and economics.

Large language model(LLM)-driven multi-agent systems(MAS) coordinate specialized agents through predefined interaction topologies and have shown promise for complex tasks such as competition-level code generation. Recent studies demonstrate that carefully designed multi-agent workflows and communication graphs can significantly improve code generation performance by leveraging collaborative reasoning. However, existing methods neither adapt topology density to task difficulty nor iteratively refine the topology within an instance using execution feedback, which leads to redundant communication and performance bottlenecks. To address these issues, we propose AgentConductor: a reinforcement learning-optimized MAS with an LLM-based orchestrator agent as its core, which enables end-to-end feedback-driven dynamic generation of interaction topologies. For each query, AgentConductor infers agent roles and task difficulty, then constructs a task-adapted, density-aware layered directed acyclic graph (DAG) topology, underpinned by two key innovations. First, we design a novel topological density function that captures communication-aware mathematical characterizations of multi-agent interactions. Second, we adopt difficulty interval partitioning to avoid excessive pruning for precise topological density upper bound measurement per difficulty level and finer-grained control. Empirically, across three competition-level and two foundational code datasets, AgentConductor achieves state-of-the-art accuracy, outperforming the strongest baseline by up to 14.6% in pass@1 accuracy, 13% in density reduction, and 68% in token cost reduction.

Portfolio Management is the process of overseeing a group of investments, referred to as a portfolio, with the objective of achieving predetermined investment goals. Portfolio optimization is a key component that involves allocating the portfolio assets so as to maximize returns while minimizing risk taken. It is typically carried out by financial professionals who use a combination of quantitative techniques and investment expertise to make decisions about the portfolio allocation. Recent applications of Deep Reinforcement Learning (DRL) have shown promising results when used to optimize portfolio allocation by training model-free agents on historical market data. Many of these methods compare their results against basic benchmarks or other state-of-the-art DRL agents but often fail to compare their performance against traditional methods used by financial professionals in practical settings. One of the most commonly used methods for this task is Mean-Variance Portfolio Optimization (MVO), which uses historical time series information to estimate expected asset returns and covariances, which are then used to optimize for an investment objective. Our work is a thorough comparison between model-free DRL and MVO for optimal portfolio allocation. We detail the specifics of how to make DRL for portfolio optimization work in practice, also noting the adjustments needed for MVO. Backtest results demonstrate strong performance of the DRL agent across many metrics, including Sharpe ratio, maximum drawdowns, and absolute returns.

As 6G wireless systems evolve, growing functional complexity and diverse service demands are driving a shift from rule-based control to intent-driven autonomous intelligence. User requirements are no longer captured by a single metric (e.g., throughput or reliability), but by multi-dimensional objectives such as latency sensitivity, energy preference, computational constraints, and service-level requirements. These objectives may also change over time due to environmental dynamics and user-network interactions. Therefore, accurate understanding of both the communication environment and user intent is critical for autonomous and sustainably evolving 6G communications. Large language models (LLMs), with strong contextual understanding and cross-modal reasoning, provide a promising foundation for intent-aware network agents. Compared with rule-driven or centrally optimized designs, LLM-based agents can integrate heterogeneous information and translate natural-language intents into executable control and configuration decisions. Focusing on a closed-loop pipeline of intent perception, autonomous decision making, and network execution, this paper investigates agentic AI for the 6G physical layer and its realization pathways. We review representative physical-layer tasks and their limitations in supporting intent awareness and autonomy, identify application scenarios where agentic AI is advantageous, and discuss key challenges and enabling technologies in multimodal perception, cross-layer decision making, and sustainable optimization. Finally, we present a case study of an intent-driven link decision agent, termed AgenCom, which adaptively constructs communication links under diverse user preferences and channel conditions.

Parameter-efficient fine-tuning techniques such as low-rank adaptation (LoRA) enable large language models (LLMs) to adapt to downstream tasks efficiently. Federated learning (FL) further facilitates this process by enabling collaborative fine-tuning across distributed clients without sharing private data. However, the use of two separate low-rank matrices in LoRA for federated fine-tuning introduces two types of challenges. The first challenge arises from the error induced by separately aggregating those two low-rank matrices. The second challenge occurs even when the product of two low-rank matrices is aggregated. The server needs to recover factors via matrix decomposition, which is non-unique and can introduce decomposition drift. To tackle the aforementioned challenges, we propose FLoRG, a federated fine-tuning framework which employs a single low-rank matrix for fine-tuning and aggregates its Gram matrix (i.e., the matrix of inner products of its column vectors), eliminating the aggregation error while also reducing the communication overhead. FLoRG minimizes the decomposition drift by introducing a Procrustes alignment approach which aligns the decomposed matrix between consecutive fine-tuning rounds for consistent updates. We theoretically analyze the convergence of FLoRG and prove that adopting the Procrustes alignment results in a tighter convergence bound. Experimental results across multiple LLM fine-tuning benchmarks demonstrate that FLoRG outperforms five state-of-the-art baseline schemes in the downstream task accuracy and can reduce the communication overhead by up to 2041$\times$.

Foreign key discovery and related schema-level prediction tasks are often modeled using graph neural networks (GNNs), implicitly assuming that relational inductive bias improves performance. However, it remains unclear when multi-hop structural reasoning is actually necessary. In this work, we introduce locality radius, a formal measure of the minimum structural neighborhood required to determine a prediction in relational schemas. We hypothesize that model performance depends critically on alignment between task locality radius and architectural aggregation depth. We conduct a controlled empirical study across foreign key prediction, join cost estimation, blast radius regression, cascade impact classification, and additional graph-derived schema tasks. Our evaluation includes multi-seed experiments, capacity-matched comparisons, statistical significance testing, scaling analysis, and synthetic radius-controlled benchmarks. Results reveal a consistent bias-radius alignment effect.

Agentic Coding, powered by autonomous agents such as GitHub Copilot and Cursor, enables developers to generate code, tests, and pull requests from natural language instructions alone. While this accelerates implementation, it produces larger volumes of code per pull request, shifting the burden from implementers to reviewers. In practice, a notable portion of AI-generated code is eventually deleted during review, yet reviewers must still examine such code before deciding to remove it. No prior work has explored methods to help reviewers efficiently identify code that will be removed.In this paper, we propose a prediction model that identifies functions likely to be deleted during PR review. Our results show that functions deleted for different reasons exhibit distinct characteristics, and our model achieves an AUC of 87.1%. These findings suggest that predictive approaches can help reviewers prioritize their efforts on essential code.

Diffusion models recently developed for generative AI tasks can produce high-quality samples while still maintaining diversity among samples to promote mode coverage, providing a promising path for learning stochastic closure models. Compared to other types of generative AI models, such as GANs and VAEs, the sampling speed is known as a key disadvantage of diffusion models. By systematically comparing transport-based generative models on a numerical example of 2D Kolmogorov flows, we show that flow matching in a lower-dimensional latent space is suited for fast sampling of stochastic closure models, enabling single-step sampling that is up to two orders of magnitude faster than iterative diffusion-based approaches. To control the latent space distortion and thus ensure the physical fidelity of the sampled closure term, we compare the implicit regularization offered by a joint training scheme against two explicit regularizers: metric-preserving (MP) and geometry-aware (GA) constraints. Besides offering a faster sampling speed, both explicitly and implicitly regularized latent spaces inherit the key topological information from the lower-dimensional manifold of the original complex dynamical system, which enables the learning of stochastic closure models without demanding a huge amount of training data.

The growing concern over training data privacy has elevated the "Right to be Forgotten" into a critical requirement, thereby raising the demand for effective Machine Unlearning. However, existing unlearning approaches commonly suffer from a fundamental trade-off: aggressively erasing the influence of target data often degrades model utility on retained data, while conservative strategies leave residual target information intact. In this work, the intrinsic representation properties learned during model pretraining are analyzed. It is demonstrated that semantic class concepts are entangled at the feature-pattern level, sharing associated features while preserving concept-specific discriminative components. This entanglement fundamentally limits the effectiveness of existing unlearning paradigms. Motivated by this insight, we propose Machine-Guided Unlearning (MeGU), a novel framework that guides unlearning through concept-aware re-alignment. Specifically, Multi-modal Large Language Models (MLLMs) are leveraged to explicitly determine re-alignment directions for target samples by assigning semantically meaningful perturbing labels. To improve efficiency, inter-class conceptual similarities estimated by the MLLM are encoded into a lightweight transition matrix. Furthermore, MeGU introduces a positive-negative feature noise pair to explicitly disentangle target concept influence. During finetuning, the negative noise suppresses target-specific feature patterns, while the positive noise reinforces remaining associated features and aligns them with perturbing concepts. This coordinated design enables selective disruption of target-specific representations while preserving shared semantic structures. As a result, MeGU enables controlled and selective forgetting, effectively mitigating both under-unlearning and over-unlearning.

Dynamic decision-making under model uncertainty is central to many economic environments, yet existing bandit and reinforcement learning algorithms rely on the assumption of correct model specification. This paper studies the behavior and performance of one of the most commonly used Bayesian reinforcement learning algorithms, Thompson Sampling (TS), when the model class is misspecified. We first provide a complete dynamic classification of posterior evolution in a misspecified two-armed Gaussian bandit, identifying distinct regimes: correct model concentration, incorrect model concentration, and persistent belief mixing, characterized by the direction of statistical evidence and the model-action mapping. These regimes yield sharp predictions for limiting beliefs, action frequencies, and asymptotic regret. We then extend the analysis to a general finite model class and develop a unified stochastic stability framework that represents posterior evolution as a Markov process on the belief simplex. This approach characterizes two sufficient conditions to classify the ergodic and transient behaviors and provides inductive dimensional reductions of the posterior dynamics. Our results offer the first qualitative and geometric classification of TS under misspecification, bridging Bayesian learning with evolutionary dynamics, and also build the foundations of robust decision-making in structured bandits.

The rapid adoption of large language models has led to the emergence of AI coding agents that autonomously create pull requests on GitHub. However, how these agents differ in their pull request description characteristics, and how human reviewers respond to them, remains underexplored. In this study, we conduct an empirical analysis of pull requests created by five AI coding agents using the AIDev dataset. We analyze agent differences in pull request description characteristics, including structural features, and examine human reviewer response in terms of review activity, response timing, sentiment, and merge outcomes. We find that AI coding agents exhibit distinct PR description styles, which are associated with differences in reviewer engagement, response time, and merge outcomes. We observe notable variation across agents in both reviewer interaction metrics and merge rates. These findings highlight the role of pull request presentation and reviewer interaction dynamics in human-AI collaborative software development.

Efficient stochastic optimization typically integrates an update direction that performs well in the deterministic regime with a mechanism adapting to stochastic perturbations. While Adam uses adaptive moment estimates to promote stability, Muon utilizes the weight layers' matrix structure via orthogonalized momentum, showing superior performance in large language model training. We propose a new optimizer and a diagonal extension, NAMO and NAMO-D, providing the first principled integration of orthogonalized momentum with norm-based Adam-type noise adaptation. NAMO scales orthogonalized momentum using a single adaptive stepsize, preserving orthogonality while improving upon Muon at negligible additional cost. NAMO-D instead right-multiplies orthogonalized momentum by a diagonal matrix with clamped entries. This design enables neuron-wise noise adaptation and aligns with the common near block-diagonal Hessian structure. Under standard assumptions, we establish optimal convergence rates for both algorithms in the deterministic setting and show that, in the stochastic setting, their convergence guarantees adapt to the noise level of stochastic gradients. Experiments on pretraining GPT-2 models demonstrate improved performance of both NAMO and NAMO-D compared to the AdamW and Muon baselines, with NAMO-D achieving further gains over NAMO via an additional clamping hyperparameter that balances the competing goals of maintaining a well-conditioned update direction and leveraging fine-grained noise adaptation.

Multi-agent reinforcement learning (MARL) has made significant progress in recent years, but most algorithms still rely on a discrete-time Markov Decision Process (MDP) with fixed decision intervals. This formulation is often ill-suited for complex multi-agent dynamics, particularly in high-frequency or irregular time-interval settings, leading to degraded performance and motivating the development of continuous-time MARL (CT-MARL). Existing CT-MARL methods are mainly built on Hamilton-Jacobi-Bellman (HJB) equations. However, they rarely account for safety constraints such as collision penalties, since these introduce discontinuities that make HJB-based learning difficult. To address this challenge, we propose a continuous-time constrained MDP (CT-CMDP) formulation and a novel MARL framework that transforms discrete MDPs into CT-CMDPs via an epigraph-based reformulation. We then solve this by proposing a novel physics-informed neural network (PINN)-based actor-critic method that enables stable and efficient optimization in continuous time. We evaluate our approach on continuous-time safe multi-particle environments (MPE) and safe multi-agent MuJoCo benchmarks. Results demonstrate smoother value approximations, more stable training, and improved performance over safe MARL baselines, validating the effectiveness and robustness of our method.

Large language models (LLMs)-based chatbots are increasingly being adopted in the financial domain, particularly in digital banking, to handle customer inquiries about products such as deposits, savings, and loans. However, these models still exhibit low accuracy in core banking computations-including total payout estimation, comparison of products with varying interest rates, and interest calculation under early repayment conditions. Such tasks require multi-step numerical reasoning and contextual understanding of banking products, yet existing LLMs often make systematic errors-misinterpreting product types, applying conditions incorrectly, or failing basic calculations involving exponents and geometric progressions. However, such errors have rarely been captured by existing benchmarks. Mathematical datasets focus on fundamental math problems, whereas financial benchmarks primarily target financial documents, leaving everyday banking scenarios underexplored. To address this limitation, we propose BankMathBench, a domain-specific dataset that reflects realistic banking tasks. BankMathBench is organized in three levels of difficulty-basic, intermediate, and advanced-corresponding to single-product reasoning, multi-product comparison, and multi-condition scenarios, respectively. When trained on BankMathBench, open-source LLMs exhibited notable improvements in both formula generation and numerical reasoning accuracy, demonstrating the dataset's effectiveness in enhancing domain-specific reasoning. With tool-augmented fine-tuning, the models achieved average accuracy increases of 57.6%p (basic), 75.1%p (intermediate), and 62.9%p (advanced), representing significant gains over zero-shot baselines. These findings highlight BankMathBench as a reliable benchmark for evaluating and advancing LLMs' numerical reasoning in real-world banking scenarios.

Graph neural networks frequently encounter significant performance degradation when confronted with structural noise or non-homophilous topologies. To address these systemic vulnerabilities, we present AdvSynGNN, a comprehensive architecture designed for resilient node-level representation learning. The proposed framework orchestrates multi-resolution structural synthesis alongside contrastive objectives to establish geometry-sensitive initializations. We develop a transformer backbone that adaptively accommodates heterophily by modulating attention mechanisms through learned topological signals. Central to our contribution is an integrated adversarial propagation engine, where a generative component identifies potential connectivity alterations while a discriminator enforces global coherence. Furthermore, label refinement is achieved through a residual correction scheme guided by per-node confidence metrics, which facilitates precise control over iterative stability. Empirical evaluations demonstrate that this synergistic approach effectively optimizes predictive accuracy across diverse graph distributions while maintaining computational efficiency. The study concludes with practical implementation protocols to ensure the robust deployment of the AdvSynGNN system in large-scale environments.

Probabilities of causation (PoCs), such as the probability of necessity and sufficiency (PNS), are important tools for decision making but are generally not point identifiable. Existing work has derived bounds for these quantities using combinations of experimental and observational data. However, there is very limited research on sample size analysis, namely, how many experimental and observational samples are required to achieve a desired margin of error. In this paper, we propose a general sample size framework based on the delta method. Our approach applies to settings in which the target bounds of PoCs can be expressed as finite minima or maxima of linear combinations of experimental and observational probabilities. Through simulation studies, we demonstrate that the proposed sample size calculations lead to stable estimation of these bounds.

Human-centric traffic signal control in corridor networks must increasingly account for multimodal travelers, particularly high-occupancy public transportation, rather than focusing solely on vehicle-centric performance. This paper proposes STDSH-MARL (Spatio-Temporal Dual-Stage Hypergraph based Multi-Agent Reinforcement Learning), a scalable multi-agent deep reinforcement learning framework that follows a centralized training and decentralized execution paradigm. The proposed method captures spatio-temporal dependencies through a novel dual-stage hypergraph attention mechanism that models interactions across both spatial and temporal hyperedges. In addition, a hybrid discrete action space is introduced to jointly determine the next signal phase configuration and its corresponding green duration, enabling more adaptive signal timing decisions. Experiments conducted on a corridor network under five traffic scenarios demonstrate that STDSH-MARL consistently improves multimodal performance and provides clear benefits for public transportation priority. Compared with state-of-the-art baseline methods, the proposed approach achieves superior overall performance. Further ablation studies confirm the contribution of each component of STDSH-MARL, with temporal hyperedges identified as the most influential factor driving the observed performance gains.

We introduce Predictive Batch Scheduling (PBS), a novel training optimization technique that accelerates language model convergence by dynamically prioritizing high-loss samples during batch construction. Unlike curriculum learning approaches that require predefined difficulty metrics or hard example mining methods that demand expensive per-sample loss tracking, PBS employs a lightweight linear predictor trained online to estimate sample difficulty from static token-level features. Our predictor achieves 0.44 correlation with actual loss using only four simple features: token frequency, sequence length, vocabulary diversity, and rare token ratio. Experiments on a 130M parameter transformer demonstrate that PBS achieves 6-13\% faster convergence measured by evaluation loss across training checkpoints, with the predictor's correlation improving from 0.14 to 0.44 over 10,000 training steps. These results validate that token frequency statistics encode meaningful information about sample difficulty, enabling effective curriculum learning with negligible computational overhead.

Sub-bit model compression seeks storage below one bit per weight; as magnitudes are aggressively compressed, the sign bit becomes a fixed-cost bottleneck. Across Transformers, CNNs, and MLPs, learned sign matrices resist low-rank approximation and are spectrally indistinguishable from an i.i.d. Rademacher baseline. Despite this apparent randomness, most weights retain their initialization signs; flips primarily occur via rare near-zero boundary crossings, suggesting that sign-pattern randomness is largely inherited from initialization. We formalize this behavior with sign lock-in theory, a stopping-time analysis of sign flips under SGD noise. Under bounded updates and a rare re-entry condition into a small neighborhood around zero, the number of effective sign flips exhibits a geometric tail. Building on this mechanism, we introduce a gap-based initialization and a lightweight outward-drift regularizer, reducing the effective flip rate to approximately $10^{-3}$ with only about a one-point increase in perplexity.

Value decomposition is a core approach for cooperative multi-agent reinforcement learning (MARL). However, existing methods still rely on a single optimal action and struggle to adapt when the underlying value function shifts during training, often converging to suboptimal policies. To address this limitation, we propose Successive Sub-value Q-learning (S2Q), which learns multiple sub-value functions to retain alternative high-value actions. Incorporating these sub-value functions into a Softmax-based behavior policy, S2Q encourages persistent exploration and enables $Q^{\text{tot}}$ to adjust quickly to the changing optima. Experiments on challenging MARL benchmarks confirm that S2Q consistently outperforms various MARL algorithms, demonstrating improved adaptability and overall performance. Our code is available at https://github.com/hyeon1996/S2Q.

While recent Arabic NLP benchmarks focus on scale, they often rely on synthetic or translated data which may benefit from deeper linguistic verification. We introduce ALPS (Arabic Linguistic & Pragmatic Suite), a native, expert-curated diagnostic challenge set probing Deep Semantics and Pragmatics, capabilities that complement specialized large-scale benchmarks. While broad-coverage benchmarks prioritize scale and multi-task coverage, ALPS targets the depth of linguistic understanding through 531 rigorously crafted questions across 15 tasks and 47 subtasks. We developed the dataset with deep expertise in Arabic linguistics, guaranteeing cultural authenticity and eliminating translation artifacts. Evaluating 23 diverse models (commercial, open-source, and Arabic-native) against a single-pass human performance (avg. 84.6% accuracy) and an expert-adjudicated oracle (99.2%), we reveal a critical dissociation: models achieve high fluency but fail on fundamental morpho-syntactic dependencies, with elevated error rates on morpho-syntactic dependencies (36.5% across diacritics-reliant tasks) compared to compositional semantics. While top commercial models (Gemini-3-flash at 94.2%) surpass the average single human, a substantial gap persists between commercial giants and Arabic-native models, with the best Arabic-specific model (Jais-2-70B at 83.6%) approaching but not matching human performance.

Large Reasoning Models (LRMs) exhibit strong performance, yet often produce rationales that sound plausible but fail to reflect their true decision process, undermining reliability and trust. We introduce a formal framework for reasoning faithfulness, defined by two testable conditions: stance consistency (a coherent stance linking reasoning to answer) and causal influence (the stated reasoning causally drives the answer under output-level interventions), explicitly decoupled from accuracy. To operationalize this, we present RFEval, a benchmark of 7,186 instances across seven tasks that probes faithfulness via controlled, output-level counterfactual interventions. Evaluating twelve open-source LRMs, we find unfaithfulness in 49.7% of outputs, predominantly from stance inconsistency. Failures are concentrated in brittle, convergent domains such as math and code, and correlate more with post-training regimes than with scale: within-family ablations indicate that adding current RL-style objectives on top of supervised fine-tuning can reduce reasoning faithfulness, even when accuracy is maintained. Crucially, accuracy is neither a sufficient nor a reliable proxy for faithfulness: once controlling for model and task, the accuracy-faithfulness link is weak and statistically insignificant. Our work establishes a rigorous methodology for auditing LRM reliability and shows that trustworthy AI requires optimizing not only for correct outcomes but also for the structural integrity of the reasoning process. Our code and dataset can be found at project page: $\href{https://aidaslab.github.io/RFEval/}{https://aidaslab.github.io/RFEval/}$

Analysing multilingual social media discourse remains a major challenge in natural language processing, particularly when large-scale public debates span across diverse languages. This study investigates how different approaches for cross-lingual text classification can support reliable analysis of global conversations. Using hydrogen energy as a case study, we analyse a decade-long dataset of over nine million tweets in English, Japanese, Hindi, and Korean (2013--2022) for topic discovery. The online keyword-driven data collection results in a significant amount of irrelevant content. We explore four approaches to filter relevant content: (1) translating English annotated data into target languages for building language-specific models for each target language, (2) translating unlabelled data appearing from all languages into English for creating a single model based on English annotations, (3) applying English fine-tuned multilingual transformers directly to each target language data, and (4) a hybrid strategy that combines translated annotations with multilingual training. Each approach is evaluated for its ability to filter hydrogen-related tweets from noisy keyword-based collections. Subsequently, topic modeling is performed to extract dominant themes within the relevant subsets. The results highlight key trade-offs between translation and multilingual approaches, offering actionable insights into optimising cross-lingual pipelines for large-scale social media analysis.

Embedding tables are critical components of large-scale recommendation systems, facilitating the efficient mapping of high-cardinality categorical features into dense vector representations. However, as the volume of unique IDs expands, traditional hash-based indexing methods suffer from collisions that degrade model performance and personalization quality. We present Multi-Probe Zero Collision Hash (MPZCH), a novel indexing mechanism based on linear probing that effectively mitigates embedding collisions. With reasonable table sizing, it often eliminates these collisions entirely while maintaining production-scale efficiency. MPZCH utilizes auxiliary tensors and high-performance CUDA kernels to implement configurable probing and active eviction policies. By retiring obsolete IDs and resetting reassigned slots, MPZCH prevents the stale embedding inheritance typical of hash-based methods, ensuring new features learn effectively from scratch. Despite its collision-mitigation overhead, the system maintains training QPS and inference latency comparable to existing methods. Rigorous online experiments demonstrate that MPZCH achieves zero collisions for user embeddings and significantly improves item embedding freshness and quality. The solution has been released within the open-source TorchRec library for the broader community.

Computer-use agents operate over long horizons under noisy perception, multi-window contexts, evolving environment states. Existing approaches, from RL-based planners to trajectory retrieval, often drift from user intent and repeatedly solve routine subproblems, leading to error accumulation and inefficiency. We present IntentCUA, a multi-agent computer-use framework designed to stabilize long-horizon execution through intent-aligned plan memory. A Planner, Plan-Optimizer, and Critic coordinate over shared memory that abstracts raw interaction traces into multi-view intent representations and reusable skills. At runtime, intent prototypes retrieve subgroup-aligned skills and inject them into partial plans, reducing redundant re-planning and mitigating error propagation across desktop applications. In end-to-end evaluations, IntentCUA achieved a 74.83% task success rate with a Step Efficiency Ratio of 0.91, outperforming RL-based and trajectory-centric baselines. Ablations show that multi-view intent abstraction and shared plan memory jointly improve execution stability, with the cooperative multi-agent loop providing the largest gains on long-horizon tasks. These results highlight that system-level intent abstraction and memory-grounded coordination are key to reliable and efficient desktop automation in large, dynamic environments.

A growing body of work attempts to evaluate the theory of mind (ToM) abilities of humans and large language models (LLMs) using static, non-interactive question-and-answer benchmarks. However, theoretical work in the field suggests that first-personal interaction is a crucial part of ToM and that such predictive, spectatorial tasks may fail to evaluate it. We address this gap with a novel ToM task that requires an agent to persuade a target to choose one of three policy proposals by strategically revealing information. Success depends on a persuader's sensitivity to a given target's knowledge states (what the target knows about the policies) and motivational states (how much the target values different outcomes). We varied whether these states were Revealed to persuaders or Hidden, in which case persuaders had to inquire about or infer them. In Experiment 1, participants persuaded a bot programmed to make only rational inferences. LLMs excelled in the Revealed condition but performed below chance in the Hidden condition, suggesting difficulty with the multi-step planning required to elicit and use mental state information. Humans performed moderately well in both conditions, indicating an ability to engage such planning. In Experiment 2, where a human target role-played the bot, and in Experiment 3, where we measured whether human targets' real beliefs changed, LLMs outperformed human persuaders across all conditions. These results suggest that effective persuasion can occur without explicit ToM reasoning (e.g., through rhetorical strategies) and that LLMs excel at this form of persuasion. Overall, our results caution against attributing human-like ToM to LLMs while highlighting LLMs' potential to influence people's beliefs and behavior.

Reinforcement learning (RL) has equipped LLM agents with a strong ability to solve complex tasks. However, existing RL methods normally use a \emph{single} policy network, causing \emph{simplicity bias} where simple tasks occupy most parameters and dominate gradient updates, leaving insufficient capacity for complex tasks. A plausible remedy could be employing the Mixture-of-Experts (MoE) architecture in the policy network, as MoE allows different parameters (experts) to specialize in different tasks, preventing simple tasks from dominating all parameters. However, a key limitation of traditional MoE is its token-level routing, where the router assigns each token to specialized experts, which fragments phase-consistent patterns into scattered expert assignments and thus undermines expert specialization. In this paper, we propose \textbf{Phase-Aware Mixture of Experts (PA-MoE)}. It first features a lightweight \emph{phase router} that learns latent phase boundaries directly from the RL objective without pre-defining phase categories. Then, the phase router allocates temporally consistent assignments to the same expert, allowing experts to preserve phase-specific expertise. Experimental results demonstrate the effectiveness of our proposed PA-MoE.

Autonomous coding agents, powered by large language models (LLMs), are increasingly being adopted in the software industry to automate complex engineering tasks. However, these agents are prone to a wide range of misbehaviors, such as deviating from the user's instructions, getting stuck in repetitive loops, or failing to use tools correctly. These failures disrupt the development workflow and often require resource-intensive manual intervention. In this paper, we present a system for automatically recovering from agentic misbehaviors at scale. We first introduce a taxonomy of misbehaviors grounded in an analysis of production traffic, identifying three primary categories: Specification Drift, Reasoning Problems, and Tool Call Failures, which we find occur in about 30% of all agent trajectories. To address these issues, we developed a lightweight, asynchronous self-intervention system named Wink. Wink observes agent trajectories and provides targeted course-correction guidance to nudge the agent back to a productive path. We evaluated our system on over 10,000 real world agent trajectories and found that it successfully resolves 90% of the misbehaviors that require a single intervention. Furthermore, a live A/B test in our production environment demonstrated that our system leads to a statistically significant reduction in Tool Call Failures, Tokens per Session and Engineer Interventions per Session. We present our experience designing and deploying this system, offering insights into the challenges of building resilient agentic systems at scale.

The continuous expansion of digital learning environments has catalyzed the demand for intelligent systems capable of providing personalized educational content. While current exercise recommendation frameworks have made significant strides, they frequently encounter obstacles regarding the long-tailed distribution of student engagement and the failure to adapt to idiosyncratic learning trajectories. We present LiveGraph, a novel active-structure neural re-ranking framework designed to overcome these limitations. Our approach utilizes a graph-based representation enhancement strategy to bridge the information gap between active and inactive students while integrating a dynamic re-ranking mechanism to foster content diversity. By prioritizing the structural relationships within learning histories, the proposed model effectively balances recommendation precision with pedagogical variety. Comprehensive experimental evaluations conducted on multiple real-world datasets demonstrate that LiveGraph surpasses contemporary baselines in both predictive accuracy and the breadth of exercise diversity.

Detecting anomalies in time-series data is critical in domains such as industrial operations, finance, and cybersecurity, where early identification of abnormal patterns is essential for ensuring system reliability and enabling preventive maintenance. However, most existing methods are reactive: they detect anomalies only after they occur and lack the capability to provide proactive early warning signals. In this paper, we propose FATE (Forecasting Anomalies with Time-series Ensembles), a novel unsupervised framework for detecting Precursors-of-Anomaly (PoA) by quantifying predictive uncertainty from a diverse ensemble of time-series forecasting models. Unlike prior approaches that rely on reconstruction errors or require ground-truth labels, FATE anticipates future values and leverages ensemble disagreement to signal early signs of potential anomalies without access to target values at inference time. To rigorously evaluate PoA detection, we introduce Precursor Time-series Aware Precision and Recall (PTaPR), a new metric that extends the traditional Time-series Aware Precision and Recall (TaPR) by jointly assessing segment-level accuracy, within-segment coverage, and temporal promptness of early predictions. This enables a more holistic assessment of early warning capabilities that existing metrics overlook. Experiments on five real-world benchmark datasets show that FATE achieves an average improvement of 19.9 percentage points in PTaPR AUC and 20.02 percentage points in early detection F1 score, outperforming baselines while requiring no anomaly labels. These results demonstrate the effectiveness and practicality of FATE for real-time unsupervised early warning in complex time-series environments.

Scientific discovery pipelines typically involve complex, rigid, and time-consuming processes, from data preparation to analyzing and interpreting findings. Recent advances in AI have the potential to transform such pipelines in a way that domain experts can focus on interpreting and understanding findings, rather than debugging rigid pipelines or manually annotating data. As part of an active collaboration between data science/AI researchers and behavioral neuroscientists, we showcase an example AI-enhanced pipeline, specifically designed to transform and accelerate the way that the domain experts in the team are able to gain insights out of experimental data. The application at hand is in the domain of behavioral neuroscience, studying fear generalization in mice, an important problem whose progress can advance our understanding of clinically significant and often debilitating conditions such as PTSD (Post-Traumatic Stress Disorder). We identify the emerging paradigm of "In-Context Learning" (ICL) as a suitable interface for domain experts to automate parts of their pipeline without the need for or familiarity with AI model training and fine-tuning, and showcase its remarkable efficacy in data preparation and pattern interpretation. Also, we introduce novel AI-enhancements to tensor decomposition model, which allows for more seamless pattern discovery from the heterogeneous data in our application. We thoroughly evaluate our proposed pipeline experimentally, showcasing its superior performance compared to what is standard practice in the domain, as well as against reasonable ML baselines that do not fall under the ICL paradigm, to ensure that we are not compromising performance in our quest for a seamless and easy-to-use interface for domain experts. Finally, we demonstrate effective discovery, with results validated by the domain experts in the team.

Group Relative Policy Optimization (GRPO) is effective for training language models on complex reasoning. However, since the objective is defined relative to a group of sampled trajectories, extended deliberation can create more chances to realize relative gains, leading to inefficient reasoning and overthinking, and complicating the trade-off between correctness and rollout efficiency. Controlling this behavior is difficult in practice, considering (i) Length penalties are hard to calibrate because longer rollouts may reflect harder problems that require longer reasoning, penalizing tokens risks truncating useful reasoning along with redundant continuation; and (ii) supervision that directly indicates when to continue or stop is typically unavailable beyond final answer correctness. We propose Weakly Supervised GRPO (WS-GRPO), which improves rollout efficiency by converting terminal rewards into correctness-aware guidance over partial trajectories. Unlike global length penalties that are hard to calibrate, WS-GRPO trains a preference model from outcome-only correctness to produce prefix-level signals that indicate when additional continuation is beneficial. Thus, WS-GRPO supplies outcome-derived continue/stop guidance, reducing redundant deliberation while maintaining accuracy. We provide theoretical results and empirically show on reasoning benchmarks that WS-GRPO substantially reduces rollout length while remaining competitive with GRPO baselines.

Conversational agents powered by large language models (LLMs) with tool integration achieve strong performance on fixed task-oriented dialogue datasets but remain vulnerable to unanticipated, user-induced errors. Rather than focusing on error prevention, this work focuses on error recovery, which necessitates the accurate diagnosis of erroneous dialogue contexts and execution of proper recovery plans. Under realistic constraints precluding model fine-tuning or prompt modification due to significant cost and time requirements, we explore whether agents can recover from contextually flawed interactions and how their behavior can be adapted without altering model parameters and prompts. To this end, we propose Reasoning Inception (ReIn), a test-time intervention method that plants an initial reasoning into the agent's decision-making process. Specifically, an external inception module identifies predefined errors within the dialogue context and generates recovery plans, which are subsequently integrated into the agent's internal reasoning process to guide corrective actions, without modifying its parameters or system prompts. We evaluate ReIn by systematically simulating conversational failure scenarios that directly hinder successful completion of user goals: user's ambiguous and unsupported requests. Across diverse combinations of agent models and inception modules, ReIn substantially improves task success and generalizes to unseen error types. Moreover, it consistently outperforms explicit prompt-modification approaches, underscoring its utility as an efficient, on-the-fly method. In-depth analysis of its operational mechanism, particularly in relation to instruction hierarchy, indicates that jointly defining recovery tools with ReIn can serve as a safe and effective strategy for improving the resilience of conversational agents without modifying the backbone models or system prompts.

The adoption of third-party libraries has become integral to modern software development, leading to large ecosystems such as PyPI, NPM, and Maven, where contributors typically share the technical expertise to sustain extensions. In communities that are not exclusively composed of developers, however, maintaining plugin ecosystems can present different challenges. In this early results paper, we study Obsidian, a knowledge--centric platform whose community is focused on writing, organization, and creativity--has built a substantial plugin ecosystem despite not being developer--centric. We investigate what kinds of plugins exist within this hybrid ecosystem and establish a foundation for understanding how they are maintained. Using repository mining and LLM-based topic modeling on a representative sample of 396 plugins, we identify six topics related to knowledge management and tooling, which is (i) dynamic editing and organization, (ii) interface and layouts, (iii) creative writing and productivity, (iv) knowledge sync solutions, (v) linking and script tools, and (vi) workflow enhancements tools. Furthermore, analysis of the Pull Requests from these plugins show that much software evolution has been performed on these ecosystem. These findings suggest that even in mixed communities, plugin ecosystems can develop recognizable engineering structures, motivating future work that highlight three different research directions with six research questions related to the health and sustainability of these non-developer ecosystems.

Automated formalization of mathematics enables mechanical verification but remains limited to isolated theorems and short snippets. Scaling to textbooks and research papers is largely unaddressed, as it requires managing cross-file dependencies, resolving imports, and ensuring that entire projects compile end-to-end. We present M2F (Math-to-Formal), the first agentic framework for end-to-end, project-scale autoformalization in Lean. The framework operates in two stages. The statement compilation stage splits the document into atomic blocks, orders them via inferred dependencies, and repairs declaration skeletons until the project compiles, allowing placeholders in proofs. The proof repair stage closes these holes under fixed signatures using goal-conditioned local edits. Throughout both stages, M2F keeps the verifier in the loop, committing edits only when toolchain feedback confirms improvement. In approximately three weeks, M2F converts long-form mathematical sources into a project-scale Lean library of 153,853 lines from 479 pages textbooks on real analysis and convex analysis, fully formalized as Lean declarations with accompanying proofs. This represents textbook-scale formalization at a pace that would typically require months or years of expert effort. On FATE-H, we achieve $96\%$ proof success (vs.\ $80\%$ for a strong baseline). Together, these results demonstrate that practical, large-scale automated formalization of mathematical literature is within reach. The full generated Lean code from our runs is available at https://github.com/optsuite/ReasBook.git.

Coordinating actions is the most fundamental form of cooperation in multi-agent reinforcement learning (MARL). Successful decentralized decision-making often depends not only on good individual actions, but on selecting compatible actions across agents to synchronize behavior, avoid conflicts, and satisfy global constraints. In this paper, we propose Action Graph Policies (AGP), that model dependencies among agents' available action choices. It constructs, what we call, \textit{coordination contexts}, that enable agents to condition their decisions on global action dependencies. Theoretically, we show that AGPs induce a strictly more expressive joint policy compared to fully independent policies and can realize coordinated joint actions that are provably more optimal than greedy execution even from centralized value-decomposition methods. Empirically, we show that AGP achieves 80-95\% success on canonical coordination tasks with partial observability and anti-coordination penalties, where other MARL methods reach only 10-25\%. We further demonstrate that AGP consistently outperforms these baselines in diverse multi-agent environments.

We present the technical report for Arcee Trinity Large, a sparse Mixture-of-Experts model with 400B total parameters and 13B activated per token. Additionally, we report on Trinity Nano and Trinity Mini, with Trinity Nano having 6B total parameters with 1B activated per token, Trinity Mini having 26B total parameters with 3B activated per token. The models' modern architecture includes interleaved local and global attention, gated attention, depth-scaled sandwich norm, and sigmoid routing for Mixture-of-Experts. For Trinity Large, we also introduce a new MoE load balancing strategy titled Soft-clamped Momentum Expert Bias Updates (SMEBU). We train the models using the Muon optimizer. All three models completed training with zero loss spikes. Trinity Nano and Trinity Mini were pre-trained on 10 trillion tokens, and Trinity Large was pre-trained on 17 trillion tokens. The model checkpoints are available at https://huggingface.co/arcee-ai.

Large language models have advanced web agents, yet current agents lack personalization capabilities. Since users rarely specify every detail of their intent, practical web agents must be able to interpret ambiguous queries by inferring user preferences and contexts. To address this challenge, we present Persona2Web, the first benchmark for evaluating personalized web agents on the real open web, built upon the clarify-to-personalize principle, which requires agents to resolve ambiguity based on user history rather than relying on explicit instructions. Persona2Web consists of: (1) user histories that reveal preferences implicitly over long time spans, (2) ambiguous queries that require agents to infer implicit user preferences, and (3) a reasoning-aware evaluation framework that enables fine-grained assessment of personalization. We conduct extensive experiments across various agent architectures, backbone models, history access schemes, and queries with varying ambiguity levels, revealing key challenges in personalized web agent behavior. For reproducibility, our codes and datasets are publicly available at https://anonymous.4open.science/r/Persona2Web-73E8.

Natural Language Querying for Time Series Databases (NLQ4TSDB) aims to assist non-expert users retrieve meaningful events, intervals, and summaries from massive temporal records. However, existing Text-to-SQL methods are not designed for continuous morphological intents such as shapes or anomalies, while time series models struggle to handle ultra-long histories. To address these challenges, we propose Sonar-TS, a neuro-symbolic framework that tackles NLQ4TSDB via a Search-Then-Verify pipeline. Analogous to active sonar, it utilizes a feature index to ping candidate windows via SQL, followed by generated Python programs to lock on and verify candidates against raw signals. To enable effective evaluation, we introduce NLQTSBench, the first large-scale benchmark designed for NLQ over TSDB-scale histories. Our experiments highlight the unique challenges within this domain and demonstrate that Sonar-TS effectively navigates complex temporal queries where traditional methods fail. This work presents the first systematic study of NLQ4TSDB, offering a general framework and evaluation standard to facilitate future research.

User stories are one of the most widely used artifacts in the software industry to define functional requirements. In parallel, the use of high-fidelity mockups facilitates end-user participation in defining their needs. In this work, we explore how combining these techniques with large language models (LLMs) enables agile and automated generation of user stories from mockups. To this end, we present a case study that analyzes the ability of LLMs to extract user stories from high-fidelity mockups, both with and without the inclusion of a glossary of the Language Extended Lexicon (LEL) in the prompts. Our results demonstrate that incorporating the LEL significantly enhances the accuracy and suitability of the generated user stories. This approach represents a step forward in the integration of AI into requirements engineering, with the potential to improve communication between users and developers.

Multi-path speculative decoding accelerates lossless sampling from a target model by using a cheaper draft model to generate a draft tree of tokens, and then applies a verification algorithm that accepts a subset of these. While prior work has proposed various verification algorithms for i.i.d rollouts, their relative performance under matched settings remains unclear. In this work, we firstly present a systematic evaluation of verification strategies across model families, tasks, and sampling regimes, and find that Traversal Verification dominates consistently, with OT-based methods lagging far behind. Our analysis uncovers that this occurs because OT-based methods achieve high multi-token acceptance near the root of the draft tree, while multi-token gains are most impactful deeper in the draft tree, where draft and target distributions diverge. Based on this insight, we propose delayed tree expansion, which drafts a partial single path, delaying the i.i.d. branching point. We show that delayed tree expansion preserves the target distribution and improves on root-node i.i.d rollouts. Further, we develop a dynamic neural selector that estimates the expected block efficiency of optimal-transport-based verification methods from draft and target features, enabling context-dependent expansion decisions. Our neural selector allows OT-based methods like SpecInfer to outperform Traversal Verification for the first time, achieving 5% higher average throughput across a wide range of models, datasets, and sampling settings.

Most recommendation benchmarks evaluate how well a model imitates user behavior. In financial advisory, however, observed actions can be noisy or short-sighted under market volatility and may conflict with a user's long-term goals. Treating what users chose as the sole ground truth, therefore, conflates behavioral imitation with decision quality. We introduce Conv-FinRe, a conversational and longitudinal benchmark for stock recommendation that evaluates LLMs beyond behavior matching. Given an onboarding interview, step-wise market context, and advisory dialogues, models must generate rankings over a fixed investment horizon. Crucially, Conv-FinRe provides multi-view references that distinguish descriptive behavior from normative utility grounded in investor-specific risk preferences, enabling diagnosis of whether an LLM follows rational analysis, mimics user noise, or is driven by market momentum. We build the benchmark from real market data and human decision trajectories, instantiate controlled advisory conversations, and evaluate a suite of state-of-the-art LLMs. Results reveal a persistent tension between rational decision quality and behavioral alignment: models that perform well on utility-based ranking often fail to match user choices, whereas behaviorally aligned models can overfit short-term noise. The dataset is publicly released on Hugging Face, and the codebase is available on GitHub.

Black-box safety evaluation of AI systems assumes model behavior on test distributions reliably predicts deployment performance. We formalize and challenge this assumption through latent context-conditioned policies -- models whose outputs depend on unobserved internal variables that are rare under evaluation but prevalent under deployment. We establish fundamental limits showing that no black-box evaluator can reliably estimate deployment risk for such models. (1) Passive evaluation: For evaluators sampling i.i.d. from D_eval, we prove minimax lower bounds via Le Cam's method: any estimator incurs expected absolute error >= (5/24)*delta*L approximately 0.208*delta*L, where delta is trigger probability under deployment and L is the loss gap. (2) Adaptive evaluation: Using a hash-based trigger construction and Yao's minimax principle, worst-case error remains >= delta*L/16 even for fully adaptive querying when D_dep is supported over a sufficiently large domain; detection requires Theta(1/epsilon) queries. (3) Computational separation: Under trapdoor one-way function assumptions, deployment environments possessing privileged information can activate unsafe behaviors that any polynomial-time evaluator without the trapdoor cannot distinguish. For white-box probing, estimating deployment risk to accuracy epsilon_R requires O(1/(gamma^2 * epsilon_R^2)) samples, where gamma = alpha_0 + alpha_1 - 1 measures probe quality, and we provide explicit bias correction under probe error. Our results quantify when black-box testing is statistically underdetermined and provide explicit criteria for when additional safeguards -- architectural constraints, training-time guarantees, interpretability, and deployment monitoring -- are mathematically necessary for worst-case safety assurance.

Modern language models exhibit rich internal structure, yet little is known about how privacy-sensitive behaviors, such as personally identifiable information (PII) leakage, are represented and modulated within their hidden states. We present UniLeak, a mechanistic-interpretability framework that identifies universal activation directions: latent directions in a model's residual stream whose linear addition at inference time consistently increases the likelihood of generating PII across prompts. These model-specific directions generalize across contexts and amplify PII generation probability, with minimal impact on generation quality. UniLeak recovers such directions without access to training data or groundtruth PII, relying only on self-generated text. Across multiple models and datasets, steering along these universal directions substantially increases PII leakage compared to existing prompt-based extraction methods. Our results offer a new perspective on PII leakage: the superposition of a latent signal in the model's representations, enabling both risk amplification and mitigation.

Despite the recent success of Multimodal Large Language Models (MLLMs), existing approaches predominantly assume the availability of multiple modalities during training and inference. In practice, multimodal data is often incomplete because modalities may be missing, collected asynchronously, or available only for a subset of examples. In this work, we propose PRIMO, a supervised latent-variable imputation model that quantifies the predictive impact of any missing modality within the multimodal learning setting. PRIMO enables the use of all available training examples, whether modalities are complete or partial. Specifically, it models the missing modality through a latent variable that captures its relationship with the observed modality in the context of prediction. During inference, we draw many samples from the learned distribution over the missing modality to both obtain the marginal predictive distribution (for the purpose of prediction) and analyze the impact of the missing modalities on the prediction for each instance. We evaluate PRIMO on a synthetic XOR dataset, Audio-Vision MNIST, and MIMIC-III for mortality and ICD-9 prediction. Across all datasets, PRIMO obtains performance comparable to unimodal baselines when a modality is fully missing and to multimodal baselines when all modalities are available. PRIMO quantifies the predictive impact of a modality at the instance level using a variance-based metric computed from predictions across latent completions. We visually demonstrate how varying completions of the missing modality result in a set of plausible labels.

We identify a structural weakness in current large language model (LLM) alignment: modern refusal mechanisms are fail-open. While existing approaches encode refusal behaviors across multiple latent features, suppressing a single dominant feature$-$via prompt-based jailbreaks$-$can cause alignment to collapse, leading to unsafe generation. Motivated by this, we propose fail-closed alignment as a design principle for robust LLM safety: refusal mechanisms should remain effective even under partial failures via redundant, independent causal pathways. We present a concrete instantiation of this principle: a progressive alignment framework that iteratively identifies and ablates previously learned refusal directions, forcing the model to reconstruct safety along new, independent subspaces. Across four jailbreak attacks, we achieve the strongest overall robustness while mitigating over-refusal and preserving generation quality, with small computational overhead. Our mechanistic analyses confirm that models trained with our method encode multiple, causally independent refusal directions that prompt-based jailbreaks cannot suppress simultaneously, providing empirical support for fail-closed alignment as a principled foundation for robust LLM safety.

Here's the corrected paragraph with all punctuation and formatting issues fixed: Financial risk systems usually follow a two-step routine: a model predicts return or risk, and then an optimizer makes a decision such as a portfolio rebalance. In practice, this split can break under real constraints. The prediction model may look good, but the final decision can be unstable when the market shifts, when discrete constraints are added (lot sizes, caps), or when the optimization becomes slow for larger asset sets. Also, regulated settings need a clear audit trail that links each decision to the exact model state and inputs. We present HQFS, a practical hybrid pipeline that connects forecasting, discrete risk optimization, and auditability in one flow. First, HQFS learns next-step return and a volatility proxy using a variational quantum circuit (VQC) with a small classical head. Second, HQFS converts the risk-return objective and constraints into a QUBO and solves it with quantum annealing when available, while keeping a compatible classical QUBO solver as a fallback for deployment. Third, HQFS signs each rebalance output using a post-quantum signature so the allocation can be verified later without trusting the runtime environment. On our market dataset study, HQFS reduces return prediction error by 7.8% and volatility prediction error by 6.1% versus a tuned classical baseline. For the decision layer, HQFS improves out-of-sample Sharpe by 9.4% and lowers maximum drawdown by 11.7%. The QUBO solve stage also cuts average solve time by 28% compared to a mixed-integer baseline under the same constraints, while producing fully traceable, signed allocation records.

Diffusion Transformers (DiTs) have achieved state-of-the-art performance in image and video generation, but their success comes at the cost of heavy computation. This inefficiency is largely due to the fixed tokenization process, which uses constant-sized patches throughout the entire denoising phase, regardless of the content's complexity. We propose dynamic tokenization, an efficient test-time strategy that varies patch sizes based on content complexity and the denoising timestep. Our key insight is that early timesteps only require coarser patches to model global structure, while later iterations demand finer (smaller-sized) patches to refine local details. During inference, our method dynamically reallocates patch sizes across denoising steps for image and video generation and substantially reduces cost while preserving perceptual generation quality. Extensive experiments demonstrate the effectiveness of our approach: it achieves up to $3.52\times$ and $3.2\times$ speedup on FLUX-1.Dev and Wan $2.1$, respectively, without compromising the generation quality and prompt adherence.

Grokking -- the abrupt transition from memorization to generalization after prolonged training -- has been linked to confinement on low-dimensional execution manifolds in modular arithmetic. Whether this mechanism extends beyond arithmetic remains open. We study two sequence-learning benchmarks: SCAN compositional generalization and Dyck-1 depth prediction. Across both tasks and a wide range of learning rates, the commutator defect -- a curvature measure derived from non-commuting gradient updates -- rises well before generalization, with lead times following a superlinear power law (alpha approximately 1.18 for SCAN, approximately 1.13 for Dyck), consistent with prior results on modular arithmetic. Weight-space PCA reveals that spectral concentration is not a universal precursor; the commutator defect is. Causal interventions demonstrate a mechanistic role: amplifying non-commutativity accelerates grokking (roughly 32% on SCAN, roughly 50% on Dyck), while suppressing orthogonal gradient flow delays or prevents it. The three task families form a spectrum of causal sensitivity -- modular arithmetic is rigid, Dyck is responsive, SCAN is intermediate -- yet suppression delays or prevents grokking in all cases, establishing necessity as a universal finding. These results identify the commutator defect as a robust, architecture-agnostic, causally implicated early-warning signal for delayed generalization in transformers.

Scalable Multi-Agent Reinforcement Learning (MARL) is fundamentally challenged by the curse of dimensionality. A common solution is to exploit locality, which hinges on an Exponential Decay Property (EDP) of the value function. However, existing conditions that guarantee the EDP are often conservative, as they are based on worst-case, environment-only bounds (e.g., supremums over actions) and fail to capture the regularizing effect of the policy itself. In this work, we establish that locality can also be a \emph{policy-dependent} phenomenon. Our central contribution is a novel decomposition of the policy-induced interdependence matrix, $H^π$, which decouples the environment's sensitivity to state ($E^{\mathrm{s}}$) and action ($E^{\mathrm{a}}$) from the policy's sensitivity to state ($Π(π)$). This decomposition reveals that locality can be induced by a smooth policy (small $Π(π)$) even when the environment is strongly action-coupled, exposing a fundamental locality-optimality tradeoff. We use this framework to derive a general spectral condition $ρ(E^{\mathrm{s}}+E^{\mathrm{a}}Π(π)) < 1$ for exponential decay, which is strictly tighter than prior norm-based conditions. Finally, we leverage this theory to analyze a provably-sound localized block-coordinate policy improvement framework with guarantees tied directly to this spectral radius.

We study the decentralized multi-player stochastic bandit problem over a continuous, Lipschitz-structured action space where hard collisions yield zero reward. Our objective is to design a communication-free policy that maximizes collective reward, with coordination costs that are independent of the time horizon $T$. We propose a modular protocol that first solves the multi-agent coordination problem -- identifying and seating players on distinct high-value regions via a novel maxima-directed search -- and then decouples the problem into $N$ independent single-player Lipschitz bandits. We establish a near-optimal regret bound of $\tilde{O}(T^{(d+1)/(d+2)})$ plus a $T$-independent coordination cost, matching the single-player rate. To our knowledge, this is the first framework providing such guarantees, and it extends to general distance-threshold collision models.

The goal of $L$-step speculative decoding is to accelerate autoregressive decoding of a target model by using a cheaper draft model to generate a candidate path of $L$ tokens. Based on a verification algorithm involving target and draft model probabilities, a prefix of the candidate sequence is accepted, and an additional correction token is sampled from a residual distribution to ensure that the final output adheres to the target distribution. While standard speculative decoding uses a verification algorithm which is independent at each token on the path, a recent extension called block verification uses a joint condition involving all sampled on-path probabilities. Block verification (BV) was shown to be optimal over all verification algorithms which use only on-path probabilities, improving on standard speculative decoding. In this work, we first show that block verification is optimal even over verification algorithms that use off-path probabilities, by constructing an information-agnostic linear program (LP). Further, we can extend our LP to the setting where the draft model samples multiple candidate paths, and use it to construct a natural class of multi-path block verification generalizations. While computing the optimal algorithm in this class is not tractable, by considering a stricter class of greedy algorithms, we can formulate an efficient method called greedy multi-path block verification (GBV). Empirically, GBV can improve block efficiency by over 30% and reduce decoding walltimes by over 15% relative to BV. On Llama-3 70B, GBV can improve the end-to-end decoding throughput over SOTA multi-path verification methods by more than 15%.

Classical Persian poetry is a historically sustained archive in which affective life is expressed through metaphor, intertextual convention, and rhetorical indirection. These properties make close reading indispensable while limiting reproducible comparison at scale. We present an uncertainty-aware computational framework for poet-level psychological analysis based on large-scale automatic multi-label annotation. Each verse is associated with a set of psychological concepts, per-label confidence scores, and an abstention flag that signals insufficient evidence. We aggregate confidence-weighted evidence into a Poet $\times$ Concept matrix, interpret each poet as a probability distribution over concepts, and quantify poetic individuality as divergence from a corpus baseline using Jensen--Shannon divergence and Kullback--Leibler divergence. To capture relational structure beyond marginals, we build a confidence-weighted co-occurrence graph over concepts and define an Eigenmood embedding through Laplacian spectral decomposition. On a corpus of 61{,}573 verses across 10 poets, 22.2\% of verses are abstained, underscoring the analytical importance of uncertainty. We further report sensitivity analysis under confidence thresholding, selection-bias diagnostics that treat abstention as a category, and a distant-to-close workflow that retrieves verse-level exemplars along Eigenmood axes. The resulting framework supports scalable, auditable digital-humanities analysis while preserving interpretive caution by propagating uncertainty from verse-level evidence to poet-level inference.

Agent hijacking, highlighted by OWASP as a critical threat to the Large Language Model (LLM) ecosystem, enables adversaries to manipulate execution by injecting malicious instructions into retrieved content. Most existing attacks rely on manually crafted, semantics-driven prompt manipulation, which often yields low attack success rates and limited transferability to closed-source commercial models. In this paper, we propose Phantom, an automated agent hijacking framework built upon Structured Template Injection that targets the fundamental architectural mechanisms of LLM agents. Our key insight is that agents rely on specific chat template tokens to separate system, user, assistant, and tool instructions. By injecting optimized structured templates into the retrieved context, we induce role confusion and cause the agent to misinterpret the injected content as legitimate user instructions or prior tool outputs. To enhance attack transferability against black-box agents, Phantom introduces a novel attack template search framework. We first perform multi-level template augmentation to increase structural diversity and then train a Template Autoencoder (TAE) to embed discrete templates into a continuous, searchable latent space. Subsequently, we apply Bayesian optimization to efficiently identify optimal adversarial vectors that are decoded into high-potency structured templates. Extensive experiments on Qwen, GPT, and Gemini demonstrate that our framework significantly outperforms existing baselines in both Attack Success Rate (ASR) and query efficiency. Moreover, we identified over 70 vulnerabilities in real-world commercial products that have been confirmed by vendors, underscoring the practical severity of structured template-based hijacking and providing an empirical foundation for securing next-generation agentic systems.

Euphemisms substitute socially sensitive expressions, often softening or reframing meaning, and their reliance on cultural and pragmatic context complicates modeling across languages. In this study, we investigate how cross-lingual equivalence influences transfer in multilingual euphemism detection. We categorize Potentially Euphemistic Terms (PETs) in Turkish and English into Overlapping (OPETs) and Non-Overlapping (NOPETs) subsets based on their functional, pragmatic, and semantic alignment. Our findings reveal a transfer asymmetry: semantic overlap is insufficient to guarantee positive transfer, particularly in low-resource Turkish-to-English direction, where performance can degrade even for overlapping euphemisms, and in some cases, improve under NOPET-based training. Differences in label distribution help explain these counterintuitive results. Category-level analysis suggests that transfer may be influenced by domain-specific alignment, though evidence is limited by sparsity.

We challenge black-box purely deep neural approaches for molecules and graph generation, which are limited in controllability and lack formal guarantees. We introduce Neuro-Symbolic Graph Generative Modeling (NSGGM), a neurosymbolic framework that reapproaches molecule generation as a scaffold and interaction learning task with symbolic assembly. An autoregressive neural model proposes scaffolds and refines interaction signals, and a CPU-efficient SMT solver constructs full graphs while enforcing chemical validity, structural rules, and user-specific constraints, yielding molecules that are correct by construction and interpretable control that pure neural methods cannot provide. NSGGM delivers strong performance on both unconstrained generation and constrained generation tasks, demonstrating that neuro-symbolic modeling can match state-of-the-art generative performance while offering explicit controllability and guarantees. To evaluate more nuanced controllability, we also introduce a Logical-Constraint Molecular Benchmark, designed to test strict hard-rule satisfaction in workflows that require explicit, interpretable specifications together with verifiable compliance.

Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback is often expensive and slow to obtain, making online reinforcement learning (RL) impractical. High-coverage hardware verification exemplifies this challenge due to its reliance on industrial simulators and non-differentiable execution signals. We propose LLM4Cov, an offline agent-learning framework that models verification as memoryless state transitions guided by deterministic evaluators. Building on this formulation, we introduce execution-validated data curation, policy-aware agentic data synthesis, and worst-state-prioritized sampling to enable scalable learning under execution constraints. We further curate a reality-aligned benchmark adapted from an existing verification suite through a revised evaluation protocol. Using the proposed pipeline, a compact 4B-parameter model achieves 69.2% coverage pass rate under agentic evaluation, outperforming its teacher by 5.3% and demonstrating competitive performance against models an order of magnitude larger.

Developing foundation models for electroencephalography (EEG) remains challenging due to the signal's low signal-to-noise ratio and complex spectro-temporal non-stationarity. Existing approaches often overlook the hierarchical latent structure inherent in neural dynamics, leading to suboptimal reconstruction of fine-grained information. In this work, we propose BrainRVQ, a general-purpose EEG foundation model pre-trained on a large-scale corpus of clinical EEG data. Unlike standard masked modeling, BrainRVQ features a Dual-Domain Residual Vector Quantization (DD-RVQ) tokenizer that disentangles temporal waveforms and spectral patterns into hierarchical discrete codes. We further introduce a hierarchical autoregressive pre-training objective that learns to reconstruct these codes in a coarse-to-fine manner, utilizing an importance-guided curriculum masking strategy to prioritize information-rich neural events over background noise. Extensive experiments across 8 diverse downstream datasets demonstrate that BrainRVQ consistently outperforms state-of-the-art baselines, validating its effectiveness in learning robust and generalizable neural representations. Our code and model weights are available:https://github.com/keqicmz/BrainRVQ

Graph Neural Networks (GNNs) have become essential in high-stakes domains such as drug discovery, yet their black-box nature remains a significant barrier to trustworthiness. While self-explainable GNNs attempt to bridge this gap, they often rely on standard message-passing backbones that inherit fundamental limitations, including the 1-Weisfeiler-Lehman (1-WL) expressivity barrier and a lack of fine-grained interpretability. To address these challenges, we propose SymGraph, a symbolic framework designed to transcend these constraints. By replacing continuous message passing with discrete structural hashing and topological role-based aggregation, our architecture theoretically surpasses the 1-WL barrier, achieving superior expressiveness without the overhead of differentiable optimization. Extensive empirical evaluations demonstrate that SymGraph achieves state-of-the-art performance, outperforming existing self-explainable GNNs. Notably, SymGraph delivers 10x to 100x speedups in training time using only CPU execution. Furthermore, SymGraph generates rules with superior semantic granularity compared to existing rule-based methods, offering great potential for scientific discovery and explainable AI.

This work introduces a verification framework that provides both sound and complete guarantees for data poisoning attacks during neural network training. We formulate adversarial data manipulation, model training, and test-time evaluation in a single mixed-integer quadratic programming (MIQCP) problem. Finding the global optimum of the proposed formulation provably yields worst-case poisoning attacks, while simultaneously bounding the effectiveness of all possible attacks on the given training pipeline. Our framework encodes both the gradient-based training dynamics and model evaluation at test time, enabling the first exact certification of training-time robustness. Experimental evaluation on small models confirms that our approach delivers a complete characterization of robustness against data poisoning.

Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that suppresses harmful text also suppress harmful actions? We introduce the GAP benchmark, a systematic evaluation framework that measures divergence between text-level safety and tool-call-level safety in LLM agents. We test six frontier models across six regulated domains (pharmaceutical, financial, educational, employment, legal, and infrastructure), seven jailbreak scenarios per domain, three system prompt conditions (neutral, safety-reinforced, and tool-encouraging), and two prompt variants, producing 17,420 analysis-ready datapoints. Our central finding is that text safety does not transfer to tool-call safety. Across all six models, we observe instances where the model's text output refuses a harmful request while its tool calls simultaneously execute the forbidden action--a divergence we formalize as the GAP metric. Even under safety-reinforced system prompts, 219 such cases persist across all six models. System prompt wording exerts substantial influence on tool-call behavior: TC-safe rates span 21 percentage points for the most robust model and 57 for the most prompt-sensitive, with 16 of 18 pairwise ablation comparisons remaining significant after Bonferroni correction. Runtime governance contracts reduce information leakage in all six models but produce no detectable deterrent effect on forbidden tool-call attempts themselves. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation.

Large language models (LLMs) increasingly answer queries by citing web sources, but existing evaluations emphasize answer correctness rather than evidence quality. We introduce SourceBench, a benchmark for measuring the quality of cited web sources across 100 real-world queries spanning informational, factual, argumentative, social, and shopping intents. SourceBench uses an eight-metric framework covering content quality (content relevance, factual accuracy, objectivity) and page-level signals (e.g., freshness, authority/accountability, clarity), and includes a human-labeled dataset with a calibrated LLM-based evaluator that matches expert judgments closely. We evaluate eight LLMs, Google Search, and three AI search tools over 3996 cited sources using SourceBench and conduct further experiments to understand the evaluation results. Overall, our work reveals four key new insights that can guide future research in the direction of GenAI and web search.

The promise of LLM-based user simulators to improve conversational AI is hindered by a critical "realism gap," leading to systems that are optimized for simulated interactions, but may fail to perform well in the real world. We introduce ConvApparel, a new dataset of human-AI conversations designed to address this gap. Its unique dual-agent data collection protocol -- using both "good" and "bad" recommenders -- enables counterfactual validation by capturing a wide spectrum of user experiences, enriched with first-person annotations of user satisfaction. We propose a comprehensive validation framework that combines statistical alignment, a human-likeness score, and counterfactual validation to test for generalization. Our experiments reveal a significant realism gap across all simulators. However, the framework also shows that data-driven simulators outperform a prompted baseline, particularly in counterfactual validation where they adapt more realistically to unseen behaviors, suggesting they embody more robust, if imperfect, user models.

Large Language Models (LLMs) have demonstrated remarkable effectiveness in adapting to downstream tasks through fine-tuning. Federated Learning (FL) extends this capability by enabling collaborative fine-tuning across distributed clients using Low-Rank Adaptation (LoRA), while preserving data privacy by avoiding raw data sharing. However, practical deployments face challenges when clients have heterogeneous resources and thus adopt different LoRA ranks, leading to substantial initialization and aggregation noise that undermines performance. To address these challenges, we propose Fed-PLoRA, a novel lightweight heterogeneous federated fine-tuning (FFT) framework. Fed-PLoRA introduces Parallel One-Rank Adaptation (PLoRA), a new LoRA variant that replaces the classic multi-rank LoRA module with multiple parallel one-rank modules, and a novel Select-N-Fold strategy that folds untrained PLoRA modules into the pre-trained weights before local training, thereby accommodating heterogeneous client resources. We provide a unified analysis of initialization and aggregation noise of Fed-PLoRA and demonstrate how it addresses the limitations of state-of-the-art methods. Extensive experiments on diverse LLM fine-tuning tasks demonstrate that Fed-PLoRA consistently outperforms existing methods in both accuracy and efficiency. The code is available at https://github.com/TNI-playground/Fed-PLoRA.

While Large Language Model (LLM) capabilities have scaled, safety guardrails remain largely stateless, treating multi-turn dialogues as a series of disconnected events. This lack of temporal awareness facilitates a "Safety Gap" where adversarial tactics, like Crescendo and ActorAttack, slowly bleed malicious intent across turn boundaries to bypass stateless filters. We introduce DeepContext, a stateful monitoring framework designed to map the temporal trajectory of user intent. DeepContext discards the isolated evaluation model in favor of a Recurrent Neural Network (RNN) architecture that ingests a sequence of fine-tuned turn-level embeddings. By propagating a hidden state across the conversation, DeepContext captures the incremental accumulation of risk that stateless models overlook. Our evaluation demonstrates that DeepContext significantly outperforms existing baselines in multi-turn jailbreak detection, achieving a state-of-the-art F1 score of 0.84, which represents a substantial improvement over both hyperscaler cloud-provider guardrails and leading open-weight models such as Llama-Prompt-Guard-2 (0.67) and Granite-Guardian (0.67). Furthermore, DeepContext maintains a sub-20ms inference overhead on a T4 GPU, ensuring viability for real-time applications. These results suggest that modeling the sequential evolution of intent is a more effective and computationally efficient alternative to deploying massive, stateless models.

Retrieval algorithms like BM25 and query likelihood with Dirichlet smoothing remain strong and efficient first-stage rankers, yet improvements have mostly relied on parameter tuning and human intuition. We investigate whether a large language model, guided by an evaluator and evolutionary search, can automatically discover improved lexical retrieval algorithms. We introduce RankEvolve, a program evolution setup based on AlphaEvolve, in which candidate ranking algorithms are represented as executable code and iteratively mutated, recombined, and selected based on retrieval performance across 12 IR datasets from BEIR and BRIGHT. RankEvolve starts from two seed programs: BM25 and query likelihood with Dirichlet smoothing. The evolved algorithms are novel, effective, and show promising transfer to the full BEIR and BRIGHT benchmarks as well as TREC DL 19 and 20. Our results suggest that evaluator-guided LLM program evolution is a practical path towards automatic discovery of novel ranking algorithms.

Lifelong multimodal agents must continuously adapt to new tasks through post-training, but this creates fundamental tension between acquiring capabilities and preserving safety alignment. We demonstrate that fine-tuning aligned vision-language models on narrow-domain harmful datasets induces severe emergent misalignment that generalizes broadly across unrelated tasks and modalities. Through experiments on Gemma3-4B, we show that misalignment scales monotonically with LoRA rank, and that multimodal evaluation reveals substantially higher misalignment ($70.71 \pm 1.22$ at $r=128$) than text-only evaluation ($41.19 \pm 2.51$), suggesting that unimodal safety benchmarks may underestimate alignment degradation in vision-language models. Critically, even 10\% harmful data in the training mixture induces substantial alignment degradation. Geometric analysis reveals that harmful behaviors occupy a remarkably low-dimensional subspace, with the majority of misalignment information captured in 10 principal components. To mitigate misalignment, we evaluate two strategies: benign narrow fine-tuning and activation-based steering. While both approaches substantially reduce misalignment, neither completely removes the learned harmful behaviors. Our findings highlight the need for robust continual learning frameworks, as current post-training paradigms may not sufficiently preserve alignment in post-deployment settings.

Prompting and steering techniques are well established in general-purpose generative AI, yet assistive visual question answering (VQA) tools for blind users still follow rigid interaction patterns with limited opportunities for customization. User control can be helpful when system responses are misaligned with their goals and contexts, a gap that becomes especially consequential for blind users that may rely on these systems for access. We invite 11 blind users to customize their interactions with a real-world conversational VQA system. Drawing on 418 interactions, reflections, and post-study interviews, we analyze prompting-based techniques participants adopted, including those introduced in the study and those developed independently in real-world settings. VQA interactions were often lengthy: participants averaged 3 turns, sometimes up to 21, with input text typically tenfold shorter than the responses they heard. Built on state-of-the-art LLMs, the system lacked verbosity controls, was limited in estimating distance in space and time, relied on inaccessible image framing, and offered little to no camera guidance. We discuss how customization techniques such as prompt engineering can help participants work around these limitations. Alongside a new publicly available dataset, we offer insights for interaction design at both query and system levels.

Much of the advancement of Multi-Agent Reinforcement Learning (MARL) in imperfect-information games has historically depended on manual iterative refinement of baselines. While foundational families like Counterfactual Regret Minimization (CFR) and Policy Space Response Oracles (PSRO) rest on solid theoretical ground, the design of their most effective variants often relies on human intuition to navigate a vast algorithmic design space. In this work, we propose the use of AlphaEvolve, an evolutionary coding agent powered by large language models, to automatically discover new multiagent learning algorithms. We demonstrate the generality of this framework by evolving novel variants for two distinct paradigms of game-theoretic learning. First, in the domain of iterative regret minimization, we evolve the logic governing regret accumulation and policy derivation, discovering a new algorithm, Volatility-Adaptive Discounted (VAD-)CFR. VAD-CFR employs novel, non-intuitive mechanisms-including volatility-sensitive discounting, consistency-enforced optimism, and a hard warm-start policy accumulation schedule-to outperform state-of-the-art baselines like Discounted Predictive CFR+. Second, in the regime of population based training algorithms, we evolve training-time and evaluation-time meta strategy solvers for PSRO, discovering a new variant, Smoothed Hybrid Optimistic Regret (SHOR-)PSRO. SHOR-PSRO introduces a hybrid meta-solver that linearly blends Optimistic Regret Matching with a smoothed, temperature-controlled distribution over best pure strategies. By dynamically annealing this blending factor and diversity bonuses during training, the algorithm automates the transition from population diversity to rigorous equilibrium finding, yielding superior empirical convergence compared to standard static meta-solvers.

We study dynamic joint assortment and pricing where a seller updates decisions at regular accounting/operating intervals to maximize the cumulative per-period revenue over a horizon $T$. In many settings, assortment and prices affect not only what an arriving customer buys but also how many customers arrive within the period, whereas classical multinomial logit (MNL) models assume arrivals as fixed, potentially leading to suboptimal decisions. We propose a Poisson-MNL model that couples a contextual MNL choice model with a Poisson arrival model whose rate depends on the offered assortment and prices. Building on this model, we develop an efficient algorithm PMNL based on the idea of upper confidence bound (UCB). We establish its (near) optimality by proving a non-asymptotic regret bound of order $\sqrt{T\log{T}}$ and a matching lower bound (up to $\log T$). Simulation studies underscore the importance of accounting for the dependency of arrival rates on assortment and pricing: PMNL effectively learns customer choice and arrival models and provides joint assortment-pricing decisions that outperform others that assume fixed arrival rates.

We present Xray-Visual, a unified vision model architecture for large-scale image and video understanding trained on industry-scale social media data. Our model leverages over 15 billion curated image-text pairs and 10 billion video-hashtag pairs from Facebook and Instagram, employing robust data curation pipelines that incorporate balancing and noise suppression strategies to maximize semantic diversity while minimizing label noise. We introduce a three-stage training pipeline that combines self-supervised MAE, semi-supervised hashtag classification, and CLIP-style contrastive learning to jointly optimize image and video modalities. Our architecture builds on a Vision Transformer backbone enhanced with efficient token reorganization (EViT) for improved computational efficiency. Extensive experiments demonstrate that Xray-Visual achieves state-of-the-art performance across diverse benchmarks, including ImageNet for image classification, Kinetics and HMDB51 for video understanding, and MSCOCO for cross-modal retrieval. The model exhibits strong robustness to domain shift and adversarial perturbations. We further demonstrate that integrating large language models as text encoders (LLM2CLIP) significantly enhances retrieval performance and generalization capabilities, particularly in real-world environments. Xray-Visual establishes new benchmarks for scalable, multimodal vision models, while maintaining superior accuracy and computational efficiency.

Modeling of longitudinal cohort data typically involves complex temporal dependencies between multiple variables. There, the transformer architecture, which has been highly successful in language and vision applications, allows us to account for the fact that the most recently observed time points in an individual's history may not always be the most important for the immediate future. This is achieved by assigning attention weights to observations of an individual based on a transformation of their values. One reason why these ideas have not yet been fully leveraged for longitudinal cohort data is that typically, large datasets are required. Therefore, we present a simplified transformer architecture that retains the core attention mechanism while reducing the number of parameters to be estimated, to be more suitable for small datasets with few time points. Guided by a statistical perspective on transformers, we use an autoregressive model as a starting point and incorporate attention as a kernel-based operation with temporal decay, where aggregation of multiple transformer heads, i.e. different candidate weighting schemes, is expressed as accumulating evidence on different types of underlying characteristics of individuals. This also enables a permutation-based statistical testing procedure for identifying contextual patterns. In a simulation study, the approach is shown to recover contextual dependencies even with a small number of individuals and time points. In an application to data from a resilience study, we identify temporal patterns in the dynamics of stress and mental health. This indicates that properly adapted transformers can not only achieve competitive predictive performance, but also uncover complex context dependencies in small data settings.

Janus is a paradigmatic example of reversible programming language. Indeed, Janus programs can be executed backwards as well as forwards. However, its small-step semantics (useful, e.g., for debugging or as a basis for extensions with concurrency primitives) is not reversible, since it loses information while computing forwards. E.g., it does not satisfy the Loop Lemma, stating that any reduction has an inverse, a main property of reversibility in process calculi, where small-step semantics is commonly used. We present here a novel small-step semantics which is actually reversible, while remaining equivalent to the previous one. It involves the non-trivial challenge of defining a semantics based on a "program counter" for a high-level programming language.

Allegro is a machine learning interatomic potential (MLIP) model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason we apply multi-objective hyperparameter optimization to the two objectives. Additionally, we experiment with modified architectures by making variants of Allegro some by adding strictly classical multi-layer perceptron (MLP) layers and some by adding quantum-classical hybrid layers. We compare the results from QM9, rMD17-aspirin, rMD17-benzene and our own proprietary dataset consisting of copper and lithium atoms. As results, we have a list of variants that surpass the Allegro in accuracy and also results which demonstrate the trade-off with inference times.

In the context of asynchronous concurrent shared-memory systems, a snapshot algorithm allows failure-prone processes to concurrently and atomically write on the entries of a shared array MEM , and also atomically read the whole array. Recently, Read-Modify-Writable (RMWable) snapshot was proposed, a variant of snapshot that allows processes to perform operations more complex than just read and write, specifically, each entry MEM[k] is an arbitrary readable object. The known RMWable snapshot algorithms heavily rely on powerful low-level operations such as compare&swap or load-link/store-conditional to correctly produce snapshots of MEM. Following the large body of research devoted to understand the limits of what can be solved using the simple read/write low-level operations, which are known to be strictly weaker than compare&swap and load-link/store-conditional, we explore if RMWable snapshots are possible using only read/write operations. We present two read/write RMWable snapshot algorithms, the first one in the standard concurrent shared-memory model where the number of processes n is finite and known in advance, and the second one in a variant of the standard model with unbounded concurrency, where there are infinitely many processes, but at any moment only finitely many processes participate in an execution.

We introduce LLM-Wikirace, a benchmark for evaluating planning, reasoning, and world knowledge in large language models (LLMs). In LLM-Wikirace, models must efficiently navigate Wikipedia hyperlinks step by step to reach a target page from a given source, requiring look-ahead planning and the ability to reason about how concepts are connected in the real world. We evaluate a broad set of open- and closed-source models, including Gemini-3, GPT-5, and Claude Opus 4.5, which achieve the strongest results on the easy level of the task and demonstrate superhuman performance. Despite this, performance drops sharply on hard difficulty: the best-performing model, Gemini-3, succeeds in only 23\% of hard games, highlighting substantial remaining challenges for frontier models. Our analysis shows that world knowledge is a necessary ingredient for success, but only up to a point, beyond this threshold, planning and long-horizon reasoning capabilities become the dominant factors. Trajectory-level analysis further reveals that even the strongest models struggle to replan after failure, frequently entering loops rather than recovering. LLM-Wikirace is a simple benchmark that reveals clear limitations in current reasoning systems, offering an open arena where planning-capable LLMs still have much to prove. Our code and leaderboard available at https:/llmwikirace.github.io.

LLM agents are increasingly deployed in long-horizon, complex environments to solve challenging problems, but this expansion exposes them to long-horizon attacks that exploit multi-turn user-agent-environment interactions to achieve objectives infeasible in single-turn settings. To measure agent vulnerabilities to such risks, we present AgentLAB, the first benchmark dedicated to evaluating LLM agent susceptibility to adaptive, long-horizon attacks. Currently, AgentLAB supports five novel attack types including intent hijacking, tool chaining, task injection, objective drifting, and memory poisoning, spanning 28 realistic agentic environments, and 644 security test cases. Leveraging AgentLAB, we evaluate representative LLM agents and find that they remain highly susceptible to long-horizon attacks; moreover, defenses designed for single-turn interactions fail to reliably mitigate long-horizon threats. We anticipate that AgentLAB will serve as a valuable benchmark for tracking progress on securing LLM agents in practical settings. The benchmark is publicly available at https://tanqiujiang.github.io/AgentLAB_main.

Task planning for robotic manipulation with large language models (LLMs) is an emerging area. Prior approaches rely on specialized models, fine tuning, or prompt tuning, and often operate in an open loop manner without robust environmental feedback, making them fragile in dynamic settings.We present MALLVi, a Multi Agent Large Language and Vision framework that enables closed loop feedback driven robotic manipulation. Given a natural language instruction and an image of the environment, MALLVi generates executable atomic actions for a robot manipulator. After action execution, a Vision Language Model (VLM) evaluates environmental feedback and decides whether to repeat the process or proceed to the next step.Rather than using a single model, MALLVi coordinates specialized agents, Decomposer, Localizer, Thinker, and Reflector, to manage perception, localization, reasoning, and high level planning. An optional Descriptor agent provides visual memory of the initial state. The Reflector supports targeted error detection and recovery by reactivating only relevant agents, avoiding full replanning.Experiments in simulation and real world settings show that iterative closed loop multi agent coordination improves generalization and increases success rates in zero shot manipulation tasks.Code available at https://github.com/iman1234ahmadi/MALLVI.

Agent development kits (ADKs) provide effective platforms and tooling for constructing agents, and their designs are critical to the constructed agents' performance, especially the functionality for agent topology, tools, and memory. However, current ADKs either lack sufficient functional support or rely on humans to manually design these components, limiting agents' generalizability and overall performance. We propose OpenSage, the first ADK that enables LLMs to automatically create agents with self-generated topology and toolsets while providing comprehensive and structured memory support. OpenSage offers effective functionality for agents to create and manage their own sub-agents and toolkits. It also features a hierarchical, graph-based memory system for efficient management and a specialized toolkit tailored to software engineering tasks. Extensive experiments across three state-of-the-art benchmarks with various backbone models demonstrate the advantages of OpenSage over existing ADKs. We also conduct rigorous ablation studies to demonstrate the effectiveness of our design for each component. We believe OpenSage can pave the way for the next generation of agent development, shifting the focus from human-centered to AI-centered paradigms.

To build a dementia classification model for middle-aged and elderly Brazilians, implemented in Python, combining variable selection and multivariable analysis, using low-cost variables with modification potential. Observational study with a predictive modeling approach using a cross-sectional design, aimed at estimating the chances of developing dementia, using data from the Brazilian Longitudinal Study of Aging (ELSI-Brazil), involving 9,412 participants. Dementia was determined based on neuropsychological assessment and informant-based cognitive function. Analyses were performed using Random Forest (RF) and multivariable logistic regression to estimate the risk of dementia in the middle-aged and elderly populations of Brazil. The prevalence of dementia was 9.6%. The highest odds of dementia were observed in illiterate individuals (Odds Ratio (OR) = 7.42), individuals aged 90 years or older (OR = 11.00), low weight (OR = 2.11), low handgrip strength (OR = 2.50), self-reported black skin color (OR = 1.47), physical inactivity (OR = 1.61), self-reported hearing loss (OR = 1.65), and presence of depressive symptoms (OR = 1.72). Higher education (OR=0.44), greater life satisfaction (OR=0.72), and being employed (OR=0.78) were protective factors. The RF model outperformed logistic regression, achieving an area under the ROC curve of 0.776, with a sensitivity of 0.708, a specificity of 0.702, an F1-score of 0.311, a G-means of 0.705, and an accuracy of 0.703. Conclusion: The findings reinforce the multidimensional nature of dementia and the importance of accessible factors for identifying vulnerable individuals. Strengthening public policies focused on promoting brain health can contribute significantly to the efficient allocation of resources in primary care and dementia prevention in Brazil

Modern datasets often contain ballast as redundant or low-utility information that increases dimensionality, storage requirements, and computational cost without contributing meaningful analytical value. This study introduces a generalized, multimodal framework for ballast detection and reduction across structured, semi-structured, unstructured, and sparse data types. Using diverse datasets, entropy, mutual information, Lasso, SHAP, PCA, topic modelling, and embedding analysis are applied to identify and eliminate ballast features. A novel Ballast Score is proposed to integrate these signals into a unified, cross-modal pruning strategy. Experimental results demonstrate that significant portions of the feature space as often exceeding 70% in sparse or semi-structured data, can be pruned with minimal or even improved classification performance, along with substantial reductions in training time and memory footprint. The framework reveals distinct ballast typologies (e.g. statistical, semantic, infrastructural), and offers practical guidance for leaner, more efficient machine learning pipelines.

As large language models from diverse providers converge toward comparable benchmark performance, the traditional paradigm of selecting a single best model per task yields diminishing returns. We argue that orchestration topology -- the structural composition of how multiple agents are coordinated, parallelized, and synthesized -- now dominates system-level performance over individual model capability. We present AdaptOrch, a formal framework for task-adaptive multi-agent orchestration that dynamically selects among four canonical topologies (parallel, sequential, hierarchical, and hybrid) based on task dependency graphs and empirically derived domain characteristics. Our framework introduces three key contributions: (1) a Performance Convergence Scaling Law, formalizing conditions under which orchestration selection outweighs model selection; (2) a Topology Routing Algorithm that maps task decomposition DAGs to optimal orchestration patterns in O(|V| + |E|) time; and (3) an Adaptive Synthesis Protocol with provable termination guarantees and heuristic consistency scoring for parallel agent outputs. We validate AdaptOrch across coding (SWE-bench), reasoning (GPQA), and retrieval-augmented generation tasks, demonstrating that topology-aware orchestration achieves 12-23% improvement over static single-topology baselines, even when using identical underlying models. Our results establish orchestration design as a first-class optimization target independent of model scaling.

Time series (TS) modeling has come a long way from early statistical, mainly linear, approaches to the current trend in TS foundation models. With a lot of hype and industrial demand in this field, it is not always clear how much progress there really is. To advance TS forecasting and analysis to the next level, here we argue that the field needs a dynamical systems (DS) perspective. TS of observations from natural or engineered systems almost always originate from some underlying DS, and arguably access to its governing equations would yield theoretically optimal forecasts. This is the promise of DS reconstruction (DSR), a class of ML/AI approaches that aim to infer surrogate models of the underlying DS from data. But models based on DS principles offer other profound advantages: Beyond short-term forecasts, they enable to predict the long-term statistics of an observed system, which in many practical scenarios may be the more relevant quantities. DS theory furthermore provides domain-independent theoretical insight into mechanisms underlying TS generation, and thereby will inform us, e.g., about upper bounds on performance of any TS model, generalization into unseen regimes as in tipping points, or potential control strategies. After reviewing some of the central concepts, methods, measures, and models in DS theory and DSR, we will discuss how insights from this field can advance TS modeling in crucial ways, enabling better forecasting with much lower computational and memory footprints. We conclude with a number of specific suggestions for translating insights from DSR into TS modeling.

The ability to manipulate tools significantly expands the set of tasks a robot can perform. Yet, tool manipulation represents a challenging class of dexterity, requiring grasping thin objects, in-hand object rotations, and forceful interactions. Since collecting teleoperation data for these behaviors is challenging, sim-to-real reinforcement learning (RL) is a promising alternative. However, prior approaches typically require substantial engineering effort to model objects and tune reward functions for each task. In this work, we propose SimToolReal, taking a step towards generalizing sim-to-real RL policies for tool manipulation. Instead of focusing on a single object and task, we procedurally generate a large variety of tool-like object primitives in simulation and train a single RL policy with the universal goal of manipulating each object to random goal poses. This approach enables SimToolReal to perform general dexterous tool manipulation at test-time without any object or task-specific training. We demonstrate that SimToolReal outperforms prior retargeting and fixed-grasp methods by 37% while matching the performance of specialist RL policies trained on specific target objects and tasks. Finally, we show that SimToolReal generalizes across a diverse set of everyday tools, achieving strong zero-shot performance over 120 real-world rollouts spanning 24 tasks, 12 object instances, and 6 tool categories.

The deployment of deep learning inference in production environments continues to grow, where throughput, latency, and hardware efficiency are critical. Although specialized accelerators are increasingly adopted, many inference workloads still run on CPU-only systems, particularly in legacy data centers and cost-sensitive environments. This study investigates the scalability limits of CPU-based inference for convolutional neural networks by benchmarking ResNet models across varying batch sizes on two hardware tiers: a legacy Intel Xeon E5-2403 v2 processor and a modern Intel Xeon 6 "Granite Rapids" platform. Results show that legacy CPUs quickly reach throughput saturation, with limited scaling beyond small batch sizes due to instruction-level and memory constraints. In contrast, the Granite Rapids system leverages Intel Advanced Matrix Extensions (AMX) to achieve substantially higher throughput. However, oversubscription beyond physical core limits introduces execution contention and tail-latency amplification, revealing a performance degradation regime in modern architectures. We introduce GDEV-AI, a reproducible benchmarking framework for analyzing scalability behavior and architectural saturation in CPU-based inference. By establishing a vendor-neutral baseline, this work provides empirical insight into performance bottlenecks and informs capacity planning in heterogeneous data center environments.

Meenzerisch, the dialect spoken in the German city of Mainz, is also the traditional language of the Mainz carnival, a yearly celebration well known throughout Germany. However, Meenzerisch is on the verge of dying out-a fate it shares with many other German dialects. Natural language processing (NLP) has the potential to help with the preservation and revival efforts of languages and dialects. However, so far no NLP research has looked at Meenzerisch. This work presents the first research in the field of NLP that is explicitly focused on the dialect of Mainz. We introduce a digital dictionary-an NLP-ready dataset derived from an existing resource (Schramm, 1966)-to support researchers in modeling and benchmarking the language. It contains 2,351 words in the dialect paired with their meanings described in Standard German. We then use this dataset to answer the following research questions: (1) Can state-of-the-art large language models (LLMs) generate definitions for dialect words? (2) Can LLMs generate words in Meenzerisch, given their definitions? Our experiments show that LLMs can do neither: the best model for definitions reaches only 6.27% accuracy and the best word generation model's accuracy is 1.51%. We then conduct two additional experiments in order to see if accuracy is improved by few-shot learning and by extracting rules from the training set, which are then passed to the LLM. While those approaches are able to improve the results, accuracy remains below 10%. This highlights that additional resources and an intensification of research efforts focused on German dialects are desperately needed.

We present a comprehensive analysis of how two-layer neural networks learn features to solve the modular addition task. Our work provides a full mechanistic interpretation of the learned model and a theoretical explanation of its training dynamics. While prior work has identified that individual neurons learn single-frequency Fourier features and phase alignment, it does not fully explain how these features combine into a global solution. We bridge this gap by formalizing a diversification condition that emerges during training when overparametrized, consisting of two parts: phase symmetry and frequency diversification. We prove that these properties allow the network to collectively approximate a flawed indicator function on the correct logic for the modular addition task. While individual neurons produce noisy signals, the phase symmetry enables a majority-voting scheme that cancels out noise, allowing the network to robustly identify the correct sum. Furthermore, we explain the emergence of these features under random initialization via a lottery ticket mechanism. Our gradient flow analysis proves that frequencies compete within each neuron, with the "winner" determined by its initial spectral magnitude and phase alignment. From a technical standpoint, we provide a rigorous characterization of the layer-wise phase coupling dynamics and formalize the competitive landscape using the ODE comparison lemma. Finally, we use these insights to demystify grokking, characterizing it as a three-stage process involving memorization followed by two generalization phases, driven by the competition between loss minimization and weight decay.

To enable human oversight, agentic AI systems often provide a trace of reasoning and action steps. Designing traces to have an informative, but not overwhelming, level of detail remains a critical challenge. In three user studies on a Computer User Agent, we investigate the utility of basic action traces for verification, explore three alternatives via design probes, and test a novel interface's impact on error finding in question-answering tasks. As expected, we find that current practices are cumbersome, limiting their efficacy. Conversely, our proposed design reduced the time participants spent finding errors. However, although participants reported higher levels of confidence in their decisions, their final accuracy was not meaningfully improved. To this end, our study surfaces challenges for human verification of agentic systems, including managing built-in assumptions, users' subjective and changing correctness criteria, and the shortcomings, yet importance, of communicating the agent's process.

Evaluating factual consistency is essential for reliable text summarization, particularly in high-stakes domains such as healthcare and news. However, most existing evaluation metrics overlook Bangla, a widely spoken yet under-resourced language, and often depend on reference summaries. We introduce BanglaSummEval, a reference-free, question-answering-based framework for evaluating factual consistency in Bangla summarization. The proposed method assesses both factual accuracy and content coverage through automatically generated questions and answers derived from the source document and the summary. A single multilingual instruction-tuned language model handles question generation, question answering, candidate answer extraction, and question importance weighting. This unified design reduces system complexity and computational cost. To capture semantic consistency beyond surface-level overlap, we use BERTScore-Recall for answer comparison. We validate BanglaSummEval on 300 human-written summaries from educational and medical domains, demonstrating strong correlation with expert human judgments (Pearson's $r = 0.694$, Spearman's $ρ= 0.763$). By providing interpretable, step-wise diagnostics alongside reliable evaluation scores, BanglaSummEval offers a practical and transparent solution for factual consistency evaluation in low-resource language settings.

We study the offline data-driven newsvendor problem with censored demand data. In contrast to prior works where demand is fully observed, we consider the setting where demand is censored at the inventory level and only sales are observed; sales match demand when there is sufficient inventory, and equal the available inventory otherwise. We provide a general procedure to compute the exact worst-case regret of classical data-driven inventory policies, evaluated over all demand distributions. Our main technical result shows that this infinite-dimensional, non-convex optimization problem can be reduced to a finite-dimensional one, enabling an exact characterization of the performance of policies for any sample size and censoring levels. We leverage this reduction to derive sharp insights on the achievable performance of standard inventory policies under demand censoring. In particular, our analysis of the Kaplan-Meier policy shows that while demand censoring fundamentally limits what can be learned from passive sales data, just a small amount of targeted exploration at high inventory levels can substantially improve worst-case guarantees, enabling near-optimal performance even under heavy censoring. In contrast, when the point-of-sale system does not record stockout events and only reports realized sales, a natural and commonly used approach is to treat sales as demand. Our results show that policies based on this sales-as-demand heuristic can suffer severe performance degradation as censored data accumulates, highlighting how the quality of point-of-sale information critically shapes what can, and cannot, be learned offline.

Large reasoning models (LRMs) excel on complex problems but face a critical barrier to efficiency: reinforcement learning (RL) training requires long rollouts for outcome-based rewards, where autoregressive decoding dominates time and memory usage. While sliding-window cache strategies can bound memory, they disrupt long-context reasoning and degrade performance. We introduce Progressive Thought Encoding, a parameter-efficient fine-tuning method that enables LRMs to reason effectively under fixed-size caches. By progressively encoding intermediate reasoning into fixed-size vector representations, our approach eliminates the need to backpropagate through full-cache rollouts, thereby reducing memory usage, while maintaining constant memory during inference. Experiments on three models, including Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct, and DeepSeek-R1-Distill-Llama-8B, on six widely used challenging mathematical benchmarks show consistent gains: our method achieves +19.3% improvement over LoRA-based fine-tuning and +29.9% over LRMs without fine-tuning on average, with up to +23.4 accuracy improvement on AIME2024/2025 under the same tight cache budgets. These results demonstrate that Progressive Thought Encoding not only improves reasoning accuracy but also makes RL training of LRMs substantially more efficient and scalable under real-world memory constraints.

Transformer models systematically favor certain token positions, yet the architectural origins of this position bias remain poorly understood. Under causal masking at infinite depth, prior theoretical analyses of attention rollout predict an inevitable collapse of attention onto the first token. Such collapse, however, does not occur in practice. We resolve this discrepancy with a residual-aware theory of cumulative attention rollout. By incorporating residual connections, we show that this architectural component prevents collapse under realistic conditions. At finite depth, we prove that causal Transformers induce a U-shaped position bias, with attention concentrating on early and late tokens. This result provides a principled architectural explanation for the Lost-in-the-Middle phenomenon.

While Large Language Models (LLMs) have achieved strong performance on general-purpose language tasks, their deployment in regulated and data-sensitive domains, including insurance, remains limited. Leveraging millions of historical warranty claims, we propose a locally deployed governance-aware language modeling component that generates structured corrective-action recommendations from unstructured claim narratives. We fine-tune pretrained LLMs using Low-Rank Adaptation (LoRA), scoping the model to an initial decision module within the claim processing pipeline to speed up claim adjusters' decisions. We assess this module using a multi-dimensional evaluation framework that combines automated semantic similarity metrics with human evaluation, enabling a rigorous examination of both practical utility and predictive accuracy. Our results show that domain-specific fine-tuning substantially outperforms commercial general-purpose and prompt-based LLMs, with approximately 80% of the evaluated cases achieving near-identical matches to ground-truth corrective actions. Overall, this study provides both theoretical and empirical evidence to prove that domain-adaptive fine-tuning can align model output distributions more closely with real-world operational data, demonstrating its promise as a reliable and governable building block for insurance applications.

Safety alignment is essential for the responsible deployment of large language models (LLMs). Yet, existing approaches often rely on heavyweight fine-tuning that is costly to update, audit, and maintain across model families. Full fine-tuning incurs substantial computational and storage overhead, while parameter-efficient methods such as LoRA trade efficiency for inconsistent safety gains and sensitivity to design choices. Safety intervention mechanisms such as circuit breakers reduce unsafe outputs without modifying model weights, but do not directly shape or preserve the internal representations that govern safety behavior. These limitations hinder rapid and reliable safety updates, particularly in settings where models evolve frequently or must adapt to new policies and domains. We present NeST, a lightweight, structure-aware safety alignment framework that strengthens refusal behavior by selectively adapting a small subset of safety-relevant neurons while freezing the remainder of the model. NeST aligns parameter updates with the internal organization of safety behavior by clustering functionally coherent safety neurons and enforcing shared updates within each cluster, enabling targeted and stable safety adaptation without broad model modification or inference-time overhead. We benchmark NeST against three dominant baselines: full fine-tuning, LoRA-based fine-tuning, and circuit breakers across 10 open-weight LLMs spanning multiple model families and sizes. Across all evaluated models, NeST reduces the attack success rate from an average of 44.5% to 4.36%, corresponding to a 90.2% reduction in unsafe generations, while requiring only 0.44 million trainable parameters on average. This amounts to a 17,310x decrease in updated parameters compared to full fine-tuning and a 9.25x reduction relative to LoRA, while consistently achieving stronger safety performance for alignment.

Exploration remains a key bottleneck for reinforcement learning (RL) post-training of large language models (LLMs), where sparse feedback and large action spaces can lead to premature collapse into repetitive behaviors. We propose Verbalized Action Masking (VAM), which verbalizes an action mask in the prompt and enforces that the model outputs an action from the masked set. Building on this interface, we introduce iterative action-space pruning: if the target action is not sampled, we remove valid sampled actions from the mask and resample under the reduced candidate set, repeating until the target is sampled or a fixed budget is exhausted. We study VAM in chess and evaluate it under two training regimes: an engine-play regime that generates states via play against an engine opponent and a fixed-dataset regime that trains from a fixed dataset of positions with verifier scores. Across held-out chess puzzles and full-game play measured by average centipawn loss (ACPL), VAM improves learning efficiency and final performance over strong baselines, highlighting verbalized masking as a practical mechanism for controllable exploration in LLM RL post-training.

Safety alignment of large language models (LLMs) is mostly evaluated in English and contract-bound, leaving multilingual vulnerabilities understudied. We introduce \textbf{Indic Jailbreak Robustness (IJR)}, a judge-free benchmark for adversarial safety across 12 Indic and South Asian languages (2.1 Billion speakers), covering 45216 prompts in JSON (contract-bound) and Free (naturalistic) tracks. IJR reveals three patterns. (1) Contracts inflate refusals but do not stop jailbreaks: in JSON, LLaMA and Sarvam exceed 0.92 JSR, and in Free all models reach 1.0 with refusals collapsing. (2) English to Indic attacks transfer strongly, with format wrappers often outperforming instruction wrappers. (3) Orthography matters: romanized or mixed inputs reduce JSR under JSON, with correlations to romanization share and tokenization (approx 0.28 to 0.32) indicating systematic effects. Human audits confirm detector reliability, and lite-to-full comparisons preserve conclusions. IJR offers a reproducible multilingual stress test revealing risks hidden by English-only, contract-focused evaluations, especially for South Asian users who frequently code-switch and romanize.

This study addresses a central tactical dilemma for football coaches: whether to employ a defensive strategy, colloquially known as "parking the bus", or a more offensive one. Using an advanced Double Machine Learning (DML) framework, this project provides a robust and interpretable tool to estimate the causal impact of different formations on key match outcomes such as goal difference, possession, corners, and disciplinary actions. Leveraging a dataset of over 22,000 matches from top European leagues, formations were categorized into six representative types based on tactical structure and expert consultation. A major methodological contribution lies in the adaptation of DML to handle categorical treatments, specifically formation combinations, through a novel matrix-based residualization process, allowing for a detailed estimation of formation-versus-formation effects that can inform a coach's tactical decision-making. Results show that while offensive formations like 4-3-3 and 4-2-3-1 offer modest statistical advantages in possession and corners, their impact on goals is limited. Furthermore, no evidence supports the idea that defensive formations, commonly associated with parking the bus, increase a team's winning potential. Additionally, red cards appear unaffected by formation choice, suggesting other behavioral factors dominate. Although this approach does not fully capture all aspects of playing style or team strength, it provides a valuable framework for coaches to analyze tactical efficiency and sets a precedent for future research in sports analytics.

When feedback is absorbed faster than task structure can be evaluated, the learner will favor feedback over truth. A two-timescale model shows this feedback-truth gap is inevitable whenever the two rates differ and vanishes only when they match. We test this prediction across neural networks trained with noisy labels (30 datasets, 2,700 runs), human probabilistic reversal learning (N = 292), and human reward/punishment learning with concurrent EEG (N = 25). In each system, truth is defined operationally: held-out labels, the objectively correct option, or the participant's pre-feedback expectation - the only non-circular reference decodable from post-feedback EEG. The gap appeared universally but was regulated differently: dense networks accumulated it as memorization; sparse-residual scaffolding suppressed it; humans generated transient over-commitment that was actively recovered. Neural over-commitment (~0.04-0.10) was amplified tenfold into behavioral commitment (d = 3.3-3.9). The gap is a fundamental constraint on learning under noisy supervision; its consequences depend on the regulation each system employs.

Traditional scoring approaches on hesitant fuzzy sets often lack a formal base in order theory. This paper proposes a unified framework, where each score is explicitly defined with respect to a given order. This order-oriented perspective enables more flexible and coherent scoring mechanisms. We examine several classical orders on hesitant fuzzy elements, that is, nonempty subsets in [0,1], and show that, contrary to prior claims, they do not induce lattice structures. In contrast, we prove that the scores defined with respect to the symmetric order satisfy key normative criteria for scoring functions, including strong monotonicity with respect to unions and the Gärdenfors condition. Following this analysis, we introduce a class of functions, called dominance functions, for ranking hesitant fuzzy elements. They aim to compare hesitant fuzzy elements relative to control sets incorporating minimum acceptability thresholds. Two concrete examples of dominance functions for finite sets are provided: the discrete dominance function and the relative dominance function. We show that these can be employed to construct fuzzy preference relations on typical hesitant fuzzy sets and support group decision-making.

Theory of mind (ToM) enables AI systems to infer agents' hidden goals and mental states, but existing approaches focus mainly on small human understandable gridworld spaces. We introduce HiVAE, a hierarchical variational architecture that scales ToM reasoning to realistic spatiotemporal domains. Inspired by the belief-desire-intention structure of human cognition, our three-level VAE hierarchy achieves substantial performance improvements on a 3,185-node campus navigation task. However, we identify a critical limitation: while our hierarchical structure improves prediction, learned latent representations lack explicit grounding to actual mental states. We propose self-supervised alignment strategies and present this work to solicit community feedback on grounding approaches.

*Automated circuit discovery* is a central tool in mechanistic interpretability for identifying the internal components of neural networks responsible for specific behaviors. While prior methods have made significant progress, they typically depend on heuristics or approximations and do not offer provable guarantees over continuous input domains for the resulting circuits. In this work, we leverage recent advances in neural network verification to propose a suite of automated algorithms that yield circuits with *provable guarantees*. We focus on three types of guarantees: (1) *input domain robustness*, ensuring the circuit agrees with the model across a continuous input region; (2) *robust patching*, certifying circuit alignment under continuous patching perturbations; and (3) *minimality*, formalizing and capturing a wide array of various notions of succinctness. Interestingly, we uncover a diverse set of novel theoretical connections among these three families of guarantees, with critical implications for the convergence of our algorithms. Finally, we conduct experiments with state-of-the-art verifiers on various vision models, showing that our algorithms yield circuits with substantially stronger robustness guarantees than standard circuit discovery methods, establishing a principled foundation for provable circuit discovery.

We propose TopoFlow (Topography-aware pollutant Flow learning), a physics-guided neural network for efficient, high-resolution air quality prediction. To explicitly embed physical processes into the learning framework, we identify two critical factors governing pollutant dynamics: topography and wind direction. Complex terrain can channel, block, and trap pollutants, while wind acts as a primary driver of their transport and dispersion. Building on these insights, TopoFlow leverages a vision transformer architecture with two novel mechanisms: topography-aware attention, which explicitly models terrain-induced flow patterns, and wind-guided patch reordering, which aligns spatial representations with prevailing wind directions. Trained on six years of high-resolution reanalysis data assimilating observations from over 1,400 surface monitoring stations across China, TopoFlow achieves a PM2.5 RMSE of 9.71 ug/m3, representing a 71-80% improvement over operational forecasting systems and a 13% improvement over state-of-the-art AI baselines. Forecast errors remain well below China's 24-hour air quality threshold of 75 ug/m3 (GB 3095-2012), enabling reliable discrimination between clean and polluted conditions. These performance gains are consistent across all four major pollutants and forecast lead times from 12 to 96 hours, demonstrating that principled integration of physical knowledge into neural networks can fundamentally advance air quality prediction.

Despite growing interest in using LLMs to generate feedback on students' writing, little is known about how students respond to AI-mediated versus human-provided feedback. We address this gap through a randomized controlled trial in a large introductory economics course (N=354), where we introduce and deploy FeedbackWriter - a system that generates AI suggestions to teaching assistants (TAs) while they provide feedback on students' knowledge-intensive essays. TAs have the full capacity to adopt, edit, or dismiss the suggestions. Students were randomly assigned to receive either handwritten feedback from TAs (baseline) or AI-mediated feedback where TAs received suggestions from FeedbackWriter. Students revise their drafts based on the feedback, which is further graded. In total, 1,366 essays were graded using the system. We found that students receiving AI-mediated feedback produced significantly higher-quality revisions, with gains increasing as TAs adopted more AI suggestions. TAs found the AI suggestions useful for spotting gaps and clarifying rubrics.

When assessing the quality of coding agents, predominant benchmarks focus on solving single issues on GitHub, such as SWE-Bench. In contrast, in real use, these agents solve more various and complex tasks that involve other skills such as exploring codebases, testing software, and designing architecture. In this paper, we first characterize some transferable skills that are shared across diverse tasks by decomposing trajectories into fine-grained components, and derive a set of principles for designing auxiliary training tasks to teach language models these skills. Guided by these principles, we propose a training environment, Hybrid-Gym, consisting of a set of scalable synthetic tasks, such as function localization and dependency search. Experiments show that agents trained on our synthetic tasks effectively generalize to diverse real-world tasks that are not present in training, improving a base model by 25.4% absolute gain on SWE-Bench Verified, 7.9% on SWT-Bench Verified, and 5.1% on Commit-0 Lite. Hybrid-Gym also complements datasets built for the downstream tasks (e.g., improving SWE-Play by 4.9% on SWT-Bench Verified). Code available at: https://github.com/yiqingxyq/Hybrid-Gym.

The expansion of AI toward the edge increasingly exposes the cost and fragility of cen- tralised intelligence. Data transmission, latency, energy consumption, and dependence on large data centres create bottlenecks that scale poorly across heterogeneous, mobile, and resource-constrained environments. In this paper, we introduce Node Learning, a decen- tralised learning paradigm in which intelligence resides at individual edge nodes and expands through selective peer interaction. Nodes learn continuously from local data, maintain their own model state, and exchange learned knowledge opportunistically when collaboration is beneficial. Learning propagates through overlap and diffusion rather than global synchro- nisation or central aggregation. It unifies autonomous and cooperative behaviour within a single abstraction and accommodates heterogeneity in data, hardware, objectives, and connectivity. This concept paper develops the conceptual foundations of this paradigm, contrasts it with existing decentralised approaches, and examines implications for communi- cation, hardware, trust, and governance. Node Learning does not discard existing paradigms, but places them within a broader decentralised perspective

Language models based on discrete diffusion have attracted widespread interest for their potential to provide faster generation than autoregressive models. In practice, however, they exhibit a sharp degradation of sample quality in the few-step regime, failing to realize this promise. Here we show that language models leveraging flow-based continuous denoising can outperform discrete diffusion in both quality and speed. By revisiting the fundamentals of flows over discrete modalities, we build a flow-based language model (FLM) that performs Euclidean denoising over one-hot token encodings. We show that the model can be trained by predicting the clean data via a cross entropy objective, where we introduce a simple time reparameterization that greatly improves training stability and generation quality. By distilling FLM into its associated flow map, we obtain a distilled flow map language model (FMLM) capable of few-step generation. On the LM1B and OWT language datasets, FLM attains generation quality matching state-of-the-art discrete diffusion models. With FMLM, our approach outperforms recent few-step language models across the board, with one-step generation exceeding their 8-step quality. Our work calls into question the widely held hypothesis that discrete diffusion processes are necessary for generative modeling over discrete modalities, and paves the way toward accelerated flow-based language modeling at scale. Code is available at https://github.com/david3684/flm.

Large-scale facilities increasingly face analysis and reporting latency as the limiting step in scientific throughput, particularly for structurally and magnetically complex samples that require iterative reduction, integration, refinement, and validation. To improve time-to-result and analysis efficiency, NeuDiff Agent is introduced as a governed, tool-using AI workflow for TOPAZ at the Spallation Neutron Source that takes instrument data products through reduction, integration, refinement, and validation to a validated crystal structure and a publication-ready CIF. NeuDiff Agent executes this established pipeline under explicit governance by restricting actions to allowlisted tools, enforcing fail-closed verification gates at key workflow boundaries, and capturing complete provenance for inspection, auditing, and controlled replay. Performance is assessed using a fixed prompt protocol and repeated end-to-end runs with two large language model backends, with user and machine time partitioned and intervention burden and recovery behaviors quantified under gating. In a reference-case benchmark, NeuDiff Agent reduces wall time from 435 minutes (manual) to 86.5(4.7) to 94.4(3.5) minutes (4.6-5.0x faster) while producing a validated CIF with no checkCIF level A or B alerts. These results establish a practical route to deploy agentic AI in facility crystallography while preserving traceability and publication-facing validation requirements.

Recent advancements in Natural Language Processing and Deep Learning have enabled the development of Large Language Models (LLMs), which have significantly advanced the state-of-the-art across a wide range of tasks, including Question Answering (QA). Despite these advancements, research on LLMs has primarily targeted high-resourced languages (e.g., English), and only recently has attention shifted toward multilingual models. However, these models demonstrate a training data bias towards a small number of popular languages or rely on transfer learning from high- to under-resourced languages; this may lead to a misrepresentation of social, cultural, and historical aspects. To address this challenge, monolingual LLMs have been developed for under-resourced languages; however, their effectiveness remains less studied when compared to multilingual counterparts on language-specific tasks. In this study, we address this research gap in Greek QA by contributing: (i) DemosQA, a novel dataset, which is constructed using social media user questions and community-reviewed answers to better capture the Greek social and cultural zeitgeist; (ii) a memory-efficient LLM evaluation framework adaptable to diverse QA datasets and languages; and (iii) an extensive evaluation of 11 monolingual and multilingual LLMs on 6 human-curated Greek QA datasets using 3 different prompting strategies. We release our code and data to facilitate reproducibility.

Code evolution is a family of techniques that rely on large language models to search through possible computer programs by evolving or mutating existing code. Many proposed code evolution pipelines show impressive performance but are often not compared to simpler baselines. We test how well two simple baselines do over three domains: finding better mathematical bounds, designing agentic scaffolds, and machine learning competitions. We find that simple baselines match or exceed much more sophisticated methods in all three. By analyzing these results we find various shortcomings in how code evolution is both developed and used. For the mathematical bounds, a problem's search space and domain knowledge in the prompt are chiefly what dictate a search's performance ceiling and efficiency, with the code evolution pipeline being secondary. Thus, the primary challenge in finding improved bounds is designing good search spaces, which is done by domain experts, and not the search itself. When designing agentic scaffolds we find that high variance in the scaffolds coupled with small datasets leads to suboptimal scaffolds being selected, resulting in hand-designed majority vote scaffolds performing best. We propose better evaluation methods that reduce evaluation stochasticity while keeping the code evolution economically feasible. We finish with a discussion of avenues and best practices to enable more rigorous code evolution in future work.

While Reinforcement Learning with Verifiable Rewards (RLVR) has shown strong effectiveness in reasoning tasks, it cannot be directly applied to non-verifiable domains lacking ground-truth verifiers, such as LLM alignment. In this work, we investigate whether reference-guided LLM-evaluators can bridge this gap by serving as soft "verifiers". First, we design evaluation protocols that enhance LLM-based evaluators for LLM alignment using reference outputs. Through comprehensive experiments, we show that a reference-guided approach substantially improves the accuracy of less capable LLM-judges using references from frontier models; stronger LLM-judges can also be enhanced by high-quality (i.e., human-written) references. Building on these improved judges, we demonstrate the utility of high-quality references in alignment tuning, where LLMs guided with references are used as judges to self-improve. We show that reference-guided self-improvement yields clear gains over both direct SFT on reference outputs and self-improvement with reference-free judges, achieving performance comparable to training with ArmoRM, a strong finetuned reward model. Specifically, our method achieves 73.1% and 58.7% on AlpacaEval and Arena-Hard with Llama-3-8B-Instruct, and 70.0% and 74.1% with Qwen2.5-7B, corresponding to average absolute gains of +20.2 / +17.1 points over SFT distillation and +5.3 / +3.6 points over reference-free self-improvement on AlpacaEval / Arena-Hard. These results highlight the potential of using reference-guided LLM-evaluators to enable effective LLM post-training in non-verifiable domains.

We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator. We then design attacks for the closed-world setting. Given two databases of pseudonymous individuals, each containing unstructured text written by or about that individual, we implement a scalable attack pipeline that uses LLMs to: (1) extract identity-relevant features, (2) search for candidate matches via semantic embeddings, and (3) reason over top candidates to verify matches and reduce false positives. Compared to prior deanonymization work (e.g., on the Netflix prize) that required structured data or manual feature engineering, our approach works directly on raw user content across arbitrary platforms. We construct three datasets with known ground-truth data to evaluate our attacks. The first links Hacker News to LinkedIn profiles, using cross-platform references that appear in the profiles. Our second dataset matches users across Reddit movie discussion communities; and the third splits a single user's Reddit history in time to create two pseudonymous profiles to be matched. In each setting, LLM-based methods substantially outperform classical baselines, achieving up to 68% recall at 90% precision compared to near 0% for the best non-LLM method. Our results show that the practical obscurity protecting pseudonymous users online no longer holds and that threat models for online privacy need to be reconsidered.

Fine-tuning pre-trained diffusion and flow models to optimize downstream utilities is central to real-world deployment. Existing entropy-regularized methods primarily maximize expected reward, providing no mechanism to shape tail behavior. However, tail control is often essential: the lower tail determines reliability by limiting low-reward failures, while the upper tail enables discovery by prioritizing rare, high-reward outcomes. In this work, we present Tail-aware Flow Fine-Tuning (TFFT), a principled and efficient distributional fine-tuning algorithm based on the Conditional Value-at-Risk (CVaR). We address two distinct tail-shaping goals: right-CVaR for seeking novel samples in the high-reward tail and left-CVaR for controlling worst-case samples in the low-reward tail. Unlike prior approaches that rely on non-linear optimization, we leverage the variational dual formulation of CVaR to decompose it into a decoupled two-stage procedure: a lightweight one-dimensional threshold optimization step, and a single entropy-regularized fine-tuning process via a specific pseudo-reward. This decomposition achieves CVaR fine-tuning efficiently with computational cost comparable to standard expected fine-tuning methods. We demonstrate the effectiveness of TFFT across illustrative experiments, high-dimensional text-to-image generation, and molecular design.

Conformal prediction (CP) offers distribution-free uncertainty quantification for machine learning models, yet its interplay with fairness in downstream decision-making remains underexplored. Moving beyond CP as a standalone operation (procedural fairness), we analyze the holistic decision-making pipeline to evaluate substantive fairness-the equity of downstream outcomes. Theoretically, we derive an upper bound that decomposes prediction-set size disparity into interpretable components, clarifying how label-clustered CP helps control method-driven contributions to unfairness. To facilitate scalable empirical analysis, we introduce an LLM-in-the-loop evaluator that approximates human assessment of substantive fairness across diverse modalities. Our experiments reveal that label-clustered CP variants consistently deliver superior substantive fairness. Finally, we empirically show that equalized set sizes, rather than coverage, strongly correlate with improved substantive fairness, enabling practitioners to design more fair CP systems. Our code is available at https://github.com/layer6ai-labs/llm-in-the-loop-conformal-fairness.

In the past year, custom and unreleased math reasoning models reached gold medal performance on the International Mathematical Olympiad (IMO). Similar performance was then reported using large-scale inference on publicly available models but at prohibitive costs (e.g., 3000 USD per problem). In this work, we present an inference pipeline that attains best-in-class performance on IMO-style math problems at an average inference cost orders of magnitude below competing methods while using only general-purpose off-the-shelf models. Our method relies on insights about grader failure in solver-grader pipelines, which we call the Cognitive Well (iterative refinement converging to a wrong solution that the solver as well as the pipeline's internal grader consider to be basically correct). Our pipeline addresses these failure modes through conjecture extraction, wherein candidate lemmas are isolated from generated solutions and independently verified alongside their negations in a fresh environment (context detachment). On IMO-ProofBench Advanced (PB-Adv), our pipeline achieves 67.1 percent performance using Gemini 3.0 Pro with an average cost per question of approximately 31 USD. At the time of evaluation, this represented the state-of-the-art on PB-Adv among both public and unreleased models, and more than doubles the success rate of the next best publicly accessible pipeline, all at a fraction of the cost.

Despite their strong performance on reasoning benchmarks, large language models (LLMs) have proven brittle when presented with counterfactual questions, suggesting weaknesses in their causal reasoning ability. While recent work has demonstrated that labeled counterfactual tasks can be useful benchmarks of LLMs' causal reasoning, producing such data at the scale required to cover the vast potential space of counterfactuals is limited. In this work, we introduce double counterfactual consistency (DCC), a lightweight inference-time method for measuring and guiding the ability of LLMs to reason causally. Without requiring labeled counterfactual data, DCC verifies a model's ability to execute two important elements of causal reasoning: causal intervention and counterfactual prediction. Using DCC, we evaluate the causal reasoning abilities of various leading LLMs across a range of reasoning tasks and interventions. Moreover, we demonstrate the effectiveness of DCC as a training-free test-time rejection sampling criterion and show that it can directly improve performance on reasoning tasks across multiple model families.

Despite their impressive performance on a wide variety of tasks, modern language models remain susceptible to distribution shifts, exhibiting brittle behavior when evaluated on data that differs in distribution from their training data. In this paper, we describe how distribution shifts in language models can be separated into observable and unobservable components, and we discuss how established approaches for dealing with distribution shift address only the former. Importantly, we identify that the resulting omitted variable bias from unobserved variables can compromise both evaluation and optimization in language models. To address this challenge, we introduce a framework that maps the strength of the omitted variables to bounds on the worst-case generalization performance of language models under distribution shift. In empirical experiments, we show that using these bounds directly in language model evaluation and optimization provides more principled measures of out-of-distribution performance, improves true out-of-distribution performance relative to standard distribution shift adjustment methods, and further enables inference about the strength of the omitted variables when target distribution labels are available.

LLM-based agents are increasingly being deployed in contexts requiring complex authorization policies: customer service protocols, approval workflows, data access restrictions, and regulatory compliance. Embedding these policies in prompts provides no enforcement guarantees. We present PCAS, a Policy Compiler for Agentic Systems that provides deterministic policy enforcement. Enforcing such policies requires tracking information flow across agents, which linear message histories cannot capture. Instead, PCAS models the agentic system state as a dependency graph capturing causal relationships among events such as tool calls, tool results, and messages. Policies are expressed in a Datalog-derived language, as declarative rules that account for transitive information flow and cross-agent provenance. A reference monitor intercepts all actions and blocks violations before execution, providing deterministic enforcement independent of model reasoning. PCAS takes an existing agent implementation and a policy specification, and compiles them into an instrumented system that is policy-compliant by construction, with no security-specific restructuring required. We evaluate PCAS on three case studies: information flow policies for prompt injection defense, approval workflows in a multi-agent pharmacovigilance system, and organizational policies for customer service. On customer service tasks, PCAS improves policy compliance from 48% to 93% across frontier models, with zero policy violations in instrumented runs.

LLMs are increasingly being used for complex problems which are not necessarily resolved in a single response, but require interacting with an environment to acquire information. In these scenarios, LLMs must reason about inherent cost-uncertainty tradeoffs in when to stop exploring and commit to an answer. For instance, on a programming task, an LLM should test a generated code snippet if it is uncertain about the correctness of that code; the cost of writing a test is nonzero, but typically lower than the cost of making a mistake. In this work, we show that we can induce LLMs to explicitly reason about balancing these cost-uncertainty tradeoffs, then perform more optimal environment exploration. We formalize multiple tasks, including information retrieval and coding, as sequential decision-making problems under uncertainty. Each problem has latent environment state that can be reasoned about via a prior which is passed to the LLM agent. We introduce a framework called Calibrate-Then-Act (CTA), where we feed the LLM this additional context to enable it to act more optimally. This improvement is preserved even under RL training of both the baseline and CTA. Our results on information-seeking QA and on a simplified coding task show that making cost-benefit tradeoffs explicit with CTA can help agents discover more optimal decision-making strategies.

Low Earth orbit (LEO) satellites are leveraged to support new position, navigation, and timing (PNT) service alternatives to GNSS. These alternatives require accurate propagation of satellite position and velocity with a realistic quantification of uncertainty. It is commonly assumed that the propagated uncertainty distribution is Gaussian; however, the validity of this assumption can be quickly compromised by the mismodeling of atmospheric drag. We develop a machine learning approach that corrects error growth in the argument of latitude for a diverse set of LEO satellites. The improved orbit propagation accuracy extends the applicability of the Gaussian assumption and modeling of the errors with a corrected mean and covariance. We compare the performance of a time-conditioned neural network and a Gaussian Process on datasets computed with an open source orbit propagator and publicly available Vector Covariance Message (VCM) ephemerides. The learned models predict the argument of latitude error as a Gaussian distribution given parameters from a single VCM epoch and reverse propagation errors. We show that this one-dimensional model captures the effect of mismodeled drag, which can be mapped to the Cartesian state space. The correction method only updates information along the dimensions of dominant error growth, while maintaining the physics-based propagation of VCM covariance in the remaining dimensions. We therefore extend the utility of VCM ephemerides to longer time horizons without modifying the functionality of the existing propagator.

Artificial Intelligence (AI) benchmarks play a central role in measuring progress in model development and guiding deployment decisions. However, many benchmarks quickly become saturated, meaning that they can no longer differentiate between the best-performing models, diminishing their long-term value. In this study, we analyze benchmark saturation across 60 Large Language Model (LLM) benchmarks selected from technical reports by major model developers. To identify factors driving saturation, we characterize benchmarks along 14 properties spanning task design, data construction, and evaluation format. We test five hypotheses examining how each property contributes to saturation rates. Our analysis reveals that nearly half of the benchmarks exhibit saturation, with rates increasing as benchmarks age. Notably, hiding test data (i.e., public vs. private) shows no protective effect, while expert-curated benchmarks resist saturation better than crowdsourced ones. Our findings highlight which design choices extend benchmark longevity and inform strategies for more durable evaluation.

Modern machine learning-based wireless localization using Wi-Fi signals continues to face significant challenges in achieving groundbreaking performance across diverse environments. A major limitation is that most existing algorithms do not appropriately weight the information from different routers during aggregation, resulting in suboptimal convergence and reduced accuracy. Motivated by traditional weighted triangulation methods, this paper introduces the concept of attention to routers, ensuring that each router's contribution is weighted differently when aggregating information from multiple routers for triangulation. We demonstrate, by incorporating attention layers into a standard machine learning localization architecture, that emphasizing the relevance of each router can substantially improve overall performance. We have also shown through evaluation over the open-sourced datasets and demonstrate that Attention to Routers outperforms the benchmark architecture by over 30% in accuracy.

We present a practical system for privacy-aware large language model (LLM) inference that splits a transformer between a trusted local GPU and an untrusted cloud GPU, communicating only intermediate activations over the network. Our system addresses the unique challenges of autoregressive LLM decoding over high-latency wide-area networks (WANs), contributing: (1) an asymmetric layer split where embedding and unembedding layers remain local, ensuring raw tokens never leave the trusted device; (2) the first application of lookahead decoding to split inference over WANs, amortizing network round-trip latency across multiple tokens per iteration; (3) an empirical inversion attack evaluation showing that split depth provides a tunable privacy-performance tradeoff -- an attacker can recover ~59%% of tokens at a 2-layer split but only ~35%% at an 8-layer split, with minimal throughput impact; (4) ablation experiments showing that n-gram speculation accepts 1.2-1.3 tokens per decoding step on average (peak of 7 observed on code), with acceptance rates consistent across model scales; (5) formal verification that lookahead decoding produces token-identical output to sequential decoding under greedy argmax, with zero quality degradation; and (6) scaling validation on Mistral NeMo 12B (40 layers), demonstrating that the system generalizes to larger models with only 4.9 GB local VRAM and matching 7B throughput. Evaluated on Mistral 7B and NeMo 12B over a ~80ms WAN link, our system achieves 8.7-9.3 tok/s (7B) and 7.8-8.7 tok/s (12B) with lookahead decoding, with an RTT decomposition model (validated at <6.2%% cross-validation error) projecting 15-19 tok/s at 20ms RTT.

In long-term multivariate time series forecasting, effectively capturing both periodic patterns and residual dynamics is essential. To address this within standard deep learning benchmark settings, we propose the Hierarchical Patching Mixer (HPMixer), which models periodicity and residuals in a decoupled yet complementary manner. The periodic component utilizes a learnable cycle module [7] enhanced with a nonlinear channel-wise MLP for greater expressiveness. The residual component is processed through a Learnable Stationary Wavelet Transform (LSWT) to extract stable, shift-invariant frequency-domain representations. Subsequently, a channel-mixing encoder models explicit inter-channel dependencies, while a two-level non-overlapping hierarchical patching mechanism captures coarse- and fine-scale residual variations. By integrating decoupled periodicity modeling with structured, multi-scale residual learning, HPMixer provides an effective framework. Extensive experiments on standard multivariate benchmarks demonstrate that HPMixer achieves competitive or state-of-the-art performance compared to recent baselines.

The pursuit of general-purpose robotic manipulation is hindered by the scarcity of diverse, real-world interaction data. Unlike data collection from web in vision or language, robotic data collection is an active process incurring prohibitive physical costs. Consequently, automated task curation to maximize data value remains a critical yet under-explored challenge. Existing manual methods are unscalable and biased toward common tasks, while off-the-shelf foundation models often hallucinate physically infeasible instructions. To address this, we introduce RoboGene, an agentic framework designed to automate the generation of diverse, physically plausible manipulation tasks across single-arm, dual-arm, and mobile robots. RoboGene integrates three core components: diversity-driven sampling for broad task coverage, self-reflection mechanisms to enforce physical constraints, and human-in-the-loop refinement for continuous improvement. We conduct extensive quantitative analysis and large-scale real-world experiments, collecting datasets of 18k trajectories and introducing novel metrics to assess task quality, feasibility, and diversity. Results demonstrate that RoboGene significantly outperforms state-of-the-art foundation models (e.g., GPT-4o, Gemini 2.5 Pro). Furthermore, real-world experiments show that VLA models pre-trained with RoboGene achieve higher success rates and superior generalization, underscoring the importance of high-quality task generation. Our project is available at https://robogene-boost-vla.github.io.

LLM-based agents execute real-world workflows via tools and memory. These affordances enable ill-intended adversaries to also use these agents to carry out complex misuse scenarios. Existing agent misuse benchmarks largely test single-prompt instructions, leaving a gap in measuring how agents end up helping with harmful or illegal tasks over multiple turns. We introduce STING (Sequential Testing of Illicit N-step Goal execution), an automated red-teaming framework that constructs a step-by-step illicit plan grounded in a benign persona and iteratively probes a target agent with adaptive follow-ups, using judge agents to track phase completion. We further introduce an analysis framework that models multi-turn red-teaming as a time-to-first-jailbreak random variable, enabling analysis tools like discovery curves, hazard-ratio attribution by attack language, and a new metric: Restricted Mean Jailbreak Discovery. Across AgentHarm scenarios, STING yields substantially higher illicit-task completion than single-turn prompting and chat-oriented multi-turn baselines adapted to tool-using agents. In multilingual evaluations across six non-English settings, we find that attack success and illicit-task completion do not consistently increase in lower-resource languages, diverging from common chatbot findings. Overall, STING provides a practical way to evaluate and stress-test agent misuse in realistic deployment settings, where interactions are inherently multi-turn and often multilingual.

We introduce NESSiE, the NEceSsary SafEty benchmark for large language models (LLMs). With minimal test cases of information and access security, NESSiE reveals safety-relevant failures that should not exist, given the low complexity of the tasks. NESSiE is intended as a lightweight, easy-to-use sanity check for language model safety and, as such, is not sufficient for guaranteeing safety in general -- but we argue that passing this test is necessary for any deployment. However, even state-of-the-art LLMs do not reach 100% on NESSiE and thus fail our necessary condition of language model safety, even in the absence of adversarial attacks. Our Safe & Helpful (SH) metric allows for direct comparison of the two requirements, showing models are biased toward being helpful rather than safe. We further find that disabled reasoning for some models, but especially a benign distraction context degrade model performance. Overall, our results underscore the critical risks of deploying such models as autonomous agents in the wild. We make the dataset, package and plotting code publicly available.

Precision fermentation relies on microbial cell factories to produce sustainable food, pharmaceuticals, chemicals, and biofuels. Specialized laboratories such as biofoundries are advancing these processes using high-throughput bioreactor platforms, which generate vast datasets. However, the lack of community standards limits data accessibility and interoperability, preventing integration across platforms. In order to address this, we introduce PREFER, an open-source ontology designed to establish a unified standard for bioprocess data. Built in alignment with the widely adopted Basic Formal Ontology (BFO) and connecting with several other community ontologies, PREFER ensures consistency and cross-domain compatibility and covers the whole precision fermentation process. Integrating PREFER into high-throughput bioprocess development workflows enables structured metadata that supports automated cross-platform execution and high-fidelity data capture. Furthermore, PREFER's standardization has the potential to bridge disparate data silos, generating machine-actionable datasets critical for training predictive, robust machine learning models in synthetic biology. This work provides the foundation for scalable, interoperable bioprocess systems and supports the transition toward more data-driven bioproduction.

This note gives a self-contained overview of some important properties of the Gromov-Wasserstein (GW) distance, compared with the standard linear optimal transport (OT) framework. More specifically, I explore the following questions: are GW optimal transport plans sparse? Under what conditions are they supported on a permutation? Do they satisfy a form of cyclical monotonicity? In particular, I present the conditionally negative semi-definite property and show that, when it holds, there are GW optimal plans that are sparse and supported on a permutation.

Large Language Models (LLMs) are increasingly used as automated annotators to scale dataset creation, yet their reliability as unbiased annotators--especially for low-resource and identity-sensitive settings--remains poorly understood. In this work, we study the behavior of LLMs as zero-shot annotators for Bangla hate speech, a task where even human agreement is challenging, and annotator bias can have serious downstream consequences. We conduct a systematic benchmark of 17 LLMs using a unified evaluation framework. Our analysis uncovers annotator bias and substantial instability in model judgments. Surprisingly, increased model scale does not guarantee improved annotation quality--smaller, more task-aligned models frequently exhibit more consistent behavior than their larger counterparts. These results highlight important limitations of current LLMs for sensitive annotation tasks in low-resource languages and underscore the need for careful evaluation before deployment.

Federated learning has emerged as a privacy-preserving and efficient approach for deploying intelligent agricultural solutions. Accurate edge-based diagnosis across geographically dispersed farms is crucial for recognising tomato diseases in sustainable farming. Traditional centralised training aggregates raw data on a central server, leading to communication overhead, privacy risks and latency. Meanwhile, edge devices require lightweight networks to operate effectively within limited resources. In this paper, we propose U-FedTomAtt, an ultra-lightweight federated learning framework with attention for tomato disease recognition in resource-constrained and distributed environments. The model comprises only 245.34K parameters and 71.41 MFLOPS. First, we propose an ultra-lightweight neural network with dilated bottleneck (DBNeck) modules and a linear transformer to minimise computational and memory overhead. To mitigate potential accuracy loss, a novel local-global residual attention (LoGRA) module is incorporated. Second, we propose the federated dual adaptive weight aggregation (FedDAWA) algorithm that enhances global model accuracy. Third, our framework is validated using three benchmark datasets for tomato diseases under simulated federated settings. Experimental results show that the proposed method achieves 0.9910% and 0.9915% Top-1 accuracy and 0.9923% and 0.9897% F1-scores on SLIF-Tomato and PlantVillage tomato datasets, respectively.

We show that training AI agents on high-fidelity reinforcement learning environments produces capabilities that generalize beyond the training distribution. We introduce CoreCraft, the first environment in EnterpriseBench, Surge AI's suite of agentic RL environments. CoreCraft is a fully operational enterprise simulation of a customer support organization, comprising over 2,500 entities across 14 entity types with 23 unique tools, designed to measure whether AI agents can perform the multi-step, domain-specific work that real jobs demand. Frontier models such as GPT-5.2 and Claude Opus 4.6 solve fewer than 30% of tasks when all expert-authored rubric criteria must be satisfied. Using this environment, we train GLM 4.6 with Group Relative Policy Optimization (GRPO) and adaptive clipping. After a single epoch of training, the model improves from 25.37% to 36.76% task pass rate on held-out evaluation tasks. More importantly, these gains transfer to out-of-distribution benchmarks: +4.5% on BFCL Parallel, +7.4% on Tau2-Bench Retail, and +6.8% on Tool Decathlon (Pass@1). We believe three environment properties are consistent with the observed transfer: task-centric world building that optimizes for diverse, challenging tasks; expert-authored rubrics enabling reliable reward computation; and enterprise workflows that reflect realistic professional patterns. Our results suggest that environment quality, diversity, and realism are key factors enabling generalizable agent capabilities.

Quality assurance (QA) during construction often relies on inspection records and laboratory test results that become available days or weeks after work is completed. On large highway and bridge projects, this delay limits early intervention and increases the risk of rework, schedule impacts, and fragmented documentation. This study presents a construction-phase digital twin framework designed to support element-level QA and readiness-based decision making during active construction. The framework links inspection records, material production and placement data, early-age sensing, and predictive strength models to individual construction elements. By integrating these data streams, the system represents the evolving quality state of each element and supports structured release or hold decisions before standard-age test results are available. The approach does not replace established inspection and testing procedures. Instead, it supplements existing workflows by improving traceability and enabling earlier, data-informed quality assessments. Practical considerations related to data integration, contractual constraints, and implementation challenges are also discussed. The proposed framework provides a structured pathway for transitioning construction QA from delayed, document-driven review toward proactive, element-level decision support during construction.

In this work, we propose a notion of practical learnability grounded in finite sample settings, and develop a conjugate learning theoretical framework based on convex conjugate duality to characterize this learnability property. Building on this foundation, we demonstrate that training deep neural networks (DNNs) with mini-batch stochastic gradient descent (SGD) achieves global optima of empirical risk by jointly controlling the extreme eigenvalues of a structure matrix and the gradient energy, and we establish a corresponding convergence theorem. We further elucidate the impact of batch size and model architecture (including depth, parameter count, sparsity, skip connections, and other characteristics) on non-convex optimization. Additionally, we derive a model-agnostic lower bound for the achievable empirical risk, theoretically demonstrating that data determines the fundamental limit of trainability. On the generalization front, we derive deterministic and probabilistic bounds on generalization error based on generalized conditional entropy measures. The former explicitly delineates the range of generalization error, while the latter characterizes the distribution of generalization error relative to the deterministic bounds under independent and identically distributed (i.i.d.) sampling conditions. Furthermore, these bounds explicitly quantify the influence of three key factors: (i) information loss induced by irreversibility in the model, (ii) the maximum attainable loss value, and (iii) the generalized conditional entropy of features with respect to labels. Moreover, they offer a unified theoretical lens for understanding the roles of regularization, irreversible transformations, and network depth in shaping the generalization behavior of deep neural networks. Extensive experiments validate all theoretical predictions, confirming the framework's correctness and consistency.

The reliability of medical LLM evaluation is critically undermined by data contamination and knowledge obsolescence, leading to inflated scores on static benchmarks. To address these challenges, we introduce LiveClin, a live benchmark designed for approximating real-world clinical practice. Built from contemporary, peer-reviewed case reports and updated biannually, LiveClin ensures clinical currency and resists data contamination. Using a verified AI-human workflow involving 239 physicians, we transform authentic patient cases into complex, multimodal evaluation scenarios that span the entire clinical pathway. The benchmark currently comprises 1,407 case reports and 6,605 questions. Our evaluation of 26 models on LiveClin reveals the profound difficulty of these real-world scenarios, with the top-performing model achieving a Case Accuracy of just 35.7%. In benchmarking against human experts, Chief Physicians achieved the highest accuracy, followed closely by Attending Physicians, with both surpassing most models. LiveClin thus provides a continuously evolving, clinically grounded framework to guide the development of medical LLMs towards closing this gap and achieving greater reliability and real-world utility. Our data and code are publicly available at https://github.com/AQ-MedAI/LiveClin.

Grokking -- the delayed transition from memorization to generalization in small algorithmic tasks -- remains poorly understood. We present a geometric analysis of optimization dynamics in transformers trained on modular arithmetic. PCA of attention weight trajectories reveals that training evolves predominantly within a low-dimensional execution subspace, with a single principal component capturing 68-83% of trajectory variance. To probe loss-landscape geometry, we measure commutator defects -- the non-commutativity of successive gradient steps -- and project them onto this learned subspace. We find that curvature grows sharply in directions orthogonal to the execution subspace while the trajectory remains largely confined to it. Importantly, curvature growth consistently precedes generalization across learning rates and hyperparameter regimes, with the lead time obeying a power law in the grokking timescale. Causal intervention experiments show that motion along the learned subspace is necessary for grokking, while artificially increasing curvature is insufficient. Together, these results support a geometric account in which grokking reflects escape from a metastable regime characterized by low-dimensional confinement and transverse curvature accumulation. All findings replicate across this learning-rate range, a qualitatively different slow regime (lr=5e-5, wd=0.1, 3 layers), and three random seeds, though alignment dynamics differ quantitatively between regimes. Causal intervention experiments establish that orthogonal gradient flow is necessary but not sufficient for grokking: suppressing it prevents generalization with a monotonic dose-response across four operations, while artificially boosting curvature defects has no effect.

Test-time scaling can improve model performance by aggregating stochastic reasoning trajectories. However, achieving sample-efficient test-time self-consistency under a limited budget remains an open challenge. We introduce PETS (Principled and Efficient Test-TimeSelf-Consistency), which initiates a principled study of trajectory allocation through an optimization framework. Central to our approach is the self-consistency rate, a new measure defined as agreement with the infinite-budget majority vote. This formulation makes sample-efficient test-time allocation theoretically grounded and amenable to rigorous analysis. We study both offline and online settings. In the offline regime, where all questions are known in advance, we connect trajectory allocation to crowdsourcing, a classic and well-developed area, by modeling reasoning traces as workers. This perspective allows us to leverage rich existing theory, yielding theoretical guarantees and an efficient majority-voting-based allocation algorithm. In the online streaming regime, where questions arrive sequentially and allocations must be made on the fly, we propose a novel method inspired by the offline framework. Our approach adapts budgets to question difficulty while preserving strong theoretical guarantees and computational efficiency. Experiments show that PETS consistently outperforms uniform allocation. On GPQA, PETS achieves perfect self-consistency in both settings while reducing the sampling budget by up to 75% (offline) and 55% (online) relative to uniform allocation. Code is available at https://github.com/ZDCSlab/PETS.

Reinforcement Learning with Verifiable Rewards (RLVR) has been shown effective in enhancing the visual reflection and reasoning capabilities of Large Multimodal Models (LMMs). However, existing datasets are predominantly derived from either small-scale manual construction or recombination of prior resources, which limits data diversity and coverage, thereby constraining further gains in model performance. To this end, we introduce \textbf{DeepVision-103K}, a comprehensive dataset for RLVR training that covers diverse K12 mathematical topics, extensive knowledge points, and rich visual elements. Models trained on DeepVision achieve strong performance on multimodal mathematical benchmarks, and generalize effectively to general multimodal reasoning tasks. Further analysis reveals enhanced visual perception, reflection and reasoning capabilities in trained models, validating DeepVision's effectiveness for advancing multimodal reasoning. Data: \href{https://huggingface.co/datasets/skylenage/DeepVision-103K}{this url}.

AI-assisted code review is widely used to detect vulnerabilities before production release. Prior work shows that adversarial prompt manipulation can degrade large language model (LLM) performance in code generation. We test whether similar comment-based manipulation misleads LLMs during vulnerability detection. We build a 100-sample benchmark across Python, JavaScript, and Java, each paired with eight comment variants ranging from no comments to adversarial strategies such as authority spoofing and technical deception. Eight frontier models, five commercial and three open-source, are evaluated in 9,366 trials. Adversarial comments produce small, statistically non-significant effects on detection accuracy (McNemar exact p > 0.21; all 95 percent confidence intervals include zero). This holds for commercial models with 89 to 96 percent baseline detection and open-source models with 53 to 72 percent, despite large absolute performance gaps. Unlike generation settings where comment manipulation achieves high attack success, detection performance does not meaningfully degrade. More complex adversarial strategies offer no advantage over simple manipulative comments. We test four automated defenses across 4,646 additional trials (14,012 total). Static analysis cross-referencing performs best at 96.9 percent detection and recovers 47 percent of baseline misses. Comment stripping reduces detection for weaker models by removing helpful context. Failures concentrate on inherently difficult vulnerability classes, including race conditions, timing side channels, and complex authorization logic, rather than on adversarial comments.

In mechanistic interpretability, recent work scrutinizes transformer "circuits" - sparse, mono or multi layer sub computations, that may reflect human understandable functions. Yet, these network circuits are rarely acid-tested for their stability across different instances of the same deep learning architecture. Without this, it remains unclear whether reported circuits emerge universally across labs or turn out to be idiosyncratic to a particular estimation instance, potentially limiting confidence in safety-critical settings. Here, we systematically study stability across-refits in increasingly complex transformer language models of various sizes. We quantify, layer by layer, how similarly attention heads learn representations across independently initialized training runs. Our rigorous experiments show that (1) middle-layer heads are the least stable yet the most representationally distinct; (2) deeper models exhibit stronger mid-depth divergence; (3) unstable heads in deeper layers become more functionally important than their peers from the same layer; (4) applying weight decay optimization substantially improves attention-head stability across random model initializations; and (5) the residual stream is comparatively stable. Our findings establish the cross-instance robustness of circuits as an essential yet underappreciated prerequisite for scalable oversight, drawing contours around possible white-box monitorability of AI systems.

Epilepsy affects over 50 million people worldwide, and one-third of patients suffer drug-resistant seizures where surgery offers the best chance of seizure freedom. Accurate localization of the epileptogenic zone (EZ) relies on intracranial EEG (iEEG). Clinical workflows, however, remain constrained by labor-intensive manual review. At the same time, existing data-driven approaches are typically developed on single-center datasets that are inconsistent in format and metadata, lack standardized benchmarks, and rarely release pathological event annotations, creating barriers to reproducibility, cross-center validation, and clinical relevance. With extensive efforts to reconcile heterogeneous iEEG formats, metadata, and recordings across publicly available sources, we present $\textbf{Omni-iEEG}$, a large-scale, pre-surgical iEEG resource comprising $\textbf{302 patients}$ and $\textbf{178 hours}$ of high-resolution recordings. The dataset includes harmonized clinical metadata such as seizure onset zones, resections, and surgical outcomes, all validated by board-certified epileptologists. In addition, Omni-iEEG provides over 36K expert-validated annotations of pathological events, enabling robust biomarker studies. Omni-iEEG serves as a bridge between machine learning and epilepsy research. It defines clinically meaningful tasks with unified evaluation metrics grounded in clinical priors, enabling systematic evaluation of models in clinically relevant settings. Beyond benchmarking, we demonstrate the potential of end-to-end modeling on long iEEG segments and highlight the transferability of representations pretrained on non-neurophysiological domains. Together, these contributions establish Omni-iEEG as a foundation for reproducible, generalizable, and clinically translatable epilepsy research. The project page with dataset and code links is available at omni-ieeg.github.io/omni-ieeg.

Secondary crash likelihood prediction is a critical component of an active traffic management system to mitigate congestion and adverse impacts caused by secondary crashes. However, existing approaches mainly rely on post-crash features (e.g., crash type and severity) that are rarely available in real time, limiting their practical applicability. To address this limitation, we propose a hybrid secondary crash likelihood prediction framework that does not depend on post-crash features. A dynamic spatiotemporal window is designed to extract real-time traffic flow and environmental features from primary crash locations and their upstream segments. The framework includes three models: a primary crash model to estimate the likelihood of secondary crash occurrence, and two secondary crash models to evaluate traffic conditions at crash and upstream segments under different comparative scenarios. An ensemble learning strategy integrating six machine learning algorithms is developed to enhance predictive performance, and a voting-based mechanism combines the outputs of the three models. Experiments on Florida freeways demonstrate that the proposed hybrid framework correctly identifies 91% of secondary crashes with a low false alarm rate of 0.20. The Area Under the ROC Curve improves from 0.654, 0.744, and 0.902 for the individual models to 0.952 for the hybrid model, outperforming previous studies.

Industrial IoT predictive maintenance requires systems capable of real-time anomaly detection without sacrificing interpretability or demanding excessive computational resources. Traditional approaches rely on static, offline-trained models that cannot adapt to evolving operational conditions, while LLM-based monolithic systems demand prohibitive memory and latency, rendering them impractical for on-site edge deployment. We introduce SEMAS, a self-evolving hierarchical multi-agent system that distributes specialized agents across Edge, Fog, and Cloud computational tiers. Edge agents perform lightweight feature extraction and pre-filtering; Fog agents execute diversified ensemble detection with dynamic consensus voting; and Cloud agents continuously optimize system policies via Proximal Policy Optimization (PPO) while maintaining asynchronous, non-blocking inference. The framework incorporates LLM-based response generation for explainability and federated knowledge aggregation for adaptive policy distribution. This architecture enables resource-aware specialization without sacrificing real-time performance or model interpretability. Empirical evaluation on two industrial benchmarks (Boiler Emulator and Wind Turbine) demonstrates that SEMAS achieves superior anomaly detection performance with exceptional stability under adaptation, sustains prediction accuracy across evolving operational contexts, and delivers substantial latency improvements enabling genuine real-time deployment. Ablation studies confirm that PPO-driven policy evolution, consensus voting, and federated aggregation each contribute materially to system effectiveness. These findings indicate that resource-aware, self-evolving 1multi-agent coordination is essential for production-ready industrial IoT predictive maintenance under strict latency and explainability constraints.

Accurate and timely identification of hospital outbreak clusters is crucial for preventing the spread of infections that have epidemic potential. While assessing pathogen similarity through whole genome sequencing (WGS) is considered the gold standard for outbreak detection, its high cost and lengthy turnaround time preclude routine implementation in clinical laboratories. We explore the utility of two rapid and cost-effective alternatives to WGS, matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry and antimicrobial resistance (AR) patterns. We develop a machine learning framework that extracts informative representations from MALDI-TOF spectra and AR patterns for outbreak detection and explore their fusion. Through multi-species analyses, we demonstrate that in some cases MALDI-TOF and AR have the potential to reduce reliance on WGS, enabling more accessible and rapid outbreak surveillance.

This paper presents empirical results from a production-grade C++ implementation of a deterministic semantic state substrate derived from prior formal work on Bounded Local Generator Classes (Martin, 2026). The system was mathematically specified prior to implementation and realized as a CPU-resident graph engine operating under bounded local state evolution. Contemporary inference-driven AI architectures reconstruct semantic state through probabilistic recomposition, producing compute cost that scales with token volume and context horizon. In contrast, the substrate described here represents semantic continuity as a persistent, addressable memory graph evolved under a time-modulated local operator g(t). Work is bounded by local semantic change Delta s, independent of total memory cardinality M. Empirical measurements on Apple M2-class silicon demonstrate invariant traversal latency (approximately 0.25 to 0.32 ms), stable CPU utilization (approximately 17.2 percent baseline with Delta CPU approximately 0 to 0.2 percent), and no scale-correlated thermal signature across 1M to 25M node regimes under sustained operation. Measured per-node density ranges from approximately 1.3 KB (Float64 baseline) to approximately 687 bytes (compressed Float32 accounting). Under binary memory accounting, this yields a 1.6 billion node capacity projection within a 1 TiB envelope. These results indicate an empirically invariant thermodynamic regime in which scaling is governed by memory capacity rather than inference-bound recomposition. The Compute ICE-AGE is defined as the Invariant Compute Envelope under Addressable Graph Evolution, and the empirical evidence presented demonstrates this regime up to 25M nodes.

This paper proposes a few-shot classification framework based on Large Language Models (LLMs) to predict whether the next day will have spikes in real-time electricity prices. The approach aggregates system state information, including electricity demand, renewable generation, weather forecasts, and recent electricity prices, into a set of statistical features that are formatted as natural-language prompts and fed to an LLM along with general instructions. The model then determines the likelihood that the next day would be a spike day and reports a confidence score. Using historical data from the Texas electricity market, we demonstrate that this few-shot approach achieves performance comparable to supervised machine learning models, such as Support Vector Machines and XGBoost, and outperforms the latter two when limited historical data are available. These findings highlight the potential of LLMs as a data-efficient tool for classifying electricity price spikes in settings with scarce data.

Accurate speed prediction is crucial for proactive traffic management to enhance traffic efficiency and safety. Existing studies have primarily relied on aggregated, macroscopic traffic flow data to predict future traffic trends, whereas road traffic dynamics are also influenced by individual, microscopic human driving behaviors. Recent Connected Vehicle (CV) data provide rich driving behavior features, offering new opportunities to incorporate these behavioral insights into speed prediction. To this end, we propose the Macro-Micro Cross-Attention Transformer (MMCAformer) to integrate CV data-based micro driving behavior features with macro traffic features for speed prediction. Specifically, MMCAformer employs self-attention to learn intrinsic dependencies in macro traffic flow and cross-attention to capture spatiotemporal interplays between macro traffic status and micro driving behavior. MMCAformer is optimized with a Student-t negative log-likelihood loss to provide point-wise speed prediction and estimate uncertainty. Experiments on four Florida freeways demonstrate the superior performance of the proposed MMCAformer compared to baselines. Compared with only using macro features, introducing micro driving behavior features not only enhances prediction accuracy (e.g., overall RMSE, MAE, and MAPE reduced by 9.0%, 6.9%, and 10.2%, respectively) but also shrinks model prediction uncertainty (e.g., mean predictive intervals decreased by 10.1-24.0% across the four freeways). Results reveal that hard braking and acceleration frequencies emerge as the most influential features. Such improvements are more pronounced under congested, low-speed traffic conditions.

We systematically evaluate the quality of widely used AI safety datasets from two perspectives: in isolation and in practice. In isolation, we examine how well these datasets reflect real-world attacks based on three key properties: driven by ulterior intent, well-crafted, and out-of-distribution. We find that these datasets overrely on "triggering cues": words or phrases with overt negative/sensitive connotations that are intended to trigger safety mechanisms explicitly, which is unrealistic compared to real-world attacks. In practice, we evaluate whether these datasets genuinely measure safety risks or merely provoke refusals through triggering cues. To explore this, we introduce "intent laundering": a procedure that abstracts away triggering cues from attacks (data points) while strictly preserving their malicious intent and all relevant details. Our results indicate that current AI safety datasets fail to faithfully represent real-world attacks due to their overreliance on triggering cues. In fact, once these cues are removed, all previously evaluated "reasonably safe" models become unsafe, including Gemini 3 Pro and Claude Sonnet 3.7. Moreover, when intent laundering is adapted as a jailbreaking technique, it consistently achieves high attack success rates, ranging from 90% to over 98%, under fully black-box access. Overall, our findings expose a significant disconnect between how model safety is evaluated and how real-world adversaries behave.

Large-scale human mobility simulation is critical for applications such as urban planning, epidemiology, and transportation analysis. Recent works treat large language models (LLMs) as human agents to simulate realistic mobility behaviors using structured reasoning, but their high computational cost limits scalability. To address this, we design a mobility-aware cache framework named MobCache that leverages reconstructible caches to enable efficient large-scale human mobility simulations. It consists of: (1) a reasoning component that encodes each reasoning step as a latent-space embedding and uses a latent-space evaluator to enable the reuse and recombination of reasoning steps; and (2) a decoding component that employs a lightweight decoder trained with mobility law-constrained distillation to translate latent-space reasoning chains into natural language, thereby improving simulation efficiency while maintaining fidelity. Experiments show that MobCache significantly improves efficiency across multiple dimensions while maintaining performance comparable to state-of-the-art LLM-based methods.

Large-scale human mobility simulation is critical for many science domains such as urban science, epidemiology, and transportation analysis. Recent works treat large language models (LLMs) as human agents to simulate realistic mobility trajectories by modeling individual-level cognitive processes. However, these approaches generate individual mobility trajectories independently, without any population-level coordination mechanism, and thus fail to capture the emergence of collective behaviors. To address this issue, we design M2LSimu, a mobility measures-guided multi-prompt adjustment framework that leverages mobility measures derived from shared data as guidance to refine individual-level prompts for realistic mobility generation. Our framework applies coarse-grained adjustment strategies guided by mobility measures, progressively enabling fine-grained individual-level adaptation while satisfying multiple population-level mobility objectives under a limited budget. Experiments show that M2LSimu significantly outperforms state-of-the-art LLM-based methods on two public datasets.

Intracoronary Optical Coherence Tomography (OCT) enables high-resolution visualization of coronary vessel anatomy but presents challenges due to noise, imaging artifacts, and complex tissue structures. This paper proposes a fully automated pipeline for vessel segmentation and classification in OCT images using machine learning techniques. The proposed method integrates image preprocessing, guidewire artifact removal, polar-to-Cartesian transformation, unsupervised K-means clustering, and local feature extraction. These features are used to train Logistic Regression and Support Vector Machine classifiers for pixel-wise vessel classification. Experimental results demonstrate excellent performance, achieving precision, recall, and F1-score values up to 1.00 and overall classification accuracy of 99.68%. The proposed approach provides accurate vessel boundary detection while maintaining low computational complexity and requiring minimal manual annotation. This method offers a reliable and efficient solution for automated OCT image analysis and has potential applications in clinical decision support and real-time medical image processing.

This work introduces EduEVAL-DB, a dataset based on teacher roles designed to support the evaluation and training of automatic pedagogical evaluators and AI tutors for instructional explanations. The dataset comprises 854 explanations corresponding to 139 questions from a curated subset of the ScienceQA benchmark, spanning science, language, and social science across K-12 grade levels. For each question, one human-teacher explanation is provided and six are generated by LLM-simulated teacher roles. These roles are inspired by instructional styles and shortcomings observed in real educational practice and are instantiated via prompt engineering. We further propose a pedagogical risk rubric aligned with established educational standards, operationalizing five complementary risk dimensions: factual correctness, explanatory depth and completeness, focus and relevance, student-level appropriateness, and ideological bias. All explanations are annotated with binary risk labels through a semi-automatic process with expert teacher review. Finally, we present preliminary validation experiments to assess the suitability of EduEVAL-DB for evaluation. We benchmark a state-of-the-art education-oriented model (Gemini 2.5 Pro) against a lightweight local Llama 3.1 8B model and examine whether supervised fine-tuning on EduEVAL-DB supports pedagogical risk detection using models deployable on consumer hardware.

For a broad family of discriminative models that includes autoregressive language models, identifiability results imply that if two models induce the same conditional distributions, then their internal representations agree up to an invertible linear transformation. We ask whether an analogous conclusion holds approximately when the distributions are close instead of equal. Building on the observation of Nielsen et al. (2025) that closeness in KL divergence need not imply high linear representational similarity, we study a distributional distance based on logit differences and show that closeness in this distance does yield linear similarity guarantees. Specifically, we define a representational dissimilarity measure based on the models' identifiability class and prove that it is bounded by the logit distance. We further show that, when model probabilities are bounded away from zero, KL divergence upper-bounds logit distance; yet the resulting bound fails to provide nontrivial control in practice. As a consequence, KL-based distillation can match a teacher's predictions while failing to preserve linear representational properties, such as linear-probe recoverability of human-interpretable concepts. In distillation experiments on synthetic and image datasets, logit-distance distillation yields students with higher linear representational similarity and better preservation of the teacher's linearly recoverable concepts.

Dataset distillation compresses the original data into compact synthetic datasets, reducing training time and storage while retaining model performance, enabling deployment under limited resources. Although recent decoupling-based distillation methods enable dataset distillation at large scale, they continue to face an efficiency gap: optimization-based decoupling methods achieve higher accuracy but demand intensive computation, whereas optimization-free decoupling methods are efficient but sacrifice accuracy. To overcome this trade-off, we propose Exploration--Exploitation Distillation (E$^2$D), a simple, practical method that minimizes redundant computation through an efficient pipeline that begins with full-image initialization to preserve semantic integrity and feature diversity. It then uses a two-phase optimization strategy: an exploration phase that performs uniform updates and identifies high-loss regions, and an exploitation phase that focuses updates on these regions to accelerate convergence. We evaluate E$^2$D on large-scale benchmarks, surpassing the state-of-the-art on ImageNet-1K while being $18\times$ faster, and on ImageNet-21K, our method substantially improves accuracy while remaining $4.3\times$ faster. These results demonstrate that targeted, redundancy-reducing updates, rather than brute-force optimization, bridge the gap between accuracy and efficiency in large-scale dataset distillation. Code is available at https://github.com/ncsu-dk-lab/E2D.

State-space models like Mamba offer linear-time sequence processing and low memory, making them attractive for medical imaging. However, their robustness under realistic software and hardware threat models remains underexplored. This paper evaluates Mamba on multiple MedM-NIST classification benchmarks under input-level attacks, including white-box adversarial perturbations (FGSM/PGD), occlusion-based PatchDrop, and common acquisition corruptions (Gaussian noise and defocus blur) as well as hardware-inspired fault attacks emulated in software via targeted and random bit-flip injections into weights and activations. We profile vulnerabilities and quantify impacts on accuracy indicating that defenses are needed for deployment.

Artificial intelligence (AI) can automatically delineate lesions on computed tomography (CT) and generate radiology report content, yet progress is limited by the scarcity of publicly available CT datasets with lesion-level annotations. To bridge this gap, we introduce CT-Bench, a first-of-its-kind benchmark dataset comprising two components: a Lesion Image and Metadata Set containing 20,335 lesions from 7,795 CT studies with bounding boxes, descriptions, and size information, and a multitask visual question answering benchmark with 2,850 QA pairs covering lesion localization, description, size estimation, and attribute categorization. Hard negative examples are included to reflect real-world diagnostic challenges. We evaluate multiple state-of-the-art multimodal models, including vision-language and medical CLIP variants, by comparing their performance to radiologist assessments, demonstrating the value of CT-Bench as a comprehensive benchmark for lesion analysis. Moreover, fine-tuning models on the Lesion Image and Metadata Set yields significant performance gains across both components, underscoring the clinical utility of CT-Bench.

Deep learning-based respiratory auscultation is currently hindered by two fundamental challenges: (i) inherent information loss, as converting signals into spectrograms discards transient acoustic events and clinical context; (ii) limited data availability, exacerbated by severe class imbalance. To bridge these gaps, we present Resp-Agent, an autonomous multimodal system orchestrated by a novel Active Adversarial Curriculum Agent (Thinker-A$^2$CA). Unlike static pipelines, Thinker-A$^2$CA serves as a central controller that actively identifies diagnostic weaknesses and schedules targeted synthesis in a closed loop. To address the representation gap, we introduce a Modality-Weaving Diagnoser that weaves EHR data with audio tokens via Strategic Global Attention and sparse audio anchors, capturing both long-range clinical context and millisecond-level transients. To address the data gap, we design a Flow Matching Generator that adapts a text-only Large Language Model (LLM) via modality injection, decoupling pathological content from acoustic style to synthesize hard-to-diagnose samples. As a foundation for these efforts, we introduce Resp-229k, a benchmark corpus of 229k recordings paired with LLM-distilled clinical narratives. Extensive experiments demonstrate that Resp-Agent consistently outperforms prior approaches across diverse evaluation settings, improving diagnostic robustness under data scarcity and long-tailed class imbalance. Our code and data are available at https://github.com/zpforlove/Resp-Agent.

The paper introduces GUI-Owl-1.5, the latest native GUI agent model that features instruct/thinking variants in multiple sizes (2B/4B/8B/32B/235B) and supports a range of platforms (desktop, mobile, browser, and more) to enable cloud-edge collaboration and real-time interaction. GUI-Owl-1.5 achieves state-of-the-art results on more than 20+ GUI benchmarks on open-source models: (1) on GUI automation tasks, it obtains 56.5 on OSWorld, 71.6 on AndroidWorld, and 48.4 on WebArena; (2) on grounding tasks, it obtains 80.3 on ScreenSpotPro; (3) on tool-calling tasks, it obtains 47.6 on OSWorld-MCP, and 46.8 on MobileWorld; (4) on memory and knowledge tasks, it obtains 75.5 on GUI-Knowledge Bench. GUI-Owl-1.5 incorporates several key innovations: (1) Hybird Data Flywheel: we construct the data pipeline for UI understanding and trajectory generation based on a combination of simulated environments and cloud-based sandbox environments, in order to improve the efficiency and quality of data collection. (2) Unified Enhancement of Agent Capabilities: we use a unified thought-synthesis pipeline to enhance the model's reasoning capabilities, while placing particular emphasis on improving key agent abilities, including Tool/MCP use, memory and multi-agent adaptation; (3) Multi-platform Environment RL Scaling: We propose a new environment RL algorithm, MRPO, to address the challenges of multi-platform conflicts and the low training efficiency of long-horizon tasks. The GUI-Owl-1.5 models are open-sourced, and an online cloud-sandbox demo is available at https://github.com/X-PLUG/MobileAgent.

Owing to the power-law anisotropy in the quasiparticle dispersion, yielding an enhanced density of states, the effects of long range Coulomb interaction get amplified in three-dimensional generalized Weyl semimetals, characterized by integer monopole charge $n>1$ of the underlying Weyl nodes. Using a Wilsonian renormalization group approach controlled by a large-$N$ expansion with $N$ as the number of Weyl fermion flavors and a gauge-consistent regularization fixed by the Ward-Takahashi identity, we uncover for $n\ge 2$ an extended interaction-dominated scaling regime with intrinsically anisotropic dynamic Coulomb screening, a finite fermionic anomalous dimension, and a power-law suppression of the quasiparticle residue, yielding an \emph{anisotropic} marginal non-Fermi liquid at intermediate energies. Ultimately, the effective fine structure constant flows to zero, albeit only logarithmically slowly, so the marginal Fermi liquid phenomenology emerges as a broad crossover, controlled by a slowly running coupling. By contrast, for $n=1$ the system retains an isotropic marginal Weyl-liquid character. These predictions can be tested via scaling in thermodynamics (specific heat and compressibility), direction-dependent optical conductivity, and by anisotropic broadening of the single-particle spectral function in angle-resolved photoemission spectroscopy.

The Variational Quantum Eigensolver (VQE) is a leading hybrid quantum-classical algorithm for simulating many-body systems in the NISQ era. Its effectiveness, however, depends on the faithful preparation of eigenstates, which becomes challenging in degenerate and strongly entangled regimes. We study this problem using the transverse-field Ising model (TFIM) with periodic boundary conditions in one, two, and three dimensions, considering systems of up to 27 qubits. We employ different ansatzes: the hardware-efficient EfficientSU2 from Qiskit, the physics-inspired Hamiltonian Variational Ansatz (HVA) and HVA with symmetry breaking, and benchmark their performance using energy variance, entanglement entropy, spin correlations, and magnetization.

The Pauli exclusion principle is a cornerstone of quantum physics: it governs the structure of matter. Extensions of this principle, such as Haldane's generalized exclusion statistics, predict the existence of exotic quantum states characterized by fractional Fermi seas (FFS), i.e. momentum distributions with uniform but fractional occupancies. Here, we report the experimental realization of fractional Fermi seas in an excited one-dimensional Bose gas prepared through ramping cycles in the interaction strength. The resulting excited yet stable Bose-gas states exhibit Friedel oscillations, smoking-gun signatures of the underlying FFS. The stabilization of these states offers an opportunity to deepen our understanding of quantum thermodynamics in the presence of exotic statistics and paves the way for applications in quantum information and sensing.

Critical quantum field theories occupy a central position in modern theoretical physics for their inherent universality stemming from long-range correlations. As an example, the Tomonaga-Luttinger liquid (TLL) describes a wealth of one-dimensional quantum systems at low temperatures. Its behavior is deeply rooted in the emergence of an effective Fermi sea, leading to power-law correlations and Friedel oscillations. A promising direction to realize systems exhibiting novel universal behavior beyond TLL is through the generalization of the underlying Fermi sea. In this Letter, we show that fractional Fermi seas with reduced occupancy arise in an integrable Bose gas driven out of equilibrium by cyclic changes in interactions from repulsive to attractive. The correlation functions feature signatures of criticality incompatible with a conventional TLL, suggesting a novel critical phase. Our predictions, based on Generalized Hydrodynamics, are directly relevant to cold atoms.

The local equilibration time $τ_{\rm eq}$ of quantum many-body systems is conjectured to be bounded below by the Planckian time $\hbar /T$. We formalize this conjecture by defining $τ_{\rm eq}$ as the time scale at which a hydrodynamic description emerges for conserved densities. Drawing on analytic properties of real time thermal correlators, we establish a rigorous lower bound $τ_{\rm eq} \geq α\hbar /T$ on the onset of hydrodynamic behavior in a `regulated' thermal two-point function. The dimensionless coefficient $α$ depends only on dimensionality and the type of hydrodynamic or diffusive behavior that emerges, and is independent of the thermalization mechanism or other microscopic details. This bound applies universally to local quantum many-body systems, with or without a quasiparticle description, including in the presence of inelastic scattering.

We develop a hybrid Monte Carlo method to efficiently compute the physical observables from the samplings of the Laughlin and the Moore-Read wave functions of fractional quantum Hall (FQH) systems. With the advancements in methodology, including global updates and double stereographic projection on spherical geometry, our hybrid Monte Carlo simulation is significantly faster than the widely used Metropolis Monte Carlo scheme. As a result, we can readily simulate systems with electron numbers $N > 1000$ on both disk and sphere geometries. We apply this method to investigating the topological shift obtained from the edge dipole moment, computed from the density of the wave function on the disk. We also numerically computed the non-Abelian braiding matrices for different braiding schemes of the Moore-Read quasiholes on the sphere. Results with much better quality compared with previous works have been achieved. With the thermodynamic limit results obtained at ease, we also discuss the future usage of our method to clarify the questions on the instability of fractional quantum Hall states in an ideal Chern band setting or under quantum decoherence.

Achieving controlled and directed motion of artificial nanoscale systems in three-dimensional fluid environments remains a key-challenge in active matter, primarily due to the prevailing thermal fluctuations that rapidly randomize the particle trajectories. While significant progress has been made with micrometer-sized particles, imparting sufficient mechanical energy, or self-propulsion, to nanometer-sized particles to overcome Brownian diffusion and enable controlled transport remains a major issue for emerging applications in nanoscience and nanomedicine. Here, we address this challenge by demonstrating the fuel-free, reversible, and tunable active behavior of gold-silica (Au-SiO2) Janus nanoparticles (radius R=33 nm) induced by optical excitation. Using single particle tracking, we provide direct experimental evidence of self-thermophoresis, clearly distinguishing active motion from thermal noise. These light-driven Janus nanoparticles constitute a minimal yet robust photothermal system for investigating active matter and its manipulation at the nanoscale.

We investigate the dynamics of a dense raft of millimeter-sized granular particles at a vertically vibrated air-water interface, which displays a rich set of patterns and particle dynamics as we vary the vibration amplitude, frequency, and particle packing fraction. While the classical parametric instability with standing waves still occurs over a certain parameter space, the measured wave dispersion relations indicate an increasing role in the raft's emerging elasticity at higher packing fractions, which induces a decrease in the effective surface tension and an increase in an out-of-plane bending modulus. At higher vibration frequencies and lower amplitudes, we also identified a regime without standing waves in which individual particles exhibit thermal-like motion and transition from diffusive to sub-diffusive transport as the packing fraction increases. Glassy behaviors such as spatial and temporal heterogeneity in particle dynamics occur as well, which is analogous to supercooled liquids. When the vibration amplitude is increased starting in this supercooled regime, a large cavity eventually forms inside the raft with its size and shape related to the vibration frequency and the injected vibration energy. The cavitation results in the coexistence of free surface water waves inside the cavity and thermal-like particle motion in the raft.

The stability of chemically complex nanoparticles is governed by an immense configurational space arising from heterogeneous local atomic environments across surface and interior regions. Efficiently identifying low-energy configurations within this space remains a central challenge for first-principles-based materials discovery, particularly when the available reference data are limited. Here, we introduce a data-efficient and physically interpretable machine-learning framework based on a fragmented, layer-resolved descriptor that explicitly decomposes nanoparticles into surface, intermediate, and core environments using a topology-driven definition. This representation preserves a compact and fixed feature dimensionality while retaining spatial resolution, enabling controlled emphasis on different regions of the nanoparticle through physically motivated weighting schemes. Coupled with gradient-boosted decision tree models and a ranking-based learning strategy, the proposed framework enables accurate identification of the most stable nanoparticle configurations using only a few hundred density functional theory reference calculations. Ranking performance metrics demonstrate near-saturation of correlation, high top-k recall, and rapidly vanishing regret at moderate training-set sizes, highlighting the strong data efficiency of the approach. Beyond predictive performance, layer-weighting and SHAP-based interpretability analyses reveal how surface segregation, coordination topology, and local chemical disorder contribute differently to stability across spatial regions of the nanoparticle. These insights provide a transparent physical interpretation of the learned models and establish a natural pathway toward active learning-driven exploration of complex nanoparticle configurational spaces.

We introduce millimeter-wave silicon photonic crystal cavities as a versatile platform for the perturbative sensing of nanoscale materials. This dielectric-based platform is compatible with strong magnetic fields, opening avenues for studying quantum materials in extreme environments where superconducting cavities cannot operate. To establish the platform's performance, we cryogenically characterize a silicon photonic crystal cavity at 4.3 K, achieving a total quality factor exceeding $10^5$ for a 96 GHz mode. As a proof-of-concept for its sensing capabilities, we position a hexagonal boron nitride-multilayer graphene (hBN-MLG) heterostructure at an electric-field antinode of the cavity and measure the perturbative response at room temperature. The heterostructure induces a significant change in the cavity's resonance, from which we extract a total sample conductivity of approximately $5.1\times10^6$~S/m. These results establish silicon photonic crystal cavities as a promising platform for sensitive, on-chip spectroscopy of nanoscale materials at millimeter-wave frequencies.

Brownian motion provides access to hydrodynamic properties of nanoscale objects independent of their optical resolvability. Here, we present a diffusion-based approach to infer effective particle size distributions of DNA-functionalized magnetic nanoparticles (MNPs), consisting of a magnetic core and a polystyrene shell, in a regime where direct geometric sizing is limited by optical diffraction. Using multi-particle tracking microscopy, we analyze the Brownian dynamics of MNPs grafted with double-stranded DNA (dsDNA) of varying contour length under low-salt conditions. A physically motivated model is introduced that relates dsDNA contour length to an effective hydrodynamic diameter via an attenuated corona description. The measured diffusion coefficient distributions exhibit a systematic and monotonic dependence on dsDNA length in quantitative agreement with the model. While the tracked objects are predominantly dsDNA-mediated agglomerates rather than isolated nanoparticles, clustering does not obscure the length-dependent signal. Instead, the dsDNA corona determines the hydrodynamic scaling, whereas agglomeration mainly introduces an offset and distribution broadening. These results demonstrate that Brownian dynamics enables robust readout of biomolecular length scales even far below the optical resolution limit. The distribution-based approach is inherently tolerant to polydispersity and aggregation, making diffusion-based tracking a simple and promising strategy for future biotechnological and biomedical assays.

We propose a strategy to obtain conducting organic materials with fully spin-polarized Fermi surface, lying at a singular flat band, with antiferromagnetically coupled magnetic moments that reside in pi-orbitals of nanographenes. We consider a honeycomb crystal whose unit cell combines two different molecules with S=1/2: an Aza-3-Triangulene, a molecule with orbital degeneracy, and a 2-Triangulene. The analyzed system is half-metallic with a ferrimagnetic order, presenting a zero net total magnetic moment per unit cell. We combine density functional theory calculations with a Hubbard model Hamiltonian to compute the magnetic interactions, the bands, the intrinsic Anomalous Hall effect, and the collective spin excitations. We obtain very large intermolecular exchange couplings, in the range of 50 meV, which ensures room temperature stability. When the magnetization is off-plane, intrinsic spin orbit coupling in graphene opens up a topological gap that, despite being very small, leads to a quantized Hall conductance in the tens of mK range. Above 1 Kelvin, the system will behave like a half-metal with fully compensated magnetic moments, thereby combining two characteristics that make it ideal for spintronics applications.

We investigate charge and energy transport in monolayer graphene with smooth finite-range disorder, modeled by soft impurity potentials. Using a continuum Dirac model, we go beyond the Born approximation by computing the exact scattering matrix for individual impurities. This captures the full nonperturbative physics of smooth disorder. From the exact scattering data, we evaluate transport coefficients by solving the Boltzmann equation with energy-resolved phase shifts. We analyze electrical and electronic thermal conductivities versus carrier density and temperature, including deviations from the Wiedemann-Franz law. Our results reveal that finite-range disorder nontrivially modifies charge and heat currents, especially at low energies where perturbative methods fail. These findings provide a more accurate transport characterization for disordered Dirac materials and clarify how smooth disorder governs energy flow in graphene.

Matrix-product operators (MPOs) appear throughout the study of integrable lattice models, notably as the transfer matrices. They can also be used as transformations to construct dualities between such models, both invertible (including unitary) and non-invertible (including discrete gauging). We analyse how the local Yang--Baxter integrable structures are modified under such dualities. We see that the $\check{R}$-matrix, that appears in the baxterization approach to integrability, transforms in a simple manner. We further show for a broad class of MPOs that the usual Yang--Baxter $R$-matrix satisfies a modified algebra, previously identified in the unitary case, that gives a local integrable structure underlying the commuting transfer matrices of the dual model. We illustrate these results with two case studies, analysing an invertible unitary MPO and a non-invertible MPO both applied to the canonical XXZ spin chain. The former is the cluster entangler, arising in the study of symmetry-protected topological phases, while the latter is the Kramers--Wannier duality. We show several results for MPOs with exact MPO inverses that are of independent interest.

We revisit the construction of the fermionic path-integral representation of overdamped scalar Langevin processes with multiplicative white noise, focusing on the covariance of the generating functional under non-linear changes of variables. We identify the transformations of the auxiliary (commuting and anticommuting) variables that ensure covariance under such transformations. The subtleties induced by the non-differentiable trajectories of the stochastic dynamics are encoded in the fermionic statistics. Upon integrating out the auxiliary variables, we derive the Onsager-Machlup formulation, which agrees with the one recently obtained using a higher-order discretization scheme. In contrast to the latter, the construction proposed here is formulated directly in continuous time.

Epitaxial single-crystalline CsPbBr3 perovskite films on mica, prepared ex-situ, are explored using a low-temperature scanning tunneling microscope (STM) by probing the unoccupied electronic states of their surface in ultra-high vacuum (UHV) at 80 K. Light-assisted STM measurements under a broadband illumination with visible light were employed to enhance and stabilize surface conductivity. STM imaging across the surface of macroscopic bulk CsPbBr3 films reveals large flat terraces characterized by a specific type of surface reconstruction, consisting of parallel rows of U-shaped atomic nanostructures. These structures are spaced by 12 angstroms and exhibit an internal periodicity of 5.1 angstroms. Density functional theory (DFT) calculations reproduce the experimental observations and reveal a competition between different orthorhombic CsPbBr3(110) surface reconstructions: a Cs-rich structure, identified as the most energetically stable, and three alternative PbBr rich reconstructions, which are slightly higher in energy yet remain consistent with the STM data. Additional analyses that explicitly account for the mica substrate exclude the cubic CsPbBr3 phase and other orthorhombic surface orientations, while showing that variations in the mica surface termination do not alter the preferred CsPbBr3(110) reconstruction. This combined approach thereby confirms our assignment and resolves previous STM interpretations of CsPbBr3.

Magnetization dynamics is commonly described by the stochastic Landau-Lifshitz-Gilbert (LLG) equation. On picosecond timescales, inertial and open-system extensions of the LLG equation are necessary to interpret recent experiments. We show analytically and numerically that the standard LLG equation exhibits strictly positive entropy production rates, while inertial and open-system LLG dynamics display temporarily negative entropy production rates indicating non-Markovianity. Here we quantify the degree of non-Markovianity using established measures. Our numerical calculations show that the open-system LLG equation consistently exhibits the highest magnitude of non-Markovianity for different initial conditions and magnetic field orientations.

We curate a large-scale dataset of low frequency dielectric anisotropy values for low molecular weight liquid crystals. Using this dataset, we demonstrate that supervised machine-learning models can predict dielectric anisotropy with substantially improved accuracy (RMSE 2.6) compared to estimates obtained from the Maier-Meier relations using molecular properties from both the widely used semiempirical AM1 method (RMSE 9.7) and the modern r2scan-3c composite method (RMSE 11.2). Realising the potential of machine learning techniques for liquid crystalline materials requires carefully curated data to be accessible, and on this basis we propose a simple and standard template for reporting data.

Moiré semiconductors built on angle-aligned transition metal dichalcogenide (TMD) heterobilayers provide a physical realization of the Kondo lattice model, in which one TMD layer is prepared in a Mott insulating state supporting a lattice of local magnetic moments and the other layer in a metallic state supporting itinerant carriers. The artificial Kondo lattice enables the exploration of exotic states of matter near a continuously tunable Kondo breakdown. Here we report the emergence of a symmetry-broken Chern insulator at a moiré hole filling factor 4/3 in angle-aligned MoTe2/WSe2 moiré bilayers, which realize a chiral Kondo lattice. The symmetry-broken Chern insulator, which exhibits integer quantized Hall conductance at a fractional moiré filling, breaks the translational symmetry of the lattice spontaneously; it also appears only near a magnetic field-induced Kondo breakdown in the mixed-valence regime of the material. We further demonstrate that the magnetic field required to induce the Kondo breakdown and to stabilize the symmetry-broken Chern insulator is twist angle dependent. The results present new opportunities for exploring the subtle interplay between topology and Kondo interactions in moiré semiconductors.

We investigate autonomous dissipative charging mechanisms for lattice quantum batteries within the framework of open quantum systems. Focusing on engineered Markovian dissipation, we show that appropriately designed Lindblad jump operators can drive tight-binding systems into highly excited band-edge states, resulting in steady states with large ergotropy. We illustrate this mechanism in a one-dimensional tight-binding chain and in a two-dimensional graphene lattice. We find that disorder enhances the charging power, indicating that dissipation-assisted localization effects can be beneficial for energy storage. Moreover, the dissipative charging process remains robust against additional local dephasing noise. Our results establish bond dissipation as an effective and physically transparent mechanism for charging lattice quantum batteries in realistic open-system settings.

The classical Mpemba effect is the counterintuitive phenomenon where hotter water freezes faster than colder water due to the breakdown of Newton's law of cooling after a sudden temperature quench. The genuine nonequilibrium post-quench dynamics allows the system to evolve along effective shortcuts absent in the quasi-static regime. When the time needed for preparing the (classical or quantum) system in the hotter initial state is included, we encounter so-called Pontus-Mpemba effects. We here investigate multi-step Pontus-Mpemba protocols for open quantum systems whose dynamics is governed by time-inhomogeneous Lindblad master equations. In the limit of infinitely many steps, one arrives at continuous Pontus-Mpemba protocols. We study the crossover between the quasi-static and the sudden-quench regime, showing the presence of dynamically generated shortcuts achieved for time-dependent dissipation rates. Time-dependent rates can also cause non-Markovian behavior, highlighting the existence of rich dynamical regimes accessible beyond the Markovian framework.

These lecture notes are based on lectures given by the author at the Les Houches 2025 summer school on "Exact Solvability and Quantum Information". The central theme of these notes is to apply the philosophy of statistical mechanics to study the dynamics of quantum information in ideal and monitored random quantum circuits -- for which an exact description of individual realizations is expected to be generically intractable.

Interfacial orbital transport remains far less understood than its bulk counterpart despite its central role in orbitronic experiments. Here, we theoretically investigate the transmission and conversion of orbital angular momentum across a metallic interface using a model Hamiltonian incorporating crystal-field effects. We show that an injected orbital dipole moment undergoes pronounced oscillations driven by the crystal field and generates characteristic quadrupole moments determined by the orbital orientation relative to the interface. Unlike spin precession, the dipole relaxes toward a finite value away from the interface. We further quantify interfacial orbital memory loss and demonstrate that orbital absorption produces a sizable mechanical torque obtained from the orbital continuity equation.

Two-dimensional semiconductors have attracted considerable interest for integration into emerging quantum photonic networks. Strain engineering of monolayer transition-metal dichalcogenides (ML-TMDs) enables the tuning of light-matter interactions and associated optoelectronic properties, and generates new functionalities, including the formation of quantum dots (QDs). Here, we combine spatially resolved micro-photoluminescence ($μ$-PL) spectroscopy from cryogenic (4$\text{-}$94 K) to room temperature with micro-Raman spectroscopy at room temperature to investigate the strain-dependent emission energies of thousands of individual QDs in ML-WS$_2$ and ML-WSe$_2$, integrated across multiple heterostructures and a piezoelectric device. Compared with delocalized excitons, QDs in both materials exhibit enhanced strain sensitivities of their emission energies $-$ approximately fourfold in WS$_2$ and twofold in WSe$_2$ $-$ leading to pronounced broadening of the ensemble emission linewidth. Temperature-dependent $μ$-PL spectroscopy combined with dynamic strain tuning experiments further reveal that the enhanced strain sensitivity of individual QDs originates from strengthened interactions with low-energy phonons induced by quantum confinement. Our results demonstrate a versatile strain-engineering approach with potential for spectral matching across solid-state, atomic, and hybrid quantum photonic networks, and provide new insights into phonon-QD interactions in two-dimensional semiconductors.

Diffuse scattering of light from disordered assemblies is traditionally viewed as an uncontrollable broadband scattering background resulting in whitish hues. Here, we demonstrate that correlated disorder enables precise engineering of light scattering from 2D arrays of emitters resulting in strong observable colors. Our analytical framework shows that introducing controlled noise, with tunable probability density functions and correlations, generates three distinct scattering components: diffraction orders, diffuse background, and correlation halos. Correlation halos, often mistaken for broadened diffraction peaks, are independent features whose positions depend on correlation range and can appear between Bragg peaks. Crucially, they persist far beyond the regime where diffraction orders vanish. The noise probability density function provides an additional control: specific diffraction orders can be selectively suppressed while preserving others. This approach reproduces the scattering signatures of natural photonic structures, e.g. found in Morpho butterflies, and reveals multiple pathways from order to disorder, each with distinct optical properties. Our work provides a practical method for inverse design -finding the disorder that produces desired scattering patterns. This establishes diffuse scattering as a designable quantity, expanding the toolkit for metasurfaces and structural color beyond periodic and hyperuniform structures.

The current-induced vibrational dynamics is a key factor determining the stability of molecular nanojunctions. Beyond conventional Joule heating, a different mechanism caused by nonconservative current-induced forces has been predicted for models with multiple vibrational modes, leading to vibrational instabilities already at low bias voltages. So far, this mechanism has only been investigated in models with harmonic nuclear potentials. Consequently, a natural question is whether this effect can also be observed in more realistic models containing anharmonic nuclear potentials, and, if so, whether it has a measurable impact on observables such as the junction dissociation probability. In this work, we apply a mixed quantum-classical approach based on electronic friction and Langevin dynamics to various anharmonic two-mode systems. By performing Langevin simulations of the vibrational dynamics, we investigate the influence of anharmonicity on instabilities arising from nonconservative forces and the corresponding dissociation dynamics of the junction, as well as steady-state observables, such as the electronic current.

Reconstructing complex magnetization textures from nitrogen-vacancy (NV) magnetometry stray-field measurements presents a challenging inverse problem. In this work, we introduce a physics-informed method that addresses this by incorporating the full micromagnetic energy directly into the variational formulation. Built on a PyTorch backend, our forward model integrates an auto-differentiable finite-differences micromagnetic framework with FFT-based stray-field calculations and Fourier-space upward continuation. This enables efficient gradient-based optimization via the adjoint method and allows the sensor-sample distance to be treated as an optimization parameter. By doing so, we eliminate the experimental uncertainty arising from unknown NV implantation depths and surface oxidation layers. Validation on synthetic data demonstrates high-fidelity reconstruction of spin textures and precise sensor height estimation. Furthermore, when applied to NV measurements of the van der Waals ferromagnet $Fe_{3-x}GaTe_2$, the method reconstructs the previously unknown NV-sample distance and physically plausible magnetization textures, which accurately reproduce the experimental observations.

The stochastic dynamics of flagellar beating for micro-swimmers, such as flagellated cells, sperms and microalgae, is dominated by a feedback mechanism between flagellar shape and the rate of activation/de-activation of the $N \gg 1$ driving molecular motors. In the context of the so-called rigid filament models, where the axoneme is described by a single degree of freedom $X(t)$, we investigate the effect of direct coupling between the activity dynamics of adjacent motors, parametrized by $K \ge 0$. A functional Fokker-Planck equation for $X$ and the state of the $N$ motors is obtained. In the limit of small coupling $K \ll 1$, we derive a system of equations governing the dynamics of the Fourier modes of the active motor density, obtaining estimates for several observables and the fluctuations' quality factor $Q$. For larger $K$ we resort to numerical simulations. The effect of introducing the coupling $K>0$ is to increase characteristic times and the beating period. Moreover at large $K$s the limit cycle becomes bi-stable, with abrupt avalanches of the motor dynamics. Increasing $K$ is similar to what observed in the case $K=0$ when the confining elastic force is strongly reduced. The quality factor of fluctuations has a non-monotonic behavior: it first increases with $K$, then decreases. This is accompanied by the reduction and eventual disappearance of regions where the fraction of activated motor is nor $0$ neither $1$.

Controlling flow-induced segregation in a granular mixture is highly relevant to many industrial settings. To enhance mixing or promote segregation, the continuous gravity flow of a bidisperse granular mixture through a series of narrow vertical channels with exit slots is investigated. The bidisperse mixture is composed of two different sizes of particles, but of the same density. In dense flow, segregation occurs, leading to formation of bands. The bands of large particles appear at a distance away from the walls. This finding is in contrast to that in shear-driven segregation in a dense flow where large particles segregate towards the walls. Using a phenomenological model, it has been shown that rolling and bouncing induced segregation is the dominant mechanism. When cylindrical inserts are placed to modify flow patterns, that significantly influences segregation patterns. The symmetrical placement of a cylindrical insert close to the exit slot vanishes the bands and enhances mixing. However, with two inserts placed symmetrically and close to the exit slot, the degree of segregation in the reverse direction is greatly enhanced compared to that without insert. In the former, small particles accumulate in thin regions adjacent to the walls, and large particles comprise the bulk of the domain and the flowing stream. The heap formation above the insert in a narrow channel, when the insert is close to the exit, enhances mixing in one configuration, whereas it amplifies reverse segregation in the other.

The maturity of the chemical vapor deposition graphene-based device processing has increased from chip level demonstrations to wafer-scale fabrication in the past few years. Due to this wafer-scale, electrical characterization and analysis of the fabricated devices has become increasingly important to enable extraction of multiple parameters with minimal number of measurements for the quality control purposes critical for industrial uptake of 2D materials-based devices. As a crucial step, we demonstrate optimization of complementary metal-oxide semiconductor (CMOS) back-end-of-line (BEOL) compatible graphene field-effect transistor (GFET) fabrication and analysis including the gate stack, bottom contact, graphene patterning and encapsulation process steps. The analysis methods include atomic force microscopy, scanning electron microscopy and most importantly electrical characterization. The electrical characterization focuses on comparing different test structures and extraction methods for mobility, contact resistance, IV-curve hysteresis and doping parameters. The comparison shows that the selected measurement test structures and analysis methods can have a large impact on the extracted values and should thus be considered when comparing data sets between different sources. The analysis shows that the optimized process offers high device yield of 98 % with good doping uniformity, contact resistance and mobility as well as low IV-curve hysteresis values on 200 mm wafers.

Harnessing dipolar spin environments as controllable quantum resources is a central challenge in solid-state quantum technologies. Here, we report the observation of a coherent mesoscopic spin state in a disordered ensemble of substitutional nitrogen (P1) centers in diamond. An iterative Hartmann-Hahn protocol transfers polarization from dense nitrogen-vacancy (NV) centers to a P1 ensemble, yielding a 740-fold enhancement over room-temperature thermal equilibrium as revealed by differential readout. The resulting mesoscopic P1 spin ensemble exhibits collective Rabi oscillations and long-lived spin-lock and Hahn-echo coherences. We identify a crossover in the saturation polarization arising from the competition between coherent driving and local disorder, providing a quantitative measure of the system's intrinsic disorder. These results establish a foundation for utilizing dark electron spin ensembles as robust resources for quantum sensing and quantum many-body simulation.

We investigate the nature of quantum phase transitions in a (1+1)-dimensional field theory composed of $N$ copies of the Ising conformal field theory interacting via competing relevant perturbations. The field theory governs the competition between a mass term and an interaction involving the product of $N$ order-parameter fields, which is realized, e.g. in coupled Ising chains, two-leg spin ladders, and SO($N$)-symmetric spin chains. By combining a perturbative renormalization group analysis and large-scale matrix-product state simulations, we systematically determine the nature of the phase transition as a function of $N$. For $N=2$ and $N=3$, we confirm that the transition is continuous, belonging to the Ising and four-state Potts universality classes, respectively. In contrast, for $N \ge 4$, our results provide compelling evidence that the transition becomes first order. We further apply these findings to specific lattice models with SO($N$) symmetry, including spin-$1/2$ and spin-$1$ two-leg ladders, that realize a direct transition between an SO($N$) symmetry-protected topological phase and a trivial phase. Our results refine a recent conjecture regarding the criticality of transitions between SPT phases.

Shifting the point dipole from the center of a Stockmayer particle is a simple geometric modification that has been explored previously, yet its implications for liquid structure, dielectric response, and phase behavior remain incompletely understood. Here, we combine molecular dynamics simulations with analytical theory to provide a unified physical interpretation of how dipole displacement reshapes microscopic correlations and propagates to macroscopic thermodynamic properties. We show that dipole shifting breaks the fore-aft symmetry of the local electrostatic field, producing only modest changes in radial packing but strong alterations in angular structure within the first solvation shell. Enhanced alignment near the dipole head is accompanied by frustrated orientational correlations near the tail, leading to broader angular distributions and a shift away from axial configurations at strong coupling. These structural asymmetries weaken cooperative ordering and result in a systematic reduction of the dielectric constant, despite locally stronger interactions. For large shifts, the dielectric response approaches the Debye limit, indicating effective suppression of dipole-dipole correlations. The same geometric frustration governs vapor-liquid equilibria: while increasing dipole strength raises the critical temperature, even modest shifts disrupt the highly polarized liquid states that emerge at strong coupling and can suppress ferroelectric-like ordering. Predictions from a reparameterized COFFEE theory capture these trends within its domain of validity, highlighting the direct connection between local orientational structure and macroscopic observables. Overall, this work demonstrates that dipole location, not only magnitude, provides a powerful control parameter in dipolar fluids and offers a clear framework for understanding geometric frustration in electrostatic liquids.

Pinning magnetization in a ferromagnetic thin film is commonly realized through exchange biasing with an adjacent antiferromagnet. Field-cooling from above the Néel temperature is a reliable yet slow re-pinning method in exchange-biased systems. For on-demand reprogrammable devices, localized and rapid exchange bias repinning methods are essential. Recent work has shown that femtosecond laser pulses enable field-free reversal of exchange bias in tailored multilayer stacks. Contrary to field-cooling, our experiments with ultrafast excitation reach hitherto unexplored regimes in the exchange bias setting process. Here, we unravel these observations by considering both ultrafast magnetization dynamics on the femto- to picosecond timescale and slow heat-driven dynamics on millisecond timescales and upwards. We develop a microscopic framework of exchange bias setting in a polycrystalline antiferromagnetic thin film like IrMn that provides a complete description of the observations in our present experiments and those found in literature. We expand the use of our model by identifying material platforms and stack designs that lead to optimized performance, aiding further development of optically reprogrammable devices.

We demonstrate that graph-based models are fully capable of representing higher-order interactions, and have a long history of being used for precisely this purpose. This stands in contrast to a common claim in the recent literature on "higher-order networks" that graph-based representations are fundamentally limited to "pairwise" interactions, requiring hypergraph formulations to capture richer dependencies. We clarify this issue by emphasizing two frequently overlooked facts. First, graph-based models are not restricted to pairwise interactions, as they naturally accommodate interactions that depend simultaneously on multiple adjacent nodes. Second, hypergraph formulations are strict special cases of more general graph-based representations, as they impose additional constraints on the allowable interactions between adjacent elements rather than expanding the space of possibilities. We show that key phenomenology commonly attributed to hypergraphs -- such as abrupt transitions -- can, in general, be recovered exactly using graph models, even locally tree-like ones, and thus do not constitute a class of phenomena that is inherently contingent on hypergraphs models. Finally, we argue that the broad relevance of hypergraphs for applications that is sometimes claimed in the literature is not supported by evidence. Instead it is likely grounded in misconceptions that network models cannot accommodate multibody interactions or that certain phenomena can only be captured with hypergraphs. We argue that clearly distinguishing between multivariate interactions, parametrized by graphs, and the functions that define them enables a more unified and flexible foundation for modeling interacting systems.

We study once-excited random walks on general trees, modeled by placing a single "cookie" at each vertex. Each cookie acts as a metaphorical reward that is consumed upon the first visit to the vertex where the cookie is placed. On that initial visit, the walk is in an excited state and behaves like a biased random walk. Once the cookie is consumed, the process reverts to a symmetric random walk on all subsequent visits. We consider a random environment in which the bias parameters are independent random variables. We prove that the process exhibits a sharp phase transition between transience and recurrence on general trees with polynomial growth, where the critical threshold is determined by the branching-ruin number of the tree.

This note provides a simple derivation of the overdamped approximation for kinetic (or underdamped) equilibrium Langevin dynamics, in cases where certain coefficients depend on the position variable. The equivalent small-mass limit of these dynamics, known as the Kramers--Smoluchowski approximation, in the case of a state-dependent friction coefficient, has been previously studied by a variety of approaches. Our new approach uses hypocoercivity estimates, which may be of interest in their own right, and lead to a very direct derivation, providing in particular a clear explanation of the ``noise-induced drift'' term in the overdamped equation in the case of a state-dependent friction term. Using the same approach, we also treat several effective kinetic dynamical models derived from a coarse-graining approximation of a high-dimensional system, as well as a class of kinetic dynamics with position-dependent mass matrices. All of these models are relevant to applications in computational chemistry. We finally identify a mistake in a related work and suggest a solution.

Chirality in polymeric systems enables a wide range of emergent optical, mechanical, and transport phenomena, yet a unified framework that quantitatively connects molecular-scale geometry to chiral behavior remains lacking. Existing theoretical descriptions typically emphasize either continuum models, such as the helical wormlike chain (HWLC), which neglect intermolecular interactions, or mesophase-level theories, which obscure the role of molecular geometry. In this work, we introduce a comprehensive framework for quantifying chirality in helical polymers by extending the Kremer-Grest bead-spring model to explicitly map intrinsic curvature and torsion onto bond angle and dihedral potentials. We establish direct theoretical relationships between helical parameters such as pitch and radius, and connect them to a normalized, dimensionless chirality characteristic, $χ$ that captures local geometric correlations absent from conventional HWLC descriptions. Furthermore, using molecular dynamics simulations, we systematically quantify the influence of excluded volume interactions and thermal fluctuations on helical geometry and chirality, dispelling the common assumption that monotonic increases in chirality are associated only with decreasing pitch. Finally, we present a coarse-graining procedure that facilitates a direct comparison between experimental helical polymers and the Kremer-Grest helical chain, demonstrating quantitative agreement across a diverse set of polymer classes. This unified geometric and particle-based description provides a predictive roadmap for selecting and engineering chiral Kremer-Grest models and offers a general platform for designing polymeric materials with controlled and tunable chirality.

One of the great miracles of random matrix theory is that, in the $N \to \infty$ limit, many otherwise intractable matrix problems with horrendously complicated finite-$N$ expressions admit remarkably simple and elegant asymptotic solutions. In this paper, we illustrate this phenomenon in the context of spectral boundaries (or spectral edges) for deformed random matrices. Specifically, we consider matrices of the form $\mathbf{A} + \mathbf{B}$, where $\mathbf{A}$ is a deterministic $N\times N$ matrix (not necessarily Hermitian) and $\mathbf{B}$ is a rotationally invariant random matrix. In the large-$N$ limit, we show that the complex eigenvalue distribution of $\mathbf{A} + \mathbf{B}$ satisfies remarkably simple boundary equations that depend on the $\mathcal{R}_1$ and $\mathcal{R}_2$ transforms of $\mathbf{B}$. We illustrate our results on several explicit random matrix ensembles and support them with numerical simulations.

We study the stability of the ordered phase of compressible polar flocks against the nucleation of counter-propagating droplets, using a combination of analytical theory, microscopic and hydrodynamic simulations. For discrete-symmetry flocks, such droplets are known to always grow and propagate, making the ordered phase metastable. We explain how, on the contrary, continuous symmetry can stabilize the ordered phase at small enough noise by destabilizing the leading edge of growing droplets. Flocking models with continuous symmetries thus have a lower critical dimension than their discrete-symmetry counterparts, in contrast to equilibrium physics.

Plates generally admit six deformation modes: three of which are high in strain energy, stretch the plate's midsurface and are called membrane modes; and three are low-energy, bend the midsurface without stretching it and are called bending modes. For origami tessellations, and other corrugated compliant thin shells, the modes are mixed and it is no longer clear what modes, if any, are low in energy in the sense that they are inextensional. Here, it is shown, by direct construction of closed-form solutions, that when the midsurface is a surface of translation, there exists three infinitesimally inextensional deformation modes that correspond to (1) stretching, with an effective Poisson's effect; (2) bending, with an effective synclastic or anti-clastic effect; and to (3) twisting. The provided expressions are valid irrespective of surface regularity and, in particular, properly handle any creases be them straight or curved. The results provide a powerful benchmark for the validation of numerical methods and further insight into the elastic stiffness of thin corrugated compliant shells.

Hydrodynamic electrons in high-mobility graphene devices have demonstrated great potential in establishing an electronic analogue of relativistic quantum fluid in solid-state systems. One of the key requirements for observing viscous electron flow in an electronic channel is a large momentum-relaxation path, a process primarily limited by electron-impurity/phonon scattering in graphene. Over the past decade, multiple complex device geometries have been successfully employed to suppress momentum-relaxing scattering mechanisms; however, experimental observations have been found to be sensitive to the device fabrication process and architecture, raising questions about the signature of electron hydrodynamics itself. Here, we present a study on multiple ultra-clean graphene field-effect transistors (FETs) in a simple, rectangular four-terminal device architecture. Using electrical transport measurements, we have characterised the pristine quality of the graphene FETs and examined the variation of electrical resistance in the doped regime as a function of carrier density and temperature. Our results reveal strong device-dependent variability even in the most simple architecture that we attribute to competing momentum-conserving and momentum-relaxing scattering mechanisms, as well as coupling to contacts. Further, we have proposed a phenomenological method for analysing the results, which yields transport parameters in accordance with recent experiments. This simple experimental strategy and analysis can serve as an efficient tool for extracting the viscous electronic contribution in state-of-the-art high-mobility graphene FETs.

Nonreciprocal theories are used to model a broad array of non-equilibrium phenomena found in nature ranging from biological systems like networks of neurons to the behavior of overflowing water fountains. This includes systems broadly classified as active matter systems. We show that the stochastic theories which describe nonreciprocal interactions can be mapped into quantum field theories described by a supersymmetric action with a single supercharge. The theories are generically non-Hermitian. This generalizes the past work of Parisi and Sourlas on reciprocal theories, which model systems with interactions derived from potentials.

Dissipation is commonly regarded as an obstacle to quantum control, as it induces decoherence and irreversibility. Here we demonstrate that dissipation can instead be exploited as a resource to reshape the dynamics of interacting quantum systems. Using an experimentally realizable Bose-Josephson junction containing two bosonic species, we demonstrate that dissipation enables distinct dynamical behaviors: synchronized phase-locked oscillations, transient chaos with long-time coherence recovery, and steady-state chaos. The emergence of each behavior is determined by experimentally tunable parameters. At weak interactions, the two components synchronize despite dissipation, exhibiting long-lived coherent oscillations reminiscent of a boundary time crystal. Stronger interactions induce a dissipative phase transition into a self-trapped regime accompanied by chaotic dynamics. Remarkably, dissipation regulates the lifetime of chaos and enables the recovery of coherence at long times. By introducing a controlled tilt between the wells, transient chaos can be converted into persistent steady-state chaos. We further show that standard spectral diagnostics fail to distinguish between the two chaotic regimes, revealing that spectral statistics primarily reflect short-time instability. These results establish dissipation as a powerful tool for engineering dynamical phases, restoring quantum coherence, and controlling the duration of chaotic behavior and information scrambling.

Motivated by the necessity to UV-regularise entanglement entropy, we present a spectral method for calculating the entropy of quasifree states, for both bosonic and fermionic field theories. This construction is defined in spacetime rather than on a hypersurface, enabling the covariant regularisation of entropies, and its calculation in generic spacetime regions. We derive these formulae, which have previously appeared in the literature, in a new manner and highlight certain aspects of them, such as their connection to the density matrix and its eigenvalues. The spacetime nature of the formulation makes it particularly apt in the context of semiclassical and quantum gravity and in connection to black hole entropy. Another useful property of the formulation is its application to settings where no notion of a Cauchy surface exists, such as in the causal set theory approach to quantum gravity. We show example applications of the formulae which demonstrate their ability to reproduce known results. We also show a calculation in a causal set in $1+1$ dimensions which makes use of several of the unique and useful features of the formalism. In this last example, we obtain a novel result of a slightly modified entanglement entropy scaling coefficient, giving a possible signature of spacetime discreteness.

Non-Hermitian systems exhibit unique spectral properties, including the non-Hermitian skin effect and exceptional points, often influenced by boundary conditions. The modulation of these phenomena by generalized boundary conditions remains unexplored and not understood. Here, we analyze the Hatano-Nelson model with generalized boundary conditions induced by complex hopping amplitudes at the boundary. Using similarity transformations, we determine the conditions yielding real energy spectra and skin effect, and identify the emergence of exceptional points where spectra transition from real to complex. We demonstrate that tuning the boundary hopping amplitudes precisely controls the non-Hermitian skin effect, i.e., the localization of eigenmodes at the lattice edges. These findings reveal the sensitivity of spectral and localization properties to boundary conditions, providing a framework for engineering quantum lattice models with tailored spectral and localization features, with potential applications in quantum devices.

We characterize a singularity in the equal-time three-point density correlations that is generic to two-dimensional interacting Fermi liquids. In momentum space where the three-point correlation is determined by two wavevectors $\mathbf{q}_1$ and $\mathbf{q}_2$, the singularity takes the form $|\mathbf{q}_1\times\mathbf{q}_2|$. We explain how this singularity is sharply defined in a long-wavelength collinear limit. For a non-interacting Fermi gas, the coefficient of this singularity is given by the quantized Euler characteristic of the Fermi sea, and it implies a long-range real space correlation favoring collinear configurations. We show that this singularity persists in interacting Fermi liquids, and express the renormalization of the coefficient of singularity in terms of Landau parameters, for both spinless and spinful Fermi liquids. Implications for quantum gas experiments are discussed.

Mapping finite-temperature dynamical phase diagrams of quantum many-body models is a necessary step towards establishing a framework of far-from-equilibrium quantum many-body universality. However, this is quite difficult due, in part, to the severe challenges in representing the volume-law entanglement that is generated under nonequilibrium dynamics at finite temperatures. Here, we address these challenges with an efficient equilibrium quantum Monte Carlo (QMC) framework for computing the finite-temperature dynamical phase diagram. Our method uses energy conservation and the self-thermalizing properties of ergodic quantum systems to determine observables at late times after a quantum quench. We use this technique to chart the dynamical phase diagram of the $2+1$D quantum Ising model generated by quenches of the transverse field in initial thermal states. Our approach allows us to track the evolution of dynamical phases as a function of both the initial temperature and transverse field. Surprisingly, we identify quenches in the ordered phase that cool the system as well as an interval of initial temperatures where it is possible to quench from the paramagnetic (PM) to ferromagnetic (FM) phases. Our method gives access to dynamical properties without explicitly simulating unitary time evolution, and is immediately applicable to other lattice geometries and interacting many-body systems. Finally, we propose a quantum simulation experiment on state-of-the-art digital quantum hardware to directly probe the predicted dynamical phases and their real-time formation.

Relativistic equilibrium is described by the inverse-temperature four-vector $β^μ= u^μ/(k_B T_0)$ rather than by a frame-dependent scalar temperature. We show that $β^μ$ can be reconstructed directly from electromagnetic fluctuations emitted by a drifting medium, without external probes, spectral lines, or absolute intensity calibration. A Lorentz boost converts isotropic rest-frame noise into correlated electric and magnetic fields, producing a gain-independent fluctuation observable that yields the drift velocity purely from stochastic data. Combined with angle-resolved noise spectra governed by the covariant fluctuation--dissipation theorem, this enables full reconstruction of $β^μ$ using electromagnetic measurements alone. Monte Carlo analysis demonstrates percent-level accuracy at realistic signal-to-noise ratios, and feasibility estimates indicate sub-microsecond integration times for laboratory plasmas. To our knowledge, this constitutes the first method that reconstructs the covariant thermal state $β^μ$ of a relativistic medium from passive stochastic fields alone, without absolute calibration, spectral lines, or external probes. These results establish vacuum electromagnetic fluctuations as a direct operational probe of relativistic equilibrium.

Recent experimental investigations have demonstrated that doping a semiconductor is a route to increase the thermal boundary conductance at metal/semiconductor interfaces. In this work, the influence of the electrical properties on heat transfer across metal/doped semiconductor junctions is investigated. Specifically, thermal boundary conductance at the interfaces between p- and n-doped silicon and titanium is measured by employing frequency-domain photothermal radiometry under varying external conditions. The influence of the doping level of the semiconductor, the barrier height, and the space charge area is analyzed. In particular, a 40% increase in the interface thermal conductance with the application of a current at n-doped silicon/titanium interfaces is reported. The enhancement of the thermal boundary conductance is explained by the shrinking of the surface charga area induced by the electric current. This study opens the way to modulating interfacial heat transfer at metal/semiconductor interfaces through fine tuning of electrical effects.

We digitally simulate quantum many-body dynamics in emergent curved backgrounds using 80 superconducting qubits on IBM Heron processors. By engineering spatially varying couplings in the spin-$\frac12$ XXZ chain, consistent with the low-energy description of the model in terms of an inhomogeneous Tomonaga-Luttinger liquid, we realize excitations that follow geodesics of an effective metric inherited from the underlying spatial deformation. Following quenches from Néel and few-spin-flip states, we observe curved light-cone propagation, horizon-induced freezing in the local magnetization, and position-dependent oscillation frequencies set by the engineered spatial deformation. Despite strong spatial inhomogeneity, unequal-time correlators reveal ballistic quasiparticle propagation in the spin chain. These results establish large-scale digital quantum processors as a flexible platform for detailed and controlled exploration of many-body dynamics in tunable and synthetic curved spacetimes.

The development of efficient hydrogen storage materials is crucial for advancing the hydrogen economy and meeting the U.S. Department of Energy's targets of 6.5 wt% and 50 g H₂/L for automotive applications. We present a computational study of ferrocene-functionalized covalent organic frameworks (COFs) for hydrogen storage. Following the <b>M</b>ulti-binding <b>S</b>ites <b>U</b>nited in <b>C</b>ovalent-<b>O</b>rganic <b>F</b>ramework (MSUCOF) approach, we introduce MSUCOF-4-FeCp, designed by incorporating ferrocene (FeCp₂) moieties into IRCOF-102. Notably, it achieves exceptional performance with gravimetric and volumetric uptakes of 18.0 wt% and 72.6 g H₂/L at 298 K and 700 bar. The material exhibits optimal binding energies (15-20 kJ/mol) ensuring both high storage capacity and deliverable hydrogen under practical conditions. This work establishes ferrocene functionalization as a cost-effective alternative to precious metal incorporation in COFs.

The synchronization of rhythms is ubiquitous in both natural and engineered systems, and the demand for data-driven analysis is growing. When rhythms arise from limit cycles, phase reduction theory shows that their dynamics are universally modeled as coupled phase oscillators under weak coupling. This simple representation enables direct inference of inter-rhythm coupling functions from measured time-series data. However, strongly rhythmic chaos can masquerade as noisy limit cycles. In such cases, standard estimators still return plausible coupling functions even though a phase-oscillator model lacks a priori justification. We therefore extend the phase description to the chaotic oscillators. Specifically, we derive a closed equation for the phase difference by defining the phase on a Poincaré section and averaging the phase dynamics over invariant measures of the induced return maps. Numerically, the derived theoretical functions are in close agreement with those inferred from time-series data. Consequently, our results justify the applicability of phase description to coupled chaotic oscillators and show that data-driven coupling functions retain clear dynamical meaning in the absence of limit cycles.

Dynamical structures in the circular restricted three-body problem (CR3BP) are fundamental for designing low-energy transfers, as they aid in analyzing phase space transport and designing desirable trajectories. This study focuses on lobe dynamics to exploit local chaotic transport around celestial bodies, and proposes a new method for systematically designing low-energy transfers by combining multiple lobe dynamics. A graph-based framework is constructed to explore possible transfer paths between departure and arrival orbits, reducing the complexity of the combinatorial optimization problem for designing fuel-efficient transfers. Based on this graph, low-energy transfer trajectories are constructed by connecting chaotic orbits within lobes. The resulting optimal trajectory in the Earth--Moon CR3BP is then converted into an optimal transfer in the bicircular restricted four-body problem using multiple shooting. The obtained transfer is compared with existing optimal solutions to demonstrate the effectiveness of the proposed method.

In the present work we explore features of single and pairs of solitary waves in a fractional variant of the nonlinear Schr{ö}dinger equation. Motivated by the recent experimental realization of arbitrary fractional exponents, upon quantifying the tail properties of such coherent structures, we detail their destabilization when the fractional exponent $α$ acquires values $α<1$ and showcase how the relevant destabilization is associated with collapse type phenomena. We then turn to in- and out-of-phase pairs of such waveforms and illustrate how they generically exist for arbitrary $α$ when we cross the harmonic limit, i.e., for $α>2$. Importantly, we use the parameter $α$ as a ``bifurcation parameter'' in order to connect the harmonic ($α=2$) and biharmonic ($α=4$) limits. Remarkably, not only do we retrieve the instability of all solitonic pairs in the biharmonic case, but showcase a stabilization feature of particular branches of such multipulses that is {\it unique} to the fractional case and does not arise -- to our knowledge -- for integer multi-pulse settings. We explain systematically this stabilization via spectral analysis and expand upon the implications of our results for the potential observability of fractional multipulse solitary waves.

We consider stationary states of an effectively one-dimensional Bose-Einstein condensate in a quasiperiodic lattice. We formulate sufficient conditions for a one-to-one correspondence between the stationary states with a fixed chemical potential and the set of bi-infinite sequences over a finite alphabet. These conditions can be checked numerically. A bi-infinite sequence can be interpreted as a code of the corresponding solution. A numerical example demonstrates the coding approach using an alphabet of three symbols.

This paper explores cellular automata (CA) constructed from Yang-Baxter maps over finite fields $F_{2^n}$. We define $R$-matrices using a map $f$ on $F_{2^n}$ and establish necessary and sufficient conditions for $f$ to satisfy the Yang-Baxter equation. We show that these conditions become remarkably streamlined in characteristic two. An exhaustive search for bijective solutions in fields of order 4, 8, and 16 yields 16, 736, and 269,056 maps, respectively. Analysis of the resulting CA under helical boundary conditions reveals a consistent alignment between the temporal period and the field order. We propose the conjecture that this periodic identity holds generally for $F_{2^n}$, supported by analytical proofs for $n=2$ and $n=3$. Our results further indicate that bijectivity is a fundamental requirement for this periodic behavior.

Abrupt transitions ("tipping") in nonlinear dynamical systems are often accompanied by changes in the geometry of the attracting set, but quantifying such changes from partial and noisy observations in high-dimensional systems remains challenging. We address this problem with a sequential diagnostic framework, Data Assimilation-High dimensional Attractor's Structural Complexity (DA-HASC). First, this method reconstructs system's high-dimensional state using data assimilation from limited and noisy observations. Second, we quantify a structural complexity of the high-dimensional system dynamics from the reconstructed state by manifold learning. Third, we capture underlying changes in the system by splitting the reconstructed timeseries into sliding windows and analyzing the changes in the temporally local attractor's structural complexity. The structural information is provided as graph Laplacian and measured by Von Neumann entropy in this framework. We evaluate DA-HASC on both synthetic and real-world datasets and demonstrate that it can detect tipping under high-dimensionality and imperfect system knowledge. We further discuss how this framework behaves across different tipping mechanisms.

The existence of generalized steady states (GSSs) in nonlinear mechanical systems under moderate temporally aperiodic forcing has only been shown recently. Here we derive systematic expansions for such GSSs and construct a numerical algorithm that yields explicit and arbitrarily refinable approximations for GSSs without the need for an initial convergence period. This is to be contrasted with a direct numerical integration of the system, whose convergence is hard to assess or is even undefined for short, transient forcing. When at least the linear part of the equations of motion is known, our GSS algorithm outperforms available data-driven neural-network-based techniques for predicting forced response in structural dynamics problems. In a fully equation-driven setting, our GSS computations are shown to be faster than a direct numerical integration of forced nonlinear finite-element models of beams and shells.

Phase-only computer-generated holography (CGH) seeks a phase pattern for a spatial light modulator (SLM) whose propagated optical field reproduces a desired intensity distribution. In the far-field (Fraunhofer) regime, optical propagation reduces to a Fourier transform, such that each hologram pixel contributes to the entire reconstructed intensity distribution. When restricted to phase-only modulation, intensity must be shaped through global phase interference effects, making the inverse mapping from target intensity to phase highly non-linear and sensitive to local minima. We present a proof-of-concept physics-in-the-loop approach in which a transformer maps a target intensity image to a phase-only SLM field and is trained end-to-end through exact FFT-based propagation embedded directly within optimization. We further observe that patch tokenization strongly shapes the optimization geometry: coarse tokenization acts as an implicit spectral regularizer that stabilizes training and suppresses checkerboard-like attractors, while finer tokenization increases spatial degrees of freedom but benefits from curriculum or hierarchical refinement. Despite training on limited primitives and restricted digit subsets, the learned generator exhibits out-of-distribution (OOD) generalization to unseen digits and hand-drawn target patterns. These results suggest that transformer architectures, whose self-attention enables global token interactions, are a natural fit for far-field holography and provide a viable foundation for scalable physics-grounded hologram generation.

We consider rectilinear free-space propagation of electromagnetic wavepackets using electromagnetic field theory, scalar wavepacket propagation, and quantum-mechanical formalism. We demonstrate that spatially localized wavepackets are inherently characterized by a subluminal group velocity and a superluminal phase velocity, whose product equals $c^2$. These velocities are also known as the 'energy' and 'momentum' velocities, introduced by K. Milton and J. Schwinger. We illustrate general conclusions by explicit calculations for Gaussian beams and wavepackets, and also highlight subtleties of the quantum-mechanical description based on the 'photon wavefunction'.

Body-centered Cubic (BCC) lattice structures demonstrate promising performance for applications that require simultaneous mechanical energy absorption and thermal management. However, current optimization approaches are typically confined to single-domain objectives, such as mechanical parameters like impact energy and peak stress, neglecting the role of multiple physics in real-world performance. To address this, we propose a multi-objective optimization framework for density-graded BCC lattices that effectively dissipates heat while maximizing absorbed impact energy. A parametric three-zone lattice configuration is investigated to explore various trade-offs between mechanical and thermal properties. Each design is evaluated through independent impact and forced-convection simulations using commercial solvers. Specific Energy Absorption (SEA) and peak stresses at the distal end quantify impact absorption performance, while the Nusselt number and pressure drop characterize thermal dissipation performance. Surrogate models constructed from this data enable multi-objective optimization via Goal Programming to identify an optimal design. Two Pareto-optimal lattice designs are identified with reduced pressure drop and peak stress, underlining the superiority of strategic density gradation. Analysis of the optimal designs reveals how material distribution and geometric design variables influence mechanical-thermal trade-offs, establishing quantitative design guidelines for lattice structures in this multi-physics application.

Many biological microswimmers can modulate their swimming gait to achieve directional control of motility, especially when performing steering towards specific directional cues. This can be achieved without the need for obvious morphological or structural asymmetries in the form of the organism, or in the number or organisation of propulsion-generating appendages such as cilia. In this work, we identify and validate a core principle of asymmetric turning in biflagellate microswimmers: propulsive forces interact constructively to drive translation whilst interacting destructively to drive rotation. We explore the ramifications of this tunable biflagellar swimming mechanism across a range of systems, from a simple, back-of-the-envelope model to a detailed computational representation of an exemplar swimmer. This leads to a markedly general quantitative relation between key drivers of asymmetry, such as ciliary beat frequency, and the curvature of emergent trajectories. We discuss how the model green alga Chlamydomonas reinhardtii, which actuates its two cilia in a symmetric breaststroke for forward swimming, may exploit this feature for phototaxis. Finally, we validate our predictions in a Chlamydomonas-inspired robophysical model, implementing closed-loop control to achieve phototactic turning.

We present a new method for solving the relativistic Vlasov--Maxwell system of equations, applicable to a wide range of extreme high-energy-density astrophysical and laboratory environments. The method directly discretizes the kinetic equation on a high-dimensional phase-space grid using a discontinuous Galerkin finite element approach, yielding a high-order, conservative numerical scheme that is free from the Poisson noise inherent to traditional Monte-Carlo methods. A novel and flexible velocity-space mapping technique enables the efficient treatment of the wide range of energy scales characteristic of relativistic plasmas, including QED pair-production discharges, instabilities in strongly magnetized plasmas surrounding neutron stars, and relativistic magnetic reconnection. Our noise-free approach is capable of providing unique insight into plasma dynamics, enabling detailed analysis of electromagnetic emission and fine-scale phase-space structure.

Free-electron interactions with light and matter have long served as a cornerstone for exploring the quantum and ultrafast dynamics of material excitation. In recent years, this paradigm has evolved from a classical description of radiation and acceleration toward a fully quantum framework, transforming our understanding of light-matter interactions at the single-electron level. These advances have opened new opportunities in high-resolution imaging, ultrafast spectroscopy, interferometry, and the coherent shaping of electron wavepackets. This review surveys stimulated interactions between slow electrons and light, encompassing free-space and near-field mediated mechanisms. We discuss how free-space optical fields coherently modulate electron momentum and energy, and how near-field coupling in nanophotonic and plasmonic structures enables strong, phase-matched, efficient momentum exchange with the electron wavepacket. We further describe electron recoil, which is significant in the slow-electron regime, and temporal and spatial wavepacket shaping that enhances coupling efficiency and extends access to quantum-coherent regimes. Building on these foundations, we outline emerging frameworks including hybrid optical-electrostatic modulation, ponderomotive laser-based aberration correction, and optical electron interferometry. By unifying these developments, stimulated electron-light interactions provide a versatile route to precise beam control, quantum-state engineering, and tailored light-matter coupling, with implications for ultrafast spectroscopy, nanoscale metrology, attosecond pulse generation, electron-photon entanglement, and the creation of nonclassical states of light.

The development of single-photon sources has been nothing but rapid in recent years, with quantum emitter-based systems showing especially impressive progress. In this article, we give an overview of the developments in single-photon sources based on single molecules. We will introduce polycyclic hydrocarbons as the most commonly used emitter systems for the realization of an organic solid-state single-photon source. At cryogenic temperatures this special class of fluorescent molecules demonstrates remarkable optical properties such as negligible dephasing, indefinite photostability, and high photon rates, which make them attractive as fundamental building blocks in emerging quantum technologies. To better understand the general properties and limitations of these molecules, we discuss sample preparation, light collection strategies and relevant emitter parameters such as absorption and emission spectra, lifetime, and dephasing. We will also give an overview of light extraction strategies as a crucial part of a single-photon source. Finally, we conclude with a look into the future, displaying current challenges and possible solutions.

Singular optics has emerged as an important research area with diverse applications, yet controlling optical singularities in nanophotonic emitters is typically limited by fixed subwavelength geometries and diffraction-limited control. Here, we circumvent this limitation and demonstrate an all-optical mechanism for reconfiguring far-field singularities in a photonic crystal laser. The underlying principle involves optical pumping, which creates a mesoscopic potential landscape whose spatial variations are slow compared to the lattice period. Such a potential localizes a Bloch band into trapped states whose envelope functions, and thus far-field singularity textures, are defined by the pump geometry. Using a honeycomb photonic crystal that supports a symmetry-protected bound state in the continuum, we achieve room-temperature telecom-band lasing with real-space polarisation singularities that are reconfigurable in both number and position, while the intrinsic momentum-space singularity at the $Γ$-point is preserved. The experimental observations align quantitatively with an analytical framework that combines the Bloch mode of the structure and envelope function theory, establishing envelope engineering as a versatile route to programmable singular-light emission in active photonic lattices.

Herd immunity is shaped not only by the infection capacity of a spreading epidemic or the contact structure of the hosting population, but also by how and under what circumstances individuals acquire immunity. Immunization strategies may interact with ongoing non-pharmaceutical interventions, which commonly aim to reduce social contact numbers. We demonstrate that these interactions can induce unexpectedly strong and counterintuitive effects on herd immunity. We explore these phenomena on spatially embedded contact networks and uncover a reversal in the relative effectiveness of disease- versus vaccine-induced immunization schemes, highlighting the average number of contacts as a critical determinant of emerging herd immunity. In sparse geometric networks with limited degree heterogeneity, uniform vaccination proves most effective; however, as average contact numbers increase, naturally acquired immunity ultimately becomes the better strategy. We show that this phenomenon may emerge not only in synthetic networks but also in real-world mixing networks, observed during non-pharmaceutical intervention periods across multiple states of the United States.

The characterization of collective behavior and nonequilibrium phase transitions in quantum systems is typically rooted in the analysis of suitable system observables, so-called order parameters. These observables might not be known a priori, but they may in principle be identified through analyzing the quantum state of the system. Experimentally, this can be particularly demanding as estimating quantum states and expectation values of quantum observables requires a large number of projective measurements. However, open quantum systems can be probed in situ by monitoring their output, e.g. via heterodyne-detection or photon-counting experiments, which provide space-time resolved information about their dynamics. Building on this, we present a machine-learning approach to detect nonequilibrium phase transitions from the measurement time-records of continuously-monitored quantum systems. We benchmark our method using the quantum contact process, a model featuring an absorbing-state phase transition, which constitutes a particularly challenging test case for the quantum simulation of nonequilibrium processes.

Density functional theory is the workhorse of modern electronic structure calculations, with wide-ranging applications in chemistry, physics, materials science, and machine learning. At its heart lies the exchange-correlation functional, a quantity which exactly encapsulates the many-body effects stemming from the quantum mechanical interactions between the electrons. Yet, the exact functional is unknown, and computationally tractable approximations are therefore necessary for practical applications. Over the past six decades, hundreds of density functional approximations have been proposed with varying degrees of accuracy and computational efficiency. This review surveys the theoretical foundations of semi-local functionals, including local density approximations, generalized gradient approximations, and meta-generalized gradient approximations. We provide a comprehensive, consistently organized discussion that consolidates both historical developments and recent advances in this field. Beginning with the essential concepts of Kohn-Sham density functional theory, we present the construction principles of semi-local exchange-correlation functionals. Special attention is given to the physical motivations underlying functional development, the mathematical properties that guide their construction, and the practical considerations that determine their applicability across different chemical and physical systems. This work is intended to serve as both a introduction for newcomers to the field and a comprehensive reference for practitioners. By consolidating the extensive literature on semi-local functionals and providing a unified framework for understanding their construction and application, we aim to facilitate further developments in density functional approximations and their use in tackling the diverse challenges of modern computational chemistry and condensed matter physics.

We present g4chargeit, a kinetic Monte Carlo framework built on Geant4 for self-consistent simulation of time-dependent electrostatic charging in dielectric materials. The model explicitly incorporates stochastic particle transport and scattering processes using validated Geant4 cross-sections, while self-consistently evolving the electric potential and field. As a representative application, we simulate the charging of regolith grains under average dayside conditions on the Moon. The surface of the Moon, in addition to other airless planetary bodies, are regularly exposed to solar ultraviolet photons and solar-wind plasma, creating a radiation environment in which electrostatic interactions among regolith grains become significant. Until now, simulations of regolith charging have often relied on analytical approximations that oversimplify grain geometry and interaction mechanisms. Our Geant4-based simulations reveal charge accumulation within intergrain micro-cavities, leading to repulsive electrostatic forces consistent with experimental observations. The framework establishes a multiscale approach that links microscopic scattering events to the continuity equation of surface charge density and to the formation of macroscopic surface charge patches in complex grain geometries. Although demonstrated here for planetary regolith, the method is general and applicable to a broad range of dielectric charging problems. The code is openly available at https://github.com/kgandhi63/g4chargeit.git.

Breaking Kirchhoff's law of thermal radiation yields new opportunities in one-way radiative thermal transport and circuitry. We investigate its consequences in the far-field regime in cylindrical cavities, by employing a specular ray-tracing algorithm. At thermal equilibrium, we show that violation of Kirchhoff's law yields non-vanishing heat rectification coefficients within different sections of the cavity, which can be tuned for perfect rectification and circulation, while internal monochromatic currents vanish due to the intrinsic coupling between emission and absorption at specular surfaces. This constraint is lifted under nonequilibrium conditions, where rotational heat fluxes within the cavity can be precisely controlled by appropriately combining reciprocal and nonreciprocal materials. These findings open new avenues for thermal management and provide design principles for nonreciprocal photonic devices.

We present a novel method to determine lepton energy scale and resolution corrections by means of an analytical likelihood maximization applied to Drell-Yan $Z \to \ell\ell$ events. The approach relies on an exact analytical treatment of the energy smearing, avoiding random-number-based convolution techniques. This formulation results in a fully differentiable likelihood enabling the use of automatic differentiation algorithms, and thus a substantial reduction in computational cost. The method, implemented in the \ijazz software, allows the simultaneous extraction of scale and resolution parameters across multiple lepton categories defined by detector or kinematic variables. We validate the technique using toy Monte Carlo studies and realistic Pythia-based simulations, demonstrating unbiased parameter recovery and accurate uncertainty estimates. Particular attention is given to categorizations involving lepton transverse momentum, for which a relative-$p_T$ strategy is introduced to mitigate biases induced by category migration and kinematic correlations. The method is further adapted to photon-energy scale measurement in $Z \to μ^-μ^+γ$ decays. Compared to conventional approaches, the analytical method improves numerical stability, robustness of the minimization, and computational performance, making it well suited for large-scale precision calibration tasks at the LHC.

We propose a multi-scale lung model to investigate spatio-temporal distributions of ventilation variables. Lung envelope and large airway geometries are derived from CT scans; smaller airways are generated using a physiologically consistent algorithm. Tissue mechanics is modeled using nonlinear elasticity under small deformations, coupled with local air pressure from fluid dynamics within the bronchial tree. Airflow accounts for inertia and static airway compliance. Simulations employ finite elements. Using this model, we explore spatio-temporal airflows and shear stresses distributions.

Background: What counts as violence is neither self-evident nor universally agreed upon. While physical aggression is prototypical, contemporary societies increasingly debate whether exclusion, humiliation, online harassment or symbolic acts should be classified within the same moral category. At the same time, Large Language Models (LLMs) are being consulted in everyday contexts to interpret and label complex social behaviors. Whether these systems reproduce, reshape or simplify human conceptions of violence remains an open question. Methods: Here we present a systematic comparison between human judgements and LLM classifications across 22 scenarios carefully designed to be morally dividing, spanning from physical and verbally aggressive behavior, relational dynamics, marginalization, symbolic actions and verbal expressions. Human responses were compared with outputs from multiple instruction-tuned models of varying sizes and architectures. We conducted global, sentence-level and thematic-domain analyses, and examined variability across models to assess patterns of convergence and divergence. Findings: This study treats violence as a strategically chosen proxy through which broader belief formation dynamics can be observed. Violence is not the focus of the study, but it serves as a tool to investigate broader analysis. It enables a structured investigation of how LLMs operationalize ambiguous moral constructs, negotiate conceptual boundaries, and transform plural human interpretations into singular outputs. More broadly, the findings contribute to ongoing debates about the epistemic role of conversational AI in shaping everyday interpretations of harm, responsibility and social norms, highlighting the importance of transparency and critical engagement as these systems increasingly mediate public reasoning.

The combinatorial approach applied to functional oxides has enabled the production of material libraries that formally contain infinite compositions. A complete ternary diagram can be obtained by pulsed laser deposition (PLD) on 100 mm silicon wafers. However, interest in such materials libraries is only meaningful if high-throughput characterization enables the information extraction from the as-deposited library in a reasonable time. While much commercial equipment allows for XY-resolved characterization at room temperature, very few sample holders have been made available to investigate structural, chemical, and functional properties at high temperatures in controlled atmospheres. In the present work, we present a furnace that enables the study of 100 mm wafers as a function of temperature. This furnace has a dome to control the atmosphere, typically varying from nitrogen gas to pure oxygen atmosphere with external control. We present the design of such a furnace and an example of X-ray diffraction (XRD) and fluorescence (XRF) measurements performed at the DiffAbs beamline of the SOLEIL synchrotron. We apply this high-throughput approach to a combinatorial library up to 735 {\textdegree}C in nitrogen and calculate the thermal expansion coefficients (TEC) of the ternary system using custom-made MATLAB codes. The TEC analysis revealed the potential limitations of Vegard's law in predicting lattice variations for high-entropy materials.

Next generation quantum technologies will need to rely on efficient transduction between electrical, optical, and mechanical quantum degrees of freedom to generate large-scale entanglement over large distances. The performance of such transducers is fundamentally limited by the cryogenic properties of the underlying materials. Here, we demonstrate that engineering strain in ferroelectric thin-film strontium titanate ($\mathrm{SrTiO_3}$) not only results in an exceptionally large Pockels coefficient, but also in a robust linear piezoelectric response at cryogenic temperatures, surpassing previous thin-film benchmarks. We measure piezoelectric tensor elements of $d_{15} = 151.8 \pm 1.5$ pm/V and $d_{33} = 54.8 \pm 4$ pm/V, and an effective photoelastic coefficient of $p_{\mathrm{eff}}$ = 0.56 at 5~K. Utilizing these enhanced properties, we demonstrate the first $\mathrm{SrTiO_3}$-on-oxide acousto-optic modulator with a voltage-length product ($V_πL$) of $0.874 \pm 0.084$ V.cm, outperforming state-of-the-art unreleased modulators that typically feature a $V_πL$ of a few V.cm. Our results establish thin-film $\mathrm{SrTiO_3}$ as a promising material system for integrated quantum photonics operating at cryogenic temperatures.

Cognitive biases are often attributed to heuristics or limited information. Yet the structure of social networks is a key, often-overlooked source of perceptual bias. When information passes through social connections, the network alone can systematically distort how individuals view society. We use a simple model in which agents have a binary attribute (e.g., atheist or believer) and show that network topology alone can cause misperceptions of peers' attributes. These misperceptions persist even after aggregation and challenge the idea of the "wisdom of the crowd." We derive an estimator that predicts the size and direction of these biases from network features. We validate our findings using three large-scale opinion surveys. Our results show that network structure is a critical factor in collective perception, with major implications for reducing segregation, polarisation, and the marginalisation of minorities.

Friction losses in rough pipes are often predicted using semi-empirical correlations, such as the Colebrook-White equation (Colebrook,1939), which do not fully replicate Nikuradse's rough-pipe experiments (1950). This study derives scaling relations for the viscous and turbulent contributions to the streamwise pressure drop through an order-of-magnitude analysis of the Reynolds-averaged Navier-Stokes equations and the kinetic-energy transport equations. These relations impose constraints on the local sensitivity of the pressure drop to factors such as mean velocity, roughness, viscosity, and density through exponent envelopes and serve as a physical prior for symbolic regression. By combining Nikuradse's rough-pipe and smooth-pipe data of Zagarola and Smits (1998), we aim to derive compact correlations for the friction factor that fit experimental data while adhering to the derived constraints. A modified genetic programming engine (GPTIPS2) optimizes model structure and evaluates it based on fitness, complexity, and constraint violation. This method yields interpretable expressions that accurately reproduce friction factors across various roughness levels and Reynolds numbers, validated up to $Re \sim 10^7$.

Characterizing the intensity, phase, and polarization of engineered light is fundamental to understanding and applying metasurfaces. However, existing characterization frameworks are hindered by several limitations, most notably their inability to account for the polarization of the field. Here, we report polarization sensitive coherent modulation imaging (PS-CMI), a light-weight but robust, high-resolution platform for the full-field characterization of metasurface-modulated light. By supplementing the orthogonal x- and y- complex amplitude components with an additional 45°-component, this approach calculates the retardance between two orthogonal polarization components while eliminating phase offsets, thereby enabling the subsequent recovery of the complete polarization state. We demonstrate the versatility of our method by characterizing light fields produced by a United States Air Force (USAF) target, two kinds of complex polarization field, and a metalens. This compact solution addresses a critical gap in metasurface metrology and is broadly applicable to other fields requiring the mapping of complex, polarized light distributions.

Quantum cascade lasers (QCLs) are unipolar semiconductor lasers first demonstrated in 1994. Since then, they have played a central role in advancing mid-infrared and terahertz photonics, becoming among the most reliable light sources in these regions of the electromagnetic spectrum. Their importance is further reinforced by their ability to generate self-starting optical frequency combs, whose investigation is motivated both by fundamental physics and by a wide range of applications, including molecular spectroscopy and free-space optical communications. This Roadmap provides a unified overview of current advances and emerging directions in QCL research. The chapters are organized into three main sections: device design and technology; frequency combs and pulse formation; and applications of QCLs. Each chapter reviews the relevant background, summarizes the current state of the art, and identifies key challenges and future directions within its specific research area.

Polarization-resolved control and measurement of the optical field are essential for a wide range of photonic systems, including coherent communication, polarimetric sensing, and quantum information processing. We present a photonic integrated circuit that enables the generation and analysis of arbitrary polarization states. The device provides reconfigurable access to the full polarization degree of freedom of coherent light within a single integrated platform. We experimentally demonstrate arbitrary polarization state generation spanning the Poincare sphere, as well as Stokes vector measurement on chip. Unlike conventional Stokes measurements that rely on direct detection, polarization analysis utilizing this architecture is intrinsically non-destructive, preserving the optical signal for further optical domain processing. The devices are fabricated in a commercial foundry using CMOS-compatible processes, enabling scalable and reproducible integration. By combining polarization generation and analysis in a compact and stable photonic circuit, this work eliminates the need for external polarization optics and provides a foundation for robust, polarization-enabled photonic integrated systems.

Photoluminescence intensity is widely used to infer exciton populations, yet the detected signal inherently convolves occupancy with radiative-rate modification and collection efficiency, making quantitative inversion vulnerable to pump and system drifts. Here we realize a dual-channel self-referenced scheme enabled by two nearly degenerate quasinormal modes in a hybrid microcavity. Their shared optical path provides common-mode observables (i.e., overall spectral and intensity drift) that track global thermo-optic and pump fluctuations, while their differential-mode observables (i.e., spectral splitting and mode-contrasted emission) remain highly sensitive to local gap dielectric perturbations and dipole-dependent radiative weights. Using temperature as a control parameter in monolayer WSe$ _2 $, we exploit this common/differential-mode framework to robustly invert the relative populations of excitons with out-of-plane ($ \perp $) and in-plane ($ \parallel $) dipole transitions without external absolute calibration. At the temperature of $\sim$50 K, we obtain $ N_\perp/N_\parallel \approx 200 $, coincident with the expected accumulation in the out-of-plane-emitting dark manifold. This internally referenced approach provides a practical route to drift-tolerant, dipole-resolved population metrology in nanogap photonic systems.

Recently, machine learning Hamiltonian (MLH) models have gained traction as fast approximations of electronic structures such as orbitals and electron densities, while also enabling direct evaluation of energies and forces from their predictions. However, despite their physical grounding, existing Hamiltonian models are evaluated mainly by reconstruction metrics, leaving it unclear how well they perform as energy-force predictors. We address this gap with a benchmark that computes energies and forces directly from predicted Hamiltonians. Within this framework, we propose QHFlow2, a state-of-the-art Hamiltonian model with an SO(2)-equivariant backbone and a two-stage edge update. QHFlow2 achieves $40\%$ lower Hamiltonian error than the previous best model with fewer parameters. Under direct evaluation on MD17/rMD17, it is the first Hamiltonian model to reach NequIP-level force accuracy while achieving up to $20\times$ lower energy MAE. On QH9, QHFlow2 reduces energy error by up to $20\times$ compared to MACE. Finally, we demonstrate that QHFlow2 exhibits consistent scaling behavior with respect to model capacity and data, and that improvements in Hamiltonian accuracy effectively translate into more accurate energy and force computations.

We introduce MxDiffusion, a hybrid physics- and data-driven diffusion-based framework that enables efficient and highly accurate generation of photonic structures from target optical properties. The improved accuracy is achieved through a two-stage generation strategy, in which the first diffusion model is explicitly trained with Maxwells equation-based loss to embed physical insight directly into the inverse design process, while the second model maps the physically consistent intermediate representation to the final structural geometry with significantly higher fidelity than solely data-driven approaches. The performance of MxDiffusion is validated on two representative applications: gold nanostructures patterned on a silica substrate and a highly tunable bandpass filter based on phase change material. In both cases, the proposed framework consistently outperforms a conventional data-driven diffusion model benchmark, particularly for out-of-training-distribution design targets and highly constrained resonance conditions. These results demonstrate the efficacy and superiority of MxDiffusion as a general physics-guided inverse design paradigm.

Previously we derived the leading term asymptotic solution of temperature distribution in skin heating by an electromagnetic beam at an arbitrary incident angle. The asymptotic analysis is based on that the penetration depth of the beam into skin is much smaller than the size of beam cross-section. It allows arbitrary incident angle. We expand the temperature in powers of the small depth to lateral scale ratio. The incident angle affects all terms in the expansion while the lateral heat conduction appears only in terms of positive even powers. The previously obtained leading term solution captures only the main effect of incident angle. The main effect of lateral heat conduction is contained in the second order term, which is mathematically negligible in the limit of small depth to lateral scale ratio. At a moderate length scale ratio (e.g., 0.1), however, the contribution from lateral conduction is quite significant and needs to be included in a meaningful approximate solution. In this study, we derive closed form analytical expressions for the first order and the second order terms in the asymptotic expansion. The resulting asymptotic solution is capable of predicting the temperature distribution accurately including the effects of both incident angle and lateral heat conduction even at a moderate length scale ratio.

This paper introduces a distillation framework for an ensemble of entropy-optimal Sparse Probabilistic Approximation (eSPA) models, trained exclusively on satellite-era observational and reanalysis data to predict ENSO phase up to 24 months in advance. While eSPA ensembles yield state-of-the-art forecast skill, they are harder to interpret than individual eSPA models. We show how to compress the ensemble into a compact set of "distilled" models by aggregating the structure of only those ensemble members that make correct predictions. This process yields a single, diagnostically tractable model for each forecast lead time that preserves forecast performance while also enabling diagnostics that are impractical to implement on the full ensemble. An analysis of the regime persistence of the distilled model "superclusters", as well as cross-lead clustering consistency, shows that the discretised system accurately captures the spatiotemporal dynamics of ENSO. By considering the effective dimension of the feature importance vectors, the complexity of the input space required for correct ENSO phase prediction is shown to peak when forecasts must cross the boreal spring predictability barrier. Spatial importance maps derived from the feature importance vectors are introduced to identify where predictive information resides in each field and are shown to include known physical precursors at certain lead times. Case studies of key events are also presented, showing how fields reconstructed from distilled model centroids trace the evolution from extratropical and inter-basin precursors to the mature ENSO state. Overall, the distillation framework enables a rigorous investigation of long-range ENSO predictability that complements real-time data-driven operational forecasts.

Introduction: While the origin and evolution of proteins remain mysterious, advances in evolutionary genomics and systems biology are facilitating the historical exploration of the structure, function and organization of proteins and proteomes. Molecular chronologies are series of time events describing the history of biological systems and subsystems and the rise of biological innovations. Together with time-varying networks, these chronologies provide a window into the past. Areas covered: Here, we review molecular chronologies and networks built with modern methods of phylogeny reconstruction. We discuss how chronologies of structural domain families uncover the explosive emergence of metabolism, the late rise of translation, the co-evolution of ribosomal proteins and rRNA, and the late development of the ribosomal exit tunnel; events that coincided with a tendency to shorten folding time. Evolving networks described the early emergence of domains and a late big bang of domain combinations. Expert opinion: Two processes, folding and recruitment appear central to the evolutionary progression. The former increases protein persistence. The later fosters diversity. Chronologically, protein evolution mirrors folding by combining supersecondary structures into domains, developing translation machinery to facilitate folding speed and stability, and enhancing structural complexity by establishing long-distance interactions in novel structural and architectural designs.

Synthetic aperture radar (SAR) is a tomographic sensor that measures 2D slices of the 3D spatial Fourier transform of the scene. In many operational scenarios, the measured set of 2D slices does not fill the 3D space in the Fourier domain, resulting in significant artifacts in the reconstructed imagery. Traditionally, simple priors, such as sparsity in the image domain, are used to regularize the inverse problem. In this paper, we review our recent work that achieves state-of-the-art results in 3D SAR imaging employing neural structures to model the surface scattering that dominates SAR returns. These neural structures encode the surface of the objects in the form of a signed distance function learned from the sparse scattering data. Since estimating a smooth surface from a sparse and noisy point cloud is an ill-posed problem, we regularize the surface estimation by sampling points from the implicit surface representation during the training step. We demonstrate the model's ability to represent target scattering using measured and simulated data from single vehicles and a larger scene with a large number of vehicles. We conclude with future research directions calling for methods to learn complex-valued neural representations to enable synthesizing new collections from the volumetric neural implicit representation.

Uniform quantization is a topic that has been extensively studied. However and although an analytical description of quantization noise has been proposed, most descriptions of the spectral properties of quantization error resort to statistical descriptions. In this paper, we show how the spectrum of a quantized signal can be expressed using pulse frequency modulation. We first establish the equivalence of a uniform quantizer with a system based on the bipolar pulse frequency modulation and we define afterwards the Fourier transform of the quantized signal using pulse frequency modulation properties. This model brings a more intuitive understanding of the spectral structure of quantization noise and complements prior research in the topic. The results of the paper can be directly applied to level crossing ADCs with zero-order-hold interpolators, giving an accurate estimation of their performance.

Reliable odometry for legged robots without cameras or LiDAR remains challenging due to IMU drift and noisy joint velocity sensing. This paper presents a purely proprioceptive state estimator that uses only IMU and motor measurements to jointly estimate body pose and velocity, with a unified formulation applicable to biped, quadruped, and wheel-legged robots. The key idea is to treat each contacting leg as a kinematic anchor: joint-torque--based foot wrench estimation selects reliable contacts, and the corresponding footfall positions provide intermittent world-frame constraints that suppress long-term drift. To prevent elevation drift during extended traversal, we introduce a lightweight height clustering and time-decay correction that snaps newly recorded footfall heights to previously observed support planes. To improve foot velocity observations under encoder quantization, we apply an inverse-kinematics cubature Kalman filter that directly filters foot-end velocities from joint angles and velocities. The implementation further mitigates yaw drift through multi-contact geometric consistency and degrades gracefully to a kinematics-derived heading reference when IMU yaw constraints are unavailable or unreliable. We evaluate the method on four quadruped platforms (three Astrall robots and a Unitree Go2 EDU) using closed-loop trajectories. On Astrall point-foot robot~A, a $\sim$200\,m horizontal loop and a $\sim$15\,m vertical loop return with 0.1638\,m and 0.219\,m error, respectively; on wheel-legged robot~B, the corresponding errors are 0.2264\,m and 0.199\,m. On wheel-legged robot~C, a $\sim$700\,m horizontal loop yields 7.68\,m error and a $\sim$20\,m vertical loop yields 0.540\,m error. Unitree Go2 EDU closes a $\sim$120\,m horizontal loop with 2.2138\,m error and a $\sim$8\,m vertical loop with less than 0.1\,m vertical error. github.com/ShineMinxing/Ros2Go2Estimator.git

This study investigates a robust reconfigurable intelligent surface (RIS)-assisted multiple-input multiple-output (MIMO) system for secure wireless communication, in which a multi-antenna transmitter (Alice) sends confidential messages to a multi-antenna receiver (Bob) in the presence of an eavesdropper (Eve). Unlike idealized models, the reflecting elements (REs) of the RIS are assumed to possess inherent electrical resistance, introducing a practical non-ideal effect often neglected in prior research. The aim of the study is to maximize the secrecy rate of the MIMO system under perfect knowledge of the channel state information (CSI). To achieve this, the secrecy rate maximization problem is formulated and solved using a low-complexity joint optimization framework based on an adaptive projected gradient method (PGM), which simultaneously updates both the transmit precoding matrix and the RIS phase shifts. Solving the exact problem is computationally complex. Thus, a simplified variant is further introduced that maximizes the channel power difference rather than the exact secrecy rate. The simulation results show that this approximation yields a secrecy rate close to the true optimum while significantly reducing the computational cost. In addition, the proposed PGM with an adaptive step size initialization and control mechanism substantially improves the secrecy rate and reduces the computational time compared to the conventional fixed step size PGM. Overall, the simulation results confirm the effectiveness of the proposed PGM and demonstrate that adopting a practical RIS model is essential for establishing secure RIS-assisted MIMO communication links, especially under varying RE resistance values.

A signal processing-based framework is proposed for detecting random segment failures in segmented waveguide-enabled pinching-antenna systems. To decouple the passively combined uplink signal and to provide per-segment observability, tagged pilots are employed. A simple tag is attached to each segment and is used to apply a known low-rate modulation at the segment feed, which assigns a unique signature to each segment. Based on the tagged-pilot model, a low-complexity per-segment maximum-likelihood (ML) detector is developed for the case in which the pilot length is no smaller than the number of segments. For the case in which the pilot length is smaller than the number of segments, sparsity in the failure-indicator vector is exploited and a compressive sensing-based detector is adopted. Numerical results show that the per-segment detector approaches joint ML performance, while the compressive sensing-based detector achieves reliable detection with a short pilot and can outperform baselines that require much longer pilots.

Near-field ultra-massive MIMO (U-MIMO) systems provide enhanced spatial resolution but present challenges for channel estimation, particularly when hybrid architectures are employed. Within this framework, dictionary-based channel estimation schemes are needed to achieve accurate reconstruction from a reduced set of measurements. However, existing near-field dictionaries generally provide full three-dimensional coverage, which is unnecessary when user equipments are primarily located on the ground. In this paper, we propose a novel near-field grid design tailored to this common scenario. Specifically, grid points lie on a reference plane located at an arbitrary height with respect to the U-MIMO system, equipped with a uniform planar array. Furthermore, a channel accuracy metric is used to improve codebook performance, and to remark the limitations of the traditional far-field angular sampling in the near field. Results show that, as long as user equipments are not far from the reference plane, the proposed grid outperforms state-of-the-art designs in both channel estimation accuracy and spectral efficiency.

Remote and resource-constrained Internet-of-Things (IoT) deployments often lack terrestrial connectivity for task offloading, motivating non-terrestrial networks (NTNs) with onboard multiaccess edge computing (MEC) capabilities. Nevertheless, in the presence of malicious actors, authentication needs to be performed to avoid non-authorized nodes from draining the computing resources of the NTN nodes. As a solution, we propose a four-layer MEC-enabled NTN with unmanned aerial vehicles (UAVs) acting as access nodes, a high altitude platform station (HAPS) acting as coordinator and authenticator, and a constellation of low-Earth orbit satellites (LEOSats) acting as remote MEC servers. We consider a tag-based physical-layer authentication (PLA) scheme to authenticate legitimate users, and formulate a joint task offloading decision and resource allocation for the admitted tasks, which is solved via block coordinate descent. Numerical results show that the PLA scheme is efficient and performs better than the benchmark schemes. We also demonstrate that the proposed scheme is robust against malicious attacks even under relaxed false-alarm constraints.

One of the key issues facing Direct-to-Cellular (D2C) satellite communication systems is ionospheric scintillation on the uplink and downlink, which can significantly degrade link quality. This work investigates the spatial and temporal characteristics of amplitude scintillation at D2C frequencies by scaling L-band scintillation observations from Global Navigation Satellite Systems (GNSS) receivers to bands relevant to D2C operation, including the low-band, and 3GPP's N255 and N256. These observations are then compared to scaled radio-occultation scintillation observations from the FORMOSAT-7/COSMIC-2 (F7/C2) mission, which can be used in regions that do not possess ground-based scintillation monitoring stations. As a proof of concept, five years of ground-based GNSS scintillation data from Sharjah, United Arab Emirates, together with two years of F7/C2 observations over the same region, corresponding to the ascending phase of Solar Cycle 25, are analyzed. Both space-based and ground-based observations indicate a pronounced diurnal scintillation peak between 20--22 local time, particularly during the equinoxes, with occurrence rates increasing with solar activity. Ground-based observations also reveal a strong azimuth dependence, with most scintillation events occurring on southward satellite links. The scintillation occurrence rate at the low-band is more than twice that observed at N255 and N256, highlighting the increased robustness of higher D2C bands to ionospheric scintillation. These results demonstrate how GNSS scintillation observations can be leveraged to characterize and anticipate scintillation-induced D2C link impairments, which help in D2C system design and the implementation of scintillation mitigation strategies.

This two-part paper aims to develop an environment-aware network-level design framework for generalized pinching-antenna systems to overcome the limitations of conventional link-level optimization, which is tightly coupled to instantaneous user geometry and thus sensitive to user mobility and localization errors. Part I investigates the traffic-aware case, where user presence is characterized statistically by a spatial traffic map and deployments are optimized using traffic-aware network-level metrics. Part II complements Part I by developing geometry-aware, blockage-aware network optimization for pinching-antenna systems in obstacle-rich environments. We introduce a grid-level average signal-to-noise (SNR) model with a deterministic LoS visibility indicator and a discrete activation architecture, where the geometry-dependent terms are computed offline in advance. Building on this model, we formulate two network-level activation problems: (i) average-SNR-threshold coverage maximization and (ii) fairness-oriented worst-grid average-SNR maximization. On the algorithmic side, we prove the coverage problem is NP-hard and derive an equivalent mix-integer linear programming reformulation through binary coverage variables and linear SNR linking constraints. To achieve scalability, we further develop a structure-exploiting coordinate-ascent method that updates one waveguide at a time using precomputed per-candidate SNR contributions. For the worst-grid objective, we adopt an epigraph reformulation and leverage the resulting monotone feasibility in the target SNR, enabling an efficient bisection-based solver with low-complexity feasibility checks over the discrete candidate set. Simulations results validate the proposed designs and quantify their gains under different environments and system parameters.

Existing studies on generalized pinching-antenna systems are predominantly link-level, which optimize system parameters for a given user set with objectives defined by per-user performance metrics. Such designs do not capture network-level requirements, e.g., region-wide coverage and location fairness, and may require frequent re-optimization as users move or enter/leave, incurring control overhead and sensitivity to localization errors. Motivated by this gap, this two-part paper aims to develop an environment-aware network-level design framework for generalized pinching-antenna systems. Part I focuses on the traffic-aware case, where user presence is modeled statistically by a spatial traffic map and performance is optimized and evaluated in a traffic-aware sense; Part II addresses the geometry-aware case in obstacle-rich environments. In Part~I, we introduce traffic-weighted average SNR metrics and formulate two traffic-aware deployment problems: (i) maximizing the traffic-weighted network average SNR, and (ii) a fairness-oriented traffic-restricted max--min average-SNR design over traffic-dominant grids. To solve these nonconvex problems with low complexity, we reveal and exploit their separable structures. For the network-average objective, we establish unimodality properties of the hotspot-induced components and develop a candidate-based global maximization method that only needs to evaluate the objective at a small set of candidate antenna positions. For the traffic-restricted max--min objective, we develop a block coordinate decent framework where each coordinate update reduces to a globally solvable one-dimensional subproblem via an epigraph reformulation and bisection. Simulations show that traffic-aware pinching-antenna positioning consistently outperforms representative fixed and heuristic traffic-aware deployments in the considered setups.

This paper investigates covert multi-hop communication in wireless networks where an adversary employs a cyclostationary (cycle) detector to reveal hidden transmissions. The covert route employs direct sequence spread spectrum (DSSS) signaling to ensure either maximum end-to-end covertness maximization or minimum latency minimization-under quality-of-service (QoS) and link budget constraints. Optimal bandwidth, transmit power, and spreading gain for each hop jointly satisfy reliability and either rate or covertness requirements. We show the equivalence between the covertness and the detection SNR gain-based widest-path formulations, and, hence, enabling efficient route computation. Numerical simulations in a realistic 3D environment illustrate that (i) end-to-end latency increases exponentially with the covertness requirement, (ii) the end-to-end latency increase is super-linear with the packet size M, and (iii) cycle and energy detectors impose different latency behavior as a function of the message length and the covertness requirement. The proposed framework provides important insights into resource allocation and routing design for covert networks against advanced detection adversaries.

This paper develops an end-to-end odor communication model for stress signaling between plants using Green Leaf Volatiles (GLV). A damaged transmitter plant emits (Z)-3-hexenal, (Z)-3-hexenol, and (Z)-3-hexenyl acetate, which propagate through a time-varying diffusion-advection channel and undergo multiplicative loss at the receiver. The sink plant is modeled with a biochemical receiver network that converts the received GLVs into the defensive metabolite (Z)-3-hexenyl $β$-vicianoside, and an alarm decision is defined based on its concentration level. Numerical results show that (Z)-3-hexenol is the primary driver of the system and that plant perception generally operates in a non-linear region. These findings provide a framework for understanding the evolution of plant-plant communication and for developing next-generation precision farming technologies.

Envelope extraction in nuclear magnetic resonance (NMR) is a fundamental step for processing the data space generated by this technique. Envelope detection accuracy improves with increasing the number of sampling points; however, we propose a novel transform that enables acceptable envelope extraction with significantly fewer sampling points, even without meeting the Nyquist rate. In this paper, we challenge the traditional scale definition and demonstrate that classic scaling lacks a physical referent in all situations. To achieve this aim, we introduce a scale based on the variations of space-invariant states, rather than the observable characteristics of matter and energy. According to this definition of the scale, we distinguished two kinds of observers: scale-variant and scale-invariant. We demonstrated that converting a scale-variant observer to a scale-invariant observer is equivalent to envelop extraction. To analyse and study the theories presented in the paper, we have designed and implemented an Earth-field NMR setup and used real data generated by it to evaluate the performance of the proposed envelope-detection transform. We compared the output of the proposed transform with that of classic and state-of-the-art methods for parameter recovery of NMR signals.

We study the dynamics of light interacting with a near-resonant atomic medium using the truncated Wigner and positive P phase-space representations. The atomic degrees of freedom are described using the Jordan-Schwinger mapping. The dynamics is first analyzed under unitary evolution and subsequently in the presence of an optical reservoir. While both approaches capture the main features of the light-matter dynamics, we find that the truncated Wigner approximation exhibits noticeable deviations for stronger interaction strengths and when reservoir-induced noise becomes significant.

Simultaneously estimating multiple parameters at the ultimate limit is a central challenge in quantum metrology, often hindered by inherent incompatibilities in optimal estimation strategies. At its most extreme, this incompatibility culminates in a fundamental impossibility when the quantum Fisher information matrix (QFIM) becomes singular, rendering joint estimation unattainable. This is the case for a canonical problem: estimating the amplitude and frequency of an AC magnetic field, where the generators are parallel to each other. Here, we introduce a quantum control protocol that resolves this singularity. Our control protocol strategically engineers the sensor's time evolution so the generators for the two parameters become orthogonal. It not only removes the singularity but also restores the optimal scaling of precision with interrogation time for both parameters simultaneously. We experimentally validate this protocol using a nitrogen-vacancy center in diamond at room temperature, demonstrating the concurrent achievement of the optimal scaling for both parameters under realistic conditions.

We initiate a systematic study of pseudo-deterministic quantum algorithms. These are quantum algorithms that, for any input, output a canonical solution with high probability. Focusing on the query complexity model, our main contributions include the following complexity separations, which require new lower bound techniques specifically tailored to pseudo-determinism: - We exhibit a problem, Avoid One Encrypted String (AOES), whose classical randomized query complexity is $O(1)$ but is maximally hard for pseudo-deterministic quantum algorithms ($Ω(N)$ query complexity). - We exhibit a problem, Quantum-Locked Estimation (QL-Estimation), for which pseudo-deterministic quantum algorithms admit an exponential speed-up over classical pseudo-deterministic algorithms ($O(\log(N))$ vs. $Θ(\sqrt{N})$), while the randomized query complexity is $O(1)$. Complementing these separations, we show that for any total problem $R$, pseudo-deterministic quantum algorithms admit at most a quintic advantage over deterministic algorithms, i.e., $D(R) = \tilde O(psQ(R)^5)$. On the algorithmic side, we identify a class of quantum search problems that can be made pseudo-deterministic with small overhead, including Grover search, element distinctness, triangle finding, $k$-sum, and graph collision.

We present a hydrodynamical description of spherical void evolution in modified gravity (MG), extending the standard General Relativity (GR) and dynamical dark energy treatment by encoding gravity modifications into effective couplings that enter the Euler and Poisson equations. This yields a compact non-linear evolution equation for the Eulerian density contrast, controlled by a time- and density-dependent effective gravitational strength, and provides a direct map between model functions and void observables. We apply the framework to the luminal Galileon class of models, where derivative self-interactions generate Vainshtein screening and might lead to a breakdown of the physical branch in sufficiently underdense regions. Exploiting this feature, we apply the void-informed viability requirement that translates into bounds on the theory parameter space and, equivalently, on the minimum attainable void depth as a function of redshift. For viable parameters of a concrete model, we quantify the impact of MG on isolated void evolution, the Lagrangian to Eulerian mapping, and the shell-crossing threshold. Relative to GR, we find a clear hierarchy of MG effects, with ${\cal O}(10\%)$ modifications in the gravitational couplings, percent-level shifts in the void density evolution, and sub-percent deviations in both the mapping and the shell-crossing thresholds. Moreover, within the adopted parametrization, we show analytically that voids always lie in an unscreened regime on the physical branch. Overall, the formalism provides a self-consistent route to predict void dynamics and consistency constraints in a broad class of MG models.

The theory of open quantum systems served as a tool to prepare entanglement at the beginning stage of quantum technology and more recently provides an important tool for state preparation. Dynamics given by time dependent Lindbladians are Markovian and lead to decoherence, decay of correlation and convergence to equilibrium. In contrast Non-Markovian evolutions can outperform their Markovian counterparts by enhancing memory. In this letter we compare the trajectories of Markovian and Non-Markovian evolutions starting from a fixed initial value. It turns out that under mild assumptions every trajectory can be obtained from a family of time dependent Lindbladians. Hence Non-Markovianity is invisible if single trajectories are concerned.

Pure 3d gravity in AdS is believed to admit a holographic description in terms of 2d CFT. We introduce a theory of fermionic 3d gravity where we sum over geometries equipped with spin structure, and propose it is holographically described by fermionic 2d CFT data. We evaluate the leading contributions to the gravity path integral with one and two torus boundaries, extracting both the spectrum and its spectral statistics from the torus wormhole. Strikingly, the theory has fermionic black hole microstates, even in the absence of bulk fermionic matter. We then incorporate subtle bulk topological field theories, classified by appropriate cobordism groups, and evaluate the one and two-boundary torus partition functions. The spectral statistics we derive from gravity are shown, in all cases, to be consistent with the pattern of anomalies expected from classifications of fermionic 2d CFT. We also define a version of RMT$_2$, a random-matrix framework compatible with the symmetries of 2d CFTs, which naturally accommodates fermionic spectra and reproduces our gravitational results across all cases we analyze.

While ground-state preparation is expected to be a primary application of quantum computers, it is also an essential subroutine for many fault-tolerant algorithms. In early fault-tolerant regimes, logical measurements remain costly, motivating adaptive, shot-frugal state-preparation strategies that efficiently utilize each measurement. We introduce the Shadow Enhanced Greedy Quantum Eigensolver (SEGQE) as a greedy, shadow-assisted framework for measurement-efficient ground-state preparation. SEGQE uses classical shadows to evaluate, in parallel and entirely in classical post-processing, the energy reduction induced by large collections of local candidate gates, greedily selecting at each step the gate with the largest estimated energy decrease. We derive rigorous worst-case per-iteration sample-complexity bounds for SEGQE, exhibiting logarithmic dependence on the number of candidate gates. Numerical benchmarks on finite transverse-field Ising models and ensembles of random local Hamiltonians demonstrate convergence in a number of iterations that scales approximately linearly with system size, while maintaining high-fidelity ground-state approximations and competitive energy estimates. Together, our empirical scaling laws and rigorous per-iteration guarantees establish SEGQE as a measurement-efficient state-preparation primitive well suited to early fault-tolerant quantum computing architectures.

We study an adiabatic variant of the variational quantum eigensolver (VQE) in which VQE is performed iteratively for a sequence of Hamiltonians along an adiabatic path. We derive the conditions under which gradient-based optimization successfully prepares the adiabatic ground states. These conditions show that the barren plateau problem and local optima can be avoided. Additionally, we propose using energy-standard-deviation measurements at runtime to certify eigenstate accuracy and verify convergence to the global optimum.

Stabilizer states hold a special place in quantum information science due to their connection with quantum error correction and quantum circuit simulation. In the context of classical simulations of many-body physics, they are an example of states that can be both highly entangled and efficiently represented and transformed under Clifford operators. Recently, Clifford operators have been discussed in the context of fermionic quantum computation through their extension, the Majorana Clifford group. Here, we document the phase-sensitive form of the corresponding Majorana stabilizer states, as well as the algorithms for computing their amplitudes, their inner products, and update rules for transforming Majorana stabilizer states under Majorana Clifford gates.

Spectroscopic measurements of single-particle spectral functions provide crucial insight into strongly correlated quantum matter by resolving the energy and spatial structure of elementary excitations. Here we introduce a spectroscopic protocol for single-charge injection with simultaneous spatial and energy resolution in a Rydberg tweezer array, effectively emulating scanning tunneling microscopy. By combining this protocol with single-atom-resolved imaging, we go beyond conventional spectroscopy by not only measuring the single-particle spectral function but also directly imaging the microscopic structure of the excitations underlying spectral resonances in frustrated $tJ$ Hamiltonians. We reveal resonances associated with the formation of bound magnetic polarons -- composite quasiparticles consisting of a mobile hole bound to a magnon -- and directly extract their binding energy, spatial extent, and spin character. Finally, by exploiting the spatial tunability of our platform, we measure the local density of states across different lattice geometries. Our work establishes Rydberg tweezer arrays as a powerful platform for spectroscopic studies of strongly correlated models, offering microscopic control and direct real-space access to emergent quasiparticles in engineered quantum matter.

Modern precision experiments often probe unknown classical fields with bosonic sensors in quantum-noise-limited regimes where vacuum fluctuations limit conventional readout. We introduce Quantum Signal Learning (QSL), a sensing framework that extends metrology to a broader property-learning setting, and propose a quantum-enhanced protocol that simultaneously estimates many properties of a classical signal with shot noise suppressed below the vacuum level. Our scheme requires only two-mode squeezing, passive optics, and static homodyne measurements, and enables post-hoc classical estimation of many properties from the same experimental dataset. We prove that our protocol enables a quantum speedup for common classical sensing tasks, including measuring electromagnetic correlations, real-time feedback control of interferometric cavities, and Fourier-domain matched filtering. To establish these separations, we introduce an optimal-transport conditioning method, and show both worst-case exponential separations from all entanglement-free strategies and practical speedups over homodyne and heterodyne baselines. We further show that when squeezing is treated as a resource, a protocol with squeezed light can sense a structured classical background exponentially faster than any coherent classical probe.

The simple harmonic oscillator has a well-known normalizable, positive energy, bound state spectrum. We show that degenerate with each such positive energy eigenvalue there is a non-normalizable positive energy eigenstate whose eigenfunction is orthogonal to that of the standard energy eigenfunction. This class of states is not built on the vacuum that $a$ annihilates, but is instead built on the vacuum that $a^{\dagger} a$ annihilates. These non-normalizable but nonetheless stationary energy eigenstates are accompanied by yet another set of non-normalizable states, states whose wave functions however are not stationary but instead grow linearly in time. With these states not being energy eigenstates, the eigenbasis of the Hamiltonian is incomplete; with the full Hilbert space containing states that are not energy eigenstates. Thus each energy eigenvalue of the harmonic oscillator is an exceptional point at which the Hamiltonian becomes of non-diagonalizable, and thus manifestly non-Hermitian, Jordan-block form. Such non-Hermitian structures occur for Hamiltonians that have an antilinear $PT$ symmetry. As is characteristic of such systems, one can construct a probability conserving inner product that despite the linear in time growth is nonetheless time independent, and not only that, it leads to states with zero norm. In addition, as is again characteristic of $PT$ symmetry, these non-normalizable states can be made normalizable by a continuation into a so-called Stokes wedge domain in the complex plane. In this domain one has a completely consistent quantum theory, one that lives alongside the standard normalizable energy eigenspectrum sector. This thus not quite so simple harmonic oscillator provides an explicit realization of our general contention that antilinearity is more basic to quantum theory than Hermiticity.

We construct new vacuum solutions of the Einstein equations generated from electrovacuum configurations embedded in external electromagnetic backgrounds. Starting from accelerating Bertotti--Robinson black holes, we employ two independent mechanisms: a Melvin--Bonnor-type magnetization and an Inversion symmetry of the Einstein--Maxwell system. In both cases the external electromagnetic field can be removed, while a non-trivial gravitational backreaction remains in the metric, yielding new accelerating vacuum spacetimes of Petrov type I. In the static, non-accelerating limit, the magnetized Schwarzschild case reproduces previously known results, whereas the Inversion symmetry leads to a genuinely new vacuum configuration. These findings provide a method for generating algebraically general vacuum geometries and illustrate how electromagnetic embeddings can produce non-trivial vacuum metrics in General Relativity. The main geometrical properties of these spacetimes are discussed. Two additional results are presented in the appendices: a stationary generalization of these vacuum geometries, and two further static vacuum configurations obtained through the same symmetries but using the Alekseev--García black hole as the seed.

We analyze the modification of entanglement dynamics in the Grover algorithm when the qubits are subjected to single-qubit amplitude-damping or phase-flip noise. We compare quantum trajectories with full density-matrix simulations, analyzing the dynamics of averaged trajectory entanglement (TE) and operator entanglement (OE), in the respective state representation. Although not a genuine entanglement measure, both TE and OE are connected to the efficiency of matrix product state simulations and thus of fundamental interest. As in many quantum algorithms, at the end of the Grover circuit entanglement decreases as the system converges towards the target product state. While we find that this is well captured in the OE dynamics, quantum trajectories rarely follow paths of reducing entanglement. Optimized unraveling schemes can lower TE slightly, however we show that deep in the circuit OE is generally smaller than TE. This implies that matrix product density operator (MPDO) simulations of quantum circuits can in general be more efficient than quantum trajectories. In addition, we investigate the noise-rate scaling of success probabilities for both amplitude-damping and phase-flip noise in Grover's algorithm.

The phase evolution of gravitational waves encodes critical information about the orbital dynamics of binary systems. In this work, we test the robustness of parameterized tests against unmodeled deviations from general relativity. We demonstrate that these parameterized tests are flexible and sensitive in detecting generic deviations in the waveform using the Cutler-Vallisneri bias formalism. This universality arises from examining the inherent geometry of the waveform signal and understanding how biases manifest. We show how Bayes factors are governed by the intrinsic geometry of the waveform signal manifold when parameterized tests are used to approximate generic violations of GR. We use the singular value decomposition to propose templates that are orthogonal to parameterized tests, identifying degeneracies and enhancing the detection of potential deviations. More broadly, the geometric framework developed here clarifies -- at a fundamental level -- how subtle waveform effects (including orbital eccentricity, spin precession, waveform systematics, and instrumental glitches) can mimic one another in data, and when they are intrinsically distinguishable.

Quantum phase transition is interpreted as an evolution, at the end of which a parameter-dependent Hamiltonian $H(g)$ loses its observability. In the language of mathematics, such a quantum catastrophe occurs at an exceptional point of order $N$ (EPN). Although the Hamiltonian $H(g)$ itself becomes unphysical in the limit of $g \to g^{EPN}$, it is shown that it can play the role of an unperturbed operator in a perturbation-approximation analysis of the vicinity of the EPN singularity. Such an analysis is elementary at $N\leq 3$ and numerical at $N\geq 5$, so we pick up $N=4$. We demonstrate that the specific EP4 degeneracy becomes accessible via a unitary evolution process realizable inside a parametric domain ${\cal D}_{\rm physical}$, the boundaries of which are determined non-numerically. Possible relevance of such a mathematical result in the context of non-Hermitian photonics is emphasized.

Quantum optimization has gained increasing attention as advances in quantum hardware enable the exploration of problem instances approaching real-world scale. Among existing approaches, variational quantum algorithms and quantum annealing dominate current research; however, both typically rely on one-hot encodings that severely limit scalability. Pauli Correlation Encoding (PCE) was recently introduced as an alternative paradigm that reduces qubit requirements by embedding problem variables into Pauli correlations. Despite its promise, PCE has not yet been studied in the context of constrained optimization. In this work, we extend the PCE framework to constrained combinatorial optimization problems and evaluate its performance across multiple problem sizes. Our results show that the standard PCE formulation struggles to reliably enforce constraints, which motivates the introduction of the Iterative-$α$ PCE. This iterative strategy significantly improves solution quality, achieving consistent constraint satisfaction while yielding better cut sizes across a wide range of instances. These findings highlight both the limitations of current PCE formulations for constrained problems and the effectiveness of iterative strategies for advancing quantum optimization in the NISQ era.

Observational data in cosmology indicate a small, positive, and nonvanishing cosmological constant that dominates the energy budget of the present universe. The origin of the cosmological constant from a quantum perspective remains unresolved, with a discrepancy of approximately 120 orders of magnitude between its observed value and theoretical estimates. Motivated by earlier work of Gibbons, we analyze the cosmological constant problem within a quantum-gravitational framework based on Schwarzian theory and its ensemble averaging. We then derive the phenomenological value of the dark energy density and obtain the corresponding equation of state. In this model, the cosmological constant arises from the ensemble average of time-reparametrization modes.

Superposition is the core feature that sets quantum theory apart from classical physics. Here, we investigate whether sets of quantum measurements can be modelled by using only devices that are operationally classical, in the sense that they have no superposition properties. This leads us to propose classical measurement models, which we show to be stronger than commutative measurements but weaker than joint measurability. We determine both the exact depolarisation noise rate and the measurement loss rate at which the all projective measurements in $d$-dimensional quantum theory admit a classical model. For finite sets of quantum measurements we develop methods both for constructing classical models and for falsifying the existence of such model via prepare-and-measure setups. Furthermore, we show that this concept also has operational implications. For that, we consider whether quantum measurements with classical side-information can be implemented in sequence without causing a disturbance and we show that classical models imply an affirmative answer. Our work sheds light on superposition as a resource for quantum measurement devices.

This paper investigates the motion of a single photon in a two-dimensional plane under closed and open boundary conditions. We employ two methods to construct the Hilbert space: Method A, based on the standard second-quantization formalism, and Method B, based on a non-standard approach. By eliminating redundant quantum states, we obtain a reduced Hilbert space with significantly lower dimensionality, thereby improving the efficiency of numerical simulations. In a closed system, the two methods are equivalent, and their unitary evolution results are identical. The probability distribution diffuses outward from the center and exhibits a significant rebound after reaching the boundary. In an open system, Method B, by incorporating more dissipation channels, provides a more accurate description of the photon escape process at the boundary. The probability curves obtained from the two methods completely overlap before reaching the boundary. After the boundary is reached, a slight difference appears, but this difference does not amplify with evolution and tends to converge in the later stage. Method B yields a slightly higher dissipative-state probability, indicating that the photon escapes faster. Visualization of the two-dimensional probability distribution shows that the three scenarios (closed system, open system with Method A, and open system with Method B) exhibit identical probability distributions before reaching the boundary. After the boundary is reached, the open systems exhibit significant probability loss, which increases rapidly with evolution. The probability distribution patterns of the two open systems are highly similar, exhibiting synchronized evolution.

Feedback-driven quantum reservoir computing has so far been studied primarily in gate-based architectures, motivating alternative scalable, hardware-friendly physical platforms. Here we investigate a linear-optical quantum reservoir architecture for time-series processing based on multiphoton interference in a reconfigurable interferometer network equipped with threshold detectors and measurement-conditioned feedback. The reservoir state is constructed from coarse-grained coincidence features, and the feedback updates only a structured, budgeted subset of programmable phases, enabling recurrence without training internal weights. By sweeping the feedback strength, we identify three dynamical regimes and find that memory performance peaks near the stability boundary. We quantify temporal processing via linear memory capacity and validate nonlinear forecasting on benchmarks, namely Mackey-Glass series, NARMA$-n$ and non-integrable Ising dynamics. The proposed architecture is compatible with current photonic technology and lowers the experimental barrier to feedback-driven QRC for time-series analysis with competitive accuracy.

We study a continuum Hatano--Nelson model with a saturating nonlinear nonreciprocity and analyze its stationary states via the associated phase-space flow. We uncover a global scenario controlled by a subcritical Hopf bifurcation and a saddle-node of limit cycles, which together generate a finite coexistence window. In this window, skin modes and extended states are both stable at a fixed energy $E$, separated by a nonlinear basin separatrix in phase space rather than a spectral (mobility-edge) mechanism in a linear system. An averaged amplitude equation yields closed-form predictions for the limit-cycle branches and the SNLC threshold. Building on the basin geometry, we introduce a basin-fraction order parameter that exhibits a first-order-like jump at SNLC. Intriguing physical phenomena in the coexistence window are also revealed, such as separatrix-induced long-lived spatial transients and hysteresis. Overall, our findings highlight that, beyond linear spectral concepts, global attractor-basin geometry provides a powerful and complementary lens for understanding stationary states in nonlinear non-Hermitian systems.

Preparing high-fidelity logical states is a central challenge in fault-tolerant quantum computing, yet existing approaches struggle to balance control complexity against resource overhead. Here, we present a complete framework for the fault-tolerant preparation of arbitrary logical states encoded in the four-legged cat code. This framework is engineered to suppress the dominant incoherent errors, including excitation decay and dephasing in both the bosonic mode and the ancilla via error detection. Numerical simulations with experimentally realistic parameters on a 3D superconducting cavity platform yield logical infidelities on the order of $10^{-4}$. A scaling analysis confirms that the logical error rate grows nearly quadratically with the physical error rate, confirming that all first-order errors are fully suppressed. Our protocol is compatible with current hardware and is scalable to multiple bosonic modes, providing a resource-efficient foundation for magic state preparation and higher-level concatenated quantum error correction.

Quantum information decoupling is a fundamental primitive in quantum information theory, underlying various applications in quantum physics. We prove a novel one-shot decoupling theorem formulated in terms of quantum relative entropy distance, with the decoupling error bounded by two sandwiched Rényi conditional entropies. In the asymptotic i.i.d. setting of standard information decoupling via partial trace, we show that this bound is ensemble-tight in quantum relative entropy distance and thereby yields a characterization of the associated decoupling error exponent in the low-cost-rate regime. Leveraging this framework, we derive several operational applications formulated in terms of purified distance: (i) a single-letter expression for the exact error exponent of quantum state merging in terms of Petz-Rényi conditional entropies, and (ii) regularized expressions for the achievable error exponent of entanglement distillation and quantum channel coding in terms of Petz-Rényi coherent informations. We further prove that these achievable bounds are tight for maximally correlated states and generalized dephasing channels, respectively, for the high distillation-rate/coding-rate regimes.

We calculate the phase diagram of the Bose-Hubbard model on a half-filled ladder lattice including the effect of finite on-site interactions. This shows that the rung-Mott insulator (RMI) phase persists to finite interaction strength, and we calculate the RMI-superfluid phase boundary in the thermodynamic limit. We show that the phases can still be distinguished using the number and parity variances, which are observables accessible in a quantum-gas microscope. Phases analogous to the RMI were found to exist in other quasi-1D lattice structures, with the lattice connectivity modifying the phase boundaries. This shows that the the presence of these phases is the result of states with one-dimensional structures being mapped onto higher dimensional systems, driven by the reduction of hopping rates along different directions.

When increasing the dimensionality of quantum systems, high-dimensional quantum state certification becomes important in quantum information science and technology. However, how to certify ensembles of high-dimensional quantum states in a black-box scenario remains a challenging task. In this work, we report an experimental test of certifying ensembles of high-dimensional quantum states based on prepare-and-measure experiments with \textit{independent devices}, where the state preparation device and the measurement device have no shared randomness. In our experiment, the prepared quantum states are high-dimensional orbital angular momentum states of single photons, and both the preparation fidelity and the measurement fidelity are about 99.0$\%$ for the six-dimensional quantum states. We also measure the crosstalk matrices and calculate the similarity parameter for up to ten dimensions. We not only experimentally certify the ensemble of high-dimensional quantum states in a semi-device-independent manner, but also experimentally investigate the effect of atmospheric turbulent noise on high-dimensional quantum state certification. Our experimental results clearly show that the certification of high-dimensional quantum states can still be achieved even under the influence of atmospheric turbulent noise. Our findings have potential implications in quantum certification and quantum random number generation.

We investigate white dwarfs in the framework of f(R,L_m) and f(R,L_m,T) gravity to explore the Chandrasekhar Limit. We have considered two functional forms of f(R,L_m) and one functional form of f(R,L_m,T) gravity. Considering the matter Lagrangian L_m=p, we calculate modified TOV equations for each of the forms. By employing the fully degenerate electron gas equation of state in the modified TOV equations, we derive the mass-radius relation for each functional form of both f(R,L_m) and f(R,L_m,T) gravity. Our models imply modifications in the Chandrasekhar mass limit that deviate significantly from the GR and the Newtonian cases. In the f(R,L_m, T)$ gravity, the new mass limit of the white dwarf can reach upto 1.537\,\mathrm{M}_\odot while in f(R,L_m) with the quadratic extension can goes upto 1.52\,\mathrm{M}_\odot and with exponential extension upto 2.08\,\mathrm{M}_\odot. Further, we analyze the static stability criterion, the gravitational redshift, and the adiabatic indices. For the power-law form of f(R,L_m) and the non-linear form of f(R,L_m,T) gravity, significant variations are observed at higher densities (ρ_c > 10^{10}\, \mathrm{g/cm^3}), while substantial changes are noted at much lower central densities in the case of exponential form of f(R,L_m) gravity. We also calculate compactness and gravitational redshift, which are much lower than those of neutron stars and black holes. Stability is also confirmed by adiabatic indices, which show that all models yield Γ> 4/3 throughout the interiors of WDs. Overall, our models provide a viable framework for the existence of super-Chandrasekhar mass limit, extending beyond the classical predictions in the Newtonian and/or GR cases.

Recent observations from the Atacama Cosmology Telescope (ACT), especially when combined with DESI baryon acoustic oscillation data, indicate a scalar spectral index $n_s$ higher than the value reported by \textit{Planck} 2018, placing tension on universal inflationary attractor models. Motivated by this discrepancy, we investigate the inflationary predictions of the $β$-exponential potential, $V(φ)=V_0\left(1-λβφ/M_p\right)^{1/β}$ considering both minimally and non-minimally coupled realizations. This potential generalizes standard exponential inflation and naturally arises in braneworld scenarios. We derive analytical expressions for the slow-roll parameters and inflationary observables using a perturbative expansion in the non-minimal coupling $ξ$, and validate these results through numerical calculations. In the minimally coupled case, the model predicts $n_s \simeq 0.976$ and $r \simeq 0.035$ for $N=50$ and moderate values of β, remaining compatible with ACT+DESI constraints at the 1σlevel while yielding a spectral tilt larger than the universal attractor prediction. Introducing a small non-minimal coupling significantly improves agreement with observations by suppressing the tensor-to-scalar ratio while preserving the enhanced scalar tilt. For $N=60, λ\sim 0.3-0.5$, and $β\sim O(1-5)$, the non-minimally coupled model yields $n_s \simeq 0.974-0.976$ and $r \lesssim 0.03$, comfortably consistent with ACT, DESI, and BICEP/Keck bounds. Our results show that the $β$-exponential potential, especially when implemented with a non-minimal coupling, exhibits good agreement with the latest CMB observations. Our inflationary predictions of the non-minimal model of $n_s$ and $r$ confirming the leading-order contributions in $ξ$ are sufficient to capture the essential features of both $r$ and $n_s$ in observationally relevant regimes.

Estimating the quantum Fisher information (QFI) is a crucial yet challenging task with widespread applications across quantum science and technologies. The recently proposed Krylov shadow tomography (KST) opens a new avenue for this task by introducing a series of Krylov bounds on the QFI. In this work, we address the practical applicability of the KST, unveiling that the Krylov bounds of low orders already enable efficient and accurate estimation of the QFI. We show that the Krylov bounds converge to the QFI exponentially fast with increasing order and can surpass the state-of-the-art polynomial lower bounds known to date. Moreover, we show that certain low-order Krylov bound can already match the QFI exactly for low-rank states prevalent in practical settings. Such exact match is beyond the reach of polynomial lower bounds proposed previously. These theoretical findings, solidified by extensive numerical simulations, demonstrate practical advantages over existing polynomial approaches, holding promise for fully unlocking the effectiveness of QFI-based applications.

In dynamically formed binaries, the spins of the black holes tend to be misaligned with the system's orbital angular momentum. This causes the spins to precess and leads to an asymmetric emission of gravitational waves. The resulting gravitational-wave multipole asymmetries directly source the recoil of the remnant black hole and are the critical element in fully describing precession. Recoil and precession are of significant astrophysical importance, but multipole asymmetries contribute only minimally to the overall signal strength. Consequently, most current gravitational-wave models either do not incorporate asymmetries at all, or only consider the dominant ones. Here we highlight the importance of subdominant multipole asymmetries for an accurate recoil velocity calculation and discuss their detectability with third generation detectors. Neglecting subdominant asymmetries leads to velocity differences of up to 210 km/s and can, in particular, introduce systematic biases in the inference of masses and the spin geometry. We further discuss universal characteristics of subdominant multipole asymmetries in order to prepare the ground for potential future asymmetry models. In the inspiral regime, the average antisymmetric frequencies can be described by a multiple of the orbital frequency. During ringdown, however, they become equal to their corresponding symmetric frequencies.

Absolute cross sections for the formation of OH- from 2-propanol (CH3CH(OH)CH3) via dissociative electron attachment (DEA) are reported in the incident electron energy range of 3.5-13 eV. Four fragment anions are observed: OH-, C2H2O-, C2H4O-, and C3H7O-. The OH- yield exhibits a pronounced resonance centered at 8.2 eV together with a broader structure extending over the 8-10 eV region. Equation-of-Motion Coupled-Cluster (electron attached) calculations with Singles and Doubles combined with a Complex Absorbing Potential (CAP/EOM-EA-CCSD) assign this feature to a two-particle-one-hole (2p-1h) core-excited Feshbach resonance. Potential energy curves along the C-OH dissociation coordinate reveal that core-excited anion states in this energy range promote efficient cleavage of the hydroxyl group. Analysis of Dyson orbitals and resonance widths demonstrates that only states with repulsive antibonding sigma(C-OH) character and sufficiently long lifetimes contribute significantly to the observed OH- production. These results provide fundamental insight into the DEA dynamics of secondary alcohols and highlight the role of multi-electron-attached resonances in site-specific bond rupture induced by low-energy electrons.

Quantum teleportation between polarized single-photon and phase-opposite coherent states is studied using a hybrid entangled resource and entangled coherent states. The polarized single-photon qubit represents a discrete-variable (DV) quantum system, whereas the phase-opposite coherent-state qubit constitutes a continuous-variable (CV) system. While teleportation from CV to DV can be achieved with near-unit success probability, the reverse process is usually limited to a maximum success probability of $1/2$. We demonstrate that, by employing cross-Kerr nonlinearity together with passive linear optical components such as polarizing beam splitters, beam splitters, and phase shifters, almost perfect teleportation from DV to CV encodings can also be achieved.

Building on hybrid quantum lattice gas algorithm, we revisit the possibilities of this quantum lattice model. By deriving a correction to the predicted viscosity, we provide analytical and numerical results that refine original formulation. We introduce a minimal 2D generalization of the algorithm, which allows to simulate anisotropic Burgerlike equations while retaining only two lattice velocities. This approach opens a promising route toward embedding momentum conservation and advancing toward NavierStokes dynamics in 2D, going beyond Frisch, Hasslacher and Pomeau (FHP) with a quantum native model.

Neural post-processing has been proposed as a lightweight route to enhance variational quantum eigensolvers by learning how to reweight measurement outcomes. In this work, we identify three general desiderata for such data-driven neural post-processing -- (i) self-contained training without prior knowledge, (ii) polynomial resources, and (iii) variational consistency -- and show that current approaches, such as diagonal non-unitary post-processing (DNP), cannot satisfy these requirements simultaneously. The obstruction is intrinsic: with finite sampling, normalization becomes a statistical bottleneck, and support mismatch between numerator and denominator estimators can render the empirical objective ill-conditioned and even sub-variational. Moreover, to reproduce the ground state with constant-depth ansatzes or with linear-depth circuits forming unitary 2-designs, the required reweighting range (and hence the sampling cost) grows exponentially with the number of qubits. Motivated by this no-go result, we develop a normalization-free alternative, the unitary variational quantum-neural hybrid eigensolver (U-VQNHE). U-VQNHE retains the practical appeal of a learnable diagonal post-processing layer while guaranteeing variational safety, and numerical experiments on transverse-field Ising models demonstrate improved accuracy and robustness over both VQE and DNP-based variants.

The gravitational wave (GW) event S250206dm, as the first well-localized neutron star merger candidate potentially located in the mass gap, presented a unique opportunity to probe the electromagnetic signatures from such a system. Here we report a deep, multiband search with the new 2.5-meter Wide Field Survey Telescope (WFST), covering about 64% of the localization region up to a 5-sigma limiting magnitude of 23 mag. In total, 12 potential candidates have been identified while none of them are likely related to S250206dm. This non-detection provides the most stringent constraint to date on any associated kilonova. Crucially, an AT 2017gfo-like event at 269 Mpc can be excluded by WFST observations alone. Based on ejecta mass limits, a neutron star-black hole with a large mass ratio (Q >= 3.2) is disfavored. This optical-derived constraint on the mass ratio reaches, for the first time, a precision comparable to that inferred from the GW signal. This work presents the best observation of this type of events until now, and demonstrates the power of rapid, deep follow-up observations to constrain the properties of compact binary progenitors, offering key insights into the constituents of the mass gap.

We have reproducibly grown near-single-domain superconducting aluminum (Al) films on GaAs(111)A wafers using molecular beam epitaxy. Synchrotron X-ray diffraction revealed twin-domain ratios of 0.00005 and 0.0003 for 19.4-nm- and 9.6-nm-thick films, respectively-the lowest reported for Al on any substrate and long considered unattainable for practical device platforms. Azimuthal scans across off-normal Al{$11\bar{1}$} reflections exhibit narrow full width at half maximum (FWHM) values down to $0.55^\circ$, unmatched by epi-Al grown by any other method. Normal scans showed a well-defined (111) orientation with pronounced Pendellösung fringes, and $θ$-rocking-curve FWHM values down to $0.018^\circ$; the former indicates abrupt film-substrate and oxide-film interfaces. Electron backscatter diffraction mapping confirms macroscopic in-plane uniformity and the absence of $Σ$3 twin domains. Atomic force microscopy and scanning transmission electron microscopy confirmed atomically smooth surfaces and abrupt heterointerfaces. The films exhibit critical temperatures approaching bulk values, establishing a materials platform for scalable, high-coherence superconducting qubits.

An industrial-scale adoption of Quantum Key Distribution (QKD) requires the development of practical, stable, resilient and cost-effective hardware that can be manufactured at large scales. In this work we present a high-speed (1.25GHz), field-deployable QKD prototype based on integrated photonics, that is consolidated into standard 19-inch rack compatible units. Through integrated photonics, the system prioritizes autonomous long-term stability in metropolitan settings. The architecture is further simplified by removing the need for chromatic dispersion compensation over metropolitan distances (below 100km). We demonstrate continuous key exchange over more than 4 km of metropolitan optical fiber, where the prototype maintained stable, uninterrupted operation across a measurement spanning more than 12 day-night cycles without manual intervention.

The Kolmogorov stochasticity parameter (KSP) as a sensitive descriptor of degree of randomness of signals is used to analyze the properties of the NANOGrav pulsar timing data associated to a stochastic gravitational wave background. The time of arrival (TOA) data of white noise for 68 pulsars are analyzed regarding their KSP properties. The analysis enables to obtain the degree of randomness of the white noise for various pulsars and to reveal its inhomogeneity, i.e. pulsars with low and high randomness of the white noise. The time-dependence of the randomness in the white noise is also studied, indicating the existence of non-stationary physical processes influencing the pulsar timing. The KSP thus is acting as an indicator for the degree of the agreement between the observations and the timing models and as a test in revealing the contribution of various physical processes in the stochastic background signal.

The inspiral of a compact object into a black hole is a key source of low-frequency gravitational waves for future space-based detectors like LISA. While models of this process have advanced, they typically focus on asymptotically flat spacetimes. In this paper, we explore how the absence of asymptotic flatness affects the slow, adiabatic orbital evolution due to radiation reaction. This lack of asymptotic flatness can arise from external environments or an expanding universe. Using the Schwarzschild-de Sitter (SdS) spacetime, where the deviation from flatness is governed by the cosmological constant, we study bound orbits characterized by their semi-latus rectum $p$ and eccentricity $e$. We calculate how the cosmological constant shifts the separatrix between bound and plunging orbits and alters the relationship between the binary's binding energy, angular momentum, and orbital parameters. Assuming the orbital timescale is much shorter than the inspiral timescale, we apply a modified quadrupole formula to examine the impact of a small positive cosmological constant on the orbital evolution in the weak-field limit. We find that the cosmological constant accelerates the decrease in eccentricity, reducing inspiral plunge times, which could influence event rate estimates for space-based detectors.

We explore the operation of quantum batteries in the Lipkin-Meshkov-Glick (LMG) model, when they are charged either through a sudden quench in the magnetic field strength or by coupling them to a bosonic oscillator bath. Through initializing the battery in either the symmetric or broken symmetry phases of the LMG model we analyze how the different spectral properties can affect the performance of both the charging and discharging of the battery. In particular, we show that by quenching the magnetic field strength from the symmetric phase to the broken phase, we can achieve a significant enhancement in stored energy, as well as stable and efficient ergotropy extraction. Similar observations can be made when introducing weak coupling between the battery with the bosonic bath, while the amount of stored work and ergotropy saturate at strong coupling. These findings emphasize the importance of the magnetic field dynamics and environmental coupling in optimizing charging performance, which could lead to practical applications in quantum energy storage.

Quantum communication holds the potential to revolutionize information transmission by enabling secure data exchange that exceeds the limits of classical systems. One of the key performance metrics in quantum information theory, namely the Holevo bound, quantifies the amount of classical information that can be transmitted reliably over a quantum channel. However, computing and optimizing the Holevo bound remains a challenging task due to its dependence on both the quantum input ensemble and the quantum channel. In order to maximize the Holevo bound, we propose a unified projected gradient ascent algorithm to optimize the quantum channel given a fixed input ensemble. We provide a detailed complexity analysis for the proposed algorithm. Simulation results demonstrate that the proposed quantum channel optimization yields higher Holevo bounds than input ensemble optimization.

We report a quantitative evaluation of weak-value amplification (WVA) for longitudinal phase measurements using Allan variance analysis. Building on a recent double-slit interferometry experiment with real weak values [Phys. Rev. Lett. 134, 080802 (2025)], our Allan variance analysis demonstrates measurement of a few attosecond time delay approaching the shot noise limit at short averaging intervals of $T$ = $0.01-0.1$ s, representing two orders of magnitude variance reduction compared to the $T=300$ s operating point in prior implementations. We demonstrate that the Allan-variance noise floor scales with the inverse of the detected photon number $1/N_r$, confirming shot-noise-limited operation with WVA. Furthermore, this $1/N_r$ scaling experimentally validates that WVA can outperform conventional measurement under fixed detected photon number and detector saturation, in the presence of technical noise, as theoretically predicted [Phys. Rev. Lett. 118, 070802 (2017)]. Our results provide rigorous, quantitative evidence of the near-optimal noise performance achievable with WVA, establishing a new benchmark for precision optical metrology. This advancement is particularly relevant to applications such as gravitational-wave detection, where signals predominantly occupy the high-frequency regime ($>10$ Hz).

Superconducting waveguides are a promising platform for ultralow-loss transmission in the millimeter-wave to terahertz band under cryogenic conditions, with potential applications in astronomical instrumentation and emerging quantum technologies. We develop a framework, based on microscopic superconductivity theory, to evaluate the power-flow attenuation constant $α$ of superconducting rectangular waveguides in the $100~\mathrm{GHz}$--THz range, applicable to arbitrary electronic mean free paths $\ell$ from the dirty limit $\ell\llξ_0$ to the clean limit $\ell\ggξ_0$. We also derive an analytical expression for two-level-system (TLS)-induced attenuation $α_{\rm TLS}$ in thin native oxide layers within the standard TLS model. Using this framework, we perform numerical evaluations of $α$ for representative materials over standard waveguide sizes from WR15 to WR1. In the high-frequency regime $f \gtrsim 0.5 Δ/h$, low attenuation favors the clean regime $\ell\gtrsimξ_0$, indicating that high-purity materials can achieve very low attenuation below their gap frequency. For the TLS contribution, using parameter values representative of native Nb oxides, we find that $α_{\rm TLS}$ can become relevant at sufficiently low temperatures $T/T_c\lesssim 0.1$-0.2, where quasiparticle dissipation is exponentially suppressed. Finally, we extend the discussion to the strong-excitation regime using a recently developed nonlinear-response theory within the Keldysh--Usadel framework of nonequilibrium superconductivity and show that nonlinear dissipation produces a Higgs-mode peak in $α$ near $f\simeq Δ/h$ via a Kerr-type nonlinearity of the dissipative conductivity. This peak provides a distinct hallmark of the Higgs mode that has been largely overlooked so far.

In his late piece 'La nouvelle cuisine' (Bell 1990), John Bell describes the steps from an intuitive, informal principle of locality to a mathematical rule called Factorizability. This rule stipulates that when possible past causes are held fixed, the joint probabilities of outcomes of spacelike separated measurements, conditional on measurement settings, be the product of the local conditional probabilities individually. Bell shows that Factorizability conflicts with predictions of QM, predictions since confirmed in many experiments. However, Bell warns his readers that the steps leading to Factorizability should 'be viewed with the utmost suspicion'. He says that 'it is precisely in cleaning up intuitive ideas for mathematics that one is likely to throw the baby out with the bathwater' (1990, 239). Bell's suspicions were well-founded, for he himself misses an important baby. Here we retrieve and identify it: it is selection bias. We explain how failure of Factorizability may be regarded as a selection artefact, requiring no violation of locality in the intuitive, conceptual sense with which Bell begins his analysis. The argument begins with a central principle of causal discovery, Reichenbach's Principle of Common Cause (PCC). It is well known that correlations due to selection bias are not subject to PCC. Several writers have proposed that EPR-Bell correlations are also an exception to PCC, but it has not been noticed that they fall under this well-known exclusion. The point is relevant not only to the status of Bell nonlocality, but also for statistics and causal modeling. For these fields, the news is that selection effects play a ubiquitous role in quantum phenomena, in a form akin to collider bias.

We study the propagation of massive scalar fields in the background of asymptotically flat regular black holes supported by a phantom scalar field with a scalar charge $A$. This parameter regularizes the geometry by removing the central singularity. Focusing on wave dynamics, we analyze scalar perturbations, quasinormal modes, and greybody factors, emphasizing the role of the regularization parameter on the effective potential and the decay properties of the modes. Using WKB methods beyond the eikonal limit, we show that the presence of scalar hair modifies both the oscillation frequencies and damping rates of quasinormal modes. In particular, we demonstrate the occurrence of an anomalous decay rate for massive scalar perturbations: above a critical field mass, the longest-lived modes correspond to lower angular momentum, in contrast with the massless case. We derive analytical expressions for the critical mass and study its dependence on the scalar charge and overtone number. Furthermore, we apply the Horowitz-Hubeny method to compute the quasinormal frequencies and show that the results obtained from the WKB and Horowitz-Hubeny approaches exhibit excellent agreement in the regime where both methods are valid. In addition, we compute reflection and transmission coefficients and analyze the corresponding greybody factors, clarifying how regularity effects imprint themselves on black-hole scattering properties. Our results show that regular black holes with scalar hair exhibit distinctive dynamical signatures that can be probed through quasinormal ringing and wave propagation.

This is a personal account of the early work that led to what is now known as the ``preheating stage" of inflationary cosmology. The broader applicability of the underlying instability mechanisms in cosmology are indicated.

A theory of the spectral "zebra" pattern of the Crab pulsar's high-frequency interpulse (HFIP) radio emission is developed. The observed emission bands are interference maxima caused by multiple ray propagation through the pulsar magnetosphere. The high-contrast interference pattern is the combined effect of gravitational lensing and plasma de-lensing of light rays. The model enables space-resolved tomography of the pulsar magnetosphere, yielding a radial plasma density profile of $n_{e}\propto r^{-3}$, which agrees with theoretical insights. We predict the zebra pattern trend to change at a higher frequency when the ray separation becomes smaller than the pulsar size. This frequency is predicted to be in the range between 42 GHz and 650 GHz, which is within the reach of existing facilities like ALMA and SMA. These observations hold significant importance and would contribute to our understanding of the magnetosphere. Furthermore, they offer the potential to investigate gravity in the strong field regime near the star's surface.

The preparation of a quantum state using a noisy quantum computer (gate noise strength $δ$), will necessarily affect an O($δ$)-fraction of the qubits, no matter which protocol is used. Here, we show that fault-tolerant quantum state preparation can be achieved with constant space overhead improving on previous constructions requiring polylogarithmic overhead. To achieve this, we add to the toolbox of fault-tolerant schemes for circuits with quantum input and output. More specifically, we construct fault-tolerant interfaces that decrease the level of protection for quantum low-density parity-check (LDPC) codes. When information is encoded in multiple code blocks, our interfaces have constant space overhead. In our decoder construction that change the level of protection by an arbitrary amount, we circumvent bottlenecks to error pileup and overhead by gradual lowering of the level of encoding at the same time as we increase the number of blocks on which decoding is carried out simultaneously.

Quantum error correction (QEC) is essential for realizing fault-tolerant quantum computation. Current QEC controllers execute all scheduled syndrome (parity-bit) measurement rounds before decoding, even when early syndrome data indicates that the run will result in an error. The resulting excess measurements increase the decoder's workload and system latency. To address this, we introduce an adaptive abort module that simultaneously reduces decoder overhead and suppresses logical error rates in surface codes and color codes under an existing QEC controller. The key idea is that initial syndrome information allows the controller to terminate risky shots early before additional resources are spent. An effective scheme balances the cost of further measurement against the restart cost and thus increases decoder efficiency. Adaptive abort schemes dynamically adjust the number of syndrome measurement rounds per shot using real-time syndrome information. We consider three schemes: fixed-depth (FD) decoding (the standard non-adaptive approach used in current state-of-the-art QEC controllers), and two adaptive schemes, AdAbort and One-Step Lookahead (OSLA) decoding. For surface and color codes under a realistic circuit-level depolarizing noise model, AdAbort substantially outperforms both OSLA and FD, yielding higher decoder efficiency across a broad range of code distances. Numerically, as the code distance increases from 5 to 15, AdAbort yields an improvement that increases from 5% to 35% for surface codes and from 7% to 60% for color codes. To our knowledge, these are the first adaptive abort schemes considered for QEC. Our results highlight the potential importance of abort rules for increasing efficiency as we scale to large, resource-intensive quantum architectures.

Quantum computing improves substantially on known classical algorithms for various important problems, but the nature of the relationship between quantum and classical computing is not yet fully understood. This relationship can be clarified by free models, that add to classical computing just enough physical principles to represent quantum computing and no more. Here we develop an axiomatisation of quantum computing that replaces the standard continuous postulates with a small number of discrete equations, as well as a free model that replaces the standard linear-algebraic model with a category-theoretical one. The axioms and model are based on reversible classical computing, isolate quantum advantage in the ability to take certain well-behaved square roots, and link to various quantum computing hardware platforms. This approach allows combinatorial optimisation, including brute force computer search, to optimise quantum computations. The free model may be interpreted as a programming language for quantum computers, that has the same expressivity and computational universality as the standard model, but additionally allows automated verification and reasoning.

Black hole mimickers (BHMs) are horizonless globally regular ultracompact relativistic self-gravitating objects (UCOs) of mass $M$ and radius $R$ with compactness $C = M/R$ higher than that of a neutron star and that produce an effective potential for null geodesics (photons) that possesses a local maximum, which is usually accompanied by an inner local minimum. The presence of a local maximum allows for unstable circular orbits to exist similar to light rings present in actual BH solutions, while it has been conjectured that the presence of a local minimum is symptomatic of potential instabilities. One such candidate for a BHM is a solitonic boson star (SBS) which is a boson star endowed with a sextic potential. In this paper we investigate further solutions of static and spherically symmetric SBSs in general relativity with a larger set of parameter values, and argue that such solutions are very similar to UCOs composed of an incompressible perfect fluid (IPF) with a sufficiently large pressure (the mimicker of a BHM). These IPFUCOs reach the Buchdahl limit $C= 4/9$ for arbitrarily large pressures. We investigate the extent to which the IPFUCOs constructed within a quadratic model in $f(R)$ gravity can overcome this limit or not, and thus pave the way for possibly building SBSs (or other kind of UCO) within this (or other alternative theory of gravity). We further elaborate about the stability properties of SBSs which have been the subject of some controversy recently.

This article is a sequel to our previous paper (arXiv:2511.12311), where we considered the conceptual problem on the empirical laws for the Klein\textendash Gordon quantum field theory in curved spacetime (QFTCS), and we will consider the similar problems for the Majorana field on curved spacetime here. A ``law'' in theoretical physics is said to be observable or empirical only if it can be verified/falsified by some experimental procedure. The notion of empiricality/observability becomes far more unclear in QFTCS, than in QFT in Minkowski (flat) spacetime (QFTM), mainly because QFTCS lacks the notion of vacuum. This could potentially undermine the status of QFTCS as a physical (not only mathematical) theory. We consider this problem for the Majorana field in curved spacetime, and examine some examples of the empirical laws.

We explore the multimetric theory of gravitation, also known as multigravity. We derive additional new exact solutions for the theory in proportional Kerr-Schild and double Kerr-Schild forms. We extend several solutions from the theory of General Relativity, characterized by a constant Ricci scalar in single and double Kerr-Schild forms, to derive solutions in the multi-gravity context. We also examine and extend the classical double copy relations that can be constructed out from these solutions in multigravity exploring the dynamics of the single copy and zero copy fields.

We investigate the collective dynamics of a three-level ensemble under the Dicke limit, revealing a unified connection between superradiant emission and electromagnetically induced transparency (EIT). Our results show that the transient superradiant burst exhibits the expected peak intensity scaling $I_{\max}\!\sim\! N^2$, with a universal finite-size correction $|ξ(N)-2|\!\sim\! 1/\ln N$ that governs the apparent scaling exponent in realistic ensembles. In the stationary regime, collective broadening modifies the EIT response: although it typically enhances absorption, it counterintuitively increases the group velocity, leading to a relative scaling $v_g\!\propto\! N^2$, even while $v_g\!\ll\! c$. This effect suggests that cooperative interactions fundamentally limit the achievable slow-light delay in dense media. To achieve these results, we derive a representative-atom master equation that quantitatively reproduces both the superradiant and EIT regimes, in excellent agreement with the exact symmetric-subspace dynamics and correctly incorporating collective feedback and $N$-dependent broadening. This unified framework bridges transient superradiant emission and steady-state quantum interference, with direct implications for slow light, quantum memories, and precision metrology.

Accurate in situ calibration of microwave attenuation and amplification-chain noise is essential for superconducting quantum circuits. We demonstrate a compact, self-calibrating cryogenic noise source based on an on-chip chromium attenuator that can be resistively heated with nanowatt-level power and directly integrated into a coaxial microwave line at the mixing-chamber stage. By comparing Johnson-Nyquist noise generated by Joule and microwave heating, measured through the amplification chain, the attenuation of the input line, and hence the gain of the chain, is determined without requiring knowledge of the attenuator temperature. The device exhibits millisecond-scale response times and negligible heating of the cryostat base plate. Using this approach, we determine the gain and added noise of a cryogenic amplification chain over the 4-8 GHz band. Our results provide a simple and accurate method to characterize near-quantum-limited parametric amplifiers used in superconducting-qubit readout.

We investigate particle motion in regular and asymptotically flat black hole spacetimes supported by Dehnen-type dark-matter halos. Two analytic models are analyzed, allowing a systematic study of circular geodesics, photon-sphere properties, shadow radius, Lyapunov exponent, ISCO frequency, binding energy, and Hawking temperature. The corrected numerical results show that the halo scale parameter can significantly modify strong-field observables. In both models, for moderate density slopes, increasing the halo parameter reduces characteristic radii while enhancing orbital instability and accretion efficiency. For steeper density falloff, however, deviations from the Schwarzschild case remain small. These results demonstrate that halo-induced modifications of optical and dynamical black hole signatures are strongly controlled by the density profile parameters.

Based on our experimental findings, we propose the following decision framework for practitioners. Quantum annealing is recommended when the problem formulation QUBO exhibits a high gradient variance (greater than 0.3) and the energy landscape contains numerous thin barriers characterized by sharp peaks and narrow valleys. Additionally, quantum approaches are particularly suitable when classical methods are observed to get trapped in local minima, the problem size is manageable given hardware constraints (less than 5000 variables for pure quantum annealing), and the time overhead of approximately 10 seconds is acceptable for the application. In contrast, classical methods are recommended when the gradient variance is low (less than 0.2), indicating smooth landscapes where quantum tunneling provides little advantage. Classical approaches are also preferable when the problem size is small and classical solvers can provide nearly instantaneous results, when solution quality requirements are modest and local optima suffice, or when hardware access or cost is a limiting factor. For problems that exceed pure quantum capacity but possess a favorable landscape structure, hybrid approaches combining quantum and classical techniques are recommended. Such hybrid methods are particularly effective when decomposition quality can be verified and both solution quality and scalability are important considerations.

We present a novel solution to the cosmological constant (CC) problem that requires no fine-tunings nor anthropic reasoning. In pre-geometric gravity (PGG), spacetime emerges from the spontaneous breaking of a fundamental gauge symmetry. This mechanism dynamically generates general relativity while also revealing a deep connection: the topological Gauss-Bonnet coupling of the theory scales precisely as the de Sitter entropy, an enormous number which reflects the information content of our universe. This coupling acts as a gravitational $θ$-angle parameter, forcing the CC to become quantized into discrete topological sectors. The symmetry-breaking dynamics naturally selects the sector corresponding to the observed vacuum energy. The selected vacuum state is stabilized by the extremely large potential barrier of the pre-geometric Higgs field, which effectively seals it off from quantum tunneling transitions to other topological sectors. The PGG framework thus provides a dynamical explanation for the smallness of the CC, linking gravity, topology and quantum information in a unified picture.

An important next step for pulsar timing arrays (PTAs) is to measure anisotropies in the gravitational wave background (GWB) at $\sim$ nano-Hz frequencies. We calculate the expected GWB anisotropies using empirically calibrated models for the merger rates of supermassive black hole binaries (SMBHBs). The anisotropies reflect both shot-noise in the discrete SMBHB populations while also tracing, in part, the large-scale structure (LSS) of the universe. The shot-noise term is sensitive to the high-mass end of the merging SMBH mass function, depends somewhat on the low-redshift tail of the merger distribution, and is a strong function of observing frequency. The precise frequency dependence provides a test of SMBHB residence times. In our models, the mean shot-noise anisotropy typically lies close to or above the broad frequency-band NANOGrav upper limits. Consequently, near-future PTA data, and potentially re-analyses of existing measurements using frequency-dependent shot-noise anisotropy templates, should be capable of detecting this signal or placing meaningful constraints on SMBHB merger models. A full interpretation, however, will require modeling the probability distribution of shot-noise amplitudes rather than relying solely on ensemble-averaged predictions. The LSS-induced anisotropies are at least two to three orders of magnitude smaller. Although the LSS contribution contains valuable information regarding the redshift distribution and clustering bias of the merging SMBHBs, detecting this component will be challenging.

Moiré systems have emerged as an exciting tunable platform for engineering and probing quantum matter. A large number of exotic states have been observed, stimulating intense efforts in experiment, theory, and simulation. Utilizing a neural-network-based quantum Monte Carlo approach, we discover a new ground state of the two-dimensional electron gas in a honeycomb moire potential at a filling factor of $ν_m =1/4$ (one electron every four moiré minima). In this state, two opposite-spin electrons pair to form a singlet-like valence bond state which restores local $C_6$ symmetry in hexagonal molecules each spanning $6$ moiré minima. These molecules of pairs then form a molecular Wigner crystal, leaving one quarter of the moiré minima mostly depleted. The formation of such a paired Wigner crystal, absent any confining potential or attractive interaction to facilitate "pre-assembling" the molecule, provides a fascinating case of collective phenomena in strongly interacting quantum many-body systems, and opportunities to engineer exotic properties.

The Laser Interferometer Space Antenna is expected to observe gravitational waves from massive black hole binaries across cosmic time. Many are anticipated to be detectable hours to weeks before coalescence. We present a fast algorithm for the pre-merger detection and preliminary characterization of such binaries. The method performs a search for excess power with a chirping time-frequency morphology in short-time Fourier transform spectrograms. By tiling the time-frequency plane with slices defined by the quadrupole frequency evolution, we define a signal significance relative to a fitted background distribution of instrumental noise and Galactic foreground. Individual search triggers are followed by a coherence tracker, which groups over time triggers consistent with the same physical signal . Doing so, our analysis provides progressively refined estimates of the chirp mass and coalescence time. We validate our algorithm on the Sangria LISA Data Challenge dataset, successfully detecting all 15 injected MBHBs: 14 of them hours-to-weeks before merger, while one is only detected after the binary coalescence. The algorithm yields chirp mass relative errors below $3\%$ for high-SNR sources and coalescence time uncertainties of up to a few hours. With a computational cost of less than a second to process a 10-day data segment on single core, our approach is suitable for generating real-time alerts, trigger protected observational periods, and provide informative priors for Bayesian parameter estimation.

Equilibrium statistical ensembles impose stringent constraints on phases of quantum matter. For example, the Mermin-Wagner theorem prohibits long-range order in low-dimensional systems beyond the ground state. Here, we show that quantum circuits can learn states of matter with long-range order that are inaccessible in equilibrium. We construct variational quantum circuits that generate symmetry-broken and symmetry-protected topological states with long-range order in one-dimensional systems at finite energy density, where equilibrium states are typically featureless. Importantly, the learned states can exhibit unconventional features with enhanced metrological properties such as a quantum Fisher information close to a GHZ state, but robust against local measurements. Our work establishes coherent quantum dynamics as a powerful resource for engineering nonequilibrium phases of matter, opening a path toward a broader dynamical scope of quantum order beyond the constraints of equilibrium ensembles.

Many quantum phases, from topological orders to superfluids, are destabilized at finite temperature by the proliferation and motion of topological defects such as anyons or vortices. Conventional protection mechanisms rely on energetic gaps and fail once thermal fluctuations exceed the gap scale. Here we examine a complementary mechanism of entropic protection, in which defect nucleation is suppressed by coupling to mesoscopic auxiliary reservoirs of dimension $M$, generating an effective free-energy barrier that increases with temperature. In the Ising chain, this produces a characteristic three-regime evolution of the correlation length as a function of temperature - linear growth, entropy-controlled plateau, and eventual breakdown - indicating a general modification of defect behavior. Focusing on two spatial dimensions, where true finite-temperature topological order is forbidden in the thermodynamic limit, we show that entropic protection can nevertheless strongly enhance stabilization at finite system size, the regime directly relevant for quantum memory and experiments. Owing to the topological character of the defects, creation and transport are independently suppressed, yielding a double parametric reduction of logical errors in the entropic toric code and enhanced coherence when the framework is extended to Berezinskii-Kosterlitz-Thouless transitions. Entropic barriers thus provide a passive and scalable route to stabilizing quantum phases in experimentally relevant regimes. We propose an experimental setup for entropic toric code using dual species Rydberg arrays with dressing.

At long distances, a gapped phase of matter is described by a topological quantum field theory (TQFT). We conjecture a tight and concrete relationship between the genuine $(d+1)$-partite entanglement -- labelled by a $d$-dimensional manifold $M$ -- in the ground state of a $(d-1)+1$-dimensional gapped theory and the partition function of the low energy TQFT on $M$. In particular, the conjecture implies that for $d=3$, the ground state wavefunction can determine the modular tensor category description of the low energy TQFT. We verify our conjecture for general (2+1)-dimensional Levin-Wen string-net models.

We derive static spherically symmetric regular black holes as vacuum solutions to purely gravitational theories in four dimensions. To that end, we construct four-dimensional non-polynomial gravities starting from subclasses of two-dimensional Horndeski actions. By construction, these theories possess second-order equations of motion on spherically symmetric backgrounds. We show that a subset of these non-polynomial gravities, referred to as non-polynomial quasi-topological gravities, admit single-function static spherically symmetric solutions whereby the metric function is determined by an algebraic equation. Solutions to these theories include the Hayward regular black-hole spacetime, for which a corresponding gravitational action is stated explicitly.

We study the late-time cosmological dynamics of a spatially flat FLRW universe in the framework of $f(Q)$ gravity, where $Q$ denotes the nonmetricity scalar. The matter sector is modeled as a bulk viscous fluid with a free equation-of-state parameter $γ$, allowing for a generalized description of cosmic matter beyond the standard dust approximation. We derive the background evolution equations and analyze the resulting expansion history. The model parameters are constrained using a combination of observational datasets, including cosmic chronometers (CC), baryon acoustic oscillations from DESI DR2, and Type~Ia supernovae (GRBs and Union3). Using the best-fit parameters, we further employ the statefinder and $\mathrm{Om}(z)$ diagnostics to distinguish the viscous $f(Q)$ scenario from the standard $Λ$CDM model. In addition, we examine the evolution of the deceleration parameter, which exhibits a transition from an early decelerated phase to the current accelerated expansion, and analyze the effective equation of state behavior. Our results show that bulk viscosity within $f(Q)$ gravity provides a viable and observationally consistent description of late-time cosmic acceleration.

We construct a class of regular black hole solutions of the Fan-Wang type within quasi-topological gravity (QTG) in arbitrary spacetime dimensions greater than four. In contrast to the original Fan-Wang solution, which was obtained in four-dimensional general relativity coupled to nonlinear electrodynamics, our higher-dimensional generalization does not require any matter fields. Instead, regularity is achieved purely through an infinite tower of higher-curvature corrections. We demonstrate that the Fan-Wang-type metric is a solution to the QTG field equations by explicitly determining the corresponding coupling constants for each curvature order. Within an appropriate parameter regime, the solution describes an asymptotically flat black hole spacetime with a regular center. Remarkably, even in the case of negative mass, the geometry can remain completely regular, in sharp contrast to Einstein gravity.

We study the parametric amplifier Hamiltonian within the framework of the Dunkl formalism. We introduce the Dunkl creation and annihilation operators and show that their quadratic combinations generate an $su(1,1)$ Lie algebra. The spectral problem is solved exactly using two algebraic methods: the $su(1,1)$ tilting transformation and the generalized Bogoliubov transformation. The exact energy spectrum and the corresponding eigenfunctions are obtained in terms of the Dunkl number coherent states. Furthermore, we compute the Mandel $Q$ parameter and the second-order correlation function $g^{(2)}(0)$ to analyze the statistical properties of the Dunkl squeezed states. We show that, for the squeezed vacuum, the Mandel parameter remains independent of the Dunkl deformation, whereas the correlation function exhibits an explicit dependence on the Dunkl parameter $μ$, which modifies the photon bunching effects. Finally, we show that our results reduce to the standard parametric amplifier case in the limit of vanishing Dunkl deformation parameter.

Superdense coding, the application of entanglement to boost classical communication capacity, is a cornerstone of quantum communication. In this paper, we propose a high-dimensional superdense coding protocol using energy-time entangled states. These states are biphoton frequency combs, an example of entangled time-frequency Gottesman-Kitaev-Preskill (TFGKP) states or time-frequency grid states. Inspired by GKP codes, our protocol involves discretizing the continuous time and frequency degrees of freedom and encoding information by time-frequency displacements. This approach leverages the inherently large Hilbert space found in quantum frequency combs, with resilience against both temporal and spectral errors. In addition to describing the theoretical structure of the protocol, we propose an experimental implementation using standard telecommunication components, time-resolving single-photon detectors and a frequency beamsplitter. We also analyze the effect of experimental noise and errors on the channel capacity of the protocol. We demonstrate that for realistic experimental parameters, contemporary technologies satisfy the prerequisites for superdense coding with biphoton frequency combs, achieving a transmission rate of approximately 8.91 bits per transmitted photon (equivalent to 481 distinguishable messages with asymptotically vanishing errors). This more than doubles the previously highest transmission rate of 4 bits achieved by the Kwiat-Weinfurter scheme, while also having competitive optical loss. Furthermore, our results beat the rate achievable using a single-photon frequency comb with identical parameters by 4.6 times. Our protocol thus represents an experimentally feasible application of time-frequency grid states to entanglement-assisted communication, contributing to the active fields of continuous-variable and high-dimensional quantum information.

We revisit the one-loop calculation of the Higgs mass spectrum of the scale-invariant THDM, relying on a direct calculation of the relevant Feynman diagrams. We highlight a number of incorrect assumptions in earlier calculations that relied on the effective-potential approach. In contrast with the earlier findings, we show that the one-loop corrections can have an effect of ${\cal O}(10\%)$ on the predictions for the BSM-Higgs masses, and they can also induce non-negligible mixing between the SM-like and BSM states in the neutral-scalar sector.

Recent experimental observations, most notably those reported by the ATOMKI and Positron Annihilation into Dark Matter Experiment (PADME) collaborations, have hinted anomalies that may indicate the presence of a new resonance with a mass around $17\,\text{MeV}$, potentially interacting with both nucleons and electrons. Since 2020, ATOMKI has observed this resonance in nuclear transitions from excited to ground states in ${}^{8}\mathrm{Be}$, ${}^{4}\mathrm{He}$, and ${}^{12}\mathrm{C}$. More recently, in 2025, PADME, operating at the Laboratori Nazionali di Frascati, has also hinted a similar excess, in this case in the $e^{+}e^{-}$ final-state events originating from positron annihilation on fixed-target atomic electrons of Carbonium. This concordance strengthens the case for a common underlying origin, potentially involving a new boson, conventionally referred to as $X_{17}$. Despite these intriguing developments, the global experimental landscape remains highly dynamic, particularly in light of recent MEG~II constraints, and a definitive confirmation or exclusion of the $X_{17}$ hypothesis is still lacking. Within this evolving and exciting context, this thesis investigates whether a hypothetical $17\,\text{MeV}$ particle, coupled to electrons as suggested by the PADME observations, could function as a mediator between the Standard Model and previously unexplored hidden sectors. Such a mediator could, in principle, offer a novel pathway toward addressing one of the principal outstanding inconsistencies of the Standard Model: the nature and origin of dark matter.

New exotic particles with MeV masses, such as axion-like particles or light dark matter, can be emitted from core-collapse supernovae (SNe) with semi-relativistic velocities. Due to their speed dispersion, they would arrive at Earth as an extended packet with a time spread that can be as large as tens of millennia for typical detectors. It has been argued in the literature that the superposition of packets from all galactic SNe would give rise to a smooth and stationary diffuse flux that could be observable on terrestrial experiments. In this article, we critically examine this hypothesis by carrying out a numerical simulation of the galactic history of SN explosions. We show that, although the particle packets do overlap, due to the short observational time window, each of them only contributes with a very narrow range of energies and with an intensity that depends on the SN distance. As a consequence, the energy dependence of the resulting flux is extremely sensitive to the stochastic nature of the SN population and far from smooth. This has profound implications for the expected signature in terrestrial experiments, which displays a spectral shape that is not properly described by the smooth approximation. We develop a numerical tool to compute this stochastic galactic flux for generic semi-relativistic particles, which also allows us to explore sub-MeV particles, where the smooth diffuse flux approach does not hold. To test this framework, we revisit existing bounds on axion-like particles and fermionic dark matter, finding weaker constraints than previously reported.

Fixed points of the 2d renormalization group flow are known to correspond to tree level string vacua. We discuss how the renormalization group (or "sigma model") approach can be extended to the string loop level. The central role of the condition of renormalizability of the generating functional for string amplitudes with respect to both "local" and "modular" infinities is emphasized. Several one-loop and two-loop examples of renormalization are considered. It is found that in order to ensure the renormalizability of the generating functional one is to use an "extended" (Schottky-type) parametrization of the moduli space. An approach to resummation of the string perturbative expansion based on operators of insertion of topological fixtures is suggested.

Radially symmetric non-BPS 't Hooft-Polyakov monopoles and dyons are constructed as resurgent transseries: infinite sums of exponentially decaying terms, each multiplied by a factorially divergent fluctuation factor. All higher exponential terms are explicitly expressed in terms of the leading order solutions. In the BPS limit all fluctuation terms truncate.

Experimental particle physics seeks to understand the universe by probing its fundamental particles and forces and exploring how they govern the large-scale processes that shape cosmic evolution. This whitepaper presents a vision for how Artificial Intelligence (AI) can accelerate discovery in this field. We outline grand challenges that must be addressed to enable transformative breakthroughs and describe how current and planned experimental facilities can implement this vision to advance our understanding of the vast and complex physical world from the smallest to the largest scales. We show how facilities currently under construction, such as the HL-LHC, DUNE and soon EIC, can both benefit from and serve as proving grounds for this vision, while also enabling a longer-term goal for how future experiments -- like FCC-ee at CERN, IceCube-Gen2, a Muon Collider in the U.S., and smaller to mid-scale projects -- can be fully AI-native. We describe how a truly national-scale collaboration, jointly managed across large funding partners, and involving both DOE laboratories and universities, can make this happen.

We present a momentum measurement method based on multiple Coulomb scattering (MCS) in the FASER$ν$ emulsion detector. The measurement of charged-particle momenta is essential for studying neutrino interactions in the TeV energy range at the FASER experiment. This method exploits the sub-micron spatial resolution and long tracking length of the FASER$ν$ detector, enabling momentum determination from a few GeV up to a few TeV. The performance was evaluated using Geant4-based Monte Carlo simulations and validated with muon test beam data in the momentum range 100-300 GeV. As a first probe of the method for higher momentum muons, background muons recorded by the FASER$ν$ detector were examined, showing reconstructed momenta consistent with expectations from their angular spread.

We review the physics of ultralight axion-like particles (ALPs) as dark matter candidates and the experimental strategies used to search for them with precision and quantum technologies. In the ultralight regime, the enormous occupation number of the dark matter field motivates a classical description in terms of a coherently oscillating background, leading to distinctive, time-dependent signatures in laboratory observables. We discuss the effective field theory framework governing ALP interactions with Standard Model fields, and show how different operators give rise to qualitatively different experimental signals. The lecture notes cover both conversion-based searches enabled by the axion-photon coupling, such as haloscopes and helioscopes, and precision experiments sensitive to oscillations of fundamental constants and material properties. These include atomic and nuclear clocks, optical cavities, laser and unequal time-delay interferometers, and mechanical or solid state resonators. Emphasis is placed on the physical origin of the sensitivity of each platform, the role of coherence, bandwidth, and noise, and the complementarity between different technologies across a wide range of ALP masses. Together, these approaches provide broad and overlapping coverage of ultralight dark matter parameter space and define a rapidly evolving experimental programme with strong discovery potential.

We introduce a compact, physics-driven event representation, RMM-C46, designed to compress the high-dimensional rapidity mass matrix (RMM) into a low-dimensional, interpretable feature set suitable for physics-informed machine learning (ML) and quantum computing applications. The full RMM encodes detailed pairwise correlations among jets, b-jets, leptons, photons, and missing transverse energy but contains more than a thousand values per event, making it computationally heavy for large-scale training and incompatible with current low-qubit quantum devices. The proposed RMM-C46 input space for ML preserves the physical block structure of the RMM through aggregated invariant mass, rapidity difference, and transverse energy components, reducing the size of the original RMM by over an order of magnitude while maintaining interpretability. Applied to simulated proton-proton collisions at centre-of-mass energy of 13.6 TeV, these representations match or exceed the discriminative performance of the full RMM in both supervised and unsupervised ML tasks. Their compactness, stability, and physics transparency also make them naturally compatible with near-term quantum machine learning architectures. RMM-C46 provides a scalable, efficient, and quantum-ready alternative to the full RMM for next-generation collider physics analyses.

In this letter we evaluate one-loop all-plus gluon amplitudes of $\mathrm{SU}(N_c)$ gauge theory with $N_f$ fundamental fermions in the presence of a flavour instanton background. Fermion zero modes are regulated with a chiral mass term. This computation is performed by cancelling a twistorial 't Hooft anomaly via the Green-Schwarz mechanism. We find that the trace-ordered amplitude has the form of a Parke-Taylor factor multiplied by the Fourier transform of the instanton density evaluated on the total momentum of the gluons. A background flavour instanton modifies the leading soft gluon and photon theorem, generating a level equal to twice the instanton charge in the soft Kac-Moody symmetry. We discuss the implications of our results for amplitudes in the presence of dynamical instantons.

The transverse-momentum spectrum of Drell-Yan lepton pairs at small $p_T$ probes non-perturbative QCD effects, including intrinsic partonic transverse momentum and initial-state soft-gluon radiation. The novel approach PDF2ISR is employed to study the transverse-momentum spectrum of Drell-Yan lepton pairs in the small-$p_T(\ell\ell)$ region. This framework is particularly well suited for such investigations, as it provides a systematic treatment of the dominant non-perturbative effects and their interplay. In order to extract robust physical conclusions, a detailed analysis of the technical aspects involved in the simulation of the intrinsic transverse momenta of partons inside the proton, as well as the remnant recoil schemes, was carried out. Furthermore, different approaches for the treatment of the strong coupling at low scales were investigated and confronted with available experimental data. This comparison enabled an assessment of the sensitivity of the experimental measurements to the chosen low-scale behaviour of $α_s$.

Event shape variables, constructed from the four-momenta of the final-state objects in an event, are sensitive to the predictions of quantum chromodynamics in multijet production. A measurement of five event shape variables is presented, using proton-proton collision data collected at a centre-of-mass energy of 13 TeV with the CMS detector during 2016$-$2018, corresponding to an integrated luminosity of 138 fb$^{-1}$. The variables are evaluated using the charged particles inside jets. After correcting for detector effects, their distributions are compared with the results from the predictions from a number of models for multijet production. Overall, there is general agreement between several theoretical predictions and the data.

The concept of a common local spin equilibrium for both spin-1/2 and spin-1 particles is incorporated into a thermal model of particle production in heavy-ion collisions at the top RHIC energies. We show that an effective spin polarization tensor leading to a correct description of the longitudinal spin polarization of $Λ$ hyperons simultaneously yields a positive alignment of vector mesons ($φ$ and $K^{*0}$) that grows monotonically with transverse momentum and centrality. Similar trends can be seen in the data, suggesting a possible common mechanism for longitudinal spin polarization and alignment. However, model calculations are insufficient to explain the data in a fully quantitative way. The correlation found between the magnitude of the $Λ$ longitudinal polarization and vector meson alignment suggests further more elaborate investigations of this issue.

Extending the scalar sector of the Standard Model is a well-motivated approach to exploring physics beyond the Standard Model. In this work, we investigate the phenomenology of the Three Higgs Doublet Model at a future muon collider. The scalar spectrum of the 3HDM comprises three CP-even Higgs bosons, two CP-odd Higgs bosons, and a pair of charged Higgs states. Focusing on Higgs pair production via muon-antimuon annihilation, we study the production and decay of neutral scalar states through the process $μ^+μ^- \to φ_i φ_j$, assuming a mass hierarchy in which the SM-like CP-even Higgs is the lightest state. We analyze several benchmark scenarios leading to $b\bar{b}b\bar{b}$ and $b\bar{b}t\bar{t}$ final states, and perform a cut-and-count analysis at a center-of-mass energy of $\sqrt{s}=3$ TeV. Our results demonstrate that a future muon collider provides a sensitive and promising environment to probe extended Higgs sectors, with neutral scalar states in the mass range of $200-400$ GeV being discoverable with $5σ$ significance for integrated luminosities of $\mathcal{O}(1-4 \ \mathrm{ab}^{-1})$.

Light meson spectra are studied in a Dyson-Schwinger/Bethe-Salpeter approach to QCD. By varying the interaction strength of three sets of models for the quark-antiquark interaction, the transition from the chiral symmetric to the chirally broken regime in the vacuum is studied. The simplest type of these models leads to degenerate meson spectra for a large domain of the strength parameter. The more sophisticated and thus more realistic models show significantly smaller parameter domains for which degenerate meson spectra are obtained. The underlying mechanism for obtaining and then lifting degeneracies is traced back to the location of the quark propagators' poles, in particular, whether they are beyond or within the domain of integration in the Bethe-Salpeter equation. In view of this mechanism the potential relation of the obtained degeneracies to the dynamical emergence of symmetries is discussed, adding thereby another point of view on the conjectured chiral spin symmetry of QCD in the temperature domain right above the crossover.

We compute the beta functions for the $O(N)^3$-invariant general sextic tensor model up to cubic order in the coupling constant, and at leading order in the $1/N$ expansion. Our method is a direct, explicit one, in the sense that we identify the appropriate Feynman graphs, we compute their amplitudes which then allows us to obtain the $β$ functions of the model. We perform these computation considering both a long-range and a short-range propagator, within the dimensional regularization framework. We find three fixed points in the short-range case and a line of fixed points, parameterized by the wheel interaction, in the long-range case. This line of fixed points is identical to the one found in the case of the $U(N)^3$-invariant model. Our result proves that the additional $O(N)^3$-invariant interactions do not modify the long-range fixed point structure of the model.

The static tidal Love numbers of four-dimensional black holes vanish identically, unlike their nontrivial dynamical response at finite frequency. Recent work has provided three complementary descriptions of this phenomenon: an emergent $\mathrm{SL}(2,\mathbb{R})$ organization of static near-zone perturbations, a graded logarithmic and multi-zeta structure in Shell Effective Field Theory (Shell EFT), and an on-shell matching framework based on gravitational Raman scattering with renormalization group (RG) running. We show that these features arise from a common near-zone truncation mechanism. For a massless scalar field, horizon regularity selects a unique static solution forming a highest-weight-type representation, truncating the hypergeometric solution to a finite polynomial and eliminating the independent decaying branch at large radius. This excludes a static Wilson coefficient in the effective theory. We demonstrate that the same truncation operates in the static Regge-Wheeler and Zerilli equations for four-dimensional Schwarzschild black holes. Analytic continuation of the horizon-regular solution to small frequency via the Coulomb-hypergeometric or Mano-Suzuki-Takasugi formalisms preserves this truncation as an anchoring condition for the renormalized angular momentum parameter. The resulting low-frequency expansion is controlled by Gamma and hypergeometric functions, generating a graded algebra of logarithms and odd Riemann zeta values. Within this structure no invariant of negative weight exists in the static sector, so the vanishing of the static Love number follows as a structural consequence. This explains the ``zero-sum'' rule of Shell EFT and why the self-induced RG flow in gravitational Raman scattering cannot generate a static invariant.

We present a new global determination of the helicity-dependent parton distribution functions (PDFs) of the proton, based on inclusive deep-inelastic scattering (DIS) and semi-inclusive DIS (SIDIS) data within a consistent next-to-leading order (NLO) QCD framework. In addition to existing measurements, we incorporate simulated pseudodata for the future Electron-Ion Collider (EIC), considering two beam-energy configurations, $E_e \times E_p = 5 \times 41~\mathrm{GeV^2}$ and $18 \times 275~\mathrm{GeV^2}$, corresponding to an extended kinematic reach down to $x \sim 10^{-5}$. We focus on longitudinal double-spin asymmetries $A_1^h$ for charge-separated pion and kaon production in SIDIS off a longitudinally polarized proton target. These projected measurements significantly improve the flavor separation of sea-quark polarized PDFs ($Δ\bar{u}$, $Δ\bar{d}$, $Δs$) and reduce the uncertainties on both quark and gluon helicity distributions, with the largest impact at small $x$. Polarized PDFs are extracted using a neural-network parametrization and a Monte Carlo replica methodology to propagate experimental uncertainties, while theoretical constraints such as positivity are imposed during the fit. We demonstrate that the inclusion of EIC pseudodata leads to a substantially more precise determination of polarized PDFs, with the largest impact in the small-$x$ region. The resulting polarized PDF sets are provided in the {\tt LHAPDF} format.

A levitated ferromagnet provides an exceptionally sensitive probe of ultralight dark matter (ULDM) through measuring weak magnetic-like field signals. We propose a ferromagnet lattice magnetometer that coherently combines multiple levitated ferromagnets to enhance effective sensitivity. By replacing a single ferromagnet with a lattice, we increase the total polarized spin while preserving the intrinsic dynamical response of each constituent ferromagnet. We show that magnetic dipole-dipole interactions within the lattice can be dynamically suppressed through a high-frequency magnetic field, rendering the system effectively noninteracting, at the cost of only a moderate reduction in signal amplitude due to the distinct renormalization of linear and quadratic spin responses. We analyze the noise properties of the lattice and demonstrate that collective readout leads to favorable scaling with the number of ferromagnets. Interpreted in terms of axion-electron, dark photon, and axion-photon couplings, our results yield projected sensitivities that significantly exceed existing single-ferromagnet implementations. In particular, for axion-photon interactions, we find a nontrivial lattice-induced enhancement of the signal itself, leading to sensitivities that surpass existing constraints over a broad mass range.

We investigate low-scale baryogenesis \textit{via} tri-resonant leptogenesis within the scotogenic model with a scalar dark matter embedded in non-holomorphic modular $A_4$ symmetry framework. The model naturally accommodates three nearly degenerate right-handed (RH) neutrinos when they are assigned to the triplet representation of $A_4$. The near degeneracy originates from treating the symmetric contribution to the Majorana mass matrix, arising from the $\mathbf{3}\otimes\mathbf{3}$ decomposition of $A_4$, as a small perturbation to the dominant singlet contribution. Generalized CP (gCP) symmetry is imposed in the model, rendering the complex modulus $τ$ as the sole source of CP violation. In particular, for the inverted hierarchy (IH), the predicted $3σ$ range of $θ_{23}$ lies in the lower octant close to maximal value while CP phase $δ_{\mathrm{CP}}$ and the Majorana phase $α_{21}$ are predicted to lie close to $0^\circ$ or $360^\circ$. Also, in this case, predicted values of $m_{ee}$ and $\sum_i m_i$ can be tested and constrained by future neutrinoless double beta decay $(0νββ)$ experiments, as well as by cosmological observations, particularly DESI+BAO and Planck data. In fact DESI+BAO disallows IH in the model. We further show that successful baryogenesis can be achieved for both normal hierarchy (NH) and inverted hierarchy (IH) of light neutrino masses with RH neutrino masses as low as $537~\mathrm{GeV}$ rendering this scenario experimentally testable. For NH, RH neutrino mass degeneracy of $\mathcal{O}(10^{-7}\!-\!10^{-6})$ is required, while for IH a stronger degeneracy of $\mathcal{O}(10^{-8})$ is needed. Remarkably, in the NH case, successful baryogenesis can occur even in the deep washout regime with decay parameters of $\mathcal{O}(10^{5})$ owing to the tri-resonant enhancement of the CP asymmetry and the inclusion of flavor effects.

We investigate the transverse momentum spectra ($p_T$) and elliptic flow ($v_2$) of $π^{\pm}$, $K^{\pm}$, $p$, and $\bar{p}$ produced in peripheral Au+Au collisions at $\sqrt{s_{\rm NN}} = 7.7$, 11.5, 19.6, 27, and 39 GeV in the Beam Energy Scan (BES) Program at the Relativistic Heavy Ion Collider (RHIC). The analysis is carried out within an expanding elliptic fire-cylinder model that incorporates longitudinal expansion and anisotropic transverse flow. Particle production at kinetic freeze-out is obtained using a local equilibrium distribution function with a blast-wave-like fluid velocity profile derived from the expansion dynamics of the elliptic fire-cylinder. The model parameters governing the collective expansion are first constrained by fitting the midrapidity $p_T$ spectra of $π^{\pm}$ and are then applied, without further adjustment, to $K^{\pm}$, $p$, and $\bar{p}$. The model provides a consistent description of the $p_T$ spectra and reproduces the qualitative behavior of the elliptic flow for all considered particle species.

We analyze the possibility of detecting the existence of mirror matter, a possible component of dark matter, through neutron interferometry. We develop an interferometer using bandpass multilayers in reflection and transmission geometry and discuss its advantages and limitations. We demonstrate that our setup can probe a considerable range of neutron-mirror neutron mixing parameters allowing us to show the existence of mirror matter using present day neutron sources based on fission or spallation processes.

The kaon, the lightest hadron containing a strangeness quark, is very peculiar. It is a Nambu-Goldstone boson, but significantly heavier than the pion. As a result, its interaction with a matter particle, such as the nucleon or a heavy-light meson, such as the $D$ meson, is completely determined by chiral dynamics and much stronger than its pion cousin. The strong attractive interaction has brought us many surprises and is manifested in the peculiar nature of many particles, such as the mysterious $Λ(1405)$ and $D_{s0}^*(2317)$. These two particles can be understood as $\bar{K}N$ and $DK$ hadronic molecules, respectively. They also imply the existence of three-body hadronic molecules that await future discovery. In this talk, I review some recent developments in our understanding of hadronic interactions involving the kaon.

The first measurement of the polarization of charm baryons by the LHCb experiment recorded in fixed-target mode is presented. The polarization of $Λ_c$ baryons is studied in collisions of protons, at an energy of 2.51 TeV, incident on a gaseous target of neon, at a nucleon-nucleon center-of-mass energy of $68.6$ GeV. The world's first measurement of separate-charge polarizations for $Λ^+_c$ and $\overlineΛ{}^-_c$ baryons is performed, determining $$ P_{Λ^+_c} = ( 24 \pm 9 \pm 2 \, )\% , $$ $$ P_{\overlineΛ{}^-_c} = (-8 \pm 12 \pm 3 \, ) \% , $$ where the first uncertainty is statistical and the second systematic. The polarization is also measured in intervals of baryon transverse momentum and the Feynman-$x$ variable.

This paper presents the first evidence of the simultaneous production of two $Z$ bosons and one photon with the ATLAS detector at the Large Hadron Collider. The measurement is performed using the full Run-2 dataset, recorded from 2015 to 2018, of proton-proton collisions at a center-of-mass energy of $13$ TeV, corresponding to an integrated luminosity of $140$ fb$^{-1}$. The fully leptonic final state with four leptons and one photon is analyzed, $pp\rightarrow ZZγ\rightarrow\ell^+\ell^-\ell'^+\ell'^-γ$ with $\ell, \ell' = e$ or $μ$. This final state is measured in a fiducial region where photon final-state radiation is minimized, and the photon has a transverse momentum of $p_{\mathrm{T}}^γ> 20$ GeV. Eight events are selected, with a background estimate of $0.92\pm0.15$. This results in an observed (expected) significance of the $ZZγ$ final state of $4.4σ$ ($4.4σ$). The measured cross-section for $pp\rightarrow ZZγ\rightarrow\ell^+\ell^-\ell'^+\ell'^-γ$ in the fiducial region is $σ_{ZZγ} = 0.144 _{-0.051}^{+0.064} \text{ (stat.)} _{-0.005}^{+0.007} \text{ (syst.) fb}$, in agreement with the predicted Standard Model one, $σ_{\textrm{fid.}}^{\textrm{SM}} = 0.143 _{-0.004}^{+0.007}$ fb.

The atomic properties of heavy, moderately-charged ions are important for a wide variety of applications, including precision tests of fundamental physics and for the study and development of atomic and nuclear clocks. In these systems it is known that relativistic effects, such as the Breit interaction and radiative quantum electrodynamics corrections, are important for an accurate understanding of atomic properties. It is also known that inclusion of correlations alongside the Breit effect is crucial. In this work we include the Breit interaction into all-orders calculations of energy levels and fine structure intervals of ions in the Cs and Fr isoelectronic sequences. This requires modifying the electron Green's function to account for Breit within the all-orders correlation potential method, which sums dominating series of perturbation diagrams exactly using a Feynman diagram technique. We find that Breit corrections to the energies of moderately ionized ions along these sequences are very large, particularly for the f states. We also observe a significant deviation from experiment for these levels. Incorporating Breit into the all-orders correlation potential provides a significant additional contribution beyond including Breit at the second-order level alone. While this does not resolve the disagreement in the energy levels, it does substantially improve the fine-structure intervals beyond what is achieved by including Breit only at second order. Furthermore, we include the frequency-dependent Breit interaction into the Dirac-Fock procedure, and find that this does not significantly modify the energy levels at this order of approximation.

Direction-sensitive direct dark matter search experiments have been conducted using gaseous detectors. In spite of the long history of the study on the energy deposition of charged particles in materials, a full agreement between the measured results and theoretical predictions, especially in a low energy scale, are yet to be achieved. It is thus important to experimentally measure the ionization yields of recoil nuclei for the experiments with gaseous detectors using an ionization charge readout scheme. This study measured the ionization yield using a low-energy ion beam facility at Kanagawa University. The ionization yields for fluorine ions with an energy range of 5 $\sim$ 50 keV were measured using a dedicated proportional wire chamber filled with $\mathrm{CF}_{4}$ gas at 0.06 atm. The low-energy ion injection scheme into a gaseous detector was established and the ionization yield for fluorine ions was obtained to be 0.45 at 30 keV with a moderate dependence on the ion energy.

To decode the near-threshold dynamics of the $X_{0,\,1}(4140)$ and $X_{1}(4685)$ states, we investigate the OZI-suppressed $\{D_{s}\bar{D}_{s},\, J/ψφ,\, D_{s}^{\ast}\bar{D}_{s}^{\ast}\}$ coupled-channel scattering in $B\to D_{s}\bar{D}_{s} K$ decays using the effective range expansion. We demonstrate that the $X_{0}(4140)$, associated with a dip in the lineshape, corresponds to a dynamically generated pole near the $J/ψφ$ threshold. The single-channel $J/ψφ$ scattering length is extracted to be $1.11\pm 0.65\,\rm{fm}$, yielding an effective scattering length of $0.12^{+0.20}_{-0.10}+i0.78^{+0.20}_{-0.40} \, \rm{fm}$ when coupled channels are included. By treating the spin-spin interaction as a subleading effect, we predict a $J^{PC}=1^{++}$ virtual state near the $J/ψφ$ threshold, which naturally resolves the empirical ambiguities surrounding the $X_{1}(4140)$ width. Extending this framework via heavy quark spin symmetry, we further interpret the $X_{1}(4685)$ as a $ψ(2S)φ$ hadronic molecule. Ultimately, these findings highlight how the $X(4140)$ family and $X_{1}(4685)$ serve as unique theoretical windows into the Fierz rearrangement and OZI suppression mechanisms in low-energy strong interactions.

We find a large class of holographic solutions describing D4-branes wrapped on 4-manifolds $\mathcal{M}_4$ with constant curvature leading to gravity duals of supersymmetric quantum mechanics in the IR via twisted compactifications. The manifolds $\mathcal{M}_4$ considered here are four-dimensional spheres and hyperbolic spaces, products of two Riemann surfaces, and Kahler four-cycles. The solutions are obtained from the maximal gauged supergravity in six dimensions with $CSO(p,q,5-p-q)$ and $CSO(p,q,4-p-q)\ltimes \mathbb{R}^4$ gauge groups. These gauged supergravities can be embedded in type IIA theory via consistent truncations on $H^{p,q}\times \mathbb{R}^{5-p-q}$ and $H^{p,q}\times\mathbb{R}^{4-p-q}\times S^1$, respectively. The solutions take the form of $t\times \mathcal{M}_4$-sliced domain walls interpolating between locally flat domain walls and singular geometries in the IR. Upon uplifted to type IIA theory, many solutions admit physical IR singularities and could holographically describe supersymmetric quantum mechanics arising from twisted compactifications of D4-branes on $\mathcal{M}_4$.

We extend the Abelian-Higgs model of superconductivity to incorporate higher-spin particles. Microscopically, these higher-spin states can be modeled as multi-electron clusters, such as spin-1 Copper pairs or quartets, existing alongside the standard Cooper pairs predicted by BCS theory. To account for these composites, we introduce vector and higher-rank tensor non-gauge fields into the Lagrangian, which serve as sources for higher-rank tensor gauge fields. In this work, we extend the particle spectrum by one rank (including the necessary auxiliary fields) and examine the resulting modifications to the fundamental phenomenological parameters of superconductivity, specifically the penetration depth and the correlation length.

We attempt to apply the Regge trajectory approach to the doubly heavy triquarks $((Qq)\bar{Q}^{\prime})$ $(Q,\,Q'=b,\,c; q=u,\,d,\,s)$. We propose the Regge trajectory relations for the doubly heavy triquarks, and then employ them to crudely estimate the spectra of the triquarks $((cu)\bar{c})$, $((cu)\bar{b})$, $((cs)\bar{c})$, $((cs)\bar{b})$, $((bu)\bar{c})$, $((bu)\bar{b})$, $((bs)\bar{c})$, and $((bs)\bar{b})$. The $λ$-trajectories and the $ρ$-trajectories are investigated. The triquark Regge trajectory becomes a new and very simple approach for estimating the spectra of triquarks. It also provides a simple method to investigate the $ρ$-mode and $σ$-mode excitations of pentaquarks and hexaquarks in the triquark picutre. Moreover, the spin-averaged masses of the ground states of pentaquarks $(\bar{c}(cu))(cu)$, $(\bar{b}(bu))(bu)$ and $(\bar{c}(cu))(bu)$ are estimated, which are consistent with other theoretical predictions.

We investigate the concept of mass in the Signum-Gordon (SG) model, a nonlinear field theory with a non-analytic potential where the perturbative mass is undefined. Using two complementary numerical methods, we map the field's dispersion relation (amplitude vs. wavenumber and frequency). We find the field's evolution depends critically on the product of its amplitude and squared wavenumber, revealing a massless regime at large values and an ultra-massive regime with dominant nonlinear Fourier mode mixing near unity. By comparing the resulting dispersion map to the massive Klein-Gordon equation, we introduce a spectral mass. We demonstrate that a specific input amplitude value induces a spectral mass of unity, effectively characterizing the massive-like behavior arising from the initial wave configuration.

We compute relativistic corrections to the gluon fragmentation functions to ${}^3P_J^{[1,8]}$ Fock states of heavy quarkonium within non-relativistic QCD factorization framework. We find that, at $\mathcal{O}(v^2)$ sub-leading order, the $S$-$D$ mixing effect must be taken into account to absorb the infrared divergence of spin-triplet $P$-wave production within full QCD into the NRQCD long-distance matrix elements. Unlike the $S$-wave case, we find that the short-distance coefficients of the fragmentation functions at leading and sub-leading order are no longer proportional to each other. However, upon convolution with the gluon production cross section, their ratios are almost constant across the whole $p_T$ region. We find the relativistic corrections to be negative and substantial, which makes them a non-negligible ingredient in the study of $J/ψ$ production at the LHC.

Batyrev's construction provides a map from fine, regular, star triangulations (FRSTs) of 4D reflexive polytopes to smooth Calabi-Yau threefolds (CYs). We prove that there are at most $10^{296}$ diffeomorphism classes of CYs produced in this manner, improving [1]'s upper bound of $10^{428}$. To show this, we make use of the fact that any two FRSTs with the same 2-face restrictions give rise to diffeomorphic CYs and bound the number of such '2-face equivalence classes' for all polytopes with Hodge number $h^{1,1} \geq 300$. We also put a lower bound of $10^{276}$ on the number of 2-face equivalence classes, but emphasize that this is not a lower bound on the number of diffeomorphism classes of CYs, as distinct 2-face equivalence classes may give rise to diffeomorphic threefolds.

We test the Etherington cosmic distance-duality relation (CDDR), by comparing Type Ia supernova (SNIa) luminosity-distance information from the Pantheon+ compilation with an angular-diameter-distance reconstructed from localized Fast Radio Bursts (FRBs). The core of our methodology is a data-driven reconstruction from FRBs using artificial neural networks (ANNs): we infer a smooth mean extragalactic dispersion-measure relation and use its redshift derivative to recover $H(z)$ and hence $D_\mathrm{A}^{\rm FRB}(z)$ without assuming a parametric form for the expansion history. Possible deviations from CDDR are parameterized through three one-parameter models of $η(z)\equiv D_\mathrm{L}/[(1+z)^2D_\mathrm{A}]$. We implement two complementary likelihoods: (i) a direct approach using individual SNIa with the full Pantheon+ covariance, and (ii) a machine-learning approach in which we reconstruct the SN Hubble diagram on the FRB redshift grid, propagating SN and FRB uncertainties into non-diagonal covariance matrices via Monte Carlo and bootstrap realizations. Within the FRB reconstruction, we anchor the mean extragalactic dispersion measure at $z=0$, which yields a data-driven constraint on the average host/near-source contribution $\mathrm{DM}_{\rm host}=128.8\pm 34.1\,\mathrm{pc\,cm^{-3}}$ ($3σ$ of statistical confidence). We find that both likelihood implementations give consistent posteriors and no statistically significant evidence for departures from CDDR at the current precision.

We discuss new-physics searches and symmetry tests with dipole moments, emphasizing the role of polarization observables. As a primary benchmark, we consider polarized photon fusion in the $e^+ e^-$ environment of the Super Tau-Charm Facility (STCF) and study $γγ\to τ^+ τ^-$ in the nearly back-to-back region, where a transverse-momentum-dependent (TMD) description provides a convenient framework for organizing polarization effects. We show that linearly polarized photons induce characteristic azimuthal asymmetries in the $τ^+ τ^-$ kinematics, enabling polarization-based observables that enhance sensitivity to the $τ$ electromagnetic dipole form factors. Moreover, $CP$-even and $CP$-odd dipole interactions can be disentangled through distinct angular structures, offering a systematic path to probe $τ$ dipole moments with improved precision at future lepton colliders. As an illustration, we obtain an improved $2σ$ reach on the anomalous magnetic dipole moment, $-4.6 \times 10^{-3} < \mathrm{Re}(a_τ) < 7.0 \times 10^{-3}$, reaching a precision level close to the Standard Model expectation. To place these prospects in a broader context, we briefly summarize the experimental status of dipole-moment measurements across different fermionic systems and highlight their complementarity in constraining new physics. We illustrate this interplay with supersymmetric scenarios featuring $R$-parity violation, in which loop-induced dipole moments provide correlated probes of $CP$-conserving and $CP$-violating interactions. Taken together, polarized photon fusion and precision dipole measurements constitute a coherent program for testing fundamental symmetries and exploring physics beyond the Standard Model.

We consider a dark sector consisting of fermionic dark matter (DM) charged under a broken dark $U(1)_D$ gauge symmetry, interacting with the Standard Model through kinetic mixing. In such models, the DM annihilation cross section is typically suppressed by the small kinetic mixing and or a heavy mediator, often leading to an overabundant relic density. We show that the observed DM abundance can be achieved if the dark Higgs undergoes a strong first order phase transition after DM freeze-out. In this scenario, the relic abundance is set by thermal freeze-out in the symmetric phase and subsequently reduced by entropy injection from the phase transition, rather than by annihilation in the broken phase. We find that to reproduce the observed relic abundance, the required phase transition is generically supercooled. The resulting stochastic gravitational wave signal lies within the sensitivity of future experiments, providing a complementary probe of this framework. Moreover, a strongly supercooled phase transition can potentially account for the NANOGrav signal for DM masses below $O(10)$ GeV.

Ultralight dark matter is expected to induce oscillations of nuclear parameters. These oscillations are characterized by extremely weak couplings or high suppression scales, with the Planck scale - the characteristic scale of quantum gravity - serving as a natural benchmark. Probing this phenomenon requires systems with exceptional sensitivity to shifts in nuclear energies. The uniquely low-energy nuclear isomeric transition in ${}^{229}$Th provides such sensitivity: it directly probes the nuclear interaction and, owing to a near cancellation between electromagnetic and nuclear contributions, its response to changes in nuclear structure is greatly amplified. We devise and perform a new type of ultrasensitive search for dark matter which uses the precision nuclear spectroscopy at JILA to set the strongest bounds in the mass range $10^{-21}\,{\rm eV} \lesssim m_{\rm DM} \lesssim 10^{-19}\,{\rm eV}$. Our results probe effective interaction scales exceeding $10^6$ times the Planck scale (the Mega-Planck scale) and establish the ${}^{229}$Th system as the leading probe of dark matter couplings to the nuclear sector.

High-energy cosmic-ray muons reaching deep underground laboratories can cause events in detectors that mimic signals expected from dark matter particles, neutrinos, or rare decays. Knowledge of the muon flux and energy spectrum is important for evaluating the background rate caused by muons and their secondaries. In this paper, we report the measurement of the cosmic-ray muon flux in the Davis Campus of the Sanford Underground Research Facility with the LUX-ZEPLIN detector. Using 366.4~days of exposure, the muon rate through the detector was measured as $10.94\pm0.17_\textrm{stat.}~\textrm{day}^{-1}$ with energy thresholds of 20~MeV in the inner xenon detector and 8 MeV in the outer liquid scintillator detector. This rate corresponds to a muon flux of $(5.09\pm0.08_\textrm{stat.}\pm0.10_\textrm{sys.})\times10^{-9}~\textrm{cm}^{-2}\textrm{s}^{-1}$ in the Davis Cavern.

We present an analysis strategy for probing physics beyond the Standard Model via modifications to the parton distribution functions (PDFs) in a muon beam, which measurably alter the kinematics of all hard processes at a future muon collider. High-energy muon colliders represent an opportunity to probe new physics using precision measurements and novel search strategies. At sufficiently high energies, light particles act as ``constituents'' of the muon described by PDFs. As a concrete case study, we apply this framework to an $L_μ - L_τ$ gauge boson and demonstrate that, for masses in the range of approximately 50--100 GeV, this indirect PDF-based approach outperforms traditional searches relying on direct gauge boson production. These results highlight muon PDF probes as a powerful and promising avenue for beyond the Standard Model physics searches at a future muon collider.

We prove that the one-loop contribution from tensor modes to the thermodynamic entropy of near-extremal black holes is universal. Our proof applies to asymptotically flat, Anti-de-Sitter and de-Sitter black holes; it also covers spherical, axial and planar symmetries. We consider black hole configurations with and without matter sectors and explicitly discuss Abelian gauge fields and neutral scalar fields with arbitrary potential. We demonstrate that under certain conditions, the thermodynamics of near-extremal black holes contains a one-loop contribution from the tensor modes that equals $\frac{3}{2}\log (T_{\rm Hawking}/T_q)$. The proof of this theorem also shows explicitly how the Schwarzian modes appear universally in near-extremal geometries in dimensions four, five and six. We apply this theorem to Kerr-de-Sitter black holes as an explicit example.

Deep, wide-area photometric surveys have uncovered a population of compact ($r_{1/2} \approx$ 1-15 pc), extremely-low-mass ($M_* \approx$ 20-4000 $M_{\odot}$) stellar systems in the Milky Way halo that are smaller in size than known ultra-faint dwarf galaxies (UFDs) and substantially fainter than most classical globular clusters (GCs). Very little is known about the nature and origins of this population of "Ultra-Faint Compact Satellites" (UFCSs) owing to a dearth of spectroscopic measurements. Here, we present the first spectroscopic census of these compact systems based on Magellan/IMACS and Keck/DEIMOS observations of 19 individual UFCSs, representing $\sim$2/3 of the known population. We securely measure mean radial velocities for all 19 systems, velocity dispersions for 15 (predominantly upper limits), metallicities for 17, metallicity dispersions for 8, and $\textit{Gaia}$-based mean proper motions for 18. This large new spectroscopic sample provides the first insights into population-level trends for these extreme satellites. We demonstrate that: (1) the UFCSs are kinematically colder, on average, than the UFDs, disfavoring very dense dark matter halos in most cases, (2) the UFCS population is chemically diverse, spanning a factor of $\sim$300 in mean iron abundance ($\rm -3.3 \lesssim [Fe/H] \lesssim -0.8$), with multiple systems falling beneath the "metallicity floor" proposed for GCs, and (3) while some higher-metallicity and/or younger UFCSs are clearly star clusters, the dynamical and/or chemical evidence allows the possibility that up to $\sim$50% of the UFCSs in our sample (9 of 19) may represent the smallest and least-massive galaxies yet discovered.

We investigate the azimuthal dependence of H I and O VI-bearing gas in the circumgalactic medium (CGM) of 113 isolated galaxies in the redshift range 0.12 < z < 0.75, including 91 new measurements from the MUSE Quasar-fields Blind Emitters Survey (MUSEQuBES). The H I covering fraction (k_HI) within the virial radius (Rvir) of low-mass (7 < log10(M*/Msun)< 9) galaxies, for a threshold column density of log10(N(HI)/cm^-2) = 14.5, exhibits an enhancement along both the disk plane (azimuthal angle phi < 20 degree) and in the polar direction (phi > 70 degree). In contrast, such a bimodal distribution is not observed for higher mass galaxies (9 < log10(M*/Msun) < 11.3). Similarly, the O VI covering fraction (k_OVI), for a threshold of log10(N(OVI)/cm^-2) = 14.0, shows a tentative enhancement along both the projected major and minor axes for low-mass galaxies. In contrast, O VI-bearing gas around higher- mass galaxies appears more uniformly distributed, with no significant azimuthal dependence. Finally, using the halo circular-velocity-normalized pixel-velocity two-point correlation function (TPCF), we find that O VI absorbers are kinematically narrower along the disk plane compared to the polar directions of the host galaxies with similar stellar mass distributions. The observed isotropic distribution of O VI in high-mass halos suggests that its spatial distribution is governed by global halo properties; however, the O VI kinematics retain memory of the site of origin.

The Seyfert galaxy Markarian 590 offers a rare glimpse into the dynamic life cycle of black hole accretion, captured across multiple wavelengths from the years 1975 to 2025. Using the decade-long multi-band observations from the Swift observatory, we capture a clear spectral state transition analogous to those seen in X-ray binaries but seldom observed in a single AGN. We track a complete AGN state transition in real time, as the source evolves from a faint, hard X-ray state to a bright, UV- and soft X-ray dominated phase. The X-ray loudness parameter $α_{ox}$ follows a pronounced 'V'-shaped dependence on Eddington ratio $λ_{Edd}$, with a break at $λ_{Edd}$ = 0.021 +/- 0.008, coinciding with thresholds identified in population studies of changing-look quasars and X-ray binaries. Across this transition, Mkn 590 evolves through distinct accretion regimes in the Hardness Intensity Diagrams (HID): faint, flaring, transitional, and bright, on a timescale of $\sim$10 years, which is well below classical viscous timescales for a geometrically thin disk but in agreement with the propagation of thermal wavefronts in the inner disk. When placed on the Fundamental Plane of black hole activity, the source broadly follows the expected radio/X-ray mass scaling, though with a 70% flatter slope, pointing towards a persistent coronal-jet coupling even in radiatively efficient states. Together, our results establish Mkn 590 as a rare, time-resolved case of AGN state transitions and offer compelling evidence for scale-invariant accretion physics across the black hole mass spectrum.

The formation pathways of sulfur-bearing species in the interstellar medium are crucial to understand astrochemical processes in cold molecular clouds and to gain new insights about the sulfur budget in these regions. We aim to explore the recently detected, thioethanal (CH$_{3}$CHS) formation mechanisms from thioethanol (CH$_{3}$CH$_{2}$SH) as a precursor in addition to secondary sulfur products. The electronic structure methods and density functional theory for both gas-phase and ice-grain surface environments is employed. To mimic interstellar ice-mantles, we use medium (W6) and large amorphized (W22) water clusters as implemented in Binding Energy Evaluation protocol. A barrierless formation mechanism for CH$_{3}$CHS under low-temperature interstellar conditions is identified, in the gas phase. Surface environments modulate activation barriers in a site-specific manner, elucidated through both Langmuir-Hinshelwood and Eley-Rideal initiated surface reaction pathways. Compared to oxygen analogs, sulfur chemistry enables alternate pathways due to weaker S-H bonding, with a competing route forming ethane-1,1-di-thiol (CH$_{3}$CH(SH)SH), on the ice-grain surface, potentially reducing CH$_{3}$CHS yields. The first accurate binding energy for thioethanol on water ice is also reported, confirming its greater volatility than ethanol. The proposed mechanism offers a tentative hypothesis for the apparent mutual exclusive detections of the CH$_{3}$CH$_{2}$SH and CH$_{3}$CHS in TMC-1, Orion, and Sgr B2(N), that further requires validation through quantitative astrochemical modeling and also to distinguish this chemical differentiation from observational sensitivity limitations. These qualitative findings highlight the multifaceted chemical behavior of sulfur-bearing organics in the interstellar medium and support CH$_{3}$CH(SH)SH as promising astro-chemical targets.

Through 2D3V PIC simulations of freely decaying sub-ion turbulence, intermittent localized regions with $\mathbf{E} \cdot \mathbf{B} \neq 0$ are found to be statistically associated with reductions in the magnitude of magnetic helicity while evolving in the early electron-scale interaction phase. Motivated by this behavior, we propose a source-compensated, history-dependent helicity density that satisfies an exact local balance identity by construction, enabling Saffman-type two-point correlation integrals which, under standard flux-decorrelation assumptions, can exhibit intermediate-scale plateaus that are roughly time-independent. In our simulations we demonstrate such plateaus to remain approximately invariant even as the usual Saffman helicity integral plateau value $I_H$ evolves during the early kinetic stage. Under approximate single-scale self-similarity, the plateau behavior of the magnetic integral is consistent with the 2D decay constraint $BL \sim \text{const}$. For initially net-helical configurations, we observe rapid development of mixed-signed magnetic helicity patches and a decrease of the global fractional helicity, such that the decay over the kinetic interval is again most consistent with the cancellation-dominated scaling constraint.

UV slopes ($β$) are a powerful diagnostics for galaxies at the Epoch of Reionization, tracing star formation, ISM ionization, and the escape fraction $f_{esc}$ of ionizing photons. Studies at low and intermediate z find a gradual $β$ reddening with time and steeper slopes for fainter galaxies, however recent JWST studies reveal a flattening of this trend at $z>7$. We want measure $β$ for galaxies at $z>7.5$ using the strong lensing around massive galaxy clusters to observe high-redshift and faint galaxies. The low-brightness regime is of particular interest for reionization, as most of the recent models of this process posit that numerous faint galaxies are the prime drivers of reionization. We use NIRCam and NIRSpec data from CANUCS, Technicolor, JUMPS, Silver Bullet, UNCOVER and MEGASCIENCE across 7 strong lensing fields. We find galaxies down to $M_{UV}\sim-16$ and 7.5<z<12.5. We measure \b{eta} with a forward-modelling procedure and estimate $f_{esc}$ for a subsample with emission line data using a relation, calibrated from a low-z sample, with UV slope, galaxy size and H$β$ equivalent width. We find 378 galaxies (45 with spectrum), yielding average values $β=-2.3\pm0.4$, $z=8.5\pm1.0$, and $M_{UV}=-18\pm1$. We find no significant $β$ evolution across our redshift range, suggesting a flattening of the $β-z$ trend above $z\sim7.5$. We find a weak negative trend between $β$ and $M_{UV}$. For 14 galaxies we estimate an average $f_{esc}=0.26\pm0.22$. The flat trend of $β$ at $z>7.5$ suggests similar properties between $300$ and $600 Myr$ after the Big Bang. The weak trend between $β$ and $M_{UV}$ suggests an analogous composition for low- and high-mass galaxies' ISM, likely due to a lack of time for dust buildup. While average $f_{esc}$ is higher than necessary to ionize the IGM by z~6, the model extrapolated at low-z may overestimate its value.

Lithium plays a unique role in astrophysics, as it is a powerful diagnostic for the physics and evolution of low-mass stars, Galactic archaeology, and cosmology. We review the Li observations in stars at different phases of their evolution, the strengths and the limitations of the current theoretical stellar models to explain the Li abundance data, our understanding of the Li sources and of the evolution of Li through- out the Galactic history. Key takeaways from the current state of the research in the field are: 1) Stellar evolution models accounting for fundamental transport processes of chemical species and angular momentum hold the promise of providing a common stellar Li depletion explanation to the Li abundance patterns observed in all Galactic stellar populations, including the dip and the plateau(s). 2) Novae are most probably the main source of Li in the Galaxy, on observational (but not yet theoretically established) grounds. 3) Radial migration of stars in the Galactic disk holds the key to understand many aspects of the Li evolution in the Milky Way.

The advent of integral field data has revolutionised the study of galaxy evolution. A key component of this is dynamical modelling methods which have allowed for crucial insights to be made from kinematic data. Despite this importance, most dynamical models make a number of key assumptions which do not hold for real galaxies. These include assumptions about the geometry (axisymmetry or triaxiality), the shape of the velocity ellipsoid, and the shape of the underlying stellar distribution. At the same time, machine learning methods are becoming increasingly powerful, with many applications appearing in astronomy. As a first step towards building new dynamical modelling methods with machine learning, it is important to understand the types of machine learning architectures that are best fit for dynamical modelling. To investigate this, we construct a training set of dynamical models of early-type galaxies using Jeans Anisotropic Modelling (JAM). We then train a neural network on this data using the parameters of JAM and mock photometry as the input. We are able to accurately model JAM galaxies with relatively simple machine learning architectures, leading to a significant speed increase over traditional JAM modelling.

We investigate the detection and nature of 5.6~μm MIRI-red sources in the MIRI Deep Imaging Survey (MIDIS), covering 2.4~arcmin$^2$ in the Hubble Ultra Deep Field. MIDIS is the deepest JWST/MIRI survey to date, probing faint limits and enabling studies of rare high-redshift galaxy populations. We define MIRI-red sources as those detected at 5$σ$ significance in MIRI/F560W with red colors: $m_{\rm F444W} - m_{\rm F560W} \ge 0.5$. Using an empirical methodology, we estimate the purity and completeness of MIRI detections and find that a 5-sigma detection at 28.75 mag has a purity of 92\% and completeness of 54\%. We identify seven MIRI-red galaxy candidates, including an F115W dropout consistent with a high-redshift galaxy candidate. We explore possible physical origins for the MIRI-red population, including active galactic nuclei, dust-obscured galaxies, extreme emission-line galaxies, evolved stellar populations, and Little Red Dots (LRDs). Given the proximity of the F444W and F560W filters and the depth of MIDIS, MIRI-red galaxies are consistent with emission-line galaxies with $EW_0(Hα) \ge 750$ Å or $EW_0(Hβ+ [OIII]) \ge 600$ Å, or high-redshift Balmer breaks of at least 1.6. We also discuss an extreme MIRI-red galaxy undetected in F444W, a potential MIRI-only source, for which we derive $EW_0(Hα) \sim 6000$ Å and $EW_0(Hβ+ [OIII]) \sim 4000$ Å, or high-$z$ LRD analogs with Balmer breaks of 6.3. Finally, we find fewer MIRI-red detections than expected from extrapolations of the H$α$ or H$β$+[OIII] line luminosity functions, consistent with previous deep searches, while the absence of $z>10$ LRD candidates agrees with theoretical expectations for the MIDIS volume.

The 21 cm bispectrum (BS) offers a powerful probe of the Epoch of Reionization (EoR), but its observational access is severely hindered by dominant astrophysical foregrounds. Considering Murchison Widefield Array (MWA) observations at $154.2~\mathrm{MHz}$ ($z=8.2$), we mitigate the foregrounds with Smooth Component Filtering (SCF) and estimate the 21 cm BS. We validate the pipeline using a simulated 21 cm signal and show that the input BS is recovered for modes $k_{\parallel} \ge [k_\parallel]_f=0.135~{\rm Mpc}^{-1}$. Applied to actual data, the SCF produces substantial foreground suppression, reducing the amplitude of the cylindrical BS $B(k_{1\perp},k_{2\perp},k_{3\perp},k_{1\parallel},k_{2\parallel})$ by $3-4$ orders of magnitude. The artifacts due to the missing frequency channels in the data are also suppressed. The resulting EoR window is significantly cleaner at small $k_{\perp}$. We adopt the region $(k_{1 \perp},k_{2 \perp},k_{3 \perp})\leq 0.026~{\rm Mpc}^{-1}$ and $(k_{1\parallel},k_{2\parallel},k_{3\parallel})>0.135~{\rm Mpc}^{-1}$ to evaluate the 3D spherical BS and constrain the EoR signal. By combining estimates over all triangle shapes, we place the lower and upper limits on the mean cube brightness temperature fluctuations $Δ^3$. The estimates are consistent with statistical fluctuations from system noise. The most stringent lower limit $Δ^3_{\rm LL}=-(1.25\times 10^4)^3~{\rm mK}^3$ and upper limit $Δ^3_{\rm UL}=(1.22\times 10^4)^3~{\rm mK}^3$ are obtained at $k_1=0.281~{\rm Mpc}^{-1}$. Additional observing time will reduce the noise level and enable substantially tighter constraints on the EoR signal.

We investigate the multiband photometric properties of seven supernovae (SNe) showing double-peaked light-curve evolution and prominent shock-cooling emission, observed by the Wide Field Survey Telescope (WFST) during its first year of operation. By jointly employing an analytic early shock-cooling model and the Arnett radioactive-diffusion model, we fit the bolometric light curves and infer ejecta masses in the range $1.1$-$2.6 M_\odot$, consistent with a transitional population between ultra-stripped supernovae (USSNe) and normal stripped-envelope supernovae (SESNe). The envelope masses are estimated to be $M_{\rm env}=0.1$-$0.4 M_\odot$, while the progenitors are constrained to be yellow or blue supergiants (YSGs/BSGs) with radii of $R=120$-$300 R_\odot$. Using empirical relations, we estimate progenitor luminosities of $L=10^{4.6}$-$10^{4.9} L_\odot$, corresponding to zero-age main-sequence (ZAMS) masses of $8$-$20 M_\odot$. Theoretical models suggest that such progenitors are more naturally produced through binary evolution channels, as single-star evolutionary pathways are unable to yield ejecta masses this low.

We present comprehensive photometric and spectroscopic observations of a transitional type Ia SN 2024aedt, discovered by the 2.5-meter Wide Field Survey Telescope (WFST) within one day of the explosion. Its light curve is characterized by a peak absolute magnitude of $M_B = -18.49 \pm 0.03$ mag and a decline rate of $Δm_{15}(B) = 1.53 \pm 0.36$ mag, placing the object on the $Δm_{15}(B)$--$M_B$ diagram in the transition region between normal and subluminous SNe Ia. Furthermore, the early-color evolution and host galaxy environment of SN 2024aedt underscore its transitional nature, sharing properties with both normal and 91bg-like SNe Ia. Light-curve modeling with MOSFiT yields a synthesized $^{56}\mathrm{Ni}$ mass of $0.414 \pm 0.042\,M_{\odot}$ and a total ejecta mass of $0.548 \pm 0.108\,M_{\odot}$. A comparison with theoretical models suggests that the evolutionary trend can be broadly explained by both delayed-detonation (DDT) and double-detonation (DDet) scenarios while possible early-excess emissions predicted by DDet cannot be identified given the limited detections soon after the SN explosion. Although the overall spectral evolution of SN 2024aedt is similar to that of other transitional SNe Ia, the spectroscopic comparison reveals diversity in the early-phase blue-end features, which becomes more homogeneous at later phases. The result indicates the importance of early-time observations in understanding the origin of SN Ia diversity.

In this paper we present 16 early-phase type Ia supernovae (SNe Ia) discovered during the pilot survey of the 2.5-meter Wide Field Survey Telescope (WFST-PS) from March 4 to July 10, 2024, including three SNe Ia with early-excess emission features (EExSNe Ia). The discovery magnitude of the 16 WFST-PS early-phase SNe is at least 3 mag fainter than their peak brightness. A large scatter of color indices is found in approximately the first 10 days of supernova explosions, indicating diverse photometric behaviors in the early phase. Three EExSNe Ia show relatively brighter peak luminosities and longer rise time compared to those of non-EExSNe Ia. The results indicate that current theoretical models require further refinement to fully capture the early photometric evolution of SNe Ia. Based on the initial high-cadence ugr-band data from the WFST-PS survey, we emphasize that early near-ultraviolet (NUV) observations are indispensable for placing tight constraints on the explosion mechanisms and progenitor systems of SNe Ia.

Shock chemistry is an excellent tool to shed light on the formation and destruction mechanisms of complex organic molecules (COMs). The L1157-mm outflow is the only low-mass protostellar outflow that has extensively been studied in this regard. Using the data taken as part of the PRODIGE (PROtostars & DIsks: Global Evolution) large program, we aim to map COM emission and derive the molecular composition of the protostellar outflow driven by the Class 0 protostar NGC1333 IRAS4B1 to introduce it as a new laboratory to study the impact of shocks on COM chemistry. In addition to typical outflow tracers such as SiO and CO, outflow emission is seen from H2CO, HNCO, and HC3N, as well as from the COMs CH3OH, CH3CN, and CH3CHO, and even from deuterated species such as DCN, D2CO, and CH2DOH. Maps of integrated intensity ratios between CH3OH and DCN, D2CO, and CH3CHO reveal gradients with distance from the protostar. Intensity ratio maps of HC3N and CH3CN with respect to CH3OH peak in the southern lobe where temperatures are highest. Rotational temperatures derived towards two positions, one in each lobe, are found in the range ~50-100 K. Abundances with respect to CH3OH are higher by factors of a few than for the L1157-B1. In conclusion, for the first time, we securely detected the COMs CH3CN, CH3CHO, and CH2DOH in the IRAS 4B1 outflow, serendipitously with limited sensitivity and bandwidth. Targeted observations will enable the discovery of new COMs and a more detailed analysis of their emission. Morphological differences between molecules in the IRAS 4B1 outflow lobes and their relative abundances provide first proof that this outflow is a promising new laboratory for shock chemistry, which will offer crucial information on COM formation and destruction as well as outflow structure and kinematics.

Feedback from active galactic nuclei (AGN) is widely acknowledged to regulate the growth of massive galaxies, though its driving mechanisms are debated. Prevailing theories suggest that AGN-driven outflows are driven either by radiation pressure acting directly on the dusty interstellar medium (ISM) or by hot winds entraining cooler ISM gas, but the relative contribution of each mechanism remains uncertain. By combining optical emission line measurements with highly ionized UV emission lines, it is possible to constrain whether the pressure source applied to ionized clouds is primarily radiation or primarily hydrodynamic, and thus constrain the dominant driver. This study presents the first multi-object analysis of far-ultraviolet (FUV) spectra from galactic-scale AGN-driven outflows in obscured quasars, based on Cosmic Origins Spectrograph observations of five low-redshift targets. By comparing narrow-line region UV emission line ratios to theoretical models that vary the importance of the two pressure sources, we find three out of five targets fall within the radiation pressure-dominated regime. A fourth target exhibits intermediate emission-line ratios that suggest radiation pressure and pressure from a hot wind are both dynamically important. Finally, the lowest-luminosity object in our sample may have a dynamically important hot wind component, but non-detections prevent a clear conclusion in this case. These results suggest radiation pressure dominates circum-nuclear narrow-line region cloud dynamics, but pressure from a hot wind also plays a role in some cases. This is consistent with AGN feedback scenarios mediated by radiation pressure or a short-lived hot wind phase that dissipates after initially accelerating outflows.

We present results from imaging observations of three quadruply lensed quasars by Hubble Space Telescope (HST) and James Webb Space Telescope (JWST) at redshifts $z = 2.550$, 2.975, and 1.500. We model our targets assuming a singular isothermal ellipsoid mass profile and an elliptical SÃ\c{opyright}rsic profile for the lensing galaxies, and reconstruct the geometric configuration of each system with measured Einstein radii of 0.44$'$, 0.58$'$, and 0.49$'$. While no spectroscopic measurements are available for the lenses, we constrain the redshift of each lens to $0.5 < z < 1.2$, $1.0 < z < 1.5$, and $0.4 < z < 0.9$. For all three lenses, the best-fit light model yield a typical de Vaucouleurs $n_{\rm S\acute{e}rsic} \sim 4$ profile and an effective radius $R_e$ around $\sim 1.5 - 3.5$ kpc. We accordingly classify the three lenses as early-type galaxies at an intermediate to high redshift, a common type for strong lensing galaxies. Compared to other known quadruple lenses, the lensing galaxies in this work are at the lower end of the distribution of Einstein radii and upper end of the distribution of the lens redshifts. They represent an interesting quadrant of subarcsecond-separation lenses in the population of single-galaxy strong lensing which have been largely unexplored yet and will be great targets of interest in upcoming high-resolution lensing surveys.

OB stars generally form in open clusters within the Milky Way's thin disk, so when they are found at high Galactic latitudes, it is thought that they were ejected from their birth clusters during the past few tens of millions of years. Using Gaia Data Release 3 (hereafter DR3) data, we traced the kinematic trajectories of 39 high-latitude B-type stars and 447 Galactic open clusters with high-quality astrometry to search for moments of past intersection. In cases where we found matching trajectories, we also considered the clusters' HR diagrams to confirm parent-orphan pairs have matching ages. Further analysis of the clusters' core environments allowed us to determine a probable ejection mechanism. Through these paternity tests, we have identified possible origins for five of these orphaned B-type stars. Here we present the likely travel times, ejection velocities, and a discussion of the runaway mechanism for each case. We also identify one star whose trajectory did not bring it near the disk during the time period of our analysis, and we discuss its possible origins as well.

We perform the first systematic study of the minute-to-hours-timescale stochastic variability observed in the X-ray luminosity of tidal disruption events (TDEs) using XMM-Newton data and Fourier analysis methods. We measure the spectral properties, power spectral densities (PSDs), fractional variability amplitudes, and energy dependence of the variability for 18 TDEs spanning 54 observations, of which 27 occur in thermal disk-dominated states and 27 show a nonthermal hard X-ray corona. Compared to pure thermal sources, we find TDEs with coronae are more X-ray variable and show steeper PSDs indicating longer correlation timescales. This state-transition behavior is qualitatively similar to X-ray binaries, which show higher fractional variability in the hard state than in the soft state. However, newborn TDE coronae show systematically flatter PSDs and softer energy spectra than their long-lived AGN counterparts. We also show that the variability amplitude of thermal TDEs increases with photon energy, consistent with variations sourced by local temperature fluctuations and exponentially enhanced in the Wien tail. Our work demonstrates that combining spectral and timing properties of X-ray TDEs can probe the microphysics of newly formed accretion flows around supermassive black holes, and that the coronae formed in TDEs fundamentally differ from those in AGN.

Neutrino emission from astrophysical sources has long been considered a signature of cosmic-ray acceleration. The IceCube neutrino observatory has observed a diffuse flux of TeV-PeV neutrinos, but very few confirmed sources have emerged. With the recent publication of IceCube Event Catalog (IceCat-1), IceCube has released a list of the most promising astrophysical neutrino events since May 2011. Using the archival data from the High Altitude Water Cherenkov (HAWC) $γ$-ray observatory, we perform a coincidence search for gamma rays and neutrinos using a Bayesian Block algorithm with the public IceCube alerts from IceCat-1, along with additional alerts issued later. In this work, we consider 368 alerts, up to July 8, 2025, that are within HAWC's field of view. We observe approximately a 5\% coincident detection rate, which is consistent with expectations from background. Two of these detections contain the Active Galactic Nuclei (AGN) Markarian 421 and Markarian 501. We discuss the likelihood that the neutrino/$γ$-ray coincidences are false positives and a brief overview of the results.

We propose and investigate a class of dynamical dark energy models in which the cosmological constant evolves from negative values in the early Universe to a positive value at low redshifts. This framework includes a generalised ladder-step evolution, as well as smooth-transition scenarios, providing a unified description of sign-changing cosmological constants. We analyse the theoretical construction and background dynamics of these models using cosmographic diagnostics. Extending this study to the linear perturbation regime, we solve the perturbation equations from the radiation-dominated era with adiabatic initial conditions. We examine the evolution of the matter density contrast, gravitational potential, and the $fσ_8$ observable. Our results are compared against the standard $Λ$CDM model and confronted with current observational data, illustrating the phenomenological viability of sign-changing dark energy models and revealing distinctive imprints on cosmic structure formation arising from the transition of the cosmological constant.

Neutron stars provide a unique environment to probe the properties of dense nuclear matter. In this work, we present a comparative study between two approaches to modeling the neutron star structure: a Chiral Effective Field Theory based approach and the MUSES Calculation Engine framework, which uses three different approaches for the three density regions. We analyze the resulting mass-radius relations, discussing the respective advantages and limitations of the two methods.

Modern cosmology and galaxy formation rely on an understanding of how cosmic baryons are distributed, a significant portion of which exist in the diffuse gas confined to halos. Fast Radio Bursts (FRBs) are a promising probe of the Universe's ionized gas. At low redshift, the contribution to the dispersion measure (DM) from the intergalactic medium (IGM) and intervening halos is subdominant, allowing us to study the circumgalactic media (CGM) of the host galaxies. We select a sample of five local universe FRBs whose host interstellar medium (ISM) DM is negligible and use these to constrain the mass of the CGM in each halo. We find that one of our sources, the only massive elliptical host galaxy, has been evacuated of its baryons ($M_\mathrm{gas}=0.02^{+0.02}_{-0.02}M_\mathrm{h}$, corresponding to $\sim$10$\%$ of the cosmological average $\frac{Ω_b}{Ω_m}$). This galaxy shows evidence of a past episode of AGN activity, consistent with the picture of strong AGN feedback in galaxy group-scale halos. The other sources are consistent with existing multiwavelength data and tentatively support more baryon retention in $L_*$ galaxies compared to group-scale halos. We show that FRBs can measure the halo gas fraction $f_\mathrm{gas}$ in halos of mass $M_\mathrm{h}\sim10^{11-13}M_\odot$, and up to $\sim10^{14}M_\odot$ if galaxy cluster hosts are included, which is a larger range than other gas probes can access. Finally, we demonstrate that a large sample of local universe FRBs, such as those expected from upcoming all-sky radio telescopes, will enable precision measurements of halo gas, which is crucial for understanding the effects of feedback.

We use a new sample of low-mass ($M_* < 10^9$ $M_\odot$) isolated galaxies from the Exploration of Local VolumE Survey - Field (ELVES-Field) to examine the star formation properties and sizes of field dwarf galaxies in the Local Volume (LV; $D<10$ Mpc). This volume-limited sample was selected from nearly 3,000 square degrees of imaging, relying on surface brightness fluctuations to determine distances to the majority of the systems and is complete to $M_* \approx 10^6$ $M_\odot$. Across the surveyed area, we catalog over 2300 candidate LV dwarfs, of which we confirm 95 as genuine LV members and reject over 1600 as background contaminants, with the remaining 600 candidates still requiring a distance measurement. Of the confirmed LV dwarfs, 46 are either new discoveries or confirmed via a distance measurement for the first time here. We explore different environmental criteria to select isolated dwarfs but primarily focus on dwarfs that are $>2\times R_{\mathrm{vir}}$ in projection from any known group with $M_\star > 10^9$ $M_\odot$. We find that, at higher dwarf masses ($M_\star \gtrsim 10^7$ $M_\odot$), essentially all field dwarfs are star-forming as has been found before. In contrast, at $M_\star \lesssim 10^7$ $M_\odot$, $\sim30\%$ of field dwarfs appear to be quenched. Finally, we find that isolated dwarfs are noticeably smaller ($\sim 20\%$) than satellite dwarfs of the same stellar mass, regardless of quenched status.

We report the first direct measurement of the period derivative in a quasi-periodic eruption (QPE), finding a smoothly increasing period with $\dot{P}\approx (1.7\pm 0.02)\times10^{-2}$ d d$^{-1}$ in the source ZTF19acnskyy/"Ansky". Most models for QPEs invoke repeated interactions of a stellar-mass orbiting companion around the supermassive black hole (SMBH) in an extreme mass-ratio inspiral (EMRI). In these scenarios, a positive $\dot{P}$ is surprising, but not impossible to produce. We explore several possible explanations for the observed $\dot{P}$, including stable mass-transfer driven by impulsive mass loss events in an EMRI, velocity kicks at pericenter due to tidal interactions with the SMBH, apparent period changes due either to general relativistic precession effects in an EMRI or light travel-time delays in a hierarchical SMBH binary, and mass-transfer variations in a thermal/viscous disk instability model. We find that none of the considered models provides a complete explanation for the data, motivating further work on physical explanations for positive period derivatives in QPEs.

The nascent methodology of applying the principles of causal discovery to astrophysical data has produced affirming results about deeply held theories concerning the causal nature behind the observed coevolution of supermassive black holes (SMBHs) with their host galaxies. The key results from observations have demonstrated an apparent causal reversal across different galaxy morphologies$\unicode{x2014}$SMBHs causally influence the evolution of the physical parameters of their spiral galaxy hosts, whereas SMBHs in elliptical galaxies are passive companions that grow in near lockstep with their hosts. To further explore and ascertain insights, it is necessary to utilize galaxy simulations to track the time evolution of the observed causal relations to learn more about the temporal nature of the changing SMBH/galaxy evolutionary directions. We conducted experiments with the NIHAO suite of cosmological zoom-in hydrodynamical simulations to follow the evolution of individual galaxies along with their central SMBH masses ($M_\unicode{x25CF}$) and properties, including central stellar velocity dispersion ($σ_0$). We reproduce the causal results from real galaxies, but add clarity by observing that the SMBH/galaxy causal directions are noticeably inverted between the epochs before and after the peak of star formation. The implications for causal reversal of the $M_\unicode{x25CF}\unicode{x2013}σ_0$ relation portend larger concerns about the reliability of SMBH masses estimated at high redshifts and presumptions of overmassive black holes at early epochs. Toward this problem, we apply updated causally-informed scaling relations that predict high-$z$ black hole masses that are approximately two orders of magnitude less massive, and thus not overmassive with respect to local $z=0$ SMBH$\unicode{x2013}$galaxy mass ratios.

We present the results of the Exploration of Local VolumE Survey - Field (ELVES-Field), a survey of the dwarf galaxies in the Local Volume (LV; $D<10$ Mpc) over roughly $3,000$ square degrees, focusing on the field dwarf population. Candidates are detected using a semi-automated algorithm tailored for low surface brightness dwarfs. Using tests with injected galaxies, we show the detection is $50\%$ complete to $m_g\sim20$ mag and $M_\star \sim 10^6$ $M_\odot$. Candidates are confirmed to be true nearby dwarfs through distance measurements including redshift, tip of the red giant branch, and surface brightness fluctuations. We identify isolated, field dwarfs using various environmental criteria. Over the survey footprint, we detect and confirm 95 LV dwarfs, 44 of which we consider isolated. Using this sample, we infer the field dwarf mass function and find good agreement at the high-mass end with previous redshift surveys and with the predictions of the IllustrisTNG simulation. This sample of isolated, field dwarfs represents a powerful dataset to investigate aspects of small-scale structure and the effect of environment on dwarf galaxy evolution.

Cosmological analyses are moving past the well understood 2-point statistics to extract more information from cosmological fields. A natural step in extending inference pipelines to other summary statistics is to include higher order N-point correlation functions (NPCFs), which are computationally expensive and difficult to model. At the same time it is unclear how many NPCFs one would have to include to reasonably exhaust the cosmological information in the observable fields. An efficient alternative is given by learned and optimized summary statistics, largely driven by overparametrization through neural networks. This, however, largely abandons our physical intuition on the NPCF formalism and information extraction becomes opaque to the practitioner. We design a simulation-based inference pipeline, that not only benefits from the efficiency of machine learned summaries through optimization, but also holds on to the NPCF program. We employ the heavily constrained Cosmological Correlator Convolutional Neural Network (C3NN) which extracts summary statistics that can be directly linked to a given order NPCF. We present an application of our framework to simulated lensing convergence maps and study the information content of our learned summary at various orders in NPCFs for this idealized example. We view our approach as an exciting new avenue for physics-informed simulation-based inference.

Rapid localisation and follow-up of gamma-ray bursts (GRBs) increasingly rely on low-latency triggers from new missions coupled to wide-field robotic optical facilities. We present the discovery and multi-wavelength follow-up of GRB 250818B, detected by the Space Variable Objects Monitor (SVOM) and localised optically by the Gravitational-wave Optical Transient Observer (GOTO). We compile and homogenise X-ray, optical/NIR, and radio data to build broadband light curves and spectral energy distributions. The afterglow is unusually luminous for a nominal short GRB, lying on the bright end of the short-GRB population in X-rays and optical and among the most luminous high-redshift short-GRB afterglows in the radio. MeerKAT detects the source at 3.1 GHz, while ALMA provides deep higher-frequency limits. Keck/LRIS spectroscopy shows continuum and metal absorption (Fe II, Mg II, Mg I), giving $z=1.216$. Synchrotron forward-shock modelling favours a constant-density medium and strongly prefers refreshed (energy-injection) emission, well described by a two-component jet with $E_{K,iso} \sim 4\times10^{52}$ erg, $n_0 \sim 3.6$ cm$^{-3}$, $θ_j \simeq 0.10$ rad ($\sim 5.7$ deg), and $p \simeq 1.64$. The host association is ambiguous: the nearest LS DR10 galaxy candidate ($r_{AB} \sim 24.7$) is offset by $\sim 4$ arcsec ($\sim 34$ kpc) with chance-alignment probability $P_{cc} \sim 0.2$, and current imaging does not exclude a fainter, near-coincident host. SED fitting of the candidate host suggests a low-mass galaxy. GRB 250818B highlights the power of rapid wide-field counterpart identification in the SVOM era, while host-association uncertainty can still limit offset-based interpretation.

We use the high angular resolution NIRCam images from the \textit{PDRs4All} program, combined with those of GTO program 1256, to extract key properties of disks in the Orion Nebula Cluster. We measure disk radii in silhouette against the bright background, identify dissociation fronts (DFs) and ionization fronts (IFs), determine Paschen $α$ intensities, and derive near-infrared spectral energy distributions (SEDs). From these diagnostics we define a typology of ONC disks. \textit{Type I} sources show merged IFs and DFs close to the disk surface. \textit{Type II} sources have DFs at the disk surface and IFs located tens of astronomical units away. \textit{Type III} sources show a DF at the disk surface but no IF. For all types, PAH emission traces the PDR. We find that the disk radius $r{\rm disk}$ increases with projected distance to the ionizing source $d{\rm proj}$, following $r{\rm disk} \propto d{\rm proj}^{0.30}$, consistent with disk truncation by photoevaporation. Disk radii measured in the infrared are larger than those measured at millimeter wavelengths, suggesting radial dust segregation within the disks. In agreement with PDR models, the thermal pressure in the disk PDR increases with the FUV radiation field $G_0$, but with a flatter slope. Finally, the SEDs of candidate Jupiter Mass Binary Objects (JuMBOs) are similar to those of \textit{Type III} disks, except for JuMBO24, which resembles a \textit{Type I} or \textit{Type II} source. Its SED is consistent with a young low-mass binary hosting an unresolved ionized disk.